The USPSTF recommends biennial screening mammography.
These recommendations do not apply to persons who have a genetic marker or syndrome associated with a high risk of breast cancer (eg, or genetic variation), a history of high-dose radiation therapy to the chest at a young age, or previous breast cancer or a high-risk breast lesion on previous biopsies.
The USPSTF recognizes that clinical decisions involve more considerations than evidence alone. Clinicians should understand the evidence but individualize decision-making to the specific patient or situation.
Table of Contents | PDF Version and JAMA Link | Archived Versions |
---|---|---|
, select . To read the evidence summary in , select . To read the modeling study in , select . |
The US Preventive Services Task Force (USPSTF) makes recommendations about the effectiveness of specific preventive care services for patients without obvious related signs or symptoms to improve the health of people nationwide.
It bases its recommendations on the evidence of both the benefits and harms of the service and an assessment of the balance. The USPSTF does not consider the costs of providing a service in this assessment.
The USPSTF recognizes that clinical decisions involve more considerations than evidence alone. Clinicians should understand the evidence but individualize decision-making to the specific patient or situation. Similarly, the USPSTF notes that policy and coverage decisions involve considerations in addition to the evidence of clinical benefits and harms.
The USPSTF is committed to mitigating the health inequities that prevent many people from fully benefiting from preventive services. Systemic or structural racism results in policies and practices, including health care delivery, that can lead to inequities in health. The USPSTF recognizes that race, ethnicity, and gender are all social rather than biological constructs. However, they are also often important predictors of health risk. The USPSTF is committed to helping reverse the negative impacts of systemic and structural racism, gender-based discrimination, bias, and other sources of health inequities, and their effects on health, throughout its work.
Among all US women, breast cancer is the second most common cancer and the second most common cause of cancer death. In 2023, an estimated 43,170 women died of breast cancer. 1 Non-Hispanic White women have the highest incidence of breast cancer (5-year age-adjusted incidence rate, 136.3 cases per 100,000 women) and non-Hispanic Black women have the second highest incidence rate (5-year age-adjusted incidence rate, 128.3 cases per 100,000 women). 2 Incidence gradually increased among women aged 40 to 49 years from 2000 to 2015 but increased more noticeably from 2015 to 2019, with a 2.0% average annual increase. 3 Despite having a similar or higher self-reported rate of mammography screening, 4 Black women are more likely to be diagnosed with breast cancer beyond stage I than other racial and ethnic groups, are more likely to be diagnosed with triple-negative cancers (ie, estrogen receptor–negative [ER–], progesterone receptor–negative [PR–], and human epidermal growth factor receptor 2–negative [HER2–]), which are more aggressive tumors, compared with White women, 5 and are approximately 40% more likely to die of breast cancer compared with White women. 6
The USPSTF concludes with moderate certainty that biennial screening mammography in women aged 40 to 74 years has a moderate net benefit .
The USPSTF concludes that the evidence is insufficient to determine the balance of benefits and harms of screening mammography in women 75 years or older.
The USPSTF concludes that the evidence is insufficient to determine the balance of benefits and harms of supplemental screening for breast cancer with breast ultrasound or MRI, regardless of breast density.
See Table 1 for more information on the USPSTF recommendation rationale and assessment. For more details on the methods the USPSTF uses to determine the net benefit, see the USPSTF Procedure Manual. 7
These recommendations apply to cisgender women and all other persons assigned female at birth (including transgender men and nonbinary persons) 40 years or older at average risk of breast cancer. This is because the net benefit estimates are driven by sex (ie, female) rather than gender identity, although the studies reviewed for this recommendation generally used the term “women.” These recommendations apply to persons who have factors associated with an increased risk of breast cancer, such as a family history of breast cancer (ie, a first-degree relative with breast cancer) or having dense breasts. They do not apply to persons who have a genetic marker or syndrome associated with a high risk of breast cancer (eg, BRCA1 or BRCA2 genetic variation), a history of high-dose radiation therapy to the chest at a young age, or previous breast cancer or a high-risk breast lesion on previous biopsies. Of note, the USPSTF has a separate recommendation on risk assessment, genetic counseling, and genetic testing for BRCA-related cancer, 8 and family history is a common feature of risk assessment tools that help determine likelihood of BRCA1 or BRCA2 genetic variation.
Both digital mammography and digital breast tomosynthesis (DBT, or “3D mammography”) are effective mammographic screening modalities. DBT must be accompanied by traditional digital mammography or synthetic digital mammography, which is a 2-dimensional image constructed from DBT data; 9 , 10 hereafter, references to DBT will imply concurrent use with digital mammography or synthetic digital mammography. In general, studies have reported small increases in positive predictive value with DBT compared with digital mammography. Trials reporting on at least 2 consecutive rounds of screening have generally found no statistically significant difference in breast cancer detection or in tumor characteristics (tumor size, histologic grade, or node status) when comparing screening with DBT vs digital mammography. 4
The Breast Cancer Surveillance Consortium (BCSC) is a network of 6 active breast imaging registries and 2 historic registries, providing a large observational database related to breast cancer screening. 11 Collaborative modeling, using inputs from BCSC data, suggests similar benefits and fewer false-positive results with DBT compared with digital mammography. 12 , 13
Available evidence suggests that biennial screening has a more favorable trade-off of benefits vs harms than annual screening. BCSC data showed no difference in detection of cancers stage IIB or higher and cancers with less favorable prognostic characteristics with annual vs biennial screening interval for any age group, 14 and modeling data estimate that biennial screening has a more favorable balance of benefits to harms (eg, life-years gained or breast cancer deaths averted per false-positive result) compared with annual screening. 12
Breast cancer treatment regimens are highly individualized according to each patient’s clinical status, cancer stage, tumor biomarkers, clinical subtype, and personal preferences. 15 Ductal carcinoma in situ (DCIS) is a noninvasive condition with abnormal cells in the breast duct lining with uncertainty regarding its prognostic significance. Consequently, there is clinical variability in the treatment approach when DCIS is identified at screening. It is unknown what proportion of screen-detected DCIS represents overdiagnosis (ie, a lesion that would not have led to health problems in the absence of detection by screening). In general, DCIS treatment, which may include surgery, radiation, and endocrine treatment, is intended to reduce the risk for future invasive breast cancer.
Mortality from breast cancer is highest for Black women, even when accounting for differences in age and stage at diagnosis; mortality is approximately 40% higher for Black women (5-year age-adjusted mortality rate, 27.6 per 100,000 women) compared with White women (5-year age-adjusted mortality rate, 19.7 per 100,000 women). 6 While the underlying causes of this disparity are complex, the National Institute of Minority Health and Disparities has developed a framework that recognizes multiple determinants, including the health care system, the sociocultural and built environments, behavioral factors, and genetic factors, that can contribute to health inequities. 16 Inequities in breast cancer mortality can be examined at each step along the cancer screening, diagnosis, treatment, and survival pathway with these factors in mind. The higher mortality rate for Black women diagnosed with breast cancer in the US aligns with other health inequities that are attributed to the effects of structural racism, which include inequalities in resources, harmful exposures, and access to and delivery of high-quality health care. 17-19 Racial and economic residential segregation driven by discriminatory housing policies has been associated with increased exposure to toxic environments such as air pollution, industrial waste, and built environments that do not support health, and stressful life conditions. Residential segregation has also been associated with both an increased risk of triple-negative breast cancer and poorer breast cancer–specific survival in Black women. 20-22
Black women have a higher incidence of breast cancer with at least 1 negative molecular marker, and the incidence of triple-negative cancers (ie, ER–, PR–, and HER2–) is twice as high in Black women compared with White women (24.2 vs 12.3 cases per 100,000 women). 5 The higher incidence of negative hormonal receptor status leads to worse outcomes because these subtypes are less readily detected through screening and less responsive to current therapy, 23 and triple-negative cancers are more likely to be aggressive and diagnosed at later stages than other subtypes. It is important to note that observed regional differences in the incidence of hormonal receptor–negative cancer within and between racial groups suggest that environmental factors and social determinants of health, including racism, are largely responsible for the differential risk of developing hormonal receptor–negative cancer. 24 , 25 Although variation in the incidence of cancer subtypes explains some of the differences in breast cancer mortality, racial differences in mortality within subtypes point to barriers to obtaining high-quality health care and disparities in screening follow-up and treatment initiation as contributors. 24
Of note, Black women have a rate of self-reported mammography screening similar to or higher than that for all women (84.5% vs 78%, respectively, in the past 2 years), based on 2020 data. 4 However, benefits from mammography screening require initiation and completion of appropriate and effective follow-up evaluation and treatment. Both screening and guideline-concordant treatment are essential for reducing breast cancer mortality, 26 highlighting the importance of timely and effective treatment at the earliest stage of diagnosis. Delays and inadequacies in the diagnostic and treatment pathway downstream from screening likely contribute to increased mortality compared with women receiving prompt, effective care.
Disparities in follow-up after screening and treatment have been observed for Asian, Black, and Hispanic women. 27-36 Adjuvant endocrine therapy reduces the risk of cancer recurrence among individuals with hormonal receptor–positive cancers, but long-term adherence can be difficult. Black women are more likely to discontinue adjuvant endocrine therapy compared with White women, in part due to greater physical (vasomotor, musculoskeletal, or cardiorespiratory) and psychological (distress or despair) symptom burdens. 35 , 36 Improvements in access to effective health care, removal of financial barriers, and use of support services to ensure equitable follow-up after screening and timely and effective treatment of breast cancer have the potential to reduce mortality for individuals experiencing disparities related to racism, rural location, 37 low income, or other factors associated with lower breast cancer survival.
Potential preventable burden.
Breast cancer incidence increases with age and peaks among persons aged 70 to 74 years, although rates in persons 75 years or older remain high (453.3 and 409.9 cases per 100,000 women aged 75 to 79 and 80 to 84 years, respectively, compared with 468.2 cases per 100,000 women aged 70 to 74 years), and mortality from breast cancer increases with increasing age. 38 , 39 However, no randomized clinical trials (RCTs) of breast cancer screening included women 75 years or older. 4 Collaborative modeling suggests that screening in women 75 years or older is of benefit, 12 but a trial emulation found no benefit with breast cancer screening in women aged 75 to 84 years. 40 Thus, there is insufficient evidence to recommend for or against screening mammography in women 75 years or older.
In women with dense breasts who have an otherwise normal mammogram result, there is insufficient evidence about the effect of supplemental screening using breast ultrasonography or magnetic resonance imaging (MRI) on health outcomes such as breast cancer morbidity and mortality. Dense breasts are associated with both reduced sensitivity and specificity of mammography and with an increased risk of breast cancer. 41 , 42 However, increased breast density itself is not associated with higher breast cancer mortality among women diagnosed with breast cancer, after adjustment for stage, treatment, method of detection, and other risk factors, according to data from the BCSC. 43
Potential harms of screening mammography include false-positive results, which may lead to psychological harms, 44 additional testing, and invasive follow-up procedures; overdiagnosis and overtreatment of lesions that would not have led to health problems in the absence of detection by screening; and radiation exposure.
Centers for Disease Control and Prevention data show that as of 2015, more than 50% of women 75 years or older reported having a mammogram within the past 2 years. 45 At present, 38 states and the District of Columbia require patient notification of breast density when mammography is performed; in some states, legislation also includes notification language informing women that they should consider adjunctive screening. 46 Starting in September 2024, the US Food and Drug Administration will require mammography centers to notify patients of their breast density, inform them that dense breast tissue increases the risk of breast cancer and makes it harder to detect on a mammogram, and that other imaging tests may help to find cancer. 47
The National Cancer Institute has information on breast cancer screening for health care professionals ( https://www.cancer.gov/types/breast/hp/breast-screening-pdq ) and for patients ( https://www.cancer.gov/types/breast/patient/breast-screening-pdq ).
The Centers for Disease Control and Prevention has information on breast cancer screening ( https://www.cdc.gov/cancer/breast/basic_info/screening.htm ).
The USPSTF has made recommendations about the use of medications to reduce women’s risk for breast cancer 48 as well as risk assessment, genetic counseling, and genetic testing for BRCA1 - or BRCA2 -related cancer. 8
This recommendation updates the 2016 recommendation on breast cancer screening. In 2016, the USPSTF recommended biennial screening mammography for women aged 50 to 74 years and individualizing the decision to undergo screening for women aged 40 to 49 years, based on factors such as individual risk and personal preferences and values. The USPSTF concluded that the evidence was insufficient to assess the benefits and harms of DBT as a primary screening method; the balance of benefits and harms of adjunctive screening for breast cancer using breast ultrasonography, MRI, or DBT in women identified to have dense breasts on an otherwise negative screening mammogram; and the balance of benefits and harms of screening mammography in women 75 years or older. 49 For the current recommendation, the USPSTF recommends biennial screening mammography for women aged 40 to 74 years. The USPSTF again finds that the evidence is insufficient to assess the balance of benefits and harms of supplemental screening for breast cancer using breast ultrasonography or MRI in women identified to have dense breasts on an otherwise negative screening mammogram and the balance of benefits and harms of screening mammography in women 75 years or older. Current evidence suggests that both digital mammography and DBT are effective primary screening modalities.
To update its 2016 recommendation, the USPSTF commissioned a systematic review 4 , 50 on the comparative effectiveness of different mammography-based breast cancer screening strategies by age to start and stop screening, screening interval, modality, use of supplemental imaging, or personalization of screening for breast cancer on the incidence of and progression to advanced breast cancer, breast cancer morbidity, and breast cancer–specific or all-cause mortality. To be included in the review, studies needed to report on detection and stage distribution of screen-detected invasive breast cancer over more than 1 round of screening, to allow assessment for evidence of stage shift (as evidence of potential benefit). Studies that reported only performance characteristics of testing (eg, sensitivity and specificity) or only detection rates were not eligible for inclusion. The review also assessed the harms of different breast cancer screening strategies. 4 Evidence from the trials that established breast cancer screening effectiveness with mammography has not been updated, as there are no new studies that include a group that is not screened. Analyses from prior reviews of that evidence were considered foundational evidence for the current recommendation.
In addition to the systematic evidence review, the USPSTF commissioned collaborative modeling studies from 6 CISNET (Cancer Intervention and Surveillance Modeling Network) modeling teams to provide information about the benefits and harms of breast cancer screening strategies that vary by the ages to begin and end screening, screening modality, and screening interval. 12 In alignment with the USPSTF’s commitment to improve health equity, the USPSTF also commissioned modeling studies from 4 CISNET teams that have developed race-specific breast cancer models for Black women, to provide information about the effectiveness and harms of these different screening strategies in Black women. The USPSTF commissions decision modeling to help inform how best to target or implement a clinical preventive service when empirical evidence supports provision of the service. 51 The modeling studies complement the evidence that the systematic review provides.
Given the documented racial disparities in breast cancer outcomes, in addition to commissioning modeling studies specific to Black women, the evidence review included contextual questions on the drivers behind and approaches to address disparities in health outcomes related to breast cancer, particularly the higher mortality in Black women.
Randomized trials that began enrolling participants more than 30 to 40 years ago have established the effectiveness of screening mammography to reduce breast cancer mortality. A meta-analysis conducted in support of the 2016 USPSTF breast cancer screening recommendation found that screening mammography was associated with relative risk (RR) reductions in breast cancer mortality of 0.88 (95% CI, 0.73-1.00; 9 trials) for women aged 39 to 49 years, 0.86 (95% CI, 0.68-0.97; 7 trials) for women aged 50 to 59 years, 0.67 (95% CI, 0.54-0.83; 5 trials) for women aged 60 to 69 years, and 0.80 (95% CI, 0.51-1.28; 3 trials) for women aged 70 to 74 years, 44 and an updated analysis of 3 Swedish screening trials reported a 15% relative reduction in breast cancer mortality for women aged 40 to 74 years (RR, 0.85 [95% CI, 0.73-0.98]). 52 Only 1 of these trials enrolled a significant proportion of Black women. 53 None of the trials nor the combined meta-analysis demonstrated a difference in all-cause mortality with screening mammography. The current USPSTF review focused on the comparative benefits of different screening strategies.
The USPSTF did not identify any RCTs designed to test the comparative effectiveness of different ages to start or stop screening that reported morbidity, mortality, or quality-of-life outcomes. One trial emulation study (n = 264,274), using a random sample from Medicare claims data, estimated the effect of women stopping screening at age 70 years compared with those who continued annual screening after age 70 years. Based on survival analysis, this study reported that continued screening between the ages of 70 and 74 years was associated with a 22% decrease in the risk of breast cancer mortality, compared with a cessation of screening at age 70 years. While collaborative modeling estimated that, compared with a stopping age of 74, screening biennially starting at age 40 years until age 79 years would lead to 0.8 additional breast cancer deaths averted, the trial emulation study found that there was no difference in the hazard ratio or absolute rates of breast cancer mortality with continued screening vs discontinued screening from ages 75 to 79 years or ages 80 to 84 years. 40
Collaborative modeling data estimated that compared with biennial screening from ages 50 to 74 years, biennial screening starting at age 40 years until 74 years would lead to 1.3 additional breast cancer deaths averted (median, 6.7 vs 8.2, respectively, across 6 models) per 1000 women screened over a lifetime of screening for all women ( Table 2 ; note that the 1.3 deaths averted is the median of the differences in each of 6 models, which is not the same as the difference of the medians noted above and in the table). Models also estimated that screening benefits for Black women are similar for breast cancer mortality reduction and greater for life-years gained and breast cancer deaths averted compared with all women. Thus, biennial screening starting at age 40 years would result in 1.8 additional breast cancer deaths averted (median, 9.2 deaths averted for screening from ages 50 to 74 vs 10.7 deaths averted, across 4 models) per 1000 women screened for Black women ( Table 2 ; note that the 1.8 deaths averted is the median of the differences in each of 4 models, which is not the same as the difference of the medians noted above and in the table). 12 Epidemiologic data has shown that the incidence rate of invasive breast cancer for 40- to 49-year-old women has increased an average of 2.0% annually between 2015 and 2019, a higher rate than in previous years. 3 These factors led the USPSTF to conclude that screening mammography in women aged 40 to 49 years has a moderate benefit by reducing the number of breast cancer deaths.
The USPSTF did not identify any randomized trials directly comparing annual vs biennial screening that reported morbidity, mortality, or quality-of-life outcomes. One trial (n = 14,765) conducted in Finland during the years 1985 to 1995 assigned participants aged 40 to 49 years to annual or triennial screening invitations based on birth year (even birth year: annual; odd birth year: triennial) and reported similar mortality from incident breast cancer and for all-cause mortality between the 2 groups, with follow-up to age 52 years. 54
A nonrandomized study using BCSC data (n = 15,440) compared the tumor characteristics of cancers detected following annual vs biennial screening intervals. 14 The relative risk of being diagnosed with a stage IIB or higher cancer and cancer with less favorable characteristics was not statistically different for biennially vs annually screened women in any of the age categories. The risk of a stage IIB or higher cancer diagnosis and of having a tumor with less favorable prognostic characteristics was higher for premenopausal women screened biennially vs annually (RR, 1.28 [95% CI, 1.01-1.63] and RR, 1.11 [95% CI, 1.00-1.22], respectively). However, this study did not conduct formal tests for interaction in the subgroup comparisons and did not adjust for multiple comparisons.
One RCT (n = 76,022) conducted between 1989 and 1996 randomized individuals to annual or triennial screening and reported on breast cancer incidence. The number of screen-detected cancers was higher in the annual screening study group (RR, 1.64 [95% CI, 1.28-2.09]). However, the total number of cancers diagnosed either clinically or with screening was similar after 3 years of screening. Cancers occurring in the annual screening group (including clinically diagnosed cancers) did not differ by prognostic features such as tumor size, node positivity status, or histologic grade compared with those in the triennial screening group. 55
Collaborative modeling estimated that biennial screening results in greater incremental life-years gained and mortality reduction per mammogram and has a more favorable balance of benefits to harms for all women and for Black women, compared with annual screening. While modeling suggests that screening Black women annually and screening other women biennially would reduce the disparity in breast cancer mortality, 12 , 13 trial or observational evidence is lacking that screening any group of women annually compared with biennial screening improves mortality from breast cancer. 4
The USPSTF did not identify any RCTs or observational studies that compared screening with DBT vs digital mammography and reported morbidity, mortality, or quality-of-life outcomes.
Three RCTs 56-58 and 1 nonrandomized study 59 compared detection of invasive cancer over 2 rounds of screening with DBT vs digital mammography. These trials screened all participants with the same screening modality at the second screening round—digital mammography in 2 trials and the nonrandomized study and DBT in 1 trial. Stage shift or differences in tumor characteristics across screening rounds could offer indirect evidence of potential screening benefit. The trials found no statistically significant difference in detection at the second screening round (pooled RR, 0.87 [95% CI, 0.73-1.05]; 3 trials [n = 105,064]). 4 , 50 The nonrandomized study (n = 92,404) found higher detection at round 1 for the group screened with DBT and higher detection at round 2 for the group screened with digital mammography at both rounds. There were no statistically significant differences in tumor diameter, histologic grade, and node status at the first or second round of screening in any of these studies.
Collaborative modeling data estimated that the benefits of DBT are similar to the estimated benefits of digital mammography (eg, approximately 5 to 6 more life-years gained per 1000 women screened). 12 , 13
The USPSTF found no studies of supplemental screening with MRI or ultrasonography, or studies of personalized (eg, risk-based) screening strategies, that reported on morbidity or mortality or on cancer detection and characteristics over multiple rounds of screening. 4 , 50 Collaborative modeling studies did not investigate the effects of screening with MRI or ultrasonography. Modeling generally estimated that the benefits of screening mammography would be greater for persons at modestly increased risk (eg, the risk of breast cancer associated with a first-degree family history of breast cancer). 12 , 13
For this recommendation, the USPSTF also reviewed the harms of screening for breast cancer and whether the harms varied by screening strategy. Potential harms of screening for breast cancer include false-positive and false-negative results, need for additional imaging and biopsy, overdiagnosis, and radiation exposure.
The most common harm is a false-positive result, which can lead to psychological harms such as anxiety or breast cancer–specific worry, 44 as well as additional testing and invasive follow-up procedures without the potential for benefit. Collaborative modeling data estimated that a strategy of screening biennially from ages 40 to 74 years would result in 1376 false-positive results per 1000 women screened over a lifetime of screening ( Table 2 ). 12 , 13
Overdiagnosis occurs when breast cancer that would never have become a threat to a person’s health, or even apparent, during their lifetime is found due to screening. It is not possible to directly observe for any individual person whether they have or do not have an overdiagnosed tumor; it is only possible to indirectly estimate the frequency of overdiagnosis that may occur across a screened population. Estimates of the percentage of cancers diagnosed in a study that represent overdiagnosed cancers from RCTs that had comparable groups at baseline, had adequate follow-up, and did not provide screening to the control group at the end of the trial range from approximately 11% to 19%. 4 , 50 Collaborative modeling data estimate that a strategy of screening biennially from ages 40 to 74 years would lead to 14 overdiagnosed cases of breast cancer per 1000 persons screened over the lifetime of screening ( Table 2 ), although with a very wide range of estimates (4 to 37 cases) across models. 12 , 13
One trial emulation (n = 264,274) compared discontinuation of mammography screening at age 70 years or older with continued annual screening beyond this age.40 Overall, the 8-year cumulative risk of a breast cancer diagnosis was higher for the continued annual screening strategy after age 70 years (5.5% overall; 5.3% in women aged 70-74 years; 5.8% in women aged 75-84 years) compared with the stop screening strategy (3.9% overall; same proportion for both age groups). Fewer cancers were diagnosed under the stop screening strategy (ages 70-84 years), resulting in a lower risk of undergoing follow-up and treatment. For women aged 75 to 84 years, additional diagnoses did not contribute to a difference in the risk of breast cancer mortality, likely due to competing causes of death, raising the possibility that the additionally diagnosed cancers represent overdiagnosis.
Collaborative modeling data estimated that lowering the age to start screening to 40 years from 50 years would result in about a 60% increase in false-positive results, and 2 additional overdiagnosed cases of breast cancer (range, 0 to 4) per 1000 women over a lifetime of screening ( Table 2 ). 12 , 13
Rates of interval cancers (cancer diagnosis occurring between screening) reported in screening studies reflect a combination of cancers that were missed during previous screening examinations (false-negative results) and incident cancers emerging between screening rounds. Evidence from studies comparing various intervals and reporting on the effect of screening interval on the rate of interval cancers is mixed. One RCT comparing annual vs triennial screening reported that the rate of interval cancers was significantly lower in the annual invitation group (1.84 cases per 1000 women initially screened) than in the triennial invitation group (2.70 cases per 1000 women initially screened) (RR, 0.68 [95% CI, 0.50-0.92]), 55 while a quasi-randomized study, also comparing annual vs triennial screening, found no difference in the number of interval cancers between the 2 groups. 54
Based on 2 studies, false-positive results were more likely to occur with annual screening compared with longer intervals between screening. 60 , 61 One of these studies, using data from the BCSC, reported that biennial screening led to a 5% absolute decrease in the 10-year cumulative false-positive biopsy rate compared with annual screening, whether screening was conducted with DBT or digital mammography. 60 Collaborative modeling estimated that annual screening results in more false-positive results and breast cancer overdiagnosis. For example, a strategy of screening annually from ages 40 to 74 years would result in about 50% more false-positive results and 50% more overdiagnosed cases of breast cancer compared with biennial screening for all women and a similar increase in false-positive results and a somewhat smaller increase in overdiagnosed cases for Black women. 12 , 13
Three RCTs did not show statistically significant differences in the risk of interval cancer following screening with DBT or digital mammography (pooled RR, 0.87 [95% CI, 0.64-1.17]; 3 trials [n = 130,196]). 4 , 50 Five nonrandomized studies generally support the RCT findings. Three of the nonrandomized studies found no significant difference in the rate of interval cancers diagnosed following screening with DBT or digital mammography, 59 , 62 , 63 while 1 study found a slight increased risk with DBT screening 64 and 1 study found an unadjusted decreased risk with DBT screening. 65
A pooled analysis of 3 RCTs (n = 105,244) comparing screening with DBT vs digital mammography did not find a difference in false-positive results at the second round of screening. 4 , 50 A nonrandomized study using BCSC data reported that the estimated cumulative probability of having at least 1 false-positive result over 10 years of screening was generally lower with DBT screening compared with digital mammography screening (annual screening: 10-year cumulative probability of a false-positive result was 49.6% with DBT and 56.3% with digital mammography; biennial screening: 10-year cumulative probability of a false-positive result was 35.7% for DBT and 38.1% for digital mammography). The risk of having a biopsy over 10 years of screening was slightly lower when comparing annual screening with DBT vs digital mammography but did not differ between DBT and digital mammography for biennial screening (annual screening: 10-year cumulative probability of a false-positive biopsy was 11.2% with DBT and 11.7% with digital mammography; biennial screening: 10-year cumulative probability of a false-positive biopsy was 6.6% for DBT and 6.7% for digital mammography). When results were stratified by breast density, the difference in false-positive result probability with DBT vs digital mammography was largest for women with nondense breasts and was not significantly different among women with extremely dense breasts. 60 Collaborative modeling, using inputs from BCSC data, estimated that screening women aged 40 to 74 years with DBT would result in 167 fewer false-positive results (range, 166-169) per 1000 persons screened, compared with digital mammography. 12 , 13
In the 3 RCTs cited above, rates of DCIS detected did not differ between persons screened with DBT and digital mammography. 56-58
Screening with DBT includes evaluation of 2-dimensional images, generated either with digital mammography or using a DBT scan to produce a synthetic digital mammography image. 9 , 10 Studies using DBT with digital mammography screening reported radiation exposure approximately 2 times higher compared with the digital mammography–only control group. 56 , 58 , 66 Differences in radiation exposure were smaller in studies using DBT/synthetic digital mammography compared with digital mammography. 67 , 68
The DENSE RCT, which compared invitation to screening with digital mammography plus MRI compared with digital mammography alone in participants aged 50 to 75 years with extremely dense breasts and a negative mammogram result, reported a significantly lower rate of invasive interval cancers—2.2 cases per 1000 women invited to screening with digital mammography plus MRI, compared with 4.7 cases per 1000 women invited to screening with digital mammography only (RR, 0.47 [95% CI, 0.29-0.77]). 69
In that trial, the rate of recall among participants who underwent additional imaging with MRI was 94.9 per 1000 screens, the false-positive rate was 79.8 per 1000 women screened, and the rate of biopsy was 62.7 per 1000 women screened. 70 In a nonrandomized study using US insurance claims data, individuals who had an MRI compared with those receiving only a mammogram were more likely in the subsequent 6 months to have additional cascade events related to extramammary findings (adjusted difference between groups, 19.6 per 100 women screened [95% CI, 8.6-30.7]), mostly additional health care visits. 71
In an RCT comparing screening with digital mammography plus ultrasonography vs digital mammography alone conducted in persons aged 40 to 49 years and not specifically among persons with dense breasts, the interval cancer rates reported were not statistically significantly different between the 2 groups (RR, 0.58 [95% CI, 0.31-1.08]); 72 similarly, in a nonrandomized study comparing digital mammography plus ultrasonography vs digital mammography alone using BCSC data, there was no difference in interval cancers (adjusted RR, 0.67 [95% CI, 0.33-1.37]), 73 although in both studies the confidence intervals were wide for this uncommon outcome. In the BCSC analysis, the rates of referral to biopsy and false-positive biopsy recommendations were twice as high and short interval follow-up was 3 times higher for the group screened with ultrasonography. 73
A draft version of this recommendation statement was posted for public comment on the USPSTF website from May 9, 2023, to June 6, 2023. The USPSTF received many comments on the draft recommendation and appreciates all the thoughtful views and perspectives that were shared. Many comments agreed with the draft recommendation. Several comments suggested that there should be no upper age limit for breast cancer screening or that an upper age should be based on life expectancy. In response, the USPSTF notes that no trials of breast cancer screening enrolled women 75 years or older and an emulated trial showed no benefit to screening women aged 75 to 79 or 80 to 84. Some comments suggested that breast cancer screening should start prior to age 40 years, either for all women or for women who are at increased risk of breast cancer. Relatedly, some comments expressed that risk-based screening should be recommended. In response, the USPSTF would like to reiterate that no trials of breast cancer screening enrolled women younger than 39 years. Additionally, the USPSTF found no evidence on the benefits or harms of individualized breast cancer screening based on risk factors. Several randomized trials of risk-based screening are underway (eg, the WISDOM trial) that may provide valuable information regarding this question.
Several comments expressed that breast cancer screening should be recommended annually. In response, the USPSTF would like to reiterate that it did not identify any randomized trials directly comparing annual vs biennial screening. Two trials conducted in the 1980s to 1990s reported no difference in breast cancer mortality or breast cancer features such as tumor size, node positivity status, or histologic grade when comparing annual vs triennial screening. The USPSTF considers both the benefits and harms of different screening intervals and notes that the modeling studies commissioned to support this recommendation found that biennial screening results in greater life-years gained and mortality reduction per mammogram and has a more favorable balance of benefits to harms compared with annual screening.
Many comments requested that the USPSTF recommend supplemental screening with MRI or ultrasound for women with dense breasts. Some comments expressed that this would improve health outcomes, while other comments requested this recommendation so that supplemental screening would be covered by insurance. In response, the USPSTF wants to restate that it found insufficient evidence on the effects of supplemental screening on health outcomes. No studies of supplemental screening reported on health outcomes or on the incidence of and progression to advanced breast cancer over more than 1 round of screening. The USPSTF wants all women to be able to get the care they need and would like to clarify that the I statement is not a recommendation for or against supplemental screening in women with dense breasts. It fundamentally means that there is insufficient evidence to assess the balance of benefits and harms, or to recommend for or against supplemental screening, and that women should talk with their clinicians about what is best given their individual circumstances. The USPSTF is also calling for more research to help close this important evidence gap.
Some comments requested clarification of the patient population included in this recommendation, particularly as it relates to women with a family history of breast cancer or those with a genetic predisposition to increased breast cancer risk. In response, the USPSTF clarified that this recommendation applies to women with a family history of breast cancer but not those who have a genetic marker or syndrome or chest radiation exposure at a young age associated with a high risk of breast cancer. The USPSTF also clarified that it has an existing recommendation on risk assessment, genetic counseling, and genetic testing for BRCA-related cancer.
Some comments expressed that racial and ethnic disparities in breast cancer outcomes, especially in Black women, need to be comprehensively addressed. Related comments expressed that the higher breast cancer mortality that Black women experience is primarily related to their not receiving follow-up evaluation and treatment of the same timeliness and quality as White women, and that starting screening at age 40 years will not remedy this inequity. The USPSTF agrees that mitigating disparities in breast cancer mortality is crucial and highlights these disparities in the Disparities in Breast Cancer Outcomes and Implementation Considerations section of this recommendation statement. The USPSTF also agrees that improvements across the entire spectrum of breast cancer care are needed to reduce mortality for individuals experiencing disparities associated with lower breast cancer survival. For this recommendation, current evidence shows that screening for breast cancer starting at age 40 years will be of significant benefit to Black women. The USPSTF is also calling for more research to understand the underlying causes of why Black women are more likely to be diagnosed with breast cancers that have biomarker patterns that confer greater risk for poor health outcomes, to understand the causes of and ways to mitigate the higher mortality from breast cancer that Black women experience.
Some comments disagreed with the USPSTF B recommendation for screening women between the ages of 40 and 49 years, questioned the evidence to support this, or expressed that the current recommendation downplays the harms of screening. In response, the USPSTF has clarified that it uses modeling to complement trial and observational evidence when there is empirical (ie, trial) evidence of the benefit of a preventive service on health outcomes, as there is for breast cancer screening. Decision modeling can assist the USPSTF in assessing the magnitude of the benefits and harms of different screening strategies. The USPSTF carefully weighs both the benefits and harms of a preventive service as it makes its recommendations and currently concludes, as it has in the past, that the benefits of breast cancer screening outweigh the harms for women between the ages of 40 and 49 years. The most recent epidemiologic data reviewed by the USPSTF show greater incidence of breast cancer at younger ages, and decision modeling shows a greater magnitude of benefit for screening women between the ages of 40 and 49 years. The USPSTF considered both these lines of evidence as it issued its current B recommendation for biennial screening mammography for women aged 40 to 74 years.
Last, in response to comments, the USPSTF added the breast cancer screening recommendations from the American College of Radiology to the Recommendations of Others section.
See Table 3 for research needs and gaps related to screening for breast cancer.
The American Cancer Society recommends that women with an average risk of breast cancer should undergo regular screening mammography starting at age 45 years. It suggests that women aged 45 to 54 years should be screened annually, that women 55 years or older should transition to biennial screening or have the opportunity to continue screening annually, that women should have the opportunity to begin annual screening between the ages of 40 and 44 years, and that women should continue screening mammography as long as their overall health is good and they have a life expectancy of 10 years or longer. 74
The American College of Obstetricians and Gynecologists recommends that women at average risk of breast cancer should be offered screening mammography starting at age 40 years, using shared decision-making, and if they have not initiated screening in their 40s, they should begin screening mammography by no later than age 50 years. It recommends that women at average risk of breast cancer should have screening mammography every 1 or 2 years and should continue screening mammography until at least age 75 years. Beyond age 75 years, the decision to discontinue screening mammography should be based on shared decision-making informed by the woman’s health status and longevity. 75
The American College of Radiology and the Society of Breast Imaging recommend annual screening mammography beginning at age 40 years for women at average risk. They recommend that screening should continue past age 74 years, without an upper age limit, unless severe comorbidities limit life expectancy. 76 The American College of Radiology also recommends breast cancer risk assessment by age 25 years for all individuals. 77
The American Academy of Family Physicians supports the 2016 USPSTF recommendation on screening for breast cancer. 78
The authors of this recommendation statement include Task Force members serving at the time of publication and former members who made significant contributions to the recommendation. Any member with a level 3 conflict of interest (COI) recusal is not included as an author (see below for relevant COI disclosures for this topic).
The US Preventive Services Task Force authors of this recommendation statement include the following individuals: Wanda K. Nicholson, MD, MPH, MBA (George Washington University, Washington, DC); Michael Silverstein, MD, MPH (Brown University, Providence, Rhode Island); John B. Wong, MD (Tufts University School of Medicine, Boston, Massachusetts); Michael J. Barry, MD (Harvard Medical School, Boston, Massachusetts); David Chelmow, MD (Virginia Commonwealth University, Richmond); Tumaini Rucker Coker, MD, MBA (University of Washington, Seattle); Esa M. Davis, MD, MPH (University of Maryland School of Medicine, Baltimore); Carlos Roberto Jaén, MD, PhD, MS (University of Texas Health Science Center, San Antonio); M. (Tonette) Krousel-Wood, MD, MSPH (Tulane University, New Orleans, Louisiana); Sei Lee, MD, MAS (University of California, San Francisco); Li Li, MD, PhD, MPH (University of Virginia, Charlottesville); Carol M. Mangione, MD, MSPH (University of California, Los Angeles); Goutham Rao, MD (Case Western Reserve University, Cleveland, Ohio); John M. Ruiz, PhD (University of Arizona, Tucson); James Stevermer, MD, MSPH (University of Missouri, Columbia); Joel Tsevat, MD, MPH (University of Texas Health Science Center, San Antonio); Sandra Millon Underwood, PhD, RN (University of Wisconsin, Milwaukee); Sarah Wiehe, MD, MPH (Indiana University, Bloomington).
Conflict of Interest Disclosures: Authors followed the policy regarding conflicts of interest described at https://uspreventiveservicestaskforce.org/uspstf/about-uspstf/conflict-interest-disclosures . All members of the USPSTF receive travel reimbursement and an honorarium for participating in USPSTF meetings. Dr Wong reported delivering numerous unpaid talks on the 2009 USPSTF breast cancer screening recommendation; serving as a paid statistical reviewer for review of the USPSTF breast cancer screening models for the Annals of Internal Medicine in 2016 and of Cancer Intervention and Surveillance Modeling Network (CISNET) breast cancer and other cancer models submitted to the Annals of Internal Medicine ; serving as an unpaid National Cancer Institute chair of a group (2 others) that performed an external evaluation of the CISNET Program to assess its past productivity and adherence to its mission in 2018 (offered payment but never received); and serving as an unpaid member of the National Cancer Institute–convened Multi-cancer Early Detection (MCED) Diagnostic Pathways Working Group. Dr Chelmow reported serving as chair of the American College of Obstetricians and Gynecologists Practice Advisory Committee; in this role, he was involved in the development of practice guidelines related to breast cancer screening. Dr Barry reported receiving grants from Healthwise, a nonprofit, outside the submitted work. Dr Lee reported receiving grants from the National Institute on Aging (K24AG066998, R01AG079982) outside the submitted work. No other disclosures were reported.
No Task Force members had a level 3 COI recusal from this topic.
Funding/Support: The USPSTF is an independent, voluntary body. The US Congress mandates that the Agency for Healthcare Research and Quality (AHRQ) support the operations of the USPSTF.
Role of the Funder/Sponsor: AHRQ staff assisted in the following: development and review of the research plan, commission of the systematic evidence review from an Evidence-based Practice Center, coordination of expert review and public comment of the draft evidence report and draft recommendation statement, and the writing and preparation of the final recommendation statement and its submission for publication. AHRQ staff had no role in the approval of the final recommendation statement or the decision to submit for publication.
Disclaimer: Recommendations made by the USPSTF are independent of the US government. They should not be construed as an official position of AHRQ or the US Department of Health and Human Services.
Copyright Notice: USPSTF recommendations are based on a rigorous review of existing peer-reviewed evidence and are intended to help primary care clinicians and patients decide together whether a preventive service is right for a patient's needs. To encourage widespread discussion, consideration, adoption, and implementation of USPSTF recommendations, AHRQ permits members of the public to reproduce, redistribute, publicly display, and incorporate USPSTF work into other materials provided that it is reproduced without any changes to the work of portions thereof, except as permitted as fair use under the US Copyright Act.
AHRQ and the US Department of Health and Human Services cannot endorse, or appear to endorse, derivative or excerpted materials, and they cannot be held liable for the content or use of adapted products that are incorporated on other Web sites. Any adaptations of these electronic documents and resources must include a disclaimer to this effect. Advertising or implied endorsement for any commercial products or services is strictly prohibited.
This work may not be reproduced, reprinted, or redistributed for a fee, nor may the work be sold for profit or incorporated into a profit-making venture without the express written permission of AHRQ. This work is subject to the restrictions of Section 1140 of the Social Security Act, 42 U.S.C. §320b-10. When parts of a recommendation statement are used or quoted, the USPSTF Web page should be cited as the source.
1. Surveillance Epidemiology and End Results Program. Cancer Stat Facts: female breast cancer. National Cancer Institute. Accessed March 5, 2024. https://seer.cancer.gov/statfacts/html/breast.html 2. Surveillance Epidemiology and End Results Program. Breast: SEER 5-year age-adjusted incidence rates, 2016-2020, by race/ethnicity, female, all ages, all stages. National Cancer Institute. Accessed March 5, 2024. https://seer.cancer.gov/statistics-network/explorer/application.html?site=55&data_type=1&graph_type=10&compareBy=race&chk_race_6=6&chk_race_5=5&chk_race_4=4&chk_ race_9=9&chk_race_8=8&series=9&sex=3&age_range=1&stage=101&advopt_precision=1&advopt_show_ci=on&hdn_view=0#resultsRegion0 3. Surveillance Epidemiology and End Results Program. SEER*Stat Database: incidence—SEER research limited-field data with delay-adjustment, 22 registries, malignant only, November 2021 submission (2000-2019)—linked to county attributes—time dependent (1990-2019) income/rurality, 1969-2020 counties. National Cancer Institute. 2022. Accessed March 26, 2024. https://seer.cancer.gov/data-software/documentation/seerstat/nov2021/ 4. Henderson JT, Webber, EM, Weyrich M, Miller M, Melnikow J. Screening for Breast Cancer: A Comparative Effectiveness Review for the U.S. Preventive Services Task Force. Evidence Synthesis No. 231. Agency for Healthcare Research and Quality; 2024. AHRQ publication 23-05303-EF-1. 5. Surveillance Epidemiology and End Results Program. Breast: SEER 5-year age-adjusted incidence rates, 2016-2020, by subtype, female, all races/ethnicities, all ages, all stages. National Cancer Institute. Accessed March 5, 2024. https://seer.cancer.gov/statistics-network/explorer/application.html?site=55&data_type=1&graph_type=10&compareBy=subtype&chk_subtype_55=55&chk_subtype_622=622&chk_subtype_623=623&chk_subtype_620=620&chk_subtype_621=621& series=9&sex=3&race=1&age_range=1&stage=101&advopt_precision=1&advopt_show_ci=on&hdn_view 6. Surveillance Epidemiology and End Results Program. Breast: SEER 5-year age-adjusted mortality rates, 2016-2020, by race/ethnicity. National Cancer Institute. Accessed March 5, 2024. https://seer.cancer.gov/statistics-network/explorer/application.html?site=55&data_type=2&graph_type=10&compareBy=race&chk_race_6=6&chk_race_5=5&chk_race_4=4&chk_race_9=9&chk_race_8=8&series=9&sex=3&age_range=1&advopt_precision=1&advopt_show_ci=on &hdn_view=0&advopt_show_apc=on&advopt_display=2#resultsRegion 7. U.S. Preventive Services Task Force. Procedure Manual. Published May 2021. Accessed March 5, 2024. https://www.uspreventiveservicestaskforce.org/uspstf/about-uspstf/methods-and-processes/procedure-manual 8. US Preventive Services Task Force. Risk assessment, genetic counseling, and genetic testing for BRCA-related cancer: US Preventive Services Task Force recommendation statement . JAMA . 2019;322(7):652-665. Medline:31429903 doi:10.1001/jama.2019.10987 9. Ciatto S, Houssami N, Bernardi D, et al. Integration of 3D digital mammography with tomosynthesis for population breast-cancer screening (STORM): a prospective comparison study. Lancet Oncol . 2013;14(7):583-589. Medline:23623721 doi:10.1016/S1470-2045(13)70134-7 10. Skaane P, Bandos AI, Eben EB, et al. Two-view digital breast tomosynthesis screening with synthetically reconstructed projection images: comparison with digital breast tomosynthesis with full-field digital mammographic images. Radiology . 2014;271(3):655-663. Medline:24484063 doi:10.1148/radiol.13131391 11. Breast Cancer Surveillance Consortium. About the BCSC. Accessed March 5, 2024. https://www.bcsc-research.org/about 12. Trentham-Dietz A, Chapman CH, Jinani J, et al. Breast Cancer Screening With Mammography: An Updated Decision Analysis for the U.S. Preventive Services Task Force . Agency for Healthcare Research and Quality; 2024. AHRQ publication 23-05303-EF-2. 13. Trentham-Dietz A, Chapman CH, Jayasekera J, et al. Collaborative modeling to compare different breast cancer screening strategies: a decision analysis for the US Preventive Services Task Force. JAMA . Published April 30, 2024. doi:10.1001/jama.2023.24766 14. Miglioretti DL, Zhu W, Kerlikowske K, et al; Breast Cancer Surveillance Consortium. Breast tumor prognostic characteristics and biennial vs annual mammography, age, and menopausal status . JAMA Oncol . 2015;1(8):1069-1077. doi:10.1001/jama.2023.24766 15. Breast Cancer Treatment (PDQ®)—Health Professional Version. National Cancer Institute. Accessed April 10, 2024. https://www.cancer.gov/types/breast/hp/breast-treatment-pdq 16. Alvidrez J, Castille D, Laude-Sharp M, Rosario A, Tabor D. The National Institute on Minority Health and health disparities research framework. Am J Public Health . 2019;109(S1):S16-S20. Medline:30699025 doi:10.2105/AJPH.2018.304883 17. Williams DR, Priest N, Anderson NB. Understanding associations among race, socioeconomic status, and health: patterns and prospects. Health Psychol . 2016;35(4):407-411. Medline:27018733 doi:10.1037/hea0000242 18. Bailey ZD, Krieger N, Agénor M, Graves J, Linos N, Bassett MT. Structural racism and health inequities in the USA: evidence and interventions. Lancet . 2017;389(10077):1453-1463. Medline:28402827 doi:10.1016/S0140-6736(17)30569-X 19. Zavala VA, Bracci PM, Carethers JM, et al. Cancer health disparities in racial/ethnic minorities in the United States. Br J Cancer . 2021;124(2):315-332. Medline:32901135 doi:10.1038/s41416-020-01038-6 20. Bemanian A, Beyer KM. Measures matter: the local exposure/isolation (LEx/Is) metrics and relationships between local-level segregation and breast cancer survival. Cancer Epidemiol Biomarkers Prev . 2017;26(4):516-524. Medline:28325737 doi:10.1158/1055-9965.EPI-16-0926 21. Goel N, Westrick AC, Bailey ZD, et al. Structural racism and breast cancer-specific survival: impact of economic and racial residential segregation. Ann Surg . 2022;275(4):776-783. Medline:35081560 doi:10.1097/SLA.0000000000005375 22. Siegel SD, Brooks MM, Lynch SM, Sims-Mourtada J, Schug ZT, Curriero FC. Racial disparities in triple negative breast cancer: toward a causal architecture approach. Breast Cancer Res . 2022;24(1):37. Medline:35650633 doi:10.1186/s13058-022-01533-z 23. Niraula S, Biswanger N, Hu P, Lambert P, Decker K. Incidence, characteristics, and outcomes of interval breast cancers compared with screening-detected breast cancers. JAMA Netw Open . 2020;3(9):e2018179. Medline:32975573 doi:10.1001/jamanetworkopen.2020.18179 24. Jatoi I, Sung H, Jemal A. The emergence of the racial disparity in U.S. breast-cancer mortality. New Engl J Med . 2022;386(25):2349-2352. Medline:35713541 doi:10.1056/NEJMp2200244 25. Davis Lynn BC, Chernyavskiy P, Gierach GL, Rosenberg PS. Decreasing incidence of estrogen receptor-negative breast cancer in the United States: trends by race and region. J Natl Cancer Inst . 2022;114(2):263-270. Medline:34508608 doi:10.1093/jnci/djab186 26. Plevritis SK, Munoz D, Kurian AW, et al. Association of screening and treatment with breast cancer mortality by molecular subtype in US women, 2000-2012. JAMA . 2018;319(2):154-164. Medline:29318276 doi:10.1001/jama.2017.19130 27. Fayanju OM, Ren Y, Stashko I, et al. Patient-reported causes of distress predict disparities in time to evaluation and time to treatment after breast cancer diagnosis. Cancer. 2021;127(5):757-768. Medline:33175437 doi:10.1002/cncr.33310 28. Selove R, Kilbourne B, Fadden MK, et al. Time from screening mammography to biopsy and from biopsy to breast cancer treatment among black and white, women Medicare beneficiaries not participating in a health maintenance organization. Womens Health Issues . 2016;26(6):642-647. Medline:27773529 doi:10.1016/j.whi.2016.09.003 29. Nguyen KH, Pasick RJ, Stewart SL, Kerlikowske K, Karliner LS. Disparities in abnormal mammogram follow-up time for Asian women compared with non-Hispanic white women and between Asian ethnic groups. Cancer . 2017;123(18):3468-3475. Medline:28603859 doi:10.1002/cncr.30756 30. Warner ET, Tamimi RM, Hughes ME, et al. Time to diagnosis and breast cancer stage by race/ethnicity. Breast Cancer Res Treat . 2012;136(3):813-821. Medline:23099438 doi:10.1007/s10549-012-2304-1 31. Kovar A, Bronsert M, Jaiswal K, et al. The waiting game: how long are breast cancer patients waiting for definitive diagnosis? Ann Surg Oncol . 2020;27(10):3641-3649. Medline:32314153 doi:10.1245/s10434-020-08484-9 32. Elmore JG, Nakano CY, Linden HM, Reisch LM, Ayanian JZ, Larson EB. Racial inequities in the timing of breast cancer detection, diagnosis, and initiation of treatment. Med Care . 2005;43(2):141-148. Medline:15655427 doi:10.1097/00005650-200502000-00007 33. Emerson MA, Golightly YM, Aiello AE, et al. Breast cancer treatment delays by socioeconomic and health care access latent classes in Black and White women. Cancer . 2020;126(22):4957-4966. Medline:32954493 doi:10.1002/cncr.33121 34. Lawson MB, Bissell MC, Miglioretti DL, et al. Multilevel factors associated with time to biopsy after abnormal screening mammography results by race and ethnicity. JAMA Oncol . 2022;8(8):1115-1126. Medline:35737381 doi:10.1001/jamaoncol.2022.1990 35. Hu X, Walker MS, Stepanski E, et al. Racial differences in patient-reported symptoms and adherence to adjuvant endocrine therapy among women with early-stage, hormone receptor-positive breast cancer. JAMA Netw Open . 2022;5(8):e2225485. Medline:35947386 doi:10.1001/jamanetworkopen.2022.25485 36. Hu X, Chehal PK, Kaplan C, et al. Characterization of clinical symptoms by race among women with early-stage, hormone receptor-positive breast cancer before starting chemotherapy. JAMA Netw Open . 2021;4(6):e2112076. Medline:34061200 doi:10.1001/jamanetworkopen.2021.12076 37. Clemons K, Blackford AL, Gupta A, et al. Geographic disparities in breast cancer mortality and place of death in the United States from 2003 to 2019. J Clin Oncol . 2022;40(16 Suppl):12034. doi:10.1200/JCO.2022.40.16_suppl.12034 38. Surveillance Epidemiology and End Results Program. Breast: SEER incidence rates by age at diagnosis, 2016-2020, by sex, delay-adjusted SEER incidence rate, all races/ethnicities. National Cancer Institute. Accessed March 5, 2024. https://seer.cancer.gov/statistics-network/explorer/application.html?site=55&data_type=1&graph_type=3&compareBy=sex&chk_sex_3=3&rate_type=2&race=1&advopt_precision=1& advopt_show_ci=on&hdn_view=0#resultsRegion0 39. Surveillance Epidemiology and End Results Program. Breast: U.S. Mortality Rates by Age at Death, 2016-2020, by Sex, All Races/Ethnicities. National Cancer Institute. Accessed March 5, 2024. https://seer.cancer.gov/statistics-network/explorer/application.html?site=55&data_type=2&graph_type=3&compareBy=sex&chk_sex_3=3&race=1 &advopt_precision=1&advopt_show_ci=on&hdn_view=0#resultsRegion0 40. García-Albéniz X, Hernán MA, Logan RW, Price M, Armstrong K, Hsu J. Continuation of annual screening mammography and breast cancer mortality in women older than 70 years. Ann Intern Med . 2020;172(6):381-389. Medline:32092767 doi:10.7326/M18-1199 41. Kerlikowske K, Zhu W, Tosteson AN, et al. Identifying women with dense breasts at high risk for interval cancer: a cohort study. Ann Intern Med . 2015;162(10):673-681. Medline:25984843 doi:10.7326/M14-1465 42. Price ER, Hargreaves J, Lipson JA, et al. The California breast density information group: a collaborative response to the issues of breast density, breast cancer risk, and breast density notification legislation. Radiology . 2013;269(3):887-892. Medline:24023072 doi:10.1148/radiol.13131217 43. Gierach GL, Ichikawa L, Kerlikowske K, et al. Relationship between mammographic density and breast cancer death in the Breast Cancer Surveillance Consortium. J Natl Cancer Inst . 2012;104(16):1218-1227. Medline:22911616 doi:10.1093/jnci/djs327 44. Nelson HD, Cantor A, Humphrey L, et al. Screening for Breast Cancer: A Systematic Review to Update the 2009 US Preventive Services Task Force Recommendation. Evidence Synthesis No. 124. Agency for Healthcare Research and Quality; 2016. AHRQ publication 14-05201-EF-1. 45. Centers for Disease Control and Prevention. Health, United States, 2018. Published 2018. Accessed March 5, 2024. https://www.cdc.gov/nchs/data/hus/hus18.pdf 46. State legislation map. Dense Breast-info. Accessed March 5, 2024. https://densebreast-info.org/legislative-information/state-legislation-map/ 47. Mammography Quality Standards Act, 21 C.F.R. § 900 (2023). 48. US Preventive Services Task Force. Medication use to reduce risk of breast cancer: US Preventive Services Task Force recommendation statement. JAMA . 2019;322(9):857-867. Medline:31479144 doi:10.1001/jama.2019.11885 49. Siu AL; US Preventive Services Task Force. Screening for breast cancer: US Preventive Services Task Force recommendation statement. Ann Intern Med . 2016;164(4):279-296. Medline:26757170 doi:10.7326/M15-2886 50. Henderson JT, Webber EM, Weyrich MS, Miller M, Melnikow J. Screening for breast cancer: evidence report and systematic review for the US Preventive Services Task Force. JAMA . Published April 30, 2024. doi:10.1001/jama.2023.25844 51. Owens DK, Whitlock EP, Henderson J, et al; US Preventive Services Task Force. Use of decision models in the development of evidence-based clinical preventive services recommendations: methods of the US Preventive Services Task Force. Ann Intern Med . 2016;165(7):501-508. Medline:27379742 doi:10.7326/M15-2531 52. Nyström L, Bjurstam N, Jonsson H, Zackrisson S, Frisell J. Reduced breast cancer mortality after 20+ years of follow-up in the Swedish randomized controlled mammography trials in Malmö, Stockholm, and Göteborg. J Med Screen . 2017;24(1):34-42. Medline:27306511 doi:10.1177/0969141316648987 53. Jones BA, Patterson EA, Calvocoressi L. Mammography screening in African American women: evaluating the research. Cancer . 2003;97(1 Suppl):258-272. Medline:12491490 doi:10.1002/cncr.11022 54. Parvinen I, Chiu S, Pylkkänen L, Klemi P, Immonen-Räihä P, Kauhava L, et al. Effects of annual vs triennial mammography interval on breast cancer incidence and mortality in ages 40-49 in Finland. Br J Cancer . 2011;105:1388-91. Medline:21934688 doi:10.1038/bjc.2011.372 55. Breast Screening Frequency Trial Group; United Kingdom Co-ordinating Committee on Cancer Research. The frequency of breast cancer screening: results from the UKCCCR randomised trial. Eur J Cancer . 2002;38(11):1458-1464. Medline:12110490 doi:10.1016/S0959-8049(01)00397-5 56. Armaroli P, Frigerio A, Correale L, et al. A randomised controlled trial of digital breast tomosynthesis versus digital mammography as primary screening tests: screening results over subsequent episodes of the Proteus Donna study. Int J Cancer . 2022;151(10):1778-1790. Medline:35689673 doi:10.1002/ijc.34161 57. Hofvind S, Moshina N, Holen ÅS, et al. Interval and subsequent round breast cancer in a randomized controlled trial comparing digital breast tomosynthesis and digital mammography screening. Radiology . 2021;300(1):66-76. Medline:33973840 doi:10.1148/radiol.2021203936 58. Pattacini P, Nitrosi A, Giorgi Rossi P, et al. A randomized trial comparing breast cancer incidence and interval cancers after tomosynthesis plus mammography versus mammography alone. Radiology . 2022;303(2):256-266. Medline:35103537 doi:10.1148/radiol.211132 59. Hovda T, Holen ÅS, Lång K, et al. Interval and consecutive round breast cancer after digital breast tomosynthesis and synthetic 2d mammography versus standard 2d digital mammography in BreastScreen Norway. Radiology . 2020;294(2):256-264. Medline:31821118 doi:10.1148/radiol.2019191337 60. Ho TH, Bissell MC, Kerlikowske K, et al. Cumulative probability of false-positive results after 10 years of screening with digital breast tomosynthesis vs digital mammography. JAMA Netw Open . 2022;5(3):e222440 Medline:35333365 doi:10.1001/jamanetworkopen.2022.2440 61. McGuinness JE, Ueng W, Trivedi MS, et al. Factors associated with false positive results on screening mammography in a population of predominantly Hispanic women. Cancer Epidemiol Biomarkers Prev . 2018;27(4):446-453. Medline:29382701 doi:10.1158/1055-9965.EPI-17-0009 62. Conant EF, Beaber EF, Sprague BL, et al. Breast cancer screening using tomosynthesis in combination with digital mammography compared to digital mammography alone: a cohort study within the PROSPR consortium. Breast Cancer Res Treat . 2016;156(1):109-116. Medline:26931450 doi:10.1007/s10549-016-3695-1 63. Kerlikowske K, Su YR, Sprague BL, et al. Association of screening with digital breast tomosynthesis vs digital mammography with risk of interval invasive and advanced breast cancer. JAMA . 2022;327(22):2220-2230. Medline:35699706 doi:10.1001/jama.2022.7672 64. Richman IB, Long JB, Hoag JR, et al. Comparative effectiveness of digital breast tomosynthesis for breast cancer screening among women 40-64 years old. J Natl Cancer Inst . 2021;113(11):1515-1522. Medline:33822120 doi:10.1093/jnci/djab063 65. Johnson K, Lang K, Ikeda DM, et al. Interval breast cancer rates and tumor characteristics in the prospective population-based Malmö breast tomosynthesis screening trial. Radiology . 2021;299(3):559-567. Medline:33825509 doi:10.1148/radiol.2021204106 66. Zackrisson S, Lång K, Rosso A, et al. One-view breast tomosynthesis versus two-view mammography in the Malmö Breast Tomosynthesis Screening Trial (MBTST): a prospective, population-based, diagnostic accuracy study. Lancet Oncol . 2018;19(11):1493-1503. Medline:30322817 doi:10.1016/S1470-2045(18)30521-7 67. Heindel W, Weigel S, Gerß J, et al. Digital breast tomosynthesis plus synthesized mammography versus digital screening mammography for the detection of invasive breast cancer (TOSYMA): a multicentre, open-label, randomised, controlled, superiority trial. Lancet Oncol . 2022;23(5):601-611. Medline:35427470 doi:10.1016/S1470-2045(22)00194-2 68. Aase HS, Holen ÅS, Pedersen K, et al. A randomized controlled trial of digital breast tomosynthesis versus digital mammography in population-based screening in Bergen: interim analysis of performance indicators from the To-Be trial. Eur Radiol. 2019;29(3):1175-1186. Medline:30159620 doi:10.1007/s00330-018-5690-x 69. Bakker MF, de Lange SV, Pijnappel RM, et al. Supplemental MRI screening for women with extremely dense breast tissue. N Engl J Med . 2019;381(22):2091-2102. Medline:31774954 doi:10.1056/NEJMoa1903986 70. Veenhuizen SG, de Lange SV, Bakker MF, et al. Supplemental breast MRI for women with extremely dense breasts: results of the second screening round of the DENSE trial. Radiology . 2021;299(2):278-286. Medline:33724062 doi:10.1148/radiol.2021203633 71. Ganguli I, Keating NL, Thakore N, Lii J, Raza S, Pace LE. downstream mammary and extramammary cascade services and spending following screening breast magnetic resonance imaging vs mammography among commercially insured women. JAMA Netw Open . 2022;5(4):e227234. Medline:35416989 doi:10.1001/jamanetworkopen.2022.7234 72. Ohuchi N, Suzuki A, Sobue T, et al; J-START Investigator Groups. Sensitivity and specificity of mammography and adjunctive ultrasonography to screen for breast cancer in the Japan Strategic Anti-cancer Randomized Trial (J-START): a randomised controlled trial. Lancet . 2016;387(10016):341-348. Medline:26547101 doi:10.1016/S0140-6736(15)00774-6 73. Lee JM, Arao RF, Sprague BL, et al. Performance of screening ultrasonography as an adjunct to screening mammography in women across the spectrum of breast cancer risk. JAMA Intern Med . 2019;179(5):658-667. Medline:26547101 doi:10.1016/S0140-6736(15)00774-6 74. Oeffinger KC, Fontham ET, Etzioni R, et al; American Cancer Society. Breast cancer screening for women at average risk: 2015 guideline update from the American Cancer Society. JAMA . 2015;314(15):1599-1614. Medline:26501536 doi:10.1001/jama.2015.12783 75. Committee on Practice Bulletins—Gynecology. Practice Bulletin Number 179: breast cancer risk assessment and screening in average-risk women. Obstet Gynecol . 2017;130(1):e1-e16. Medline:28644335 doi:10.1097/AOG.0000000000002158 76. Monticciolo DL, Malak SF, Friedewald SM, et al. Breast cancer screening recommendations inclusive of all women at average risk: update from the ACR and Society of Breast Imaging. J Am Coll Radiol . 2021;18(9):1280-1288. Medline:34154984 doi:10.1016/j.jacr.2021.04.021 77. Monticciolo DL, Newell MS, Moy L, Lee CS, Destounis SV. Breast cancer screening for women at higher-than-average risk: updated recommendations from the ACR. J Am Coll Radiol . 2023;20(9):902-914. Medline:37150275 doi:10.1016/j.jacr.2023.04.002 78. American Academy of Family Physicians. Clinical Preventive Service Recommendation: Breast Cancer. Accessed March 5, 2024. https://www.aafp.org/family-physician/patient-care/clinical-recommendations/all-clinical-recommendations/breast-cancer.html
Rationale | Assessment |
---|---|
Benefits of screening for breast cancer | |
Harms of screening for breast cancer | |
USPSTF assessment |
Abbreviations: MRI, magnetic resonance imaging; USPSTF, US Preventive Services Task Force.
|
|
|
| ||
---|---|---|---|---|---|
All women (across 6 models) | |||||
Biennial (40-74) | 16,116 | 8.2 | 165.2 | 1376 | 14 |
Biennial (50-74) | 11,208 | 6.7 | 120.8 | 873 | 12 |
Black women (across 4 models) | |||||
Biennial (40-74) | 15,801 | 10.7 | 228.9 | 1253 | 18 |
Biennial (50-74) | 10,905 | 9.2 | 176.7 | 814 | 16 |
To fulfill its mission to improve health by making evidence-based recommendations for preventive services, the USPSTF routinely highlights the most critical evidence gaps for creating actionable preventive services recommendations. The USPSTF often needs additional evidence to create the strongest recommendations for everyone, especially those with the greatest burden of disease. In some cases, clinical preventive services have been well studied, but there are important evidence gaps that prevent the USPSTF from making recommendations for specific populations. In this table, the USPSTF summarizes the gaps in the evidence for screening for breast cancer and emphasizes health equity gaps that need to be addressed to advance the health of the nation. Although the health equity gaps focus on Black women because they have the poorest health outcomes from breast cancer, it is important to note that all studies should actively recruit enough women of all racial and ethnic groups, including Asian, Black, Hispanic, Native American/Alaska Native, and Native Hawaiian/Pacific Islander participants, to investigate whether the effectiveness of screening, diagnosis, and treatment vary by group. For additional information on research needed to address these evidence gaps, see the Research Gaps Taxonomy table on the USPSTF website ( ). |
Research is needed to determine the benefits and harms of screening for breast cancer in women age 75 years or older. |
Research is needed to help clinicians and patients understand the best strategy for breast cancer screening in women found to have dense breasts on a screening mammogram, which occurs in more than 40% of women screened. |
Research is needed to understand and address the higher breast cancer mortality among Black women. |
Research is needed to identify approaches to reduce the risk of overdiagnosis leading to overtreatment of breast lesions identified through screening that may not be destined to cause morbidity and mortality, including DCIS. |
Abbreviations: DBT, digital breast tomosynthesis; DCIS, ductal carcinoma in situ; MRI, magnetic resonance imaging; USPSTF, US Preventive Services Task Force.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Let’s Talk About It
We all want better ways to find breast cancer early and save lives from this disease. Breast cancer screening can help to detect cancer early, when it’s most treatable. This guide is meant to help you and your health care professional understand the benefits and risks of breast cancer screening, including what age to start screening and how often people should be screened. This guide is not for women who have a BRCA gene variant, a history of chest radiation, or who have had breast cancer. These women should talk to their health care professional about how best to stay healthy.
Breast Cancer and Its Impact
Breast cancer is the second most common cause of cancer death for women in the US. Each year, there are about 240 000 cases diagnosed and nearly 43 000 women die from breast cancer.
Notably, Black women are 40% more likely to die from breast cancer than White women, even though they get breast cancer at a roughly similar rate. Black women more often get aggressive cancers at younger ages.
The Good News Is That Getting Screened for Breast Cancer Every Other Year Can Reduce Your Risk of Dying From This Disease
This guidance is for women and people assigned female at birth who
have no signs or symptoms of breast cancer
are aged 40 to 74 years
are at average risk of developing breast cancer, as well as those who have dense breasts or a family history of breast cancer
For Some Women, There’s More to Consider
While the evidence is clear that all women aged 40 to 74 years should have a mammogram every other year, there are some areas where the research is limited and clinical judgment and patient medical history, values, and preferences play a role in decision-making.
Women With Dense Breasts
Women find out if they have dense breasts after a mammogram. Women with dense breasts have a higher chance of getting breast cancer and that risk increases the more dense breasts are. While having dense breasts means that mammograms may not work as well, it’s important to still get screened. More research is needed on how to better find breast cancer in women with dense breasts, whether that is by adding an ultrasound, an MRI, or something else entirely.
What Can Be Done?
Health care professionals can share information on the benefits and harms of additional screening methods to help women with dense breasts decide what is best for them.
Women Aged 75 Years or Older
Studies very rarely included women aged 75 years or older, so the evidence is not clear about whether they should continue or stop screening.
Health care professionals and women aged 75 years or older may consider factors such as overall health and previous screening history when deciding whether to continue or stop screening.
So What Does This Mean?
All women should get screened for breast cancer every other year starting at age 40. There is not enough evidence to decide whether to continue or stop screening in women 75 years or older and what more should be done to screen for breast cancer in women with dense breasts.
To learn more, view the full USPSTF breast cancer screening recommendation . The USPSTF also has recommendations on BRCA -related cancer prevention and medication use to reduce risk of breast cancer .
The US Preventive Services Task Force is an independent, volunteer panel of national experts that works to improve the health of people nationwide by making evidence-based recommendations about clinical preventive services.
To learn more about the recommendation, visit https://www.uspreventiveservicestaskforce.org/ or view or download the USPSTF Prevention Task Force app.
To find this and other JAMA Patient Pages, go to the Patient Information collection at http://jamanetworkpatientpages.com .
Published Online: April 30, 2024. doi:10.1001/jama.2024.5535
Conflict of Interest Disclosures: None reported.
Sources: US Preventive Services Task Force ( https://uspreventiveservicestaskforce.org/uspstf/recommendation-topics/lets-talk-about-it-discussion-guides )
US Preventive Services Task Force. Screening for breast cancer: US Preventive Services Task Force Recommendation Statement. JAMA. Published online April 30, 2024. doi:10.1001/jama.2024.5534
US Preventive Services Task Force. Screening for Breast Cancer. JAMA. 2024;331(22):1973–1974. doi:10.1001/jama.2024.5535
© 2024
Artificial Intelligence Resource Center
Cardiology in JAMA : Read the Latest
Browse and subscribe to JAMA Network podcasts!
The page you recommended will be added to the "what others are reading" feed on "My ACR".
The page you bookmarked will be added to the "my reading list" feed on "My ACR".
New American College of Radiology ® (ACR ® ) breast cancer screening guidelines now call for all women — particularly Black and Ashkenazi Jewish women — to have risk assessment by age 25 to determine if screening earlier than age 40 is needed. The ACR continues to recommend annual screening starting at age 40 for women of average risk, but earlier and more intensive screening for high-risk patients. The new ACR guidelines for high-risk women were published online May 3 in the Journal of the American College of Radiology ( JACR ). Other Notable Updates:
“The latest scientific evidence continues to point to earlier assessment as well as augmented and earlier-than-age-40 screening of many women — particularly Black women and other minority women,” said Debra Monticciolo, MD, FACR, primary author of the new guidelines and division chief, Breast Imaging, Massachusetts General Hospital, Boston. “These evidence-based updates should spur more-informed doctor-patient conversations and help providers save more lives.” Factors that contributed to the ACR reclassification of Black women and other minorities to high-risk include that, compared to non-Hispanic white women:
“Since 1990, breast cancer death rates in Black women, who develop and die from the disease earlier, have only dropped approximately half as fast as in white women,” said Stamatia Destounis, MD, FACR, co-author of the new guidelines, chair of the American College of Radiology Breast Imaging Commission and managing partner at Elizabeth Wende Breast Care in Rochester, NY. “We continue to regularly examine the latest evidence and update our recommendations to help save more Black women and others at high risk from this deadly disease.” According to National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) data, since mammography became widespread in the 1980s, the U.S. breast cancer death rate in women, unchanged for the previous 50 years, has dropped 43%. Breast cancer deaths in men, who have the same treatment as women but are not screened, have not declined. For more information regarding the proven effectiveness of regular mammography screening at reducing breast cancer deaths, please visit RadiologyInfo.org , MammographySavesLives.org and EndTheConfusion.org.
Website maintenance is scheduled for Saturday, September 7, and Sunday, September 8. Short disruptions may occur during these days.
MUNEEZA KHAN, MD, AND ANNA CHOLLET, MD, MPH
This is a corrected version of the article that appeared in print.
Am Fam Physician. 2021;103(1):33-41
Patient information: See related handout on mammogram screening for breast cancer , written by the authors of this article.
Author disclosure: No relevant financial affiliations.
Breast cancer is the most common nonskin cancer in women and accounts for 30% of all new cancers in the United States. The highest incidence of breast cancer is in women 70 to 74 years of age. Numerous risk factors are associated with the development of breast cancer. A risk assessment tool can be used to determine individual risk and help guide screening decisions. The U.S. Preventive Services Task Force (USPSTF) and American Academy of Family Physicians (AAFP) recommend against teaching average-risk women to perform breast self-examinations. The USPSTF and AAFP recommend biennial screening mammography for average-risk women 50 to 74 years of age. However, there is no strong evidence supporting a net benefit of mammography screening in average-risk women 40 to 49 years of age; therefore, the USPSTF and AAFP recommend individualized decision-making in these women. For average-risk women 75 years and older, the USPSTF and AAFP conclude that there is insufficient evidence to recommend screening, but the American College of Obstetricians and Gynecologists and the American Cancer Society state that screening may continue depending on the woman's health status and life expectancy. Women at high risk of breast cancer may benefit from mammography starting at 30 years of age or earlier, with supplemental screening such as magnetic resonance imaging. Supplemental ultrasonography in women with dense breasts increases cancer detection but also false-positive results.
Breast cancer is the most common nonskin cancer in women and accounts for 30% of all new cancers in the United States. 1 From 2001 to 2016, more than 2.3 million women in the United States were diagnosed with breast cancer. 2 The incidence of breast cancer increases after 25 years of age, peaking between 70 and 74 years. 2 Approximately one in eight women will develop invasive breast cancer (12.8% lifetime risk). 1
WHAT'S NEW ON THIS TOPIC
Breast Cancer Screening
A 2016 meta-analysis calculated that per 10,000 women screened with mammography, three breast cancer deaths are avoided over 10 years in women 40 to 49 years of age, eight deaths are avoided in women 50 to 59 years, 21 deaths are avoided in women 60 to 69 years, and 13 deaths are avoided in women 70 to 74 years. [ corrected ]
One out of every eight women 40 to 49 years of age who has a screening mammogram will subsequently undergo additional imaging, and for every case of invasive breast cancer detected by screening mammography in this age group, 10 women will have had a biopsy.
In a large, multicenter trial, women with dense breasts and a negative standard mammogram result had two-year screening with MRI or standard mammography. The interval cancer rate was lower in the MRI group than in the mammography group; however, MRI had a high false-positive rate with hundreds of negative breast biopsy results among the 4,738 women who underwent MRI screening.
MRI = magnetic resonance imaging.
, , | Consensus expert opinion; lack of evidence supporting breast self-examination | |
, | Meta-analysis of eight randomized trials; other organizations recommend considering annual screening | |
, , , | Consensus expert opinion; the risks vs. benefits in this age group have not been well-studied |
Do not routinely use breast magnetic resonance imaging for breast cancer screening in average-risk women. | Society of Surgical Oncology |
Do not perform screening mammography in asymptomatic patients with normal examination findings who have less than a five-year life expectancy. | American Society of Breast Surgeons – Benign Breast Disease |
Do not recommend screening for breast cancer if life expectancy is estimated to be less than 10 years. | Society for Post-Acute and Long-Term Care Medicine |
Do not recommend screening for breast, colorectal, prostate, or lung cancers without considering life expectancy and the risks of testing, overdiagnosis, and overtreatment. | American Geriatrics Society |
The overall mortality rate in U.S. women with breast cancer is about 20 per 100,000. Mortality rates are highest in women 85 years and older (170 per 100,000). 2 White women have the highest rate of breast cancer diagnosis, whereas Black women have the highest rate of breast cancer–related death. 2 Breast cancer is also the most common cause of cancer-related death in Hispanic women and the second leading cause of cancer-related death behind lung cancer among all women. 2
Cancer screening recommendations are determined by the patient's current anatomy. Transgender females with breast tissue and transgender males who have not undergone complete mastectomy should receive screening mammography based on guidelines for cisgender persons (see https://www.aafp.org/afp/2018/1201/p645.html#sec-4 ).
The strongest risk factors are a history of childhood chest radiation, older age, increased breast density, family history of breast cancer, and certain genetic mutations ( Table 1 ). 3 – 16 However, most women who develop invasive breast cancer do not have any of these risk factors . 3
mutations by age group | |
20 to 29 years | : 106, : 46 |
30 to 39 years | : 44, : 38 |
40 to 49 years | : 15, : 13 |
50 to 59 years | : 8.2, : 9.8 |
60 to 69 years | : 2.8, : 2.1 |
Overall | SIR = 21.9 |
Whole lung radiation | SIR = 43.6 |
Age ≥ 65 years vs. < 65 years | > 4.0 |
≥ 75% | 4.64 |
50% to 74% | 2.92 |
25% to 49% | 2.11 |
5% to 24% | 1.79 |
< 5% | 1.0 |
First-degree relative with breast cancer | HR = 1.61; OR = 1.64 |
Personal history of breast cancer | 1.42 |
Proliferative disease with atypia | 3.93 |
Proliferative disease without atypia | 1.76 |
Nonproliferative disease | 1.17 |
45 to 49 years | 0.86 |
≥ 55 years | 1.12 |
11 years | 1.09 |
15 years | 0.92 |
Nulliparity | 1.0 |
First pregnancy at 35 years | 1.16 |
First pregnancy at < 30 years | 0.73 |
1 to 4 years of combined estrogen/progestin | 1.60 |
1 to 4 years of estrogen only | 1.17 |
5 to 14 years of combined estrogen/progestin | 2.08 |
5 to 14 years of estrogen only | 1.33 |
Obesity (body mass index > 30 kg per m ) | 1.48 |
Moderate alcohol consumption | 1.1 |
A retrospective cohort study demonstrated a standardized incidence ratio (i.e., the ratio of observed to expected number of cases) of 21.9 for breast cancer in women who received chest radiation during childhood. 4 Higher doses of radiation were associated with higher risk, and the highest risk was in those who received whole lung radiation (standardized incidence ratio = 43.6). The overall cumulative risk of developing breast cancer by 50 years of age was 30%. 4
Increasing age is another strong risk factor. Invasive breast cancer will be diagnosed in one out of 42 women 50 to 59 years of age, and this rate increases to one out of 14 in women 70 years and older. 5
Breast density is the amount of glandular and stromal tissue compared with adipose tissue shown on a mammogram. A systematic review and meta-analysis found that compared with women who do not have dense breasts, the relative risk of developing breast cancer is 1.79 for women with breast density between 5% and 24% and 4.64 for those with breast density of 75% or higher. 6
Data from the Breast Cancer Surveillance Consortium and the Collaborative Breast Cancer Study showed that having a first-degree relative with breast cancer increases a woman's personal risk by a hazard ratio of 1.61 and odds ratio of 1.64. 7 For patients with BRCA mutations, the risk of developing breast cancer by 80 years of age is 60% to 63%, regardless of family history. 8
Several validated risk assessment tools are available to stratify breast cancer risk ( Table 2 ). 17 These tools can assist physicians and patients in developing individualized plans regarding screening, genetic testing, or chemoprevention .
Age (years) | Any | 35 to 90 | 19 to 85 | 35 to 74 | 20 to 79 |
Family history of breast cancer | Any | First-degree relative | First-, second-, or third-degree relative | First-degree relative | First- and second-degree relative |
Age at onset of breast cancer | Yes | No | Yes | No | Yes |
Bilateral breast cancer | Yes | No | Yes | No | No |
Personal or family history of ovarian cancer | Yes | No | Yes | No | Yes |
Atypical hyperplasia | No | Yes | Yes | Yes | No |
LCIS | No | No | Yes | Yes | No |
History of breast biopsy | No | Yes | Yes | Yes | No |
Hormonal factors | |||||
Age at menarche | No | Yes | Yes | No | No |
Age at menopause | No | No | Yes | No | No |
Age at first live birth | No | Yes | Yes | No | No |
Parity | No | No | Yes | No | No |
Hormone therapy | No | No | Yes | No | No |
Body mass index | No | No | Yes | Yes | No |
Breast density | No | No | Yes | Yes | No |
Race/ethnicity | Yes | Yes | Yes | Yes | No |
Predicted outcome(s) | Invasive breast cancer | Invasive breast cancer | Invasive breast cancer | Invasive breast cancer | Invasive breast cancer or DCIS |
Prediction period | Any number of years up to 110 years | Any number of years up to 90 years | 1 to 20 years or lifetime risk up to 85 years | 5-year or 10-year risk | 10-year intervals |
Exclusion criteria | None | History of breast cancer, mutation, history of chest irradiation, DCIS, LCIS | None | History of breast cancer, DCIS, mastectomy, breast augmentation, no breast density measurement | No family history of breast cancer or ovarian cancer |
A large retrospective cohort study compared the six-year accuracy of five validated risk assessment tools among 35,921 women 40 to 84 years of age who underwent screening mammography in the United States from 2007 to 2009. 17 The models were BRCAPRO ( https://projects.iq.harvard.edu/bayesmendel/bayesmendel-r-package ); Breast Cancer Risk Assessment Tool, or Gail model ( https://bcrisktool.cancer.gov , https://www.mdcalc.com/gail-model-breast-cancer-risk ); Tyrer-Cuzick model, or International Breast Cancer Intervention Study model ( http://www.ems-trials.org/riskevaluator ); Breast Cancer Surveillance Consortium model ( https://tools.bcsc-scc.org/BC5yearRisk/calculator.htm ); and Claus model (computer program).
Based on overall performance, the positive predictive values were 2.6% for BRCAPRO and the Tyrer-Cuzick model, 2.9% for the Breast Cancer Risk Assessment Tool and Breast Cancer Surveillance Consortium model, and 3.9% for the Claus model. The negative predictive values were high at 97% or more for all of the models. 17
Screening mammography reduces breast cancer–related mortality, with larger reductions as women get older .
Modeling studies estimate that in women 40 to 49 years of age, the number needed to screen (NNS) with annual mammography to prevent one breast cancer death is 746. The NNS decreases to 351 in women 50 to 59 years and to 233 in women 60 to 69 years. The NNS is 377 in women 70 to 79 years of age. 18 However, randomized controlled trials have demonstrated a substantially higher NNS. A meta-analysis performed for the U.S. Preventive Services Task Force (USPSTF) calculated that per 10,000 women screened with mammography, only three breast cancer deaths are avoided over 10 years in women 40 to 49 years of age, eight deaths are avoided in women 50 to 59 years, 21 deaths are avoided in women 60 to 69 years, and 13 deaths are avoided in women 70 to 74 years. 19 [ corrected ]
Between 2008 and 2017, yearly rates of newly diagnosed breast cancer increased by 0.3%, and rates of breast cancer death fell by 1.5%. 20 This may be partly attributable to early detection of small, curable breast cancers that have a five-year relative survival rate of 98.8% posttreatment. 20 Studies have shown a reduction in the incidence of large tumors, which is also likely because of early detection of smaller tumors by mammography. 21
Lower death rates, however, may also reflect improved treatments. With older treatments, the reduction in mortality after screening mammography was approximately 12 deaths per 100,000 women. With improved treatments, the reduction in mortality after screening mammography is now about eight deaths per 100,000 women. 21
False-positive results are common with screening mammography, especially in younger women, leading to further imaging and radiation exposure and subsequent breast biopsies that can be painful, can cause anxiety, and usually yield benign results. Furthermore, screening can lead to overdiagnosis and overtreatment of cancers that may never have become symptomatic or life-threatening .
According to the USPSTF, the false-positive rate of mammography is highest in women 40 to 49 years of age at 121 per 1,000 and decreases with age to 69.6 per 1,000 women 70 to 79 years of age. 22 About one of every eight women 40 to 49 years of age who has a screening mammogram will subsequently undergo additional imaging, and for every case of invasive breast cancer detected by screening mammography in this age group, 10 women will have had a biopsy, compared with only three women in their 70s. 22
False-positive results are associated with increased antidepressant and anxiolytic prescriptions, with a relative risk of 1.13 to 1.19. 23 Women at highest risk of needing antidepressant and anxiolytic therapy are those 40 to 49 years of age who underwent multiple tests, including a biopsy, and who had to wait more than one week to be told the results were false-positive. 23
Systematic reviews have found that screening mammography leads to an overdiagnosis rate of 10% to 30%. 24 , 26 [ corrected ] Overdiagnosis can lead to unnecessary treatments for screening-detected breast cancers. Sometimes this involves treating ductal carcinoma in situ that would have been inconsequential over a woman's lifetime. 3 A study based on a large U.S. cancer registry reported that out of 297,000 women 40 years and older who had a mastectomy in 2013, 18% may not have needed one. 25 Thus, the USPSTF concludes that there is no strong evidence supporting mammography screening of average-risk women in their 40s. 26
Recommendations for breast self-examinations, clinical breast examinations, and mammography vary among organizations . Table 3 summarizes recommendations from the USPSTF, the American Academy of Family Physicians (AAFP), the American College of Obstetricians and Gynecologists (ACOG), the American College of Radiology (ACR), the American Cancer Society (ACS), and the National Comprehensive Cancer Network (NCCN) . 3 , 26 – 33
Breast self-examination | Recommends against teaching patients | Encourages breast self-awareness | — | Encourages breast self-awareness | Encourages breast self-awareness |
Clinical breast examination | Insufficient evidence | May be offered every 1 to 3 years from 25 to 39 years of age and then annually | — | Not recommended | Recommends every 1 to 3 years from 25 to 39 years of age and then annually |
40 to 44 years | Individual decision | Offer annual or biennial (individual decision) | Annual | Offer annual (individual decision) | Annual |
45 to 49 years | Individual decision | Offer annual or biennial (individual decision) | Annual | Annual | Annual |
50 to 54 years | Biennial | Annual or biennial | Annual | Annual | Annual |
55 to 74 years | Biennial | Annual or biennial | Annual | Biennial, option to continue annually | Annual |
When to stop mammography | Insufficient evidence for continued screening in women 75 years and older | Shared decision-making in women 75 years and older | Discontinue when life expectancy is < 5 to 7 years | Discontinue when life expectancy is ≤ 10 years | Discontinue when life expectancy is ≤ 10 years |
Adjunct MRI for high-risk women | Insufficient evidence | Offer annual mammography and MRI starting at 30 years | Annual mammography and MRI starting at 30 years | Offer annual mammography and MRI starting at 30 years | Annual mammography, clinical breast examination every six to 12 months, consider annual MRI starting at 30 years |
Women with dense breasts | Insufficient evidence | Insufficient evidence | Consider ultrasonography plus mammography | Insufficient evidence | Counsel on risks and benefits of supplemental screening |
Breast Self-Examination . The USPSTF and AAFP recommend against teaching patients to perform breast self-examinations because of a lack of supporting evidence. 26 , 27 ACOG, the NCCN, and the ACS encourage breast self-awareness (i.e., patient familiarity with how her breasts usually feel and look) and advise women to seek medical attention if they notice breast changes. 3 , 31 , 33 There may be some rationale for breast self-awareness based on the frequency of self-detection cited in some studies. For example, out of 361 breast cancer survivors who participated in the 2003 National Health Interview Survey, 43% reported detecting their own cancers. 34
Clinical Breast Examination . The USPSTF and AAFP state that there is insufficient evidence to assess the benefits and harms of clinical breast examinations. 26 , 28 The ACS recommends against these examinations because of insufficient evidence of benefit and a high rate of false-positive results (55 false-positives for every breast cancer detected). 31 , 35 For average-risk women 40 years and older, ACOG says that annual clinical breast examinations may be offered, and the NCCN recommends annual clinical breast examinations. 3 , 33
Mammography . Evidence of benefit varies with a woman's age. The USPSTF found lower mortality rates and a reduced risk of advanced breast cancer in women 50 years and older who had mammography screening (relative risk = 0.62; 95% CI, 0.46 to 0.83) but not in women 39 to 49 years of age (relative risk = 0.98; 95% CI, 0.74 to 1.37). 19 The number of breast cancer deaths prevented with screening over 10 years was 12.5 per 10,000 women 50 years and older but only 2.9 per 10,000 women in their 40s. 19 Overall, women 50 to 59 years of age have the best balance of risks and benefits from mammography. 3 , 19
ACS data, however, showed improved mortality benefit across all age groups, although the benefit was lower in younger women. The NNS to reduce mortality rates by 20% was 1,770 for women in their 40s, 1,087 for women in their 50s, and 835 for women in their 60s. 31
The USPSTF recommends biennial screening mammography for women 50 to 74 years of age. 26 This recommendation excludes women 40 to 49 years of age because the number needed to invite (NNI) of 1,904 and the NNS of 1,034 to detect one case of breast cancer with screening mammography were considered too high. The NNI of 1,339 and NNS of 455 in women 50 to 59 years of age and the NNI of 377 and NNS of 233 for women 60 to 69 years of age were considered acceptable. 18 The AAFP supports the USPSTF recommendation. 29
The ACS recommends annual screening mammography starting at 45 years of age and transitioning to biennial screening at 55 years of age. 31 This recommendation is based on multivariable analyses suggesting that women in the younger age group are more likely than older women to have advanced stage cancer when screened biennially rather than annually. 31
The NCCN recommends annual screening mammography. 33 , 36 ACOG recommends shared decision-making based on a discussion of benefits and harms when deciding between annual and biennial screening intervals. 3
Women at average risk should continue screening mammography through 74 years of age . 3 , 26 , 29 – 31 , 33 Starting at 75 years of age, women should be involved in shared decision-making based on overall health status and life expectancy according to ACOG recommendations . 3 The ACS and NCCN recommend continued screening after 75 years of age if life expectancy is at least 10 years, and the ACR recommends continued screening if life expectancy is at least five to seven years . 30 , 31 , 33 The USPSTF states that there is insufficient evidence to assess the benefits and harms of screening past 74 years of age, and the AAFP supports this finding . 26 , 29
Randomized controlled trials have shown that when mammography screening prevents a death, the death would have occurred within five to seven years after screening; thus, screening women with limited life expectancy is not warranted. 36 In addition, the number of life-years gained from screening decreases from 7.8 to 11.4 per 1,000 mammograms at 74 years of age to 4.8 to 7.8 per 1,000 at 80 years and to 1.4 to 2.4 per 1,000 at 90 years. 37 When adjusted for quality of life, the number of life-years gained decreases even further, and by 90 to 92 years of age, all life-years gained are counter-balanced by a loss in quality of life, presumably because of treatment adverse effects. 37 Yet, despite these data and the corresponding recommendations, 62% of women 75 to 79 years of age and 50% of women 80 years or older get mammograms, and 70% to 86% of physicians recommend mammography for 80-year-old women. 38 , 39
ACS recommends that women with a 20% or higher lifetime risk of breast cancer (assessed using a risk assessment tool [ Table 2 17 ] ) be offered annual mammography and magnetic resonance imaging (MRI), typically starting at 30 years of age . 32 For high-risk women 25 to 29 years of age, ACOG recommends a clinical breast examination every six to 12 months and annual breast MRI with contrast. For patients 30 years and older, ACOG recommends annual mammography and MRI with contrast . 40 The NCCN recommends that women with a lifetime risk of more than 20% have breast self-awareness and receive a clinical breast examination every six to 12 months starting at 21 years of age. Annual breast MRI is recommended starting at 25 years of age with annual screening mammography starting at 30 years . 33 Women younger than 25 years with a history of chest radiation should have breast self-awareness and receive a clinical breast examination every six to 12 months starting 10 years after radiation therapy. Once these women are 25 years old, annual breast MRI is recommended, then screening mammography starting at 30 years of age . 33 The USPSTF states that there is insufficient evidence to assess the benefits and harms of using MRI for breast cancer screening, and the AAFP supports this finding . 26 , 29
The evidence for adding annual MRI screening to mammography and clinical breast examinations in women with more than a 20% lifetime risk of breast cancer is based on nonrandomized screening trials and observational studies from the 1990s. 32 These studies showed that MRI has a sensitivity of 71% to 100% for detecting breast cancer in high-risk women vs. mammography's sensitivity of 16% to 40% in the same population. However, MRI is less specific (81% to 99%) compared with mammography (93% to more than 99%), resulting in higher rates of false-positives, subsequent medical appointments, and biopsies, with a positive predictive value of 20% to 40%. No data were collected on survival rates with MRI screening or on the optimal MRI screening interval. 32
Almost 50% of women 40 to 74 years of age have dense breasts, which is a risk factor for breast cancer and for false-negative results on standard mammography . 41 Ultrasonography, MRI, and digital breast tomosynthesis (also known as 3D mammography) have been proposed as methods to detect breast cancers that might be missed on mammography in women with dense breasts .
The ACR recommends considering ultrasonography in addition to screening mammography based on a randomized multicenter trial showing improved cancer detection rates compared with mammography alone (1.9 vs. 4.2 per 1,000). 30 , 42 Ultrasonography may be particularly useful for women who have a 15% to 20% lifetime risk of breast cancer and dense breasts but no additional risk factors. 43
Data from the Connecticut Experiments showed an additional 2.3 cancers detected per 1,000 women with dense breasts who were screened with ultrasonography in addition to mammography. 43 By the fourth year of the study, the positive predictive value had increased from 7.3% to 20.1%, indicating an improved learning curve for the radiologists regarding which lesions to biopsy. Another study, involving 2,662 women with dense breasts plus one other risk factor for breast cancer, showed that adding ultrasonography to mammography increased the sensitivity of breast cancer detection compared with mammography alone (52% vs. 76%). 42
It is important to note, however, that the increased sensitivity comes at the cost of increasing false-positives. An observational cohort study of 6,081 women with varying risk of breast cancer showed that the false-positive rate was 22.2 per 1,000 screens for mammography alone vs. 52 per 1,000 screens for mammography plus ultrasonography (relative risk = 2.23). 44
MRI has also been studied as a screening option in women with dense breasts. A large multicenter trial randomized women with dense breasts and a negative result on standard mammography to two-year screening with either MRI or standard mammography. 45 The cancer detection rate during the two years was lower in the MRI group than in the mammography group (2.5 vs. 5 per 1,000 screens). More than 90% of MRI-detected cancers, however, were stage 0 or 1, and MRI screening resulted in a high false-positive rate (79.8 per 1,000 screens) with hundreds of negative breast biopsy results among the 4,738 women who underwent MRI screening.
MRI has also been compared with digital breast tomosynthesis. There were higher rates of cancer detection with MRI (11.8 per 1,000 screens) than with digital breast tomosynthesis (4.8 per 1,000 screens), but no data are available on long-term outcomes. 46 A study comparing standard mammography with digital breast tomosynthesis is underway. 47
The long-term survival of women whose breast cancers were detected with supplemental imaging modalities has not been studied.
This article updates previous articles on this topic by Tirona , 48 Knutson and Steiner , 49 and Apantaku . 50
Data Sources: A PubMed search was completed in Clinical Queries using the key terms breast cancer, breast cancer screening, risk factors for breast cancer, breast cancer risk assessment tools, breast cancer screening recommendations, breast density, mammography, supplemental screening. The search included meta-analyses, randomized controlled trials, clinical trials, and reviews. Also searched were the Agency for Healthcare Research and Quality Effective Healthcare Reports, the Cochrane database, and Essential Evidence Plus. Search date: April 2020.
American Cancer Society. Cancer facts and figures 2020. Accessed April 4, 2020. https://www.cancer.org/content/dam/cancer-org/research/cancer-facts-and-statistics/annual-cancer-facts-and-figures/2020/cancer-facts-and-figures-2020.pdf
Centers for Disease Control and Prevention. United States cancer statistics: data visualizations. Accessed August 2, 2020. https://gis.cdc.gov/Cancer/USCS/DataViz.html
American College of Obstetricians and Gynecologists. Practice bulletin no. 179. Breast cancer risk assessment and screening in average-risk women. Obstet Gynecol. 2017;130(1):e1-e16.
Moskowitz CS, Chou JF, Wolden SL, et al. Breast cancer after chest radiation therapy for childhood cancer. J Clin Oncol. 2014;32(21):2217-2223.
Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. 2020;70(1):7-30.
McCormack VA, dos Santos Silva I. Breast density and parenchymal patterns as markers of breast cancer risk: a meta-analysis. Cancer Epidemiol Biomarkers Prev. 2006;15(6):1159-1169.
Shiyanbola OO, Arao RF, Miglioretti DL, et al. Emerging trends in family history of breast cancer and associated risk. Cancer Epidemiol Biomarkers Prev. 2017;26(12):1753-1760.
Metcalfe KA, Lubinski J, Gronwald J, et al.; Hereditary Breast Cancer Clinical Study Group. The risk of breast cancer in BRCA1 and BRCA2 mutation carriers without a first-degree relative with breast cancer. Clin Genet. 2018;93(5):1063-1068.
Chen S, Iversen ES, Friebel T, et al. Characterization of BRCA1 and BRCA2 mutations in a large United States sample. J Clin Oncol. 2006;24(6):863-871.
American Cancer Society. Breast cancer facts and figures 2019–2020. Accessed August 2, 2020. https://www.cancer.org/content/dam/cancer-org/research/cancer-facts-and-statistics/breast-cancer-facts-and-figures/breast-cancer-facts-and-figures-2019-2020.pdf
Schacht DV, Yamaguchi K, Lai J, et al. Importance of a personal history of breast cancer as a risk factor for the development of subsequent breast cancer. AJR Am J Roentgenol. 2014;202(2):289-292.
Dyrstad SW, Yan Y, Fowler AM, et al. Breast cancer risk associated with benign breast disease. Breast Cancer Res Treat. 2015;149(3):569-575.
Liu Y, Colditz GA, Rosner B, et al. Alcohol intake between menarche and first pregnancy. J Natl Cancer Inst. 2013;105(20):1571-1578.
Collaborative Group on Hormonal Factors in Breast Cancer. Type and timing of menopausal hormone therapy and breast cancer risk: individual participant meta-analysis of the worldwide epidemiological evidence. Lancet. 2019;394(10204):1159-1168.
Collaborative Group on Hormonal Factors in Breast Cancer. Menarche, menopause, and breast cancer risk. Lancet Oncol. 2012;13(11):1141-1151.
Colditz GA, Rosner B. Cumulative risk of breast cancer to age 70 years according to risk factor status. Am J Epidemiol. 2000;152(10):950-964.
McCarthy AM, Guan Z, Welch M, et al. Performance of breast cancer risk-assessment models in a large mammography cohort. J Natl Cancer Inst. 2020;112(5):489-497.
Hendrick RE, Helvie MA. Mammography screening. AJR Am J Roentgenol. 2012;198(3):723-728.
Nelson HD, Fu R, Cantor A, et al. Effectiveness of breast cancer screening: systematic review and meta-analysis to update the 2009 U.S. Preventive Services Task Force recommendation. Ann Intern Med. 2016;164(4):244-255.
National Cancer Institute. Female breast cancer. Accessed August 2, 2020. https://seer.cancer.gov/statfacts/html/breast.html
Welch HG, Prorok PC, O'Malley AJ, et al. Breast-cancer tumor size, overdiagnosis, and mammography screening effectiveness. N Engl J Med. 2016;375(15):1438-1447.
Nelson HD, O'Meara ES, Kerlikowske K, et al. Factors associated with rates of false-positive and false-negative results from digital mammography screening. Ann Intern Med. 2016;164(4):226-235.
Segel JE, Balkrishnan R, Hirth RA. The effect of false-positive mammograms on antidepressant and anxiolytic initiation. Med Care. 2017;55(8):752-758.
Monticciolo DL, Helvie MA, Hendrick RE. Current issues in the overdiagnosis and overtreatment of breast cancer. AJR Am J Roentgenol. 2018;210(2):285-291.
Harding C, Pompei F, Burmistrov D, et al. Use of mastectomy for overdiagnosed breast cancer in the United States. J Cancer Epidemiol. ;2019:5072506.
U.S. Preventive Services Task Force. Breast cancer: screening. January 11, 2016. Accessed July 20, 2019. https://www.uspreventiveservicestaskforce.org/uspstf/recommendation/breast-cancer-screening
American Academy of Family Physicians. Breast cancer, breast self exam (BSE). Accessed July 20, 2019. https://www.aafp.org/patient-care/clinical-recommendations/all/breast-cancer-self-bse.html
American Academy of Family Physicians. Breast cancer, clinical breast examination (CBE). Accessed July 20, 2019. https://www.aafp.org/patient-care/clinical-recommendations/all/breast-cancer-cbe.html
American Academy of Family Physicians. Breast cancer. Accessed May 31, 2020. https://www.aafp.org/patient-care/clinical-recommendations/all/breast-cancer.html
Lee CH, Dershaw DD, Kopans D, et al. Breast cancer screening with imaging: recommendations from the Society of Breast Imaging and the ACR on the use of mammography, breast MRI, breast ultrasound, and other technologies for the detection of clinically occult breast cancer. J Am Coll Radiol. 2010;7(1):18-27.
Oeffinger KC, Fontham ETH, Etzioni R, et al.; American Cancer Society. Breast cancer screening for women at average risk [published correction appears in JAMA . 2016;315(13):1406]. JAMA. 2015;314(15):1599-1614.
Saslow D, Boetes C, Burke W, et al. American Cancer Society guidelines for breast screening with MRI as an adjunct to mammography [published correction appears in CA Cancer J Clin . 2007;57(3):185]. CA Cancer J Clin. 2007;57(2):75-89.
Bevers TB, Helvie M, Bonaccio E, et al. NCCN clinical practice guidelines in oncology. Breast cancer screening and diagnosis. Version 3.2019. May 17, 2019. Accessed May 31, 2019. https://www.nccn.org/professionals/physician_gls/pdf/breast-screening.pdf
Roth MY, Elmore JG, Yi-Frazier JP, et al. Self-detection remains a key method of breast cancer detection for U.S. women. J Womens Health (Larchmt). 2011;20(8):1135-1139.
Meyers ER, Moorman P, Gierisch JM, et al. Benefits and harms of breast cancer screening: a systematic review [published correction appears in JAMA . 2016;315(13):1406]. JAMA. 2015;314(15):1615-1634.
Helvie MA, Bevers TB. Screening mammography for average-risk women. J Natl Compr Canc Netw. 2018;16(11):1398-1404.
van Ravesteyn NT, Stout NK, Schechter CB, et al. Benefits and harms of mammography screening after age 74 years. J Natl Cancer Inst. 2015;107(7):djv103.
Bellizzi KM, Breslau ES, Burness A, et al. Prevalence of cancer screening in older, racially diverse adults. Arch Intern Med. 2011;171(22):2031-2037.
Leach CR, Klabunde CN, Alfano CM, et al. Physician over-recommendation of mammography for terminally ill women. Cancer. 2012;118(1):27-37.
American College of Obstetricians and Gynecologists. Hereditary breast and ovarian cancer syndrome. Practice bulletin no. 182. September 2017. Accessed August 2, 2020. https://bit.ly/37Kl2M4
Sprague BL, Gangnon RE, Burt V, et al. Prevalence of mammographically dense breasts in the United States. J Natl Cancer Inst. 2014;106(10):dju255.
Berg WA, Zhang Z, Lehrer D, et al.; ACRIN 6666 Investigators. Detection of breast cancer with addition of annual screening ultrasound or a single screening MRI to mammography in women with elevated breast cancer risk. JAMA. 2012;307(13):1394-1404.
Thigpen D, Kappler A, Brem R. The role of ultrasound in screening dense breasts. Diagnostics (Basel). 2018;8(1):20.
Lee JM, Arao RF, Sprague BL, et al. Performance of screening ultrasonography as an adjunct to screening mammography in women across the spectrum of breast cancer risk [published correction appears in JAMA Intern Med . 2019;179(5):733]. JAMA Intern Med. 2019;179(5):658-667.
Bakker MF, de Lange SV, Pijnappel RM, et al.; DENSE Trial Study Group. Supplemental MRI screening for women with extremely dense breast tissue. N Engl J Med. 2019;381(22):2091-2102.
Comstock CE, Gatsonis C, Newstead GM, et al. Comparison of abbreviated breast MRI vs digital breast tomosynthesis for breast cancer detection among women with dense breasts undergoing screening [published correction appears in JAMA . 2020;323(12):1194]. JAMA. 2020;323(8):746-756.
National Cancer Institute. TMIST (Tomosynthesis Mammographic Imaging Screening Trial). Accessed January 10, 2020. https://www.cancer.gov/about-cancer/treatment/clinical-trials/nci-supported/tmist
Tirona MT. Breast cancer screening update. Am Fam Physician. 2013;87(4):274-278. Accessed September 15, 2020. https://www.aafp.org/afp/2013/0215/p274.html
Knutson D, Steiner E. Screening for breast cancer. Am Fam Physician. 2007;75(11):1660-1666. Accessed September 15, 2020. https://www.aafp.org/afp/2007/0601/p1660.html
Apantaku LM. Breast cancer diagnosis and screening. Am Fam Physician. 2000;62(3):596-602. Accessed September 15, 2020. https://aafp.org/afp/2000/0801/p596.html
More in pubmed.
Copyright © 2021 by the American Academy of Family Physicians.
This content is owned by the AAFP. A person viewing it online may make one printout of the material and may use that printout only for his or her personal, non-commercial reference. This material may not otherwise be downloaded, copied, printed, stored, transmitted or reproduced in any medium, whether now known or later invented, except as authorized in writing by the AAFP. See permissions for copyright questions and/or permission requests.
Copyright © 2024 American Academy of Family Physicians. All Rights Reserved.
A .gov website belongs to an official government organization in the United States.
A lock ( ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.
Related Topics:
Breast cancer screening means checking a woman's breasts for cancer before there are signs or symptoms of the disease. All women need to be informed by their health care provider about the best screening options for them. When you are told about the benefits and risks of screening and decide with your health care provider whether screening is right for you—and if so, when to have it—this is called informed and shared decision-making.
Although breast cancer screening cannot prevent breast cancer, it can help find breast cancer early, when it is easier to treat. Talk to your doctor about which breast cancer screening tests are right for you, and when you should have them.
The US Preventive Services Task Force is an organization made up of doctors and disease experts who look at research on the best way to prevent diseases and make recommendations on how doctors can help patients avoid diseases or find them early.
The Task Force recommends that women who are 40 to 74 years old and are at average risk for breast cancer get a mammogram every 2 years. Women should weigh the benefits and risks of screening tests (see below).
CDC's Dr. Lisa Richardson talks about the best time for women to start getting mammograms in this video.
A mammogram is an x-ray of the breast. For many women, mammograms are the best way to find breast cancer early, when it is easier to treat and before it is big enough to feel or cause symptoms. Having regular mammograms can lower the risk of dying from breast cancer. At this time, a mammogram is the best way to find breast cancer for most women of screening age.
A breast MRI uses magnets and radio waves to take pictures of the breast. Breast MRI is used along with mammograms to screen women who are at high risk for getting breast cancer. Because breast MRIs may appear abnormal even when there is no cancer, they are not used for women at average risk.
Having a clinical breast exam or doing a breast self-exam has not been found to lower the risk of dying from breast cancer.
Every screening test has benefits and risks, which is why it's important to talk to your doctor before getting any screening test, like a mammogram.
The benefit of screening is finding cancer early, when it's easier to treat.
Harms can include false positive test results, when a doctor sees something that looks like cancer but is not. This can lead to more tests, which can be expensive, invasive, and time-consuming, and may cause anxiety.
Tests also can lead to overdiagnosis, when doctors find a cancer that would not have gone on to cause symptoms or problems, or even may go away on its own. Treatment of these cancers is called overtreatment. Overtreatment can include treatments recommended for breast cancer, such as surgery or radiation therapy. These can cause unnecessary and unwanted side effects. Other potential harms from breast cancer screening include pain during procedures and radiation exposure from the mammogram test itself. While the amount of radiation in a mammogram is small, there may be risks with having repeated x-rays.
Mammograms may also miss some cancers, called false negative test results, which may delay finding a cancer and getting treatment.
You can get screened for breast cancer at a clinic, hospital, or doctor's office. If you want to be screened for breast cancer, call your doctor's office. They can help you schedule an appointment.
Most health insurance plans are required to cover screening mammograms every 1 to 2 years for women beginning at age 40 with no out-of-pocket cost (like a co-pay, deductible, or co-insurance).
Find a mammography facility near you.
Breast cancer.
Talk to your doctor about when to start and how often to get a mammogram.
Public health.
For the best browsing experience please enable JavaScript. Instructions for Microsoft Edge and Internet Explorer , other browsers
Screening aims to find breast cancers early, when they have the best chance of being successfully treated.
Contact your breast screening service if you haven’t received your invitation or breast screening test results.
Cancer screening involves testing apparently healthy people for early signs of cancer.
Breast screening uses a test called mammography which involves taking x-rays of the breasts. Screening can help to find breast cancers early when they are too small to see or feel. These cancers are usually easier to treat than larger ones.
It is important to remember that screening will not prevent you from getting breast cancer but aims to find early breast cancers.
Overall, the breast screening programme finds cancer in around 9 out of every 1,000 women having screening.
The NHS Breast Screening Programme invites all women from the age of 50 to 70 registered with a GP for screening every 3 years. This means that some people may not have their first screening mammogram until they are 52 or 53 years.
If you are older than 70
In England, Wales and Northern Ireland you can still have screening every 3 years but you won't automatically be invited. To continue to have screening contact your GP or your local breast screening unit.
In Scotland you can continue to have breast screening after you are 71 years of age up until your 75th birthday, but you won’t automatically be invited, you have to contact your local breast screening service.
If you are younger than 50
Your risk of breast cancer is generally very low. Mammograms are more difficult to read in younger women because their breast tissue is denser. So the patterns on the mammogram don't show up as well. There is little evidence to show that regular mammograms for women below the screening age would reduce deaths from breast cancer.
Breast screening for transgender or non-binary people
Breast screening is also for some trans or non-binary people. This includes:
The screening invitations you automatically receive depend on how your sex is registered with your GP. Any hormones or surgeries you’ve had will impact which screenings are relevant for you.
If you haven’t had a breast screening invitation when you think you should, or you have any further questions, speak to your GP or gender identity clinic.
Breast screening takes 2 x-rays of each breast. The x-rays are called mammograms.
You have one mammogram from above and one from the side on each breast.
A mammogram is still the best way to detect early breast cancer, even if you have breast implants. But a small amount of the breast tissue might be hidden by the implant.
This means that it is not as easy to see all the breast tissue, and you may have more x-rays taken. This will help the doctor see as much of the breast tissue as possible.
It is useful to let the screening unit staff know that you have implants before your mammogram.
You should get your results within 2 to 3 weeks. The radiographer can tell you when to expect yours. Most people have a normal reading.
You will receive a letter to let you know your mammogram does not show any signs of cancer. Your next screening appointment will be in 3 years’ time. Do contact your GP or local screening unit if you haven’t received an appointment and think you are due one.
It is important to see your GP If you notice any symptoms between your screening mammograms.
If the x-ray isn't clear enough or shows any abnormal areas, the clinic staff will call you back for more tests. You might need to have the x-rays taken again.
Around 4 out of 100 women (around 4%) are called back for more tests. If this happens, you might feel very worried. But many of these women won’t have cancer.
If you are called back because your mammogram showed an abnormal area, you might have a magnified mammogram. This can show up particular areas of the breasts more clearly. These mammograms show the borders of any lump or thickened area. They can also show up areas of calcium (calcification).
Breast cancers found by screening are generally at an early stage. Very early breast cancers are usually easier to treat, may need less treatment, and are more likely to be successfully treated.
The current evidence suggests that breast screening reduces the number of deaths from breast cancer by about 1,300 a year in the UK.
Almost all women diagnosed with breast cancer at the earliest possible stage in England survive their disease for at least 5 years after diagnosis.
Although breast screening can find many cancers early, it isn't perfect. There are some risks, and some people may have a false positive or false negative result.
Screening doesn't always find a cancer that is there. So some people with breast cancer will be missed. This is called a false negative result.
In some women, the test picks up something even though they don't have breast cancer. This is called a false positive result and can lead to anxiety and further tests such as a breast biopsy.
As well as finding cancers that need treating, screening can also pick up breast cancers that won't ever cause any problems.
At the moment it isn't possible to know whether a breast cancer will grow quickly and need treatment, or will grow slowly, or not at all. So, almost all women diagnosed have surgery to remove the cancer. Many also have radiotherapy, hormone therapy or chemotherapy.
For some people the treatment is unnecessary but at the moment doctors can't tell who needs treatment and who doesn't.
Screening can also pick up changes in the lining of the breast ducts called ductal carcinoma in situ (DCIS). It isn't possible to tell whether DCIS will develop into a cancer or not. So, many women with DCIS also have surgery and radiotherapy or hormone therapy.
A 2012 breast screening review found that screening leads to around 4,000 women overdiagnosed in the UK each year.
Each mammogram exposes a person to small amounts of radiation from the x-rays. But the amount of radiation is very small.
X-rays can very rarely cause cancer. Having mammograms every 3 years for 20 years very slightly increases the chance of getting cancer over a woman’s lifetime.
An NHS digital report found that more than 20,100 breast cancers were diagnosed through screening in England between 2021 and 2022.
Of those breast cancer cases detected most of these were found at an early stage.
Treatment is likely to be more successful if the cancer is an early stage.
Speak to your GP if you think you might be at an increased risk. They can refer you to a genetic specialist, who can assess your risk. Not everyone with a family history of cancer is at increased risk themselves.
UK guidelines recommend that women with a moderate or high risk of breast cancer because of their family history should start having screening mammograms every year in their forties.
If you have had tests that showed a mutation that increases the risk of breast cancer, the recommendations are slightly different.
UK guidelines recommend yearly MRI scans from:
It’s important that you have access to enough information about the benefits and harms of breast screening to make the decision.
You can talk to your own doctor or nurse. Or you can contact the Cancer Research UK information nurses on freephone 0808 800 4040. The lines are open from 9am to 5pm, Monday to Friday.
Even if you are having mammograms every 3 years it is important to make sure that you know how your breasts normally look and feel. Cancers can develop between mammograms. This is known as an interval cancer. Mammograms can also miss some cancers.
If you notice any unusual changes in your breast don’t wait until your next mammogram. See your GP straight away.
You can watch a video about women with learning disabilities who are going to have breast screening. The video was produced by Avon Breast Screening. It is about 11 minutes long.
For people with a learning disability NHS England have an easy read leaflet about breast screening as well as links to other resources for those that cannot read.
There is BSL information about breast screening in the different UK nations. Public Health Wales have 3 BSL videos about breast screening. They are about 35 minutes, 9 minutes and 7 minutes long.
NHS information have 2 videos about breast screening in Scotland. These are about 9 minutes and 11 minutes long.
The Public Health Agency in Northern Ireland have a British sign language and Irish sign language video. You can find these towards the bottom of the page and are both about 5 minutes long.
JoC was diagnosed when she attended a screening appointment.
"I actually didn't have any idea that I had cancer."
Symptoms of breast cancer.
Symptoms of breast cancer include a lump or thickening in the breast. Find out more about this and other possible symptoms and when you should see your GP.
You have a number of tests to check for breast cancer. This includes a breast examination, a mammogram, a biopsy and scans.
Get information about how doctors stage and grade breast cancer. In the UK, doctors use the TNM system to stage breast cancer. You may also be told about the number staging system.
Treatment for breast cancer depends on a number of factors. Find out about breast cancer treatments, where and how you have them, and how to cope with possible side effects.
Breast cancer is cancer that starts in the breast tissue. Find out about who gets breast cancer and where it starts.
Find out about breast cancer, including symptoms, diagnosis, treatment, survival, and how to cope with the effects on your life and relationships.
It’s a worrying time for many people and we want to be there for you whenever - and wherever - you need us. Cancer Chat is our fully moderated forum where you can talk to others affected by cancer, share experiences, and get support. Cancer Chat is free to join and available 24 hours a day.
Visit the Cancer Chat forum
About Cancer generously supported by Dangoor Education since 2010.
Search our clinical trials database for all cancer trials and studies recruiting in the UK
Talk to other people affected by cancer
Questions about cancer? Call freephone 9 to 5 Monday to Friday or email us
An official website of the United States government
The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.
The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.
Email citation, add to collections.
Your saved search, create a file for external citation management software, your rss feed.
Affiliations.
In most developed countries, both organized screening (OrgS) and opportunistic screening (OppS) coexist. The literature has extensively covered the impact of organized screening on women's survival after breast cancer. However, the impact of opportunistic screening has been less frequently described due to the challenge of identifying the target population. The aim of this study was to describe the net survival and excess mortality hazard (EMH) in each screening group (OrgS, OppS, or No screening) and to determine whether there is an identical social gradient in each groups. Three data sources (cancer registry, screening coordination centers, and National Health Data System [NHDS]) were used to identify the three screening groups. The European Deprivation Index (EDI) defined the level of deprivation. We modeled excess breast cancer mortality hazard and net survival using penalized flexible models. We observed a higher EMH for "No screening" women compared with the other two groups, regardless of level of deprivation and age at diagnosis. A social gradient appeared for each group at different follow-up times and particularly between 2 and 3 years of follow-up for "OrgS" and "OppS" women. Net survival was higher for "OrgS" women than "OppS" women, especially for the oldest women, and regardless of the deprivation level. This study provides new evidence of the impact of OrgS on net survival and excess mortality hazard after breast cancer, compared with opportunistic screening or no screening, and tends to show that OrgS attenuates the social gradient effect.
Keywords: breast cancer; deprivation; excess mortality hazard; opportunistic screening; organized screening.
© 2024 UICC.
PubMed Disclaimer
NCBI Literature Resources
MeSH PMC Bookshelf Disclaimer
The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.
Emeritus Professor of Medicine, Boston University
Assistant Professor of Health Policy and Clinical Practice, Dartmouth College
Professor of Radiology, Boston University
Associate Director of Cancer Care Equity, Yale Cancer Center, Yale University
Nancy Kressin received funding from The American Cancer Society.
Christine M. Gunn receives funding from National Cancer Institute (1K07CA221899).
Priscilla Slanetz is site Principal Investigator for the Tomosynthesis Mammographic Imaging Screening Trial (TMIST). She also serves on the Practice Parameter Committee of the American College of Radiology (ACR), is Chair of the ACR Commission of Publications and Lifelong Learning, and serves on the ACR Board of Chancellors and the Association of Academic Radiology's Board of Directors and Research & Education Fund. Finally, she is co-author on the ACR Appropriateness Criteria Supplemental Breast Cancer Screening Based on Breast Density document which was first published in November 2021 and will be updated in late fall 2024.
Tracy A. Battaglia received grant funding from the American Cancer Society. She is co-author on the ACR Appropriateness Criteria® Supplemental Breast Cancer Screening Based on Breast Density published in 2021.
Boston University provides funding as a founding partner of The Conversation US.
View all partners
The Food and Drug Administration implemented a rule to go into effect on Sept. 10, 2024, requiring mammography facilities to notify women about their breast density. The goal is to ensure that women nationwide are informed about the risks of breast density, advised that other imaging tests might help find cancers and urged to talk with their doctors about next steps based on their individual situation.
The FDA originally issued the rule on March 10, 2023 , but extended the implementation date to give mammography facilities additional time to adhere.
The Conversation U.S. asked a team of experts in social science and patients’ health behaviors , health policy , radiology and primary care and health services research to explain the FDA’s new regulations about these health communications and what women should consider as they decide whether to pursue additional imaging tests, often called supplemental screening.
Breast density is categorized into four categories: fatty, scattered tissue, heterogeneously dense or extremely dense.
Dense breasts are composed of more fibrous, connective tissue and glandular tissue – meaning glands that produce milk and tubes that carry it to the nipple – than fatty tissue. Because fibroglandular tissue and breast masses both look white on mammographic images, greater breast density makes it more difficult to detect cancer. Nearly half of all American women are categorized as having dense breasts.
Having dense breasts also increases the risk of getting breast cancer , though the reason for this is unknown.
Because of this, decisions about breast cancer screening get more complicated . While evidence is clear that regular mammograms save lives , additional testing such as ultrasound, MRI or contrast-enhanced mammography may be warranted for women with dense breasts.
The FDA now requires specific language to ensure that all women receive the same “accurate, complete and understandable breast density information.” After a mammogram, women must be informed:
– Whether their breasts are dense or not dense
– That having dense breasts increases the risk of breast cancer
– That having dense breasts makes it harder to find breast cancer on mammograms
– That for those with dense breasts, additional imaging tests might help find cancer
They must also be advised to discuss their individual situation with their health care provider, to determine which, if any, additional screening might be indicated.
Prior to the federal rule, 38 U.S. states required some form of breast density notification . But some states had no notification requirements, and among the others there were many inconsistencies that raised concerns by advocates , including women with dense breasts whose advanced cancer had not been detected on a mammogram.
The FDA standardized the information women must receive. It is written at an eighth grade reading level and may address racial and literacy-level differences in women’s knowledge about breast density and reactions to written notifications.
For instance, our research team found disproportionately more confusion and anxiety among women of color, those with low literacy and women for whom English was not their first language. And some women with low literacy reported decreased future intentions to undergo mammographic screening.
Standard mammograms use X-rays to produce two-dimensional images of the breast. A newer type of mammography imaging called tomosynthesis produces 3D images, which find more cancers among women with dense breasts . So, researchers and doctors generally agree that women with dense breasts should undergo tomosynthesis screening when available.
There is still limited scientific evidence to guide recommendations for supplemental breast screening beyond standard mammography or tomosynthesis for women with dense breast tissue. Data shows that supplemental screening with ultrasound , MRI or contrast-enhanced mammography may detect additional cancers, but there are no prospective studies confirming that such additional screening saves more lives.
So far, there is no data from randomized clinical trials showing that supplemental breast MRIs, the most often-recommended supplemental screening, reduce death from breast cancer.
However, more early stage – but not late-stage – cancers are found with MRIs , which may require less extensive surgery and less chemotherapy.
Various professional organizations and experts interpret the available data about supplemental screening differently, arriving at different conclusions and recommendations. An important consideration is the woman’s individual level of risk, since emerging evidence suggests that women whose personal risk of developing breast cancer is high are most likely to benefit from supplemental screening.
Some organizations have concluded that current evidence is too limited to make a recommendation for supplemental screening , or they do not recommend routine use of supplemental screening for women based solely on breast density . Others recommend additional screening for women with extremely or heterogeneously dense breasts, even when their risk is at the intermediate level.
Because personal risk of breast cancer is a crucial consideration in deciding whether to undergo supplemental screening, women should understand their own risk.
The American College of Radiology recommends that all women undergo risk assessment by age 25. Women and their providers can use risk calculators such as Tyrer-Cuzick, which is free and available online .
Women should also understand that breast density is only one of several risks for breast cancer, and some of the others can be modified. Engaging in regular physical activity, maintaining a healthy weight, limiting alcohol use and eating a healthy diet rich in vegetables can all decrease breast cancer risk .
Amid the debate about the benefits of supplemental breast screening , there is less discussion about its possible harms. Most common are false alarms : results that suggest a finding of cancer that require follow-up testing. Less commonly, a biopsy is needed, which may lead to short-term fear and anxiety, medical bills or potential complications from interventions.
Supplemental screening can also lead to overdiagnosis and overtreatment – the small risk of identifying and treating a cancer that would have never posed a problem.
MRI screening also involves use of a chemical substance called gadolinium to improve imaging. Although tiny amounts of gadolinium are retained in the body, the FDA considers the contrast agent to be safe when given to patients with normal kidney function.
MRIs may also identify incidental findings outside the breast – such as in the lungs – that warrant additional concern, testing and cost. Women should consider their tolerance for such risks, relative to their desire for the benefits of additional screening.
The out-of-pocket cost of additional screening beyond a mammogram is also a consideration; only 29 states plus the District of Columbia require insurance coverage for supplemental breast cancer screening , and only three states – New York, Connecticut and Illinois – mandate insurance coverage with no copays.
Though the FDA urges women to talk with their providers, our research found that few women have such conversations and that many providers lack sufficient knowledge about breast density and current guidelines for breast screening.
It’s not yet clear why, but providers receive little or no training about breast density and report little confidence in their ability to counsel patients on this topic.
To address this knowledge deficit in some health care settings, radiologists, whose screening guidelines are more stringent than some other organizations, sometimes provide a recommendation for supplemental screening as part of their mammography report to the provider who ordered the mammogram.
Learning more about the topic in advance of a discussion with a provider can help a woman better understand her options.
Numerous online resources can provide more information, including the American Cancer Society , the website Dense Breast-info and the American College of Radiology .
Armed with information about the complexities of breast density and its impact on breast cancer screening, women can discuss their personal risk with their providers and evaluate the options for supplemental screening, with consideration of how they value the benefits and harms associated with different tests.
Previously, it said women could choose to start breast cancer screening as young as 40, with a stronger recommendation that women get the exams every two years starting at age 50, by carla k. johnson | the associated press • published april 30, 2024 • updated on april 30, 2024 at 8:11 pm.
Regular mammograms to screen for breast cancer should start younger, at age 40, according to an influential U.S. task force. Women ages 40 to 74 should get screened every other year, the group said.
Previously, the task force had said women could choose to start breast cancer screening as young as 40, with a stronger recommendation that they get the exams every two years from age 50 through 74.
📺 Watch News4 now: Stream NBC4 newscasts for free right here, right now.
The announcement Tuesday from the U.S. Preventive Services Task Force makes official a draft recommendation announced last year. The recommendations were published in the Journal of the American Medical Association.
“It’s a win that they are now recognizing the benefits of screening women in their 40s,” said Dr. Therese Bevers of MD Anderson Cancer Center in Houston. She was not involved in the guidance.
Other medical groups, including the American College of Radiology and the American Cancer Society, suggest mammograms every year — instead of every other year — starting at age 40 or 45, which may cause confusion, Bevers said, but “now the starting age will align with what many other organizations are saying.”
Breast cancer death rates have fallen as treatment continues to improve. But breast cancer is still the second-most common cause of cancer death for U.S. women. About 240,000 cases are diagnosed annually and nearly 43,000 women die from breast cancer.
The nudge toward earlier screening is meant to address two vexing issues: the increasing incidence of breast cancer among women in their 40s — it's risen 2% annually since 2015 — and the higher breast cancer death rate among Black women compared to white women, said task force vice chair Dr. John Wong of Tufts Medical Center in Boston.
“Sadly, we know all too well that Black women are 40% more likely to die from breast cancer than white women,” Wong said. Modeling studies predict that earlier screening may help all women, and have “even more benefit for women who are Black,” he said.
Here are more details on what’s changed, why it’s important and who should pay attention.
Age 40 is when mammograms should start for women, transgender men and nonbinary people at average risk. They should have the X-ray exam every other year, according to the new guidance. Other groups recommend annual mammograms, starting at 40 or 45.
The advice does not apply to women who've had breast cancer or those at very high risk of breast cancer because of genetic markers. It also does not apply to women who had high-dose radiation therapy to the chest when they were young, or to women who've had a lesion on previous biopsies.
It's not clear whether older women should continue getting regular mammograms. Studies rarely include women 75 and older, so the task force is calling for more research.
Bevers suggests that older women talk with their doctors about the benefits of screening, as well as harms like false alarms and unnecessary biopsies.
Mammograms don't work as well for women with dense breasts, but they should still get the exams.
The task force would like to see more evidence about additional tests such as ultrasounds or MRIs for women with dense breasts. It's not yet clear whether those types of tests would help detect cancer at an earlier, more treatable stage, Wong said.
Congress already passed legislation requiring insurers to pay for mammograms for women 40 and older without copays or deductibles. In addition, the Affordable Care Act requires insurers to cover task force recommendations with an “A” or “B” letter grade. The mammography recommendation has a “B” grade, meaning it has moderate net benefit.
Roche advances ai-driven cancer diagnostics by expanding its digital pathology open environment.
Tucson, Arizona
Roche (SIX: RO, ROG; OTCQX: RHHBY) announced today the expansion of its digital pathology open environment with the integration of more than 20 advanced artificial intelligence (AI) algorithms from eight new collaborators. These strategic collaborations aim to support pathologists and scientists in cancer research and diagnosis by leveraging cutting-edge AI technology.
The seamless integration is facilitated through Roche’s navify® Digital Pathology enterprise software, an application for the pathologist’s workflow, which now incorporates a diverse range of AI-driven algorithms, creating easy access to third-party innovation. These AI tools are designed to enhance pathology insights, helping benefit cancer patients through precision medicine and enabling targeted treatments.
"We are excited to welcome these new collaborators into our digital pathology ecosystem," said Jill German, Head of Pathology Lab for Roche Diagnostics. "By combining our leadership in tissue diagnostics with a broad offering of state-of-the-art AI technology, we aim to revolutionise cancer research, diagnostics and treatment, ultimately helping clinicians improve the lives of patients worldwide.”
The collaborators are:
With these new collaborations and integrations, Roche emphasises its commitment to improving patient outcomes and advancing personalised healthcare by providing scientists and clinicians with the resources they need to deliver precise and effective cancer diagnoses.
As the leading provider of pathology lab solutions, Roche delivers an end-to-end digital pathology solution from tissue staining to producing high-quality digital images that can be reliably assessed using automated AI-based clinical image analysis algorithms. Roche minimises variables that can impact analysis, and it is this end-to-end development that produces the quality results healthcare providers and researchers can depend on. With the acceleration of immunotherapy and the development of more complex assays, Roche is moving these traditionally research-oriented AI tools into routine clinical practice and is committed to investing in and shaping the future of pathology. The Roche Digital Pathology Open Environment* serves as a collaborative platform that brings together innovative AI-based pathology tools.
Founded in 1896 in Basel, Switzerland, as one of the first industrial manufacturers of branded medicines, Roche has grown into the world’s largest biotechnology company and the global leader in in-vitro diagnostics. The company pursues scientific excellence to discover and develop medicines and diagnostics for improving and saving the lives of people around the world. We are a pioneer in personalised healthcare and want to further transform how healthcare is delivered to have an even greater impact. To provide the best care for each person we partner with many stakeholders and combine our strengths in Diagnostics and Pharma with data insights from the clinical practice.
In recognising our endeavour to pursue a long-term perspective in all we do, Roche has been named one of the most sustainable companies in the pharmaceuticals industry by the Dow Jones Sustainability Indices for the fifteenth consecutive year. This distinction also reflects our efforts to improve access to healthcare together with local partners in every country we work.
Genentech, in the United States, is a wholly owned member of the Roche Group. Roche is the majority shareholder in Chugai Pharmaceutical, Japan.
For more information, please visit www.roche.com .
All trademarks used or mentioned in this release are protected by law.
* Some algorithms are Research Use Only and not for use in diagnostic procedures. Please consult with local markets on the regulatory status of these algorithms.
Jo Lynn Garing, +1 317-363-7286 or [email protected]
Latest press releases, you might be interested in, modal title.
Oops, you're using an old version of your browser so some of the features on this page may not be displaying properly.
MINIMAL Requirements: Google Chrome 24+ , Mozilla Firefox 20+ , Internet Explorer 11 , Opera 15–18 , Apple Safari 7 , SeaMonkey 2.15-2.23
Aired live on 04 Sep 2024
Learn more about how you can join ESMO
Member access only
Nancy u. lin, volkmar müller, giuseppe curigliano, barbara pistilli, description.
This ESMO Deep Dive Breast Cancer Webinar focused on HER2 positive metastatic breast cancer.
This webinar was developed with the aim to provide a second, deeper layer of educational experience in emerging data related to: molecular biology and classification, translational research and biomarkers for precision medicine, and unknowns, hypotheses, and ongoing research in areas of controversy or unmet need for optimal management of patients.
It began with a brief welcome and introduction, then four lectures were delivered on: brain metastases, optimal treatment sequences after guideline-based early breast cancer therapy and other potential targets, as well as the role of the molecular tumour board emerging concepts. The lectures were followed by a live discussion with the speakers’ panel.
This programme was designed to provide an overview of the latest understanding of molecular biology and the impact of targeting different pathways in the management of patients with breast cancers, offer expert opinion exchange, and provide some important considerations in this field.
You may also be interested in....
ESMO Deep Dive: Breast Cancer
This site uses cookies. Some of these cookies are essential, while others help us improve your experience by providing insights into how the site is being used.
For more detailed information on the cookies we use, please check our Privacy Policy .
Necessary cookies enable core functionality. The website cannot function properly without these cookies, and you can only disable them by changing your browser preferences.
Warning: The NCBI web site requires JavaScript to function. more...
An official website of the United States government
The .gov means it's official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you're on a federal government site.
The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
IARC Working Group on the Evaluation of Cancer-Preventive Interventions. Breast cancer screening. Lyon (FR): International Agency for Research on Cancer; 2016.
5. effectiveness of breast cancer screening.
This section considers measures of screening quality and major beneficial and harmful outcomes. Beneficial outcomes include reductions in deaths from breast cancer and in advanced-stage disease, and the main example of a harmful outcome is overdiagnosis of breast cancer. The absolute reduction in breast cancer mortality achieved by a particular screening programme is the most crucial indicator of a programme’s effectiveness. This may vary according to the risk of breast cancer death in the target population, the rate of participation in screening programmes, and the time scale observed ( Duffy et al., 2013 ). The technical quality of the screening, in both radiographic and radiological terms, also has an impact on breast cancer mortality. The observational analysis of breast cancer mortality and of a screening programme’s performance may be assessed against several process indicators. The major indicators of both the screening process and the clinical outcome, and the associated analytical methodologies, are described below.
5.1.1. performance indicators.
As a general principle, the most important indicator of the effectiveness of a screening programme is its effect on breast cancer mortality. Nevertheless, the performance of a screening programme should be monitored to identify and remedy shortcomings before enough time has elapsed to enable observation of mortality effects.
The randomized trials performed during the past 30 years have enabled the suggestion of several indicators of quality assurance for screening services ( Day et al., 1989 ; Tabár et al., 1992 ; Feig, 2007 ; Perry et al., 2008 ; Wilson & Liston, 2011 ), including screening participation rates, rates of recall for assessment, rates of percutaneous and surgical biopsy, and breast cancer detection rates. Detection rates are often classified by invasive/in situ status, tumour size, lymph-node status, and histological grade.
Table 5.1 and Table 5.2 show selected quality standards developed in England by the National Health Service (NHS) ( Wilson & Liston, 2011 ; Department of Health, 2013 ) and in the USA by the Agency for Health Care Policy and Research and endorsed by the American College of Radiology, respectively ( Bassett et al., 1994 ; D’Orsi et al., 2013 ). Similar sets of standards exist for screening in Australia, Canada, and Europe ( National Quality Management Committee of BreastScreen Australia, 2008 ; Perry et al., 2008 ; CPAC, 2013 ) (see Section 3.2). The programmes specify standards – related mainly to the screening process and not directly to technical aspects of image quality – that all units should attain, as well as achievable targets at which units should aim.
Minimum quality standards and targets considered in the National Health Service breast screening programme in England.
Minimum quality standards for mammography in the USA a .
Table 5.1 pertains to a programme that targets women aged 50–70 years with a maximum screening interval of 36 months in high-incidence countries. In the example in England, two-view mammography is used, and the programme changed from film to digital mammography during 2010–2014.
Minimum standards are specified for screening attendance and detection rates, in particular detection rates of small cancers, which are expected to be high in an effective screening programme. Maximum standards are specified for adverse effects of screening, such as radiation dose, and for rates of interval cancers, repeat examinations, and recalls for assessment. In addition, maximum times to events in the screening, diagnostic, and treatment processes are specified; these are important for the patient’s experience and quality of life, although they do not necessarily reflect clinical or radiological quality.
Some of the criteria and standards are very specific to the programme. For example, the randomized trials of breast screening observe a higher rate of breast cancer detection at the prevalent (first) screen than at incident (subsequent) screens (see, for example, Tabár et al., 1992 ). However, the detection rate standards are expected to be higher for incident screens because these values are based not on observations of a cohort recruited at the prevalent screen and followed up thereafter but on a programme in which prevalent screens usually take place at about age 50 years and incident screens on average at about age 60 years (when the underlying risk of cancer is higher).
Another measure that is used in the United Kingdom is the standardized detection ratio, obtained by comparing the observed detection rates of invasive cancers by age with those of the Swedish Two-County trial ( Tabár et al., 1992 ), on which the United Kingdom breast screening programme was modelled. At present, the standard is almost invariably exceeded ( NHSBSP, 2009 ), probably at least partly due to the fact that breast cancer incidence in the United Kingdom in the 21st century is higher than that in Sweden in the 1970s and 1980s. This example implies that standards should be revised over time, although it has also been observed that lower standards followed by remedial action have conferred substantial improvements in programme performance ( Blanks et al., 2002 ). Wallis et al. (2008) gave a demonstration of how careful surveillance of audit standards can lead to changes in practice and improved performance at the local and national levels.
Indicators such as detection rates are typically part of the monitoring system of most screening programmes, but the actual target values will vary according to the screening regimen, the target population, the underlying incidence in the programme’s location, and possibly aspects of the health-care delivery systems and the medicolegal environment ( Klabunde et al., 2001 ).
Table 5.2 shows selected standards developed in the USA. These standards include acceptable ranges for positive predictive values (PPVs) of recall for assessment and for recommendation for biopsy. They specify that the proportion of cases recalled for assessment that result in diagnosis of cancer should be 5–10%, and that the proportion of biopsies that result in diagnosis of cancer should be 25–40%. These are powerful measures of the process since they reflect detection rates, recall rates, and biopsy rates.
In a screening setting, the prevalence of the disease in screened subjects, expressed as a proportion, is usually very low; a very small number of those screened at each screening round are diagnosed with cancer, whereas thousands of women are screened negative. Typically, in European screening programmes, per 10 000 women screened, about 9500 will have a normal initial result and about 500 will be recalled for further assessment, of whom about 70 will have breast cancer. After the screen, about 10–30 will present with symptomatic interval cancer.
Components of the quality monitoring data listed above can be useful to estimate some important attributes of the screening programme, notably the specificity and sensitivity (the correct classification of negative and positive subjects) and the PPV. Specificity estimates the false-positives, or the complement of the proportion of screened-negative cases that are recalled for further assessment. The classic definition of test sensitivity is the probability that if the screening test is applied to someone with the disease, a positive diagnosis will result. PPV is the proportion of test-positive subjects who are diagnosed as cases at the end of the screening episode and is a function of the prevalence of the lesion. There are costs, both human and economic, to achieving a good balance of these performance parameters.
Other parameters of cancer detection have been defined by Hakama et al. (2007) : test sensitivity, programme sensitivity, and episode sensitivity.
In a clinical setting, test sensitivity is usually measured by comparison with a “gold standard”. This is rarely possible in a screening setting, where the objective of the test is the detection of a lesion in the preclinical detectable phase, and where only those with suspicious initial screening findings receive further investigation. Test sensitivity is the number of cancers detected at a screen divided by the sum of those detected at the screen plus the false-negatives. In principle, the false-negatives can be identified by a radiological audit of the original screening mammograms in those screened negative and subsequently diagnosed with interval breast cancer ( Houssami et al., 2006 ; Perry et al., 2006 ). This method of estimation involves assumptions about the audit quality, and the audit itself consumes resources, but it is a crucial learning tool and has the potential to improve the programme’s ability to detect early-stage cancers.
In the past, a common convention has been to estimate sensitivity as the number of cancers detected at a screen divided by the sum of those detected at the screen plus the interval cancers arising within 1 year. Two main sources of error have been identified: first, the interval cancers arising within 1 year will include true negatives that have entered the preclinical detectable phase during that year, and, second, they will not exclude those cancers missed at the screen but taking longer than 1 year to arise symptomatically ( Day, 1985 ). The reasoning implies that interval cancers are a mixture of missed and newly arising cancers, which tend to be more rapidly developing tumours. This, in turn, suggests that interval cancers will also be a mixture with respect to the aggressive potential of the cancers. In the epoch of film mammography, test sensitivity was reported to range from 83% to 95%, with the higher values observed for screening women older than 50 years ( Mushlin et al., 1998 ). In the epoch of digital mammography, the difference in sensitivity between age groups may be smaller ( Vinnicombe et al., 2009 ).
Programme sensitivity may be defined as the proportion of cancers diagnosed among women attending a screening programme or as the proportion of cancers diagnosed in the screening-eligible population. The first definition is the number of screen-detected cases divided by the sum of the screen-detected cancers plus the interval cancers. The second definition includes in the denominator cancers diagnosed among those who were invited but did not attend screening. Programme sensitivity is often described as the ability of the programme to detect cancers. It is generally estimated from steady-state screening, from the numbers of cancers diagnosed at several incident screens (not from prevalent screening) and the symptomatic cancers occurring in the same number of intervals between screens.
Programme sensitivity depends on the test sensitivity, the screening interval, and (depending on which measure is used) the attendance rate. It is typically estimated to be 50–60% ( Anttila et al., 2002 ; Zorzi et al., 2010 ). This means that in organized programmes, about half of the cancers in the target population are detected by screening. Of course, this will depend strongly on the rate of participation in screening.
Hakama et al. (2007) defined episode sensitivity as the incidence reduction in a specified period after screening compared with the expected incidence in the absence of screening, that is 1 − ( P 1 / P 0 ), where P 1 is the incidence among the screened subjects in the specified period after screening and P 0 is the expected incidence in the absence of screening (which, in practice, is difficult to estimate).
Taylor et al. (2002 , 2004 ) reviewed estimates of the proportional incidence in the first year of the screening interval, comparing international data published since 1975 and including results from randomized trials and service screening programmes in Australia, Canada, Italy, the Netherlands, Scandinavia, the United Kingdom, and the USA (Health Insurance Plan study). A large variability was reported, with an overall point estimate of the proportional incidence of 18.5% from all randomized trials and 27.3% from service screening programmes, corresponding to episode sensitivity estimates of 91.5% for the randomized trials and 72.7% for service screening.
A pooled analysis in the service screening centres of six European countries ( Törnberg et al., 2010 ) reported a large variation in screening sensitivity and performance, with a proportional incidence of 46% (episode sensitivity, 54%) in the 24 months after screening. The European standards ( Perry et al., 2006 ) were 30% and 50% for the proportional incidence at the prevalent screen and at subsequent screenings, respectively, corresponding to recommended episode sensitivities of 70% and 50%, respectively.
Note that all three measures discussed above require an estimation of interval cancer incidence. This illustrates the crucial nature of interval cancers in programme evaluation. Whereas screen detection rates are important, the future cancer risk in those screened negative is at least equally informative about the programme’s ability to detect cancer in the preclinical phase.
Bennett et al. (2011) noted the complexity of the evaluation of interval cancers on a large scale. They analysed 26 475 interval cancers in the NHS Breast Screening Programme (England, Wales, and Northern Ireland) and found a large variability in the regional estimates, with an estimate of a higher level than expected on the basis of the randomized trial experience. The conclusion was that comparison of different programmes is possible only if the methodology used is very thorough and guidelines are agreed upon in advance, with accurate follow-up and homogeneous reporting.
Table 5.1 includes standards for maximum interval cancer rates, that is, rates of symptomatic cancers that are diagnosed after a screen with negative findings and before the next scheduled screen (usually a period of 1–3 years). Together with prompt and nearly complete cancer registration, the interval cancer rate can be a powerful indicator of screening quality ( Bennett et al., 2011 ). The observation that interval cancer rates were very high in the early years of the United Kingdom programme in the East of England prompted a radiological audit, which consisted of re-reading previous screening mammograms, both of interval cancers and of non-cancers, without knowledge of the diagnostic result ( Day et al., 1995 ). This identified issues of sensitivity, which were later remedied, and served as a learning resource for quality improvement in other regions of England ( Duncan & Wallis, 1995 ). Interval cancer rates are now considerably lower in the East of England and similar to those in the rest of the United Kingdom ( Bennett et al., 2011 ; Offman & Duffy, 2012 ). The radiological audit of advanced disease may be suggested in health-care settings where cancer registration systems do not sufficiently identify interval cancers.
Interval cancer rates can also yield inferences about the effect of changes to the screening regimen. The policy of two-view mammography for incident screens was shown first to increase detection rates ( Blanks et al., 2005 ) and subsequently to reduce interval cancer rates by almost exactly the same absolute numbers ( Dibden et al., 2014 ). The concomitant reduction in interval cancer rates gave some assurance that the increased detection capability was not an overdiagnosis phenomenon.
Estimates and characteristics of interval cancers in national and regional screening programmes have been published, confirming the need for surveillance and improvement of service screening ( Ganry et al., 2001 ; Wang et al., 2001 ; Hofvind et al., 2006 ; Bucchi et al., 2008 ; Domingo et al., 2013a ; Carbonaro et al., 2014 ; Dibden et al., 2014 ; José Bento et al., 2014 ; Renart-Vicens et al., 2014 ).
The relationship between detection modality and tumour characteristics of breast cancers has been investigated ever since the first randomized trials ( Duffy et al., 1991 ). Recently, the renewed interest in interval cases and their radiological classification ( Houssami et al., 2006 ) has enabled the analysis of tumour characteristics by detection mode and interval type in terms of new biomolecular classifications and mammographic breast density at screening. Such analyses, along with recent findings with respect to genetic predisposition, have raised interest in personalized screening ( Hall & Easton, 2013 ). Although personalized screening is not simple to incorporate into existing programmes ( Paci & Giorgi Rossi, 2010 ), such interest does indicate that investigation of interval cancers can inform hypotheses to potentially improve screening policy.
As noted above, the most telling indicator of the effectiveness of a screening programme is its effect on breast cancer mortality. However, estimating this effect is not straightforward ( Duffy et al., 2007 ; Otten et al., 2008 ; Broeders et al., 2012 ; Independent UK Panel on Breast Cancer Screening, 2012 ). Temporal and geographical comparisons are potentially confounded with other parameters that influence breast cancer mortality; simultaneous temporal and geographical control yields more directly interpretable results ( Otto et al., 2003 ; Olsen et al., 2005 ). The introduction of breast screening as in Finland, with date-of-birth clusters randomized to receive screening first, yields results that may be interpretable directly as estimates of the efficacy of the programme ( Hakama et al., 1997 ). It is worth noting that such designs do not obviate the need for sufficient follow-up. In absolute terms, in the early years of a programme the adverse effects are enumerable, but the benefits in terms of numbers of breast cancer deaths avoided are not.
Arguably the most important issue for observational evaluation of screening and breast cancer mortality is the diagnostic period. Because of the generally good breast cancer survival rates, unrefined mortality (used hereafter to denote breast cancer mortality regardless of the time of diagnosis) in the epoch of screening will be contaminated by a substantial numbers of deaths from cancers diagnosed before screening was initiated ( Duffy et al., 2007 ). This will tend to bias results against screening. The bias can be avoided by using refined or incidence-based mortality (IBM), where mortality is ascertained specific to the diagnostic period ( Olsen et al., 2005 ; Swedish Organised Service Screening Evaluation Group, 2006a , b ). Alternatively, the bias can be minimized by estimating the mortality effect in a period beginning some years after the start of screening, albeit with some qualifications on interpretation ( Duffy et al., 2010 ).
Epoch of diagnosis also has implications for treatment and management of breast cancer, so that the before–after comparisons of mortality are almost invariably confounded with changes in treatment, as with the expansion in use of adjuvant systemic therapies in the 1980s and 1990s. This is considered further in Section 5.1.2 .
Concerns have been expressed with respect to ascertainment of cause of death ( Gøtzsche & Jørgensen, 2013 ). Results suggest that this is not a serious cause of bias ( Goldoni et al., 2009 ; Holmberg et al., 2009 ), partly because the number of women with advanced breast cancer who do not die of breast cancer is limited ( de Koning et al., 1992 ). In any case, it can be addressed by estimating the effect of screening on excess mortality in breast cancer cases, which does not require individual determination of cause of death ( Jonsson et al., 2007 ).
Methods and results in terms of breast cancer screening and mortality are dealt with in more detail in Section 5.1.2 , and possible surrogate indicators of breast cancer mortality are considered in Section 5.1.3 .
(a) general principles.
Attempts to estimate exact proportions of recent reductions in breast cancer mortality are subject to difficulties in modelling and interpreting the dynamism of incidence, behaviour, screening policy, treatment policy, and the correlations among these. In addition, there are always difficulties in interpreting directions of causality in changes, particularly in breast cancer incidence.
The main observational methods to assess the effect of screening are: (i) analysis of temporal trends in unrefined breast cancer mortality, reporting annual percentage changes in screening and pre-screening periods and change points when trends are estimated to change in magnitude or direction; (ii) comparison of unrefined mortality rates in screening or invited exposed populations with temporal, geographical, or other demographic control; (iii) the same comparison using IBM; and (iv) case–control studies where women who have died of breast cancer are compared with women who have not, with respect to screening histories before diagnosis of the case. In addition, modelling studies can provide information on outcomes beyond the limits of observational studies. This section outlines the principles and practice of each method, illustrating them with published results. First, two commonly occurring biases, and possible methods for their correction, are described.
Any estimate of the effect of being screened might be biased by factors influencing self-selection, such as the risk of death from breast cancer. In the Swedish breast screening trials, women not attending screening had a 36% higher risk of death from breast cancer compared with the uninvited control group ( Duffy et al., 2002a ). This was a combination of a lower incidence of breast cancer and a considerably higher case fatality rate ( Duffy et al., 1991 ). A difference of this nature would induce a bias in favour of screening if not addressed by design or analysis.
Cuzick et al. (1997) developed a method to correct for this bias in randomized controlled trials (RCTs), assuming a latent non-attender population in the control group. Duffy et al. (2002a) adapted this for case–control studies and later for other designs of observational studies ( Swedish Organised Service Screening Evaluation Group, 2006a ). The correction depends crucially on an estimate of the relative risk of breast cancer death in non-attenders compared with an uninvited population. Although this can be readily estimated within a given trial, in observational studies this is not generally the case. In the past, observational studies have relied on a relative risk estimate of 1.36 from the Swedish trials (e.g. Allgood et al., 2008 ) and, more recently, on estimates from the target population ( Paap et al., 2011 ). Paap et al. (2011) noted that in the Netherlands, the non-participant population had, if anything, a lower a priori risk of breast cancer death compared with the participant population. Table 5.3 shows the odds ratios (with and without correction for self-selection bias) for breast cancer mortality associated with screening, and the relative risks for non-participants in screening, in five regions of the Netherlands. Those regions with a non-participant relative risk greater than 1 had a corrected odds ratio that was less extreme than the uncorrected one, whereas those regions with a non-participant relative risk less than 1 had a more extreme corrected odds ratio. This leads to the observation that in the organized screening in the Netherlands, self-selection bias appeared to have only a minor effect ( Otto et al., 2012a ).
Odds ratios, with and without correction for self-selection bias, for breast cancer mortality associated with screening in five regions of the Netherlands.
Differences in prognosis between attenders and non-attenders could be explained by: a different underlying risk of disease; different help-seeking habits for symptoms, which lead, in turn, to differences in stage at presentation; varying compliance with treatment; or different comorbidities, which have a bearing on outcome ( Aarts et al., 2011 ). Socioeconomic status has been suggested as the major confounder of both outcome and participation in screening ( Palli et al., 1986 ; Aarts et al., 2011 ), although adjustment for it made almost no difference to the estimated effect of attending screening ( Palli et al., 1986 ).
There is greater uncertainty about the appropriate correction in observational studies with respect to randomized trials when estimating the effect of actually being screened. However, Duffy et al. (2002a) illustrated that the relative risk of breast cancer death may differ a priori between attenders and non-attenders, in ways that are not related to screening and thus completely annul the benefit observed among the screened population. The authors first considered a Swedish case–control study with an uncorrected relative risk of 0.50 for being screened, and then calculated that the a priori risk of breast cancer death among non-attenders would have to be 1.53 to be entirely due to self-selection bias, in a programme with 70% attendance. For a true (i.e. often suggested by trials’ meta-analyses) relative risk of 0.80 associated with invitation to screening, the relative risk would have to be 1.23. Such reverse calculation of the required size of the bias to annul the result, or to give a result consistent with the trials, may provide some assistance in interpreting the results of observational estimates of the effect of actually being screened.
Screening opportunity bias pertains particularly to case–control studies, where controls can only be exposed to screening if they attended their last screen, whereas cases can be exposed to screening if they attended their last screen or were screen-detected ( Walter, 2003 ). This means that if the screens at which any screen-detected cases were detected are included as exposure, there is a bias against screening, and if they are excluded, there is a bias in favour of screening. Duffy et al. (2008) developed a method that estimates the additional opportunity for screening exposure among the cases and yields a correction to the odds ratio for this, obtaining an estimate that lies between the odds ratios including and excluding the detection screen.
A common evaluation technique consists of comparing rates of unrefined mortality (i.e. regardless of time of diagnosis) in a screened versus an unscreened population (whether historical or contemporaneous or both). An early but very clear example of this approach is the estimation of the effect of the NHS Breast Screening Programme in England and Wales by Blanks et al. (2000) . The authors fitted age-cohort models to breast cancer mortality data recorded over the period 1971–1989, before the advent of substantial screening coverage, and projected these to estimate the expected mortality in the absence of screening for the period 1990–1998, in which the screening programme was achieving high coverage. The authors compared the observed reductions in mortality with expected rates for the age groups 55–69, 50–54, and 75–79 years. The observed reductions in breast cancer mortality were 21.3% in the age group 55–69 years and 14.9% in the age groups 50–54 years and 75–79 years, age groups that might reasonably be expected to be unaffected by breast screening. The estimated reduction in breast cancer mortality associated with the NHS Breast Screening Programme was 6.4%. The authors noted that the inclusion of deaths from cancers diagnosed before the screening started would dilute the observed benefit of screening. Duffy et al. (2002b) subsequently showed that more than half of the breast cancer deaths in a given 10-year period are from cancers diagnosed before screening started, and consequently that the effect on mortality from cancers diagnosed in the screening epoch is likely to be twice as high as the 6.4% mortality reduction estimated. For this and other reasons, the full effect of the screening programme was unlikely to be seen until between 2005 and 2010.
As with any temporal comparison, the issue of confounding with treatment arises. Although the age groups above the screening range might not have benefited fully from the therapeutic changes, it is reasonable to suppose that the age groups below the screening range would have done so. The greater mortality reduction in 1998 in the age group 50–54 years compared with the age group 75–79 years (17.0% vs 12.8%) appears to bear this out.
Incidence-based mortality studies are cohort studies in which the incidence-based mortality from breast cancer diagnosed after the first invitation to screening is compared with an estimate of expected breast cancer mortality in the absence of screening. The breast cancer mortality expected in a situation without screening can be estimated using breast cancer mortality rates in a cohort not (yet) invited to screening, or using historical data on breast cancer mortality patterns from the same region. Ideally, historical and current data on breast cancer mortality from a region in which screening is absent are included, to account for possible temporal changes that affect breast cancer mortality (e.g. improvements in breast cancer treatment). Incidence-based mortality studies have several methodological advantages, including avoidance of lead-time bias and achieving appropriate correspondence in time of the breast cancer incidence and mortality between the study and control cohorts.
Suppose a screening programme started in 1990, in a stable target population of 100 000 women aged 50–69 years. One might have available data to compare breast cancer mortality in the 1 000 000 person–years of eligible follow-up in 1990–1999 with the same mortality in the corresponding 1 000 000 person–years of observation in 1980–1989, before the screening was initiated. However, such a comparison of deaths from breast cancer regardless of time of diagnosis would include in 1990–1999 deaths from breast cancers diagnosed before 1990 and so with no potential for exposure to screening. The IBM approach would include only deaths from cancers diagnosed at ages 50–69 years during either 1990–1999 or 1980–1989. Although this approach may incur some conservative bias due to lead time, this would be outweighed by the correct classification of exposure to invitation to screening ( Swedish Organised Service Screening Evaluation Group, 2006a ). Since the risk of breast cancer death may change with time since diagnosis, it is desirable that the observation periods with and without screening be of equal duration.
A real instance of this approach is now considered. The study of Olsen et al. (2005) compared changes in incidence-based breast cancer mortality in the period 1991–2001 in the Copenhagen screening programme with changes in the rest of Denmark (which was without a screening programme and was consequently taken as the national control group). Incidence-based breast cancer mortality rates declined from 69 per 100 000 in the pre-screening period to 52 per 100 000 in the screening period in the Copenhagen area, and almost no change (from 52 to 53 per 100 000) was observed in the national control group. This observation led to an estimated relative risk of breast cancer death of 0.75 (95% confidence interval [CI], 0.63–0.89). Any changes in therapy in the Copenhagen area over the period would also have been seen in the national control group, given the standardization of treatment performed in accordance with the Danish Breast Cancer Cooperative Group ( Fischerman & Mouridsen, 1988 ). Since the only deaths included were those from cancers diagnosed during the relevant periods, there was no dilution of the effect of the screening due to deaths from cancers diagnosed before screening started.
In a case–control study, exposure to screening (history of breast cancer screening attendance) is compared between women who died of breast cancer (cases) and women who did not die of breast cancer (controls). Potentially important biases associated with case–control studies include selection bias and information bias related to the time at which exposure is defined. Because screening attendance is used as the exposure measure, selection bias plays an important role, as women attending screening might be more health-conscious than women not attending screening. Selection bias influences the estimated effect of the study in favour of screening but may be corrected, at least partially, using statistical methods (adaptation by Duffy et al., 2002a of the correction of Cuzick et al., 1997 for RCTs). For a correct estimate of selection bias, it is crucial to have data available on the variables that influence breast cancer mortality, or on breast cancer mortality between attenders and non-attenders ( Paap et al., 2014 ).
Generally speaking, the definition of exposure to screening can lead to bias both in favour of screening and against screening. If exposure is defined as “ever screened” versus “never screened”, bias will occur in favour of screening. Because all cases have died of breast cancer and were therefore very likely to have been diagnosed with breast cancer some time before death, most will have stopped being invited to screening some time before death. In contrast, controls (most of whom were not diagnosed with breast cancer) would have continued to be invited to screening up to near the time of their death, and would thus have been more likely to be exposed to screening. This difference in the probability of having been screened would lead to bias in favour of screening. This bias in favour of screening is eliminated if exposure is defined as screening attendance to the time of the case’s breast cancer diagnosis, so that exposure stops simultaneously for cases and controls. Although in this design the bias in favour of screening is eliminated, bias against screening is likely to occur because a case is eligible to be screened until cancer is detected either clinically or by screening, whereas controls matched to a case with a cancer detected by screening are eligible to be screened only until the cancer of their matched case is detected by screening. This bias can be corrected by defining exposure for controls matched to cases with a screen-detected cancer to the time at which cases with a screen-detected cancer would have been clinically diagnosed (in the absence of screening), but this requires an estimate of the screening lead time for each case ( Connor et al., 2000 ). Exposure of controls matched to cases with a clinical diagnosis remains unchanged.
Essential elements in performing case–control studies are: (i) sampling cases and controls from the same population (i.e. controls that would have had the same probability of becoming cases); (ii) qualitatively equal information on the primary outcome measure; and (iii) correct definition of (population-based) mammography screening exposure. In countries with complete population registries and full coverage of cancer registries and vital statistics, such case–control studies approximate nested case–control studies. Examples of this type of study are the case–control studies done in the Netherlands (e.g. Paap et al . , 2014 ).
Case–control studies consistently report a greater breast cancer mortality reduction associated with screening (up to 50%) compared with the RCTs ( Walter, 2003 ; Broeders et al., 2012 ). Only a small part of this difference in breast cancer mortality reduction can be explained by differences in study design. RCTs compare breast cancer mortality in women offered screening with that in women not offered screening. The estimated effect is influenced by the participation rate (women who decline the invitation to screening are included in the screened group) and by contamination of the control group. In contrast, most case–control studies estimate breast cancer mortality reduction in women who are screened compared with women who are not screened, thereby excluding women who decline the invitation to screening from the case group and avoiding contamination of the control group. Therefore, the effect estimate assessed in case–control studies can be expected to be stronger, even if adjusted for selection effects.
The independent United Kingdom panel on breast cancer screening reviewed the usefulness of case–control studies in estimating breast cancer mortality reduction associated with screening and considered that bias could inflate the estimate of benefit and that the RCTs provide more reliable evidence for mortality reduction ( Marmot et al., 2013 ). However, the number of screens performed in current screening programmes outnumbers the women screened in the RCTs by hundreds of millions. Therefore, studies conducted in high-quality organized invitation systems, which have almost complete follow-up data and high acceptance rates, can best estimate whether currently implemented programmes are of benefit to women invited (effectiveness).
The case–control approach is a relatively quick and inexpensive one, based on the principle that if the screening is reducing mortality, women who have died of breast cancer will be characterized by lesser screening histories than those who have not. It does have specific complexities and risks of bias ( Walter, 2003 ; Duffy, 2007 ; Verbeek & Broeders, 2010 ). However, these can to some extent be addressed by design and analytical tactics. Within opportunistic, rather than organized, screening, the case–control approach is one of the few evaluation options available. In some health-care environments, it may not be possible to link screening and mortality records, in which case the advanced disease status might be used to define cases (with the possibility to be interviewed with respect to screening status in the absence of screening records).
A notable feature of the case–control evaluation is that its primary comparison is made between participants and non-participants in the screening programme, and this option thus introduces the possibility of self-selection bias. Duffy et al. (2002a) developed a correction for this bias that requires a reliable estimate of the relative risk of breast cancer death in non-attenders versus those not invited to screening. This may be difficult to estimate; however, the method also provides an estimate of how large this relative risk would have to be for the observed benefit to be entirely due to self-selection bias.
An example of a case–control evaluation is the study of the effect of participation in the BreastScreen Australia programme, which has been inviting women aged 50–69 years to 2-yearly mammography since the mid-1990s ( Nickson et al., 2012 ). The 427 breast cancer deaths occurring at some time during 1995–2006 were compared with 3650 controls who were alive. A variable number of controls, selected by incidence density sampling, were matched by month and year of birth to cases ( Greenland & Thomas, 1982 ). In each case–control matched set, a date of first diagnosis of breast cancer (in the majority, the date of diagnosis of the case) was defined as the reference date. The primary definition of exposure to screening was having had a mammogram between the woman’s 50th birthday and the case–control set reference date. Exposure to screening was less common in cases than in controls (39% vs 56%). The odds ratio associated with screening, adjusted for remoteness of residence and socioeconomic status, was 0.48 (95% CI, 0.38–0.59). A series of sensitivity analyses yielded a range of 0.44 to 0.52.
This result may be affected by self-selection bias, despite the adjustment for socioeconomic status and the various sensitivity analyses performed. However, to be entirely due to self-selection bias, the a priori risk of breast cancer death in non-participants compared with uninvited women would have to be at least 1.80, which seems unlikely given the evidence that participants are at a higher risk of breast cancer than non-participants ( Thompson et al., 1994 ; van Schoor et al., 2010 ; Beckmann et al., 2013 ). Clearly, the self-selection bias can act in either direction. However, the results do indicate that case–control evaluations appear to be less conservative compared with prospective evaluation approaches.
An ecological study makes use of aggregated data for exposure or outcome identification, or both, rather than individual-level assessment of the association of the exposure with the outcome.
Ecological studies are generally accorded a lower status than randomized trials or studies using individual data, such as case–control and cohort studies. However, there may be cases where a well-conducted ecological study is more pertinent than a poorly conducted cohort or case–control study. In fact, for population interventions such as mammography breast cancer screening, the distinctions between these study types may be blurred, making it more important to consider the studies on a case-by-case basis, or at least according to a finer subdivision of types.
Two factors limit the ability to interpret findings in ecological studies. First, the ecological fallacy relates to the uncertain relationship between the mean and the median of characteristics of individuals in cells of aggregated data. Thus, the average use of screening in region A may be higher than that in region B, but if this average is due to very intensive use by a small number of women, one would not expect to see an overall mortality advantage for the women in region A. Second, differences in outcomes may be explained by other risk factors that differ between two regions. These may not be adjusted for, because they are unknown, are unmeasured, or are measured only on average (which returns one to the ecological fallacy). Adequate treatment of these two issues is a necessary condition for considering an ecological study as informative with respect to the effectiveness of mammography screening.
Ecological studies for breast cancer mortality compare data in countries or areas before and after the introduction of screening (interrupted time series), or concurrently between areas with and without screening (geographical comparisons). In the first type of study, extrapolation of time trends means that decisions must be made, for example about the linearity or otherwise of the trend, the choice of time periods considered as “before” and “after” screening, and the age groups included. In the second type of study, choices must be made about the areas to include, the time period considered, and the age groups included. Such decisions, which can appear to have been made rather arbitrarily, can have a profound impact on the estimates obtained. Lack of comparability and different time trends in the groups being contrasted could lead to substantial bias.
Ecological studies that use temporal trends fit regression models to national or regional published mortality data, commonly to estimate annual rates of change in mortality over time and to assess whether and to what extent breast cancer screening affects them. The change points are either dictated by the date of introduction of screening programmes or estimated from the data using joinpoint regression models ( Mukhtar et al., 2013 ). Studies comparing the levels of mortality rates between screening and non-screening periods are not included in this definition (please refer to Sections 5.1.2b and c ).
Mukhtar et al. (2013) analysed unrefined breast cancer mortality data (i.e. regardless of epoch of diagnosis) from 1971 or 1979 to 2009 in England, using log-linear models with joinpoint regression. They estimated similar contemporaneous downward trends in mortality during the screening epoch for women younger than 50 years and for those older than 50 years, the lower age limit for screening in Oxford. The joinpoint regression estimated no changes in trends for women aged 64 years or younger but significant changes in the late 1980s in older women. In England as a whole, the authors estimated the largest decreasing relative trend in women younger than 40 years. Years of peak mortality were observed in the mid- to late 1980s, before an effect of screening would be expected.
The authors concluded that screening was unlikely to have affected breast cancer mortality. Problems with this interpretation include the following. (i) The greatest mortality reduction in the most recent period was observed for the youngest age group. Rates were rising in the screening age group until the mid-1980s and falling thereafter. (ii) Because of the methodology’s choice of discontinuities at different ages, the calendar periods comparing the screening and non-screening age groups are not the same. (iii) Screening was mostly confined to ages 50–64 years, and the effect on mortality would be quite substantial in the late sixties and early seventies rather than in the early fifties. (iv) The emphasis on individual years of peak mortality and year-to-year trends loses sight of the more stable mortality estimates as a whole. The level of mortality was considerably lower in the screening epoch than in the pre-screening epoch, and this difference was most pronounced in the screening age group. (v) The maximum number of change points allowed should be specified. This will also affect their estimated occurrence.
Usually, it is most difficult to anticipate the occurrence of a change point, or its magnitude, based on year-to-year trends in unrefined mortality. This may influence the subjective decision about the number of joinpoints and about whether trends of decreasing mortality would have continued unabated in the absence of screening. Nevertheless, despite the significant complexities of analysis and interpretations, trend studies can be informative, such as the Otto et al. (2003) study.
Formally, RCTs answer one specific outcome question, namely whether mammography screening reduces breast cancer mortality, given the exact design features, like fixed interval, starting age, and stopping age, and given the background situation of the control group to compare with. Modelling studies are generally intended to predict outcomes beyond the (limited) end of the trial follow-up, and for different schedules of screening. They seek to avoid possible overestimation of the effect of screening on breast cancer mortality, due to lead-time and length bias, by modelling the breast cancer process more directly. The essence of modelling is simulating the natural history of disease, based on the best available data. This is realized by incorporating variables associated with the disease process and with detection and treatment of breast cancer, including the mean duration of the preclinical detectable phase, the probability of transition to the next tumour stage, age- and stage-specific sensitivity of mammography, and stage-specific response to treatment ( Berry et al., 2005 ; Groenewoud et al., 2007 ). As an example, the number and the time frame of interval cancers being diagnosed give estimates of sensitivity, whereas the detection rates (by stage, age, calendar year, etc.) and interval cancers together give information on the sojourn times of disease (duration of period when cases are screen-detectable). Modelling produces estimates of these unobservable phenomena, and thus there is sometimes scepticism about the evidence coming from modelling studies. Modelling tries to incorporate all available screen and non-screen data and to give the best estimate of the natural history of disease and of what would have happened if no screening had been implemented. In the evaluation of screening, when it is already being introduced, such model predictions are valuable to evaluate and steer the programme, and they are also advisable before implementation for estimating the optimal programme of screening with its benefits and harms as well as its cost–effectiveness. With good estimates, especially of the screen-detectable period, overdiagnosis can be estimated ( van Ravesteyn et al., 2015 ).
However, all good modelling analyses that predict the consequences of treating earlier in the natural history of disease are dependent on efficacy measures, from RCTs or high-quality observational studies, to estimate such results. Therefore, high-quality models are calibrated to such high-quality data ( de Koning et al., 1995 ). The advantage is that differences in protocol, for example attendance and referral rates, and in follow-up period can specifically be taken into account.
In such modelling, the natural history of breast cancer in the absence of screening is first modelled. Some women in the simulated population may develop breast cancer, which develops from a small preclinical lesion to a symptomatic cancer, possibly leading to breast cancer death. In each stage, a lesion may grow to the next stage, regress, or be clinically diagnosed because of symptoms. The natural course of the disease may be interrupted by screening, at which a preclinical lesion can become screen-detected. Screen detection can result in the detection of smaller tumours, which may entail a survival benefit. Each screen-detected or clinically diagnosed tumour may be treated with adjuvant systemic therapy, which may also improve survival. Critical components of such models are the assumed natural history component, the effects of interrupting by screening or treatment, and extrapolating lifetime harms and benefits ( Heijnsdijk et al., 2012 ). In principle, such elements are calibrated and validated against data from trials and observational studies, and criteria to evaluate models have been proposed ( Habbema et al., 2014 ).
As noted above, although in principle the main indicator of the effectiveness of a screening programme is its impact on breast cancer mortality, to estimate this impact in practice can be complicated. The population incidence of advanced-stage disease ( Smith et al., 2004 ; Autier et al., 2011 ) or predicted mortality from the stage of disease diagnosed have been suggested as surrogates for mortality. Randomized trials show that screening that results in a reduction in the incidence of node-positive breast cancer is also accompanied by a reduction in mortality ( Smith et al., 2004 ). A review confirmed this strong inverse association of exposure to screening and of screen detection with nodal status and tumour size ( Nagtegaal & Duffy, 2013 ). To consider potential confounding, the incidence of disease should be compared before and after the introduction of screening, to account for changes in treatment as well as more complete pathological staging and reporting (e.g. the implementation of sentinel node biopsy) in the screening epoch. This gives rise to further complexities of analysis and interpretation of data ( Swedish Organised Service Screening Evaluation Group, 2007 ).
Another possible confounder is the increase in breast cancer incidence recorded in almost all parts of the world in the second half of the 20th century, which is related to mortality and incidence of advanced disease as well as to the introduction of screening. Thus, there are methodological problems when trying to estimate the expected incidence of disease by stage in the absence of screening.
Despite these problems, the rates of advanced-stage disease are still a very direct measure of the impact of early detection by screening, as several studies have reported. To estimate the potential beneficial effect, not simply the proportion of cases with advanced-stage disease but also the reduction in the absolute rate of advanced-stage disease should be reported.
Thus, the incidence of advanced-stage disease might be used as a surrogate for the effect of screening on mortality, but the above-mentioned limitations should be considered. Other indicators include the detection rate of interval cancers and of small tumours, which are necessary but not sufficient indicators of the success of screening ( Day et al., 1989 , 1995 ; Tabár et al., 1992 ). Although they are less direct, these indicators are often more generally observable than the absolute population incidence of advanced-stage disease.
5.2.1. incidence-based cohort mortality studies.
IBM studies are the most methodologically robust studies for evaluating the effectiveness of service mammography in reducing breast cancer mortality (see Section 5.1.1 ). They are cohort studies usually conducted in association with a population-based mammography screening programme. Their defining feature is the observation of deaths from breast cancer in women diagnosed after their first invitation to (or attendance to) mammography screening, that is, at a time when their risk of breast cancer death could have been affected by screening. The expected number of breast cancer deaths is estimated in women diagnosed with breast cancer but not invited to screening compared with a matching cohort of women over a similar period of time.
The screening and non-screening cohorts can be fixed or dynamic, most commonly dynamic. For those invited to screening, the date of first invitation is taken from screening records or is estimated from the cohort member’s residence location and the history of the roll-out of screening in the study area and period. For those not invited to screening, the date of first invitation may be allocated to correspond in age and time to those invited, or at about the midpoint of the first screening round for those invited. The two cohorts’ age distributions are usually matched, as are the periods over which their breast cancer experience is recorded. In most cases, incident breast cancers during the accrual period for the study (which begins at the date of first invitation to screening for each woman) and the associated breast cancer deaths are identified in a population-based cancer registry, and deaths from other causes in a regional or national death register. In some studies, one or both cohorts have also been identified in national registers and individual women tracked into and out of the cohorts for accurate estimation of person–years of experience; otherwise, the person–years are estimated using aggregated population data.
This description of the results of IBM studies is based on studies correctly characterized as IBM studies, mostly covered by two recent systematic reviews. The first of these, the Euroscreen review, systematically searched for relevant studies published up to February 2011 in women aged 50–69 years covered by European population-based screening mammography programmes ( Broeders et al., 2012 ; Njor et al., 2012 ). The second had a similar search strategy to the Euroscreen review but without age restriction or limitation to European populations, and included studies published up to January 2013 ( Irvin & Kaplan, 2014 ). Additional IBM studies were found in an unrestricted systematic search that covered literature published between March 2011 and 22 July 2014. One study published after July 2014 ( Coldman et al., 2014 ) and two early studies not identified in the searches ( Morrison et al., 1988 and Thompson et al., 1994 ) were also known to the Working Group.
Four analyses that were excluded from the Njor et al. (2012) review report were also excluded by the Working Group, on the grounds that they were based exclusively on some or all of the data used for previous reports. However, there remains significant overlap among several studies, which is detailed below.
In almost all instances, the studies reviewed were conducted in areas where population-based service mammography screening had been implemented. There is, in principle, no reason for not conducting such studies within a population exposed only to opportunistic screening, but they are more readily conducted in areas of population-based screening and the Working Group knew of no IBM studies that had been conducted in an area with exclusively opportunistic screening.
The following summary of results of IBM studies is organized into two broad sections: studies that report on breast cancer mortality reduction after mammography screening of women in age groups that include most or all of the age range 50–69 years, and studies that report on mortality reduction from screening in an age group that lies mainly below or above that age range (i.e. women younger than 50 years or older than 69 years).
The results of studies of mammography screening mainly in women aged 50–69 years are summarized in Table 5.4 and Table 5.5 . Table 5.4 covers estimates of relative risk of breast cancer death in women invited to mammography screening relative to women not invited. Table 5.5 does the same for women who were invited and attended screening relative to women who were invited but did not attend. Studies are ordered in the table by the country in which they were conducted (with countries in the order in which their mammography screening programmes were first introduced) and within each country by the earliest date of mammography screening that was included in the analysis.
Incidence-based mortality studies of the effectiveness of invitation to mammography screening a mainly in women aged 50–69 years, by country and follow-up period.
Incidence-based mortality studies of the effectiveness of participation in mammography screening a mainly in women aged 50–69 years, by country and follow-up period.
All analyses reviewed here included women in the age group 50–69 years, with the exception of four analyses in which the women invited or otherwise targeted for screening were aged up to 59 or 64 years and one in which only women from age 55 years were invited. Eight analyses included women invited to screening before age 50 years, and five analyses included women invited to screening beyond age 69 years.
The six reports based on population-based mammography screening in Sweden have multiple overlaps in space and time; that is, they drew on geographical mammography experience for more than a year that overlapped with that drawn on by at least one other study. The experiences in the reports of Duffy et al. (2002a , b ) are almost completely a subset within that of the Swedish Organised Service Screening Evaluation Group (2006a , b ) reports; however, the reports of Duffy et al. (2002a , b ) provide valuable additional results and so are included separately in Table 5.4 and below. The whole mammography experience of Jonsson et al. (2007) is also included in that of Swedish Organised Service Screening Evaluation Group (2006a , b ), but it does provide some independent information since it uses contemporary and not historical control areas. Most of the screening experience in two of the seven screening areas of Jonsson et al. (2001) overlaps with that in Swedish Organised Service Screening Evaluation Group (2006a , b ), and two of the control counties overlap more than 50% of the time with the control counties in Jonsson et al. (2007) . The screening experience of the one screening county in Jonsson et al. (2003a) overlaps by 2 years that of Duffy et al. (2002a , b ) and by 1 year that of Swedish Organised Service Screening Evaluation Group (2006a , b ). The screening experience of one of the two counties included in Tabár et al. (2001) is also included in Duffy et al. (2002a , b ) and Swedish Organised Service Screening Evaluation Group (2006a , b ).
Sweden’s first population-based mammography screening programme was introduced in 1974 to cover women aged 40–64 years in Gävleborg County. Jonsson et al. (2003a) primarily compared IBM in Gävleborg County with an age-matched control population from four neighbouring counties without mammography screening programmes. Cohorts of women were defined in Gävleborg County according to the date at which invitation to screening began in their district, and corresponding cohorts were created in the control counties. Incident breast cancers and their dates of diagnosis were identified, and their date and cause of death obtained from the Swedish Cancer Registry; aggregated population data were used to estimate person–years at risk. The study also included a reference period (1964–1973), in which any pre-existing difference in breast cancer mortality between Gävleborg County and the control counties could be estimated and adjusted for in the analysis. Incident breast cancers were accrued for 10 years, and the follow-up period for breast cancer mortality was 22 years; cases were accrued only in the age group 40–64 years, and follow-up extended to age 79 years. [These differences in accrual and follow-up periods and age groups created the possibility of lead-time bias in the results. Also, bias due to inclusion of some cases of breast cancer that occurred early in the roll-out of screening and before the first invitation to screening (inclusion bias) was possible.] The estimated IBM relative risk for death from breast cancer was 0.86 (95% CI, 0.71–1.05) based on breast cancer deaths ascertained as the underlying cause of death from the death certificate and adjusted for age, follow-up time, county, and period (study or reference). Corresponding relative risks were 0.82 (no CI stated) after adjustment for lead-time and inclusion biases, 0.82 (95% CI, 0.65–1.03) when based on an estimate of excess mortality due to breast cancer, which does not require use of the certified underlying cause of death, and 0.93 (95% CI, 0.77–1.11) when based on the “rest of Sweden” as the control group.
The relative risk of 0.86 (95% CI, 0.71–1.05) was chosen from the alternatives listed above to be reported in the table. This choice was made a priori on the grounds that: (i) the relative risk was based on the underlying cause of death (the excess mortality measure is not consistently reported in the studies reviewed); (ii) it was the most fully adjusted relative risk that also included its 95% confidence interval; (iii) the Working Group considered four neighbouring counties to be a more nearly similar control group for the study group than the whole of the rest of Sweden; and (iv) this study overlapped the least with other Swedish studies.
Jonsson et al. (2001) and Jonsson et al. (2007) had fundamentally the same design as Jonsson et al. (2003a) , except that Jonsson et al. (2007) made historical rather than geographical comparisons of breast cancer mortality in women invited to screening in a later period with that in women in the same population not invited to screening in an earlier period. [ Jonsson et al. (2007) is the weakest of the three, because of its overlaps with Jonsson et al. (2001) and Swedish Organised Service Screening Evaluation Group (2006a , b ) and because of the difference in the length of the follow-up periods in women invited and not invited to screening.] The IBM relative risk in invited women aged 50–69 years was 0.90 (95% CI, 0.74–1.10) (0.87 adjusted for inclusion bias, and with lead-time bias estimated to be −0.4%) in Jonsson et al. (2001) and 0.86 (95% CI, 0.86–1.17) (lead-time bias estimated to be −2%) in Jonsson et al. (2007) .
Tabár et al. (2001) estimated post-RCT effectiveness of mammography screening in the Swedish Two-County study by comparing post-RCT experience with a balanced period of pre-RCT experience. [The reporting of this analysis is limited; there is uncertainty as to whether the result may be affected by lead-time bias and whether there is any statistical adjustment of the relative risks.] To obtain the IBM relative risk for breast cancer mortality in women invited to screening, the authors first estimated the IBM relative risk for attendance to screening (by comparing breast cancer mortality in women aged 40–69 years who attended screening in 1988–1996 with that in women aged 40–69 years in 1968–1977, before any screening) and then adjusted this for self-selection bias to obtain an adjusted relative risk for invitation to screening of 0.52 (95% CI, 0.43–0.63). [However, this estimate was not adjusted for the underlying trend in breast cancer mortality between 1968–1977 and 1988–1996.]
In a similar historical control-design IBM study based in seven Swedish counties, Duffy et al. (2002a , b ) estimated an IBM relative risk of 0.74 (95% CI, 0.68–0.81) for screening in women aged 40–69 years based on 5–20 years of screening and follow-up until 1997 or 1998, and adjusted for lead-time bias and the underlying time trend in breast cancer mortality. For counties with 10 years or less of screening, the estimated relative risk was 0.82 (95% CI, 0.72–0.94), and for counties with more than 10 years of screening, it was 0.68 (95% CI, 0.60–0.77).
The Swedish Organised Service Screening Evaluation Group (2006a , b ) analysis was of a similar design but expanded to 13 areas of Sweden and had 11–22 years of screening experience of women aged 40–69 years or 50–69 years and followed up until 2001. The IBM relative risk for screening at age 40–69 years was 0.73 (95% CI, 0.69–0.77) after adjustment for the underlying trend in breast cancer mortality.
Peer et al. (1995) compared breast cancer mortality in women born in 1925–1939 who were resident in Nijmegen and were offered mammography screening every 2 years from 1975 until the end of 1990 with that of age-matched women resident in Arnhem and not offered screening. Cause of death was ascertained from clinical records and was considered to be breast cancer if metastases had been diagnosed and other causes of death could be ruled out. The IBM relative risk for breast cancer mortality in Nijmegen women relative to Arnhem women was 0.94 (95% CI, 0.68–1.29). [Breast cancer mortality in women aged 35–64 years had been reported to be lower in Nijmegen than that in Arnhem in 1970–1974. This difference was observed not to persist in the period 1975–1979. No adjustment was made for possible differences or trends in underlying breast cancer mortality rates.]
The United Kingdom Trial of Early Detection of Breast Cancer ( UK Trial of Early Detection of Breast Cancer Group, 1999 ) was a non-randomized trial that began in 1979 and preceded population-based mammography screening in the United Kingdom by 10 years. IBM to 16 years of follow-up was compared between two health service areas in which women aged 45–64 years were invited to be screened by mammography and clinical breast examination (CBE) every 2 years for four rounds, with CBE only in the intervening years, and two areas in which women received the usual care. The relative risk was 0.73 (95% CI, 0.63–0.84).
Five studies have reported IBM analyses of mammography screening in Finland. [Overlaps are not accurately identifiable from published reports but seem likely.] The study of Hakama et al. (1997) overlaps minimally with the studies of Anttila et al. (2008) and Sarkeala et al. (2008a , b ) because Hakama et al. (1997) covered screening in 1987–1992 and the other three covered screening from 1992 to 2002 or 2003. Anttila et al. (2008) and Sarkeala et al. (2008a , b ), which cover 410 and 260 municipalities, respectively, appear to overlap substantially; each of these two studies also overlaps with that of Parvinen et al. (2006) , in which the intervention group primarily covered the “entry” cohort in the city of Turku in 1987. The study of Anttila et al. (2002) , which included screening in Helsinki in the period 1986–1997, does not overlap with that of Hakama et al. (1997) or with that of Sarkeala et al. (2008a , b ) but is assumed to overlap with that of Anttila et al. (2008) in the period 1992–1997. On these bases, it appears that Hakama et al. (1997) , Anttila et al. (2002) , and Sarkeala et al. (2008a , b ) give nearly complete coverage of screening in Finland from 1986 to 2003 with minimal overlap.
Hakama et al. (1997) compared IBM in women aged 50–64 years invited and not invited to mammography screening in 84% of municipalities in 1987–1992, the first 6 years of nationwide screening in Finland. Individual year-of-birth cohorts of women were progressively invited for the first time during this period and experienced up to three screening rounds. The estimated relative risk of breast cancer death was 0.76 (95% CI, 0.53–1.09). The analysis of Anttila et al. (2002) of screening of women in Helsinki over the period 1986–1997 compared IBM in women born in 1935–1939, who had been invited to screening, with that in women born in 1930–1934, who had not. The estimated relative risk of breast cancer death was 0.81 (95% CI, 0.62–1.05) after adjustment for age at death and the estimated trend in breast cancer mortality from the trend across the two cohorts at age 40–49 years. [There may be lead-time bias in this result.] Using data from 260 Finnish municipalities and modelling the time trend in breast cancer mortality in the absence of screening, with mortality data from 1974–1985 providing estimated pre-screening mortality, Sarkeala et al. (2008a) estimated an IBM relative risk for invitation to screening in 1992–2003 of 0.78 (95% CI, 0.70–0.87) at age 50–69 years. All municipalities regularly invited only women aged 50–59 years. In those municipalities that had regularly invited women aged 50–69 years (and up to 74 years in some of these) throughout the study period, the corresponding IBM relative risk was 0.72 (95% CI, 0.51–0.97). Incidence and death were measured at age 60–79 years, whereas no impact was observed in municipalities that had stopped screening at age 59 years ( Sarkeala et al., 2008b ). Studies with variable screening policies provided no clear evidence for a difference in the relative risk for screening between the first 5 years ( Hakama et al., 1997 ) and the next 10 years ( Anttila et al., 2002 ; Sarkeala et al., 2008a , b ). In addition, the results of Parvinen et al. (2006) demonstrated a significant effect in women screened regularly at age 55–74 years since 1987 in the “entry” cohort of the screening programme in the municipality of Turku ( Table 5.4 ).
Paci et al. (2002) estimated the IBM relative risk for women aged 50–69 years invited to screening in the first 7 years of population-based mammography screening in Florence over the period 1990–1999. The expected number of deaths in the absence of invitation to screening was estimated from the expected number of incident breast cancers in women not yet invited to screening in each half-year of the period 1990–1996 and the estimated number of breast cancer deaths to 1999 (from estimated case fatality rates for up to 9.5 years after diagnosis) in women expected to be diagnosed with breast cancer in each of these half-year cohorts. The estimated relative risk was 0.81 (95% CI, 0.64–1.01). [The nature of 13 breast cancer deaths classified as “other” (neither invited nor not invited, and treated as not invited in the analysis) is unclear. If they had been treated as invited, the relative risk would have been 0.83.]
Based on a population-based mammography screening programme targeting women aged 45–64 years in Navarre, Ascunce et al. (2007) reported an IBM relative risk of 0.58 (95% CI, 0.44–0.75) for invitation to screening of women aged 50–69 years in 1997–2001. There was no adjustment for the overall trend in breast cancer mortality; the corresponding relative risk was 1.07 (95% CI, 0.66–1.74) in women aged 30–44 years and 1.03 (95% CI, 0.77–1.37) in those aged 75 years and older (outside the target age group). The relative risk adjusted for the average of these two trends was 0.56 (95% CI, 0.39–0.80).
Based on linked screening registry, cancer registry, cause of death registry, and population register data for individual women, Olsen et al. (2005) analysed IBM for invitation to screening in the first 10 years (1991–2001) of population-based mammography screening offered every 2 years to women aged 50–69 years in Copenhagen. Three comparison groups, Copenhagen in 1981–1991 and Denmark (except Copenhagen and two other areas with population-based screening before 2001) in 1991–2001 and 1981–1991 (secondary control groups to provide data on the underlying trend in breast cancer mortality), were constructed from women’s individual records in the population register, and the women were allocated pseudo-dates of first invitation. In all cases, women with prevalent breast cancer before their real date or pseudo-date of invitation were excluded. Analysis was done by way of a Poisson regression model of breast cancer mortality with age, whether invited or not, period, region, and interaction between period and region as covariates, thus adjusting the estimate of effect of invitation for differences in age, place, and time between invited and not invited women. The estimated IBM relative risk for invitation to screening was 0.75 (95% CI, 0.63–0.89).
A population-based programme that offers mammography screening every 2 years to women aged 50–69 years began as a pilot programme in four of the 19 Norwegian counties in 1996; roll-out to the rest of the country began 2 years later and was completed in 2005 ( Hofvind et al., 2013 ). Population-based screening was preceded by widespread opportunistic screening, to the extent that 38% of women who had their first mammogram within the programme in 1996–2006 had received a mammogram within the preceding 3 years, and 64% had ever had a mammogram ( Hofvind et al., 2013 ). Also, importantly, the roll-out of population-based screening in Norway was accompanied by or preceded by the establishment of multidisciplinary breast cancer care units in each county, in which all women being investigated or treated for breast cancer (whether screen-detected or not) were managed ( Kalager, 2011 ).
Three studies have reported on IBM in women invited to screening in the Norwegian population-based programme. One included population-based screening experience accumulated to 2001–2002 in women in the four pilot study counties ( Olsen et al., 2013 ). The second included the experience in the whole country to the end of 2005 ( Kalager et al., 2010 ), thus fully with overlapping the first. The third included the experience in the whole country to the end of 2009 ( Weedon-Fekjær et al., 2014 ), thus fully overlapping with both of the others.
Olsen et al. (2013) compared mortality from breast cancer diagnosed after screening began in women in the four pilot screening counties with the corresponding mortality in women in these counties over the 6 years before screening began. They adjusted their comparison for the underlying trend in breast cancer mortality by estimating it in five non-screening counties in similar periods before and after the beginning of 1996. The authors linked individual data obtained from the central population register, cancer registry, and cause of death registry for all women within the scope of their analysis; aggregated data were not required. However, they did not have individual screening data, so women in the screening counties during the screening period were allocated the date of first invitation to screening in their municipality as their first invitation date. Women included in the 6-year control period for the screening counties were allocated pseudo-dates of invitation 6 years before those in the screening period. The maximum period of screening was 6 years. The authors estimated the IBM relative risk for invitation to screening to be 0.89 (95% CI, 0.71–1.12). [This relative risk includes lead-time bias. Also, the underlying downtrend may have been greater in screening counties than in non-screening counties, due to the introduction of multidisciplinary breast cancer care units along with screening.]
The analysis of Kalager et al. (2010) used a similar approach to that of Olsen et al. (2013) except that it covered mammography screening in the period 1996–2005 and had individual data only for women who had been diagnosed with breast cancer. To address effects of the underlying trends in breast cancer mortality, comparisons were made between women invited to screening in 1996–2005 and corresponding women not invited to screening in 1986–1995, and vice versa. The comparisons were made primarily in women aged 50–69 years at diagnosis of breast cancer. [Balanced breast cancer accrual and follow-up periods and age groups avoided lead-time bias. However, as a consequence of the manner of roll-out of population-based screening in Norway, the group invited to screening and its historical comparison were concentrated in the second halves of the compared periods and the group not invited to screening and its historical comparison were concentrated in the first halves, making the latter a potentially inaccurate estimate of the underlying trend in breast cancer mortality in the group offered screening.] The authors estimated the relative risk comparing IBM for the group invited to screening relative to its historical comparison group to be 0.72 (95% CI, 0.63–0.81) and the corresponding relative risk in the group not invited to screening to be 0.82 (95% CI, 0.71–0.93). [From these relative risks, the Working Group estimated the IBM for invitation to screening adjusted for the underlying mortality trend to be 0.88 (95% CI, 0.73–1.05). The Working Group noted, in agreement with Olsen et al. (2013) , that the mortality trend in areas without screening may not accurately indicate the trend in areas with screening.]
Weedon-Fekjær et al. (2014) obtained individually linked data for all women, as Olsen et al. (2013) had, and in addition obtained individual dates of screening invitations. Unusually, however, they based their analysis of invitation to screening over the period 1996–2009 on the complete, dynamic population of Norwegian women aged 50–79 years in 1986–2009. Thus, their population of women unexposed to screening included women from 10 years before the implementation of population-based screening; as a result, they drew on nearly 13 million person–years of experience before invitation to screening and only 2.4 million after. The IBM relative risk for invitation to screening was estimated to be 0.72 (95% CI, 0.64–0.79) using a complex Poisson regression modelling approach. [The authors noted that they could not exclude possible effects of the establishment of multidisciplinary breast cancer care centres in parallel with the roll-out of the screening programme.]
[The relative risks for invitation to screening of these three, overlapping studies of the Norwegian experience are compatible to the extent that their 95% confidence intervals overlap, although the upper limit for the Weedon-Fekjær et al. (2014) study is less than the point estimates for the other two studies, suggesting that it could be lower. In principle, a lower relative risk in Weedon-Fekjær et al. (2014) would be expected because: it includes a later 4 years of the population-based programme’s experience than the other two studies; it would be based, on average, on longer periods of individual women’s experience in the programme; and it would be less affected by the previous high level of opportunistic screening. It might also, perhaps, be affected by the inclusion of a large volume of pre-screening breast cancer mortality experience, which, in the event of a falling trend in underlying breast cancer mortality, might produce an artificially lower relative risk. There is evidence of such a trend ( Kalager et al., 2010 ), and it could be sufficient to explain the difference between the estimate of Weedon-Fekjær et al. (2014) and those of the other two studies. Although the adjustment for period should have addressed this issue, the statistical dominance of person–years before 1996 may have compromised the effectiveness of this adjustment.]
The IBM relative risks for invitation to screening ranged overall from 0.58 to 0.94, with a median value of 0.78. Lead-time bias was the most common residual bias and would be expected to be conservative. If the Swedish, Finnish, and Norwegian studies that are overlapped substantially or fully by other studies ( Duffy et al., 2002b ; Jonsson et al., 2007 ; Anttila et al., 2008 ; Kalager et al., 2010 ; Olsen et al., 2013 ) are removed, the range of the remaining 14 studies is the same as for all 19 studies and the median is little changed, at 0.77. Furthermore, if all Norwegian studies are removed because of the introduction of multidisciplinary breast care centres in parallel with screening, the range remains the same and the median is 0.76. The United Kingdom Trial of Early Detection of Breast Cancer (relative risk [RR], 0.73; 95% CI, 0.63–0.84) included annual CBE in the intervention.
The design and results of studies reviewed are summarized in Table 5.5 . Studies are ordered in the table by the country in which they were conducted (with countries in the order in which their mammography screening programmes were first introduced) and within each country by the earliest date of mammography screening that was included in the analysis.
Most of the studies in Table 5.5 were based on the same mammography experience as was used for analyses of the outcomes of invitation to screening. Self-selection for attendance is an important issue in these analyses because the numerator for the IBM relative risk for breast cancer mortality is based on the experience of women attending screening while the denominator is based on all women in a different era or area who were not invited to screening or on women in the same area and era who chose not to attend screening. Self-selection may bias the IBM relative risk estimate if it creates a difference in the underlying risk of breast cancer death between women attending screening and all women, or women not attending screening.
Tabár et al. (2001) reported an estimate of the IBM relative risk for women aged 40–64 years attending screening of 0.37 (95% CI, 0.30–0.46). [The estimate appears not to have been adjusted either for self-selection or for the underlying time trend in breast cancer mortality. However, data on this trend in women aged 20–39 years in 1968–1977 or 1988–1996 were reported, and the Working Group used this trend to obtain an adjusted IBM relative risk of 0.46 (95% CI, 0.21–0.97). This relative risk may still be affected by self-selection bias.]
The other three Swedish studies that estimated IBM relative risk for attendance to screening ( Duffy et al., 2002a , b ; Swedish Organised Service Screening Evaluation Group, 2006a , b ; Jonsson et al., 2007 ) overlapped substantially with one another in their coverage of the screening experience, and Swedish Organised Service Screening Evaluation Group (2006a , b ) included the experience of one of the counties analysed in Tabár et al. (2001) . These three studies variously covered screening of women aged 40–74 years, but mostly aged 50–69 years, and screening during various parts of the period 1978–2001. The results, each adjusted for self-selection bias, were reasonably similar ( Table 5.5 ): the IBM relative risks were 0.61 (95% CI, 0.55–0.68) for Duffy et al. (2002a , b ), 0.57 (95% CI, 0.53–0.62) for Swedish Organised Service Screening Evaluation Group (2006a , b ), and 0.70 (95% CI, 0.57–0.86) for Jonsson et al. (2007) . The methods of adjustment for self-selection bias were, respectively, that of Duffy et al. (2002a , b ), a refinement of that method as reported in Swedish Organised Service Screening Evaluation Group (2006a , b ), and the method of Cuzick et al. (1997) .
One Finnish study has estimated the IBM relative risk for attendance to screening ( Sarkeala et al., 2008b ). Based on the same data set as used for Sarkeala et al. (2008a) , this study was designed primarily to assess the effect of different screening centre policies on screening effectiveness. Screened women attended between 1992 and 2003; unscreened women included those residing in the same areas in 1974–1985 and women who were invited in 1992–2003 but did not attend. The IBM relative risk for attendance to screening was 0.63 (95% CI, 0.53–0.75), adjusted for self-selection bias using the method of Cuzick et al. (1997) .
Based on a similar population of women invited to the first screening round (1991–1993) in Florence described in Paci et al. (2002) ( Table 5.4 ), Puliti & Zappa (2012) followed up women invited to mammography screening every 2 years at age 50–69 years for incidence of breast cancer to 2007 and mortality from breast cancer and other causes to 2008 ( Table 5.5 ). The estimated IBM relative risk for women who had ever been screened relative to those who had never been screened was 0.51 (95% CI, 0.40–0.66). This estimate was adjusted for marital status and small-area deprivation index in the hope of reducing self-selection bias. [There would also have been some lead-time bias because the mortality follow-up period was 1 year longer than the period of incident breast cancer accrual.]
Based on data obtained from seven of the 12 provincial mammography screening programmes established in or after 1988 under the Canadian Breast Cancer Screening Initiative, Coldman et al. (2014) reported an IBM relative risk of 0.60 (95% CI, 0.52–0.67) for women who were screened at least once in the period 1990–2009 ( Table 5.5 ). For the seven individual provinces, the relative risk ranged from 0.41 (95% CI, 0.33–0.48) in New Brunswick to 0.73 (95% CI, 0.68–0.78) in Ontario. The analysis was based on 20.2 million person–years of experience. Population data from Statistics Canada indicated that 32.4% (Ontario) to 53.0% (New Brunswick) of women aged 50–69 years attended screening in 2005–2006 and that 56.1% (Manitoba) to 64.3% (Quebec) reported undergoing bilateral mammography during the same period. An ad hoc method (described fully in the authors’ online supplementary methods) was used to adjust the relative risk in British Columbia for self-selection.
Olsen et al. (2005) , who estimated the IBM relative risk for women invited to screening in the Copenhagen population-based mammography programme ( Table 5.4 ), also estimated the IBM relative risk for women screened relative to those not screened, which was 0.60 (95% CI, 0.52–0.67) unadjusted for self-selection for screening. The relative risk adjusted for self-selection using an ad hoc approach was estimated to be 0.63 (95% CI not reported).
In a study of women invited to attend the Norwegian Breast Cancer Screening Program, Hofvind et al. (2013) compared breast cancer mortality in women who accepted the invitation with that in women who did not ( Table 5.5 ). This study is based entirely on linked unit record data of individual women invited to attend a population-based mammography screening programme, which included screening history, cancer registrations, and death records. Women could contribute person–years of experience to both the unscreened and the screened group. Overall, 84% of women attended screening for 1–15 years, with a median of 4.5 years. Accrual of incident breast cancers ended in 2009, and emigration and mortality follow-up continued until the end of 2010. The relative risk of death from breast cancer in screened relative to unscreened women was estimated to be 0.57 (95% CI, 0.51–0.64) adjusted for age at breast cancer diagnosis, calendar year, time since inclusion in the unscreened or screened group, and self-selection bias estimated using the average estimate of the breast cancer mortality relative risk for non-attenders relative to uninvited women (1.36; 95% CI, 1.11–1.67, from Duffy et al. 2002a , b ) and the study estimate of attendance in response to a screening invitation. The authors noted that 38% of women first attending the Norwegian Breast Cancer Screening Program in 1996–2006 reported having had a mammogram within the preceding 3 years, which could have biased the estimate of programme effectiveness. They also noted that the contemporaneous introduction of multidisciplinary breast care centres should not have biased their relative risk estimates because only women who were invited to the programme were included in the analysis. [No adjustment was made for lead-time bias.]
Morrison et al. (1988) examined breast cancer mortality within the Breast Cancer Detection Demonstration Project, which was initiated in 1973 by the American Cancer Society and the National Cancer Institute to demonstrate the feasibility of large-scale screening for breast cancer ( Beahrs et al., 1979 ; Baker, 1982 ). Screening was initially with two-view mammography, CBE, and thermography, but in later years thermography was dropped and mammography use was reduced, particularly in women younger than 50 years. Morrison et al. (1988) estimated the ratios (observed to expected) for death from breast cancer to be 0.80 overall and 0.89, 0.76, and 0.74, respectively, for women aged 35–49, 50–59, and 60–74 years at entry. [No confidence intervals or P values were reported.]
A case–cohort study approach was used by Thompson et al. (1994) to evaluate the effect of a mammography screening programme offered from 1985 to eligible members of a health maintenance organization in Washington State. Women aged 40–49 years were offered screening in the programme only if they had a risk factor for breast cancer, and women aged 50 years and older were invited every 1–3 years, depending on their risk factors; all were recommended to have annual CBE. A randomly selected age-stratified sample representing 2.4% of women was selected as a subcohort to represent the experience of all women in the cohort in the analysis. The formal screening programme began in 1985 and included mammography every 1–3 years depending on risk and annual CBE. About 10% of the women had been screened before implementation of the programme. By 1988 (3.5 years after implementation of the programme), about 34–56% of women (depending on age) had been screened. The IBM relative risk adjusted for mother’s history of breast cancer, nulliparity, and history of previous breast biopsy was 0.61 (95% CI, 0.23–1.62) for women aged 50 years and older.
The IBM relative risks for attendance to screening ranged from 0.51 to 0.80 after adjustment for self-selection. The lower value of 0.46 of Tabár et al. (2001) was not adjusted for selection bias, and it is likely that the value of 0.51 of Puliti & Zappa (2012) was incompletely adjusted for self-selection bias. The relative risks for the remaining studies ranged from 0.57 to 0.80 (median, 0.60) when including only the largest of the substantially overlapping Swedish studies ( Swedish Organised Service Screening Evaluation Group, 2006a , b ). The two studies in the USA (RR, 0.80 for each) included CBE in the intervention.
Only studies designed to separate the effect of screening on breast cancer mortality in a specified age group were considered to be informative. To study effectiveness of screening in women younger than 50 years, the analysis of breast cancer mortality should be limited to deaths in women whose breast cancer was diagnosed when they were younger than 50 years, unless screening was offered only to women while they were younger than 50 years (see Section 4.2.1 for discussion of age creep). Similarly, to study effectiveness of screening in women older than 69 years, the analysis should be limited to women first offered screening when they were older than 69 years and to breast cancer deaths that followed a diagnosis of breast cancer when the women were older than 69 years. Only results of studies that meet these criteria are included in this section. Studies are not included that presented age-specific results for women younger than 50 years but included deaths from breast cancers diagnosed at later ages ( UK Trial of Early Detection of Breast Cancer Group, 1999 ; Coldman et al., 2014 ) or for women older than 69 years at death from breast cancer who had not been offered screening ( Ascunce et al., 2007 ; Sarkeala et al., 2008b ; Kalager et al., 2010 ; Weedon-Fekjær et al., 2014 ) or had not been first offered screening in this age group ( Jonsson et al., 2007 ).
The design and results of studies reviewed for this section are summarized in Table 5.6 , by age (younger than 50 years or older than 69 years) and by country (in the order in which their mammography screening programmes were first introduced), and within each country by the earliest date of mammography screening that was included in the analysis.
Incidence-based mortality studies of the effectiveness of invitation to mammography screening a mainly in women younger than 50 years or older than 69 years.
Jonsson et al. (2000) compared IBM in women with breast cancer diagnosed at age 40–49 years in 14 Swedish study-group areas in which population-based mammography screening was offered from age 40 years and 15 control-group areas in which it was offered from age 50 years. These areas excluded five in which RCTs of screening had been conducted, one in which screening had been introduced very early, and one that offered screening from age 45 years. Women in the study group entered the study when screening started in their area. In both groups, mortality follow-up was to age 59 years, creating the possibility of lead-time bias in the result. A geographically identical, historical reference period (1976–1986) was defined for the study group and for the control group. The estimated IBM relative risk for women invited to screening at age 40 years was 0.91 (95% CI, 0.72–1.15), compared with the geographical areas that started screening at age 50 years, and adjusting for year of follow-up, geographical area, and time period. [Geographical area, as included in the model, was not defined but is likely to have been highly correlated with invitation to screening; therefore, the reported relative risk may be unreliable.]
The mammography screening experience of Jonsson et al. (2007) overlaps almost completely with that analysed by Jonsson et al. (2000) , and also compares IBM in women invited and not invited to screening over unbalanced time periods. The IBM relative risk for invitation to screening in women aged 40–49 years was 0.64 (95% CI, 0.43–0.97). [The Working Group estimated the IBM relative risk to be 0.51 (95% CI, 0.29–0.90) after adjustment for the difference in underlying breast cancer mortality with reference to results in the authors’ Table 3. Lead-time bias was estimated to be −5%.]
Hellquist et al. (2011) updated the analysis of Jonsson et al. (2000) and extended the period of accrual of breast cancer cases from 1997 to 2005. Women in 34 Swedish counties or screening areas were considered invited to screening if they resided when aged 40–49 years in an area that invited women of this age to screening (the same logic was applied for uninvited women in control areas during 1986–2005, with the same average follow-up time and mid-calendar year of follow-up). Such areas were required to have offered screening to women aged 40–49 years for at least 6 years from 1986 to 2005 (mean, 15.8 years). Only breast cancers incident at age 40–49 years were included. The IBM relative risk adjusted for misclassification of breast cancer cases in women invited to screening was 0.74 (95% CI, 0.66–0.83). Assuming 1 month and 1 year of lead time produced estimates of lead-time bias of −0.01% and −0.05%, respectively. Adjusted relative risks for breast cancer deaths in women diagnosed at ages 40–44 years and 45–49 years were estimated to be 0.83 (95% CI, 0.70–1.00) and 0.68 (95% CI, 0.59–0.78), respectively. Adjusted relative risks in women who attended screening were 0.71 (95% CI, 0.62–0.80), 0.82 (95% CI, 0.67–1.00), and 0.63 (95% CI, 0.54–0.75) for the age groups 40–49, 40–44, and 45–49 years, respectively. These estimates were made by adjusting the estimates for invitation to screening using the method of Cuzick et al. (1997) . The above estimates were not adjusted for a pre-screening difference in breast cancer mortality (RR, 0.94; 95% CI, 0.85–1.05) between screening and non-screening areas; taking this into account, the Working Group calculated an IBM relative risk of [0.79 (95% CI, 0.67–0.92)] for invited women and [0.76 (95% CI, 0.64–0.89)] for women who were ever screened, using a method developed by Altman & Bland (2003) .
Mammography was initiated on a pilot basis in Finland in the early 1980s. Women born in 1940 or 1942 were invited to attend screening with mammography and CBE in 1982; women born in 1936 or 1938 were invited in 1983, and thus they were aged 40–47 years at entry. They were re-invited every 2 years until 1990 (a total of four or five invitations), and women were considered to be non-attenders if they did not attend the first round. Women born in alternate years from 1935 to 1943 were used as a control cohort. The IBM relative risk was 0.11 (95% CI, 0.00–0.71) for invitation to screening and 0.10 (95% CI, 0.00–0.53) for attendance to screening ( Hakama et al., 1995 ). [The Working Group agreed with the authors’ opinion that an estimated programme sensitivity of 25% was too low for programme effectiveness to be the sole explanation for the very low relative risk.]
The Swedish study of Hellquist et al. (2011) encompassed the whole screening experience covered by Jonsson et al. (2000) and Jonsson et al. (2007) and provided IBM relative risks of 0.74 (95% CI, 0.66–0.83) for being invited to screening and 0.71 (0.62–0.80) for being ever screened. No weight was given to the very low relative risk that Hakama et al. (1995) observed, because it was based on only one death and appears incompatible with the estimated screening programme sensitivity of 25%.
The results of Jonsson et al. (2003b) are similar to those of Jonsson et al. (2000) , except that the analysis was based on first invitation to screening of women aged 65–74 years and covered 23 areas (16 study-group areas and 7 control-group areas) and not 29; the additional exclusions were principally counties in which screening did not begin until after 1990. The mean follow-up time was 10.1 years in the study group (8.1 years if estimated individual date of first screening was used, and not date of start of the screening programme in each area) and 9.3 years in the control group. Breast cancer deaths included in the analysis were only those that followed a diagnosis of breast cancer at age 70–74 years. The IBM relative risk for invitation to screening was 0.96 (95% CI, 0.73–1.25) when breast cancer mortality was based on underlying cause of death and adjusted for the difference in underlying mortality between the study-group and control-group areas. With further adjustment for inclusion bias and lead-time bias, the relative risk was 0.93 (95% CI not reported). The authors argued that the underlying cause of death may have been a particularly inaccurate classifier of mortality due to breast cancer in older women and that an excess mortality estimate would be more accurate. The corresponding excess mortality estimate of the relative risk was 0.84 (95% CI, 0.59–1.19) adjusted for the difference in underlying mortality between the study-group and control-group areas; with further adjustment for inclusion bias and lead-time bias, the relative risk was 0.78.
Jonsson et al. (2007) also reported on IBM associated with invitation to screening at age 70–74 years. However, the two screening counties in this study were also study-group (screening) counties in the Jonsson et al. (2003b) study, and the periods covered by the two studies were nearly the same. Therefore, Jonsson et al. (2007) was not considered to provide independent evidence.
Van Dijck et al. (1997) reported on IBM in women first invited to mammography screening at age 68–83 years in the city of Nijmegen compared with that in the city of Arnhem over an accrual and follow-up period of 1977–1990. Attendance rates in Nijmegen fell sharply with age, from approximately 70% in women in their late sixties to about 40% in those in their seventies and to less than 20% for the first round and less than 10% for the second and later rounds in women in their eighties and nineties. Screening began in Arnhem in 1989. The IBM relative risk for invitation to screening over the whole study period was estimated to be 0.80 (95% CI, 0.53–1.22), which became [0.89 (95% CI, 0.56–1.40)] when adjusted for the estimated difference in underlying breast cancer mortality between Nijmegen and Arnhem (see Table 5.6 ). For the period 9–13 years after the start of screening, the IBM relative risk estimate was 0.53 (95% CI, 0.27–1.04), and 0.59 (95% CI, 0.30–1.16) after adjusting for the difference in underlying breast cancer mortality.
In the Canadian provincial mammography screening programmes ( Coldman et al., 2014 ), the relative risk for women first screened at age 70–79 years was 0.65 (95% CI, 0.56–0.74) in the four provinces that offered screening to women in this age group. The province-specific relative risks varied from 0.63 (95% CI, 0.49–0.76) to 0.84 (95% CI, 0.36–1.31). The authors estimated that self-selection bias was conservative (−9% in an analysis limited to women aged 40–49 years in British Columbia). [This estimate may not be applicable to screening of women aged 70–79 years. Also, opportunistic breast screening before first screening in the provincial programmes could have affected the reported results, particularly in the age group 70–79 years.]
Three studies reported potentially valid estimates of IBM relative risks for breast cancer mortality in women older than 69 years: one for the age group 68–83 years ( Van Dijck et al., 1997 ), one for 65–74 years ( Jonsson et al., 2003b ), and one for 70–79 years ( Coldman et al., 2014 ). The reported relative risks, of 0.89 (95% CI, 0.56–1.40) by Van Dijck et al. (1997) , 0.96 (95% CI, 0.73–1.25) by Jonsson et al. (2003b) , and 0.65 (95% CI, 0.56–0.74) by Coldman et al. (2014) , are heterogeneous. However, the heterogeneity is reduced if the excess mortality estimate of the relative risk, 0.84 (95% CI, 0.59–1.19), of Jonsson et al. (2003b) is accepted as the more accurate estimate from that study. Lack of adjustment for self-selection bias and lack of consideration of possible effects of previous opportunistic screening limit the weight that can be given to the result of Coldman et al. (2014) .
The reported case–control studies are presented by country in the text and tables. All case–control studies are based on defined populations, but some of these are specific cohorts, with the methods of analysis being a case–control study nested within the cohort. In many case–control studies, the risk estimates are calculated for women who participated in screening compared with women who had been invited (or to whom screening was otherwise offered) but who did not participate. The non-participating women may have a different risk of death from breast cancer compared with the average population ( Cuzick et al., 1997 ; Duffy et al., 2002a ; Swedish Organised Service Screening Evaluation Group, 2006a ; Sarkeala et al., 2008a , b ), so this may result in selection bias. If the case–control study is based on systematic historical databases on screening, information bias can be considered minimal. However, in other case–control studies, information bias may be a problem. Rather few case–control studies have assessed screening impact compared with expectation in the absence of screening (or invitation) in the average population, as is usually done in cohort mortality studies. There are further limitations in the reported case–control studies in taking into account full screening histories in the risk estimates, and consequently there is wide variation in the follow-up windows for incidence and mortality after index screening. This potentially affects the magnitude of the estimates, even though these follow-up details are not always reported in connection with the individual studies. Some studies used only age at death in matching, whereas most studies also matched on residence at the time of diagnosis of the case. In addition, since the risk of breast cancer could be different among women who attend screening after receiving an invitation compared with those who are invited but do not attend, selection factors may confound the estimates of efficacy. A potential asset in case–control studies is that an adjustment for sociodemographic factors can also be attempted.
See Table 5.7 .
Case–control studies of the effectiveness of mammography screening within service screening programmes, by country.
Allgood et al. (2008) performed a case–control study in the East Anglia region. The cases were deaths from breast cancer in women diagnosed between the ages of 50 years and 70 years, after the initiation of the East Anglia Breast Screening Programme in 1989. The controls were women (two per case) who had not died of breast cancer, from the same area, matched by date of birth to the cases. Each control was known to be alive at the date of death of her matched case. All women were known to the breast screening programme and had been invited, at least once, to be screened. The unadjusted odds ratio for risk of death from breast cancer in women who attended at least one routine screen compared with those who did not attend was 0.35 (95% CI, 0.24–0.50), and 0.65 (95% CI, 0.48–0.88) after adjusting for self-selection bias using the more conservative intention-to-treat analysis ( Duffy et al., 2002a ).
Fielder et al. (2004) conducted a case–control study to estimate the effect of service screening, as provided by the NHS Breast Screening Programme, on breast cancer mortality in Wales. The 419 cases were deaths from breast cancer in women aged 50–75 years at diagnosis who were diagnosed after the start of screening in 1991 and who died after 1998. The 717 controls were women who had not died of breast cancer or any other condition during the study period. The aim was to select one control from the same general practitioner’s practice and another from a different general practitioner’s practice within the same district, matched by year of birth. The unadjusted odds ratio for risk of death from breast cancer in women who attended at least one routine screen compared with those who had never been screened was 0.62 (95% CI, 0.47–0.82), and 0.75 (95% CI, 0.49–1.14) after excluding cases diagnosed before 1995 and adjusting for self-selection bias.
Gabe et al. (2007) conducted a case–control study to evaluate the impact of the Icelandic breast screening programme, which was initiated in November 1987 in Reykjavik and covered the whole country from December 1989, comprising biennial invitation to mammography screening for women aged 40–69 years. The cases were deaths from breast cancer matched by age and screening area to population-based controls. The unadjusted odds ratio for risk of death from breast cancer in women who attended at least one screen compared with those who had never been screened was 0.59 (95% CI, 0.41–0.84), and 0.65 (95% CI, 0.39–1.09) after correction for both self-selection bias and screening opportunity bias.
Broeders et al. (2002) conducted a case–control study to describe the effect of population-based mammography screening in Nijmegen on breast cancer mortality, based on a 20-year follow-up period. The risk of death from breast cancer was calculated per 10-year moving age group for women who had attended the index screening (the screening immediately before diagnosis of breast cancer) versus those who had not. Odds ratios were presented by age group for both participation in index screening (see Table 5.7 ) and participation in either the index screening or the previous screening, or both; none showed a statistically significant effect. The youngest 10-year age group that showed an effect was women aged 45–54 years at their index screening; the odds ratio in women aged 45–49 years was 0.56 (95% CI, 0.20–1.61). The odds ratios for women aged 40–49 years were 0.90 (95% CI, 0.38–2.14) for participation in the index screening and 0.84 (95% CI, 0.30–2.29) for participation in the index screening and the previous screening. The corresponding odds ratios for women aged 70–79 years were 1.13 (95% CI, 0.50–2.58) and 0.70 (95% CI, 0.32–1.54). There was no limitation in these analyses as to age at first attendance to screening. [This analysis overlaps partly with that of van Schoor et al. (2010) (see Section 5.2.2b ).]
By 2008, 55 529 women had received an invitation to screening in Nijmegen, and another case–control study was performed ( van Schoor et al., 2011 ). The odds ratio for breast cancer death in the screened group over the complete period was 0.65 (95% CI, 0.49–0.87). Analyses were also performed by calendar period of index invitation to screening (see Table 5.7 ). [It is unclear why the numbers analysed for the two screening periods are so much less than the overall total of cases and controls included in this study.]
Paap et al. (2010) designed a case–control study to investigate the effect of mammography screening at the individual level. The study population included all women aged 50–75 years in Limburg Province who had been invited to the screening programme in 1989–2006. The unadjusted odds ratio for the screened versus the unscreened women was 0.30 (95% CI, 0.14–0.63), and 0.24 (95% CI, 0.10–0.58) after adjustment for self-selection. [This analysis includes only deaths in the most recent screening years. Deaths in the period from inception of the programme in 1989 until 2003 were not included.]
Paap et al. (2014) estimated the effect of the Dutch screening programme on breast cancer mortality by means of a large multiregion case–control study. They identified all breast cancer deaths in 2004 and 2005 in women aged 50–75 years who had received at least one invitation to the service screening programme in five participating screening regions. Cases were individually matched to controls from the population invited to screening. Conditional logistic regression was used to estimate the odds ratio of breast cancer death according to individual screening history. The unadjusted odds ratio for breast cancer death in screened versus unscreened women was 0.48 (95% CI, 0.40–0.58), and 0.42 (95% CI, 0.33–0.53) after adjustment for self-selection bias using regional correction factors for the difference in the baseline risk of breast cancer death between screened and unscreened women.
Otto et al. (2012b) conducted a case–control study in the south-western region of the Netherlands for the period 1995–2003, including women aged 49–75 years. There was no restriction with respect to age at first invitation. The all-age odds ratio for the association between attending screening at the index invitation and risk of breast cancer death was 0.56 (95% CI, 0.44–0.71), and 0.51 (95% CI, 0.40–0.66) for women attending any of the three screening examinations (for analyses by age at the index invitation, see Table 5.7 ).
Puliti et al. (2008) conducted a case–control study to evaluate the impact of service screening programmes on breast cancer mortality in five regions of Italy. The odds ratio for invited women compared with not-yet-invited women was 0.75 (95% CI, 0.62–0.92). When the analyses were restricted to invited women, the odds ratio for screened women compared with never-respondent women, corrected for self-selection bias, was 0.55 (95% CI, 0.36–0.85).
Roder et al. (2008) conducted a case–control study of women in South Australia aged 45–80 years during 2002–2005 (diagnosed after the start of BreastScreen Australia) and live controls (three per death) randomly selected from the state electoral roll after date-of-birth matching. The programme has provided biennial screening, with two-view mammography and double reading, since its inception. It actively targets women aged 50–69 years and allows access to women aged 40–49 years and those aged 70 years and older. The odds ratio for breast cancer death in all BreastScreen participants compared with non-participants was 0.59 (95% CI, 0.47–0.74). The corresponding odds ratio in women younger than 50 years at diagnosis was 1.18 (95% CI, 0.70–1.98) and in those aged 70 years and older at diagnosis was 0.43 (95% CI, 0.25–0.72). Compared with non-participants, the odds ratio was 0.70 (95% CI, 0.47–1.05) for women last screened through BreastScreen more than 3 years before diagnosis of the index case, and 0.57 (95% CI, 0.44–0.72) for women screened more recently.
Nickson et al. (2012) conducted another case–control study within BreastScreen Australia, in which women aged 50–69 years on the electoral roll (98.9% of the eligible population) are invited to attend screening. Eligible women were those aged 50 years and older on the Western Australian electoral roll between 1995 and 2006. The cases were women from this population who died of breast cancer between 1995 and 2006. Controls (10 per case) were selected by incidence density sampling from the source population (those with a breast cancer diagnosis were not excluded). Exposure to screening was defined as receipt of a screening mammogram from BreastScreen at any point between the woman’s 50th birthday and the case–control set reference date (the date of earliest breast cancer diagnosis for that set; for 89%, this was the date of diagnosis of the case); 56% of controls and 39% of cases attended screening. The odds ratio from the primary analyses (adjusted for remoteness and relative socioeconomic disadvantage) was 0.48 (95% CI, 0.38–0.59). The odds ratio was found to vary little by reference age group or year of death and was robust to sensitivity analyses.
See Table 5.8 .
Other case–control studies of the effectiveness of mammography screening.
In 1974, de Waard et al. (1984a) set up a population-based study of periodic screening by xeromammography of women aged 50–64 years in Utrecht; 72% of invited women attended the first of four rounds. The effect of the programme on breast cancer mortality was evaluated in a nested case–control study, which showed an odds ratio for breast cancer mortality in women who had ever been screened of 0.30 (95% CI, 0.13–0.70) compared with those who had never been screened ( Collette et al., 1984 ). The odds ratios for women aged 50–54, 55–59, 60–64, and 65–69 years at diagnosis were 1.13, 0.31, 0, and 0.10, respectively. [These estimates were based on small numbers, and no confidence intervals were given.]
An updated case–control analysis 17 years after the initiation of this project was reported by Miltenburg et al. (1998) . Controls (three for each case) were defined as women with the same year of birth as the case, living in the city of Utrecht at the time the case died, and having had the opportunity to be screened in the DOM project. The odds ratio for breast cancer mortality for screening in the period 1975–1992 was 0.54 (95% CI, 0.37–0.79). Stratification by birth cohort is given in Table 5.8 .
In 1975, a population-based screening programme was set up in Nijmegen, a city with about 150 000 inhabitants ( Peeters et al., 1989a ). The first screening round, in 1975–1976, involved 23 000 women born in 1910–1939, who were thus aged 35–64 years. In the subsequent screening rounds, the same birth cohort was invited, as well as 7700 women born before 1910. The odds ratio for death from breast cancer estimated in a case–control analysis was 1.2 (95% CI, 0.31–4.8) for women aged 35–49 years, 0.26 (95% CI, 0.10–0.67) for those aged 50–64 years, and 0.81 (95% CI, 0.23–2.8) for those aged 65 years and older ( Verbeek et al., 1985 ).
In a further case–control study based on the Nijmegen population, Van Dijck et al. (1996) selected women who were 65 years or older when first invited to screening. The rate ratio of breast cancer mortality in women who had participated regularly (i.e. in the two most recent screening rounds before diagnosis) compared with those who had not participated in screening was 0.56 (95% CI, 0.28–1.13). The rate ratio for women aged 65–74 years at the most recent invitation was 0.45 (95% CI, 0.20–1.02), and for women aged 75 years and older it was 1.05 (95% CI, 0.27–4.14). [The Working Group estimated rate ratios for women who had ever been screened by combining, using fixed effects meta-analysis, reported relative risks for women who had been screened regularly and women who had been screened “otherwise” relative to women who had not been screened. The estimates rate ratios were 0.68 (95% CI, 0.44–1.05) for all ages, 0.54 (95% CI, 0.31–0.95) for ages 65–74 years, and 0.94 (95% CI, 0.45–1.88) for ages 75 years and older. Forty of the 82 deaths from breast cancer included in this study were included in a separate IBM analysis of effectiveness of screening in women aged 68–83 years at entry into the Nijmegen screening programme ( Van Dijck et al., 1997 ).]
van Schoor et al. (2010) designed a case–control study to investigate the effect of biennial mammography screening on breast cancer mortality in women aged 40–69 years between 1975 and 1990 in Nijmegen. In women aged 40–49 years at their index screening (in cases, the last screening before diagnosis of breast cancer), the odds ratio for screening was 0.50 (95% CI, 0.30–0.82). Similarly, an odds ratio of 0.54 (95% CI, 0.35–0.85) was reported for women aged 50–59 years, and an odds ratio of 0.65 (95% CI, 0.38–1.13) for those aged 60–69 years.
Between 1970 and 1980, women aged 40–70 years living in 24 municipalities in Florence were invited to mammography screening with craniocaudal and mediolateral oblique views every 2.5 years. In 1989, the screening area was extended to include the city of Florence. Palli et al. (1986 , 1989 ) conducted a case–control study within this population to estimate the impact on breast cancer mortality. The odds ratios for women aged 40–49 years and for those aged 50 years and older at diagnosis of breast cancer were estimated to be 0.63 (95% CI, 0.24–1.6) and 0.51 (95% CI, 0.29–0.89), respectively.
Elmore et al. (2005) conducted a matched case–control study among women enrolled in six health plans in the states of California, Massachusetts, Minnesota, Oregon, and Washington and examined the efficacy of screening by mammography and/or CBE among women in two age cohorts (40–49 years and 50–65 years) and in two levels of breast cancer risk (in women at average risk and women with a family history and/or previous breast biopsy) until 1983–1998. The effect of screening with mammography, or of screening with mammography and CBE, during the 3 years before the index date (defined as the date of first suspicion of breast abnormalities in case subjects, with the same date used for matched control subjects) was evaluated. For women aged 40–49 years at diagnosis of breast cancer, the odds ratio was 0.85 (95% CI, 0.65–1.23), and for women aged 50–65 years, the odds ratio was 0.47 (95% CI, 0.35–0.63) for screening with mammography alone. The odds ratio for women at an increased risk was 0.74 (95% CI, 0.50–1.03) and for women at average risk was 0.96 (95% CI, 0.80–1.14); however, the difference was not statistically significant ( P = 0.17).
Norman et al. (2007) used data from a subset of the Women’s Contraceptive and Reproductive Experiences (CARE) Study, a population-based multicentre case–control study of risk factors for breast cancer among White and Black women conducted in metropolitan Atlanta, Georgia; Detroit, Michigan; Los Angeles, California; Philadelphia, Pennsylvania; and Seattle, Washington, to estimate the relative mortality rates from invasive breast cancer among women with at least one screening mammogram in the 2 years before a baseline reference date compared with unscreened women, adjusting for potential confounding. The odds ratio for breast cancer death within 5 years after diagnosis was 0.89 (95% CI, 0.65–1.23) for ages 40–49 years at diagnosis and 0.47 (95% CI, 0.35–0.63) for ages 50–64 years at diagnosis.
A meta-analysis was performed of some of the earlier case–control studies ( Demissie et al., 1998 ), and Broeders et al. (2012) conducted a meta-analysis of seven more recent case–control studies. The combined unadjusted odds ratio in women who were screened versus those who were not screened was 0.46 (95% CI, 0.40–0.54), and 0.52 (95% CI, 0.42–0.65) when adjusted for self-selection using the method of Duffy et al. (2002a) . The crude odds ratio for breast cancer mortality reduction, translated to intention-to-treat estimates for women who were invited versus those who were not invited was 0.69 (95% CI, 0.57–0.83).
Several of the case–control studies summarized above reported results in several age groups, including those that lie below or above the age range 50–69 years. Such results can be validly used to infer the effectiveness, or otherwise, of screening women younger than 50 years, provided they are based only on deaths from breast cancer of women whose breast cancer was diagnosed when they were younger than 50 years. The results that permit this inference are those of Palli et al. (1989) , Broeders et al. (2002) , Elmore et al. (2005) , Norman et al. (2007) , and Roder et al. (2008) (see Table 5.8 ).
The use of results from case–control studies to infer effectiveness at ages older than 69 years is less straightforward because, even if they are based only on deaths from breast cancer of women whose breast cancer was diagnosed when they were older than 69 years, the relative risk of death calculated will have been influenced by screening at age 69 years and younger, assuming screening effectiveness ( Otto et al., 2012b ). This influence can only be removed by limiting the analysis to women first offered screening after age 69 years. No case–control study has been done in a context in which this limitation could be applied; however, that of Van Dijck et al. (1996) was limited to women first offered screening from age 65 years.
In assessing the effectiveness of breast cancer screening, the Working Group considered that accurate information on standards of breast cancer treatment in different regions analysed and careful matching of regions by treatment standards or adjustment for differences between regions in treatment standards are minimum criteria for validity of ecological studies. Therefore, simple comparisons of trends between unmatched regions or without potentially effective statistical adjustment, or in a single region over time, were excluded.
Correcting for differences in underlying incidence is a challenge. Differences in incidence between regions, or across time, may indicate an important difference in baseline risk that must be adjusted for, or they may indicate overdiagnosis and should not be adjusted for. These studies were therefore excluded, as were any that measured differences in survival, due to the well-recognized issue of lead time.
Studies of population-based screening in Europe were reviewed to assess the value of trend analyses in population breast cancer mortality ( Moss et al., 2012 ). A literature review identified 17 reports, of which 12 provided quantitative estimates of the impact of screening. Due to differences in comparisons and outcome measures, no pooled estimate of effectiveness was calculated. Overall, this approach proved to be of limited value for assessment of screening impact.
For the purpose of selecting studies to review, the Working Group defined the following subcategories:
A total of 87 studies were identified by the Working Group through literature searches and were reviewed for initial categorization according to the above criteria. After the initial exclusion of studies in categories 1 ( n = 25) or 2 ( n = 20), studies of other designs (9 case–control studies, 4 cohort studies, and 3 studies based on RCTs), and studies with other limitations ( n = 12), 14 studies were further considered. Eight of these were then identified as IBM studies ( Tabár et al., 2001 ; Duffy et al., 2002b ; Jonsson et al., 2003a , b ; Parvinen et al., 2006 ; Anttila et al., 2008 ; Sarkeala et al., 2008b ; Kalager et al., 2010 ) and were therefore excluded. Of the remaining six ecological studies, two were judged to be uninformative: Das et al. (2005) used correlation as the measure of association, and Autier et al. (2011) may have been biased by the evolution of staging data over the study period; the remaining four studies were found to be informative. One additional informative study was identified separately ( Otto et al., 2003 ) and was included in the review.
Otto et al. (2003) reviewed mortality trends in the Netherlands from 1980 to 1998, using clustered municipality-level data in 1-month bands, including the progressive introduction of screening from 1989 until 1997. Four age bands were compared to detect changes in treatment effectiveness: 45–54, 55–64, 65–74, and 75–84 years. Rates of change and cumulative changes were estimated in both the pre-screening and screening eras. Analysis was via linear splines (i.e. a single joinpoint). There was a downturn in mortality for the middle two age bands (55–64 years and 65–74 years) coincident with the introduction of screening, with an accumulated mortality reduction by 1999 estimated to be 19.1%. The annual rate of decline (annual percentage change) was 1.7% (95% CI, 1–2.4%) in these two age groups combined and 1.2% (95% CI, 0.1–2.4%) in the younger age group (45–54 years). There was no significant change in the older age group (75–84 years). Before screening, the trend was upward at 0.3% per year.
Törnberg et al. (2006) compared time trends in breast cancer incidence and mortality after the introduction of mammography screening in Copenhagen, Helsinki, Stockholm, and Oslo. In Helsinki, screening was offered to women aged 50–59 years, starting in 1986, and in the other three capitals, screening was offered to women aged 50–69 years, starting between 1989 and 1996. Peaks in breast cancer incidence depended on the age groups covered by the screening, the length of the implementation of screening, and the extent of background opportunistic screening. No mortality reduction after the introduction of screening was visible after 7–12 years of screening in any of the capitals. [No visible effect on mortality reduction was expected in Oslo, due to too short an observation period.]
Jørgensen et al. (2010) compared breast cancer mortality trends in Denmark, between Copenhagen (where screening was introduced in 1991) and Funen County (where screening started in 1993) and the rest of Denmark (which served as an unscreened control group). Unscreened age groups were used to further control for effects of changing treatment. Screening was offered to women aged 55–74 years, and mortality was evaluated in three age bands: 35–54, 55–74, and 75–84 years. The pre-screening period was 1982–1991, and the post-screening period was restricted to 1997–2006, to allow for a lag in benefit. The annual percentage change in breast cancer mortality was evaluated by Poisson regression. For the likely-to-benefit age band (55–74 years), the annual percentage change changed from +1 to −1% in the screening areas and from +2 to −2% in the non-screening areas. For the younger age band (35–54 years), the annual percentage change changed from +2% to −5% in the screening areas and from 0% to −6% in the non-screening areas. No significant changes were observed in the older age band.
The mortality benefit of attending screening was estimated using a Markov model of disease progression based on three regions in France ( Uhry et al., 2011 ). Attempts were made to correct for opportunistic screening, and overdiagnosis was included as an explicit assumption, at either 10% or 20%. The corresponding estimates of mortality reduction were 23% (95% CI, 4% to 38%) and 19% (95% CI, −3% to 35%). [Problems of model fit were reported.]
Poisson regression was used in a study reanalysing population data from the era of Swedish screening trials ( Haukka et al., 2011 ). [The data used were from NORDCAN ( Engholm et al., 2010 ), which had variable levels of agreement with trial data where it could be compared.] The model assumed a delayed step change due to screening after the staggered introduction by region, with different lead times tested for best fit. Using the 3-year lead time estimate, breast cancer mortality decreased by 16% (RR, 0.84; 95% CI, 0.78–0.91) in the screening age group 40–69 years and by 11% (RR, 0.89; 95% CI, 0.80–0.98) in the age group 70–79 years.
See Table 5.9 .
Studies using stage or indicators of stage at diagnosis of breast cancer as measures of screening performance.
Hofvind et al. (2012c) compared incidence of advanced breast cancer cases diagnosed among screened and unscreened women aged 50–69 years in Norway. A total of 11 569 breast tumours (1670 ductal carcinoma in situ [DCIS] and 9899 invasive cancer) were diagnosed among 640 347 women who were invited to the screening programme during the study period. Participants in the screening programme accounted for 9726 breast tumours (1517 DCIS and 8209 invasive cancer) and non-participants accounted for 1843 breast tumours (153 DCIS and 1690 invasive cancer). When cases were compared between participants and non-participants, a significant reduction was observed in stage III (RR, 0.5; 95% CI, 0.4–0.7) and stage IV (RR, 0.3; 95% CI, 0.2–0.4) cancers, in tumours larger than 50 mm (RR, 0.4; 95% CI, 0.4–0.6), and in distant metastasis (RR, 0.3; 95% CI, 0.2–0.4). Distributions by stage, size, and nodal status were similar in women who did not attend screening and those who were not invited.
Domingo et al. (2013b) analysed data on invitation to organized screening programmes in Copenhagen (first eight invitations rounds, 1991–2008) and in Funen (first six invitation rounds, 1993–2005) ( Table 5.10 ). Both programmes offered biennial screening to women aged 50–69 years. The Working Group calculated the rate ratios and 95% confidence intervals for tumour size and nodal status of screen-detected breast cancers versus those diagnosed in women who were not screened, for Copenhagen and Funen together. Among screen-detected cancers, a significant increase in detection of tumours of size 0–10 mm [RR, 2.91; 95% CI, 2.47–3.44] and 11–20 mm [RR, 1.27; 95% CI, 1.14–1.41] and a reduction in detection of tumours of size 21–30 mm [RR, 0.47; 95% CI, 0.40–0.55] and larger than 30 mm [RR, 0.26; 95% CI, 0.21–0.33] and in node–positive cancers [RR, 0.61; 95% CI, 0.54–0.67] were estimated. The rates of large screen-detected cancers were significantly lower, and screen-detected cancers were significantly less frequently lymph node-positive.
Number of breast cancers (invasive and carcinoma in situ) detected at screening in participants, diagnosed as interval cancers in participants, or diagnosed in unscreened women (Copenhagen and Funen screening programmes, Denmark).
Foca et al. (2013) analysed data from 700 municipalities in Italy, with a total population of 692 824 women aged 55–74 years targeted by organized mammography screening from 1991 to 2005. The effect of the screening was evaluated from year 1 (the year screening started at the municipal level) to year 8 (based on the decreasing number of available municipalities). The study was based on a total of 14 447 incident breast cancers. The observed 2-year, age-standardized (Europe) incidence rate ratio (ratio of the incidence rate to the expected rate) was calculated. Expected rates were estimated assuming that the incidence of breast cancer was stable and equivalent to that in the 3 years before year 1. The incidence rate ratio for pT2–pT4 breast cancers was 0.97 (95% CI, 0.90–1.04) in years 1 and 2, 0.81 (95% CI, 0.75–0.88) in years 3 and 4, 0.79 (95% CI, 0.73–0.87) in years 5 and 6, and 0.71 (95% CI, 0.64–0.79) in years 7 and 8. A significant and stable decrease in the incidence of late-stage breast cancer was observed from the third year of screening onward.
Nederend et al. (2012) analysed a consecutive series of 351 009 screening mammograms of 85 274 women aged 50–75 years, who underwent biennial screening in a breast screening region in the Netherlands in 1997–2008. A total of 1771 screen-detected cancers and 669 interval cancers were diagnosed in 2440 women. The authors observed, as expected, no decline in detection rates of advanced breast cancer during each round of 12 years of biennial screening mammography in the screened population. In the source population (data from a cancer registry), no decline in advanced breast cancer has been reported.
Autier & Boniol (2012) estimated incidence trends in advanced breast cancer from 1989 to 2004 in the West Midlands (United Kingdom), where breast screening started in 1988 for women aged 50–64 years ( Fig. 5.1 ). The authors extracted numbers of breast cancer cases from the Cancer Incidence in Five Continents database ( Ferlay et al., 2014 ). They used published data ( Lawrence et al., 2009 ; Nagtegaal et al., 2011 ) for the annual percentage change (APC) in the incidence rates of lymph node-positive/node-negative breast cancer and of tumours larger than 50 mm for the screening period. According to their analysis, the incidence rates of node-positive breast cancer increased from 1989 to 1992 and then decreased below the pre-screening level in 1993–1995 but returned to pre-screening levels in 1996–2000 and then stabilized. From 1989 to 2004, the APC was 2.2% (95% CI, 1.2% to 3.1%) for node-negative cancers and −0.7% (95% CI, −1.8% to 0.3%) for node-positive cancers. The incidence of tumours larger than 50 mm remained stable from 1989 to 2004 (APC, 0.2%; 95% CI, −2.2% to 2.7%).
Annual incidence rates from 1989 to 2004 of advanced breast cancer in women aged 50–64 years in the West Midlands, United Kingdom
Eisemann et al. (2013) reported data from 2008–2009 in Germany, where breast cancer screening started in 2005, biennially, for women aged 50–69 years. From 2002 to 2007, the absolute number of breast cancer diagnoses (including in situ cases) increased markedly, by 15%: for in situ tumours, by +94%; for T1 tumours, by +18%; for T2 tumours, by +11%; for T3 tumours, by +14%; and for tumours of unknown stage, by +24%. A decrease of about −10% was observed for T4 tumours. [No comparison of rates of advanced cancers was reported in the screened or invited population versus the population not screened or not invited.]
Elting et al. (2009) assessed the association between in-county mammography facilities (in 2002–2004) and mammography screening and breast cancer diagnosis at a late stage among women in Texas older than 40 years. Half of the 254 counties had no mammography facility. In 2004, a total of 12 469 of the 4 639 842 women in Texas older than 40 years were diagnosed with either invasive breast cancer or DCIS (risk per 10 000 women aged > 40 years, 26.87; 95% CI, 26.4–27.3). The risk of diagnosis at early and late stages varied significantly between counties with and without mammography facilities. After accounting for confounding by age, race, and ethnicity, multivariate analysis showed that women who lived in counties with facilities were more likely to be diagnosed with DCIS (odds ratio [OR], 1.32; 95% CI, 0.98–1.77; P = 0.06) and significantly less likely to be diagnosed at an advanced stage (OR, 0.36; 95% CI, 0.26–0.51; P < 0.001) than their counterparts who lived in counties without a facility. These differences were observed despite adjustment for higher probabilities of advanced disease among African-American and Hispanic women.
Bleyer & Welch (2012) used data from the Surveillance, Epidemiology, and End Results (SEER) Program of the United States National Cancer Institute to examine trends from 1976 to 2008 in the incidence of early-stage and late-stage breast cancer among women aged 40 years and older. The 3-year period 1976–1978 was chosen to obtain the estimate of the baseline incidence of breast cancer detected without mammography. During this period, the incidence of breast cancer was stable and few cases of DCIS were detected (findings compatible with the very limited use of screening mammography). The estimate of the current incidence of breast cancer was based on the 3-year period 2006–2008. To eliminate the effect of use of hormone replacement therapy, the observed incidence was truncated if it was higher than the estimate of the current incidence (the annual incidence per 100 000 women of DCIS was not allowed to exceed 56.5 cases, of localized disease to exceed 177.5 cases, of regional disease to exceed 77.6 cases, and of distant disease to exceed 16.6 cases, during the period 1990–2005). A substantial increase in the use of screening mammography during the 1980s and early 1990s among women aged 40 years and older in the USA, a substantial concomitant increase in the incidence of early-stage breast cancer among these women, and a small decrease in the incidence of late-stage breast cancer were observed. A large increase in cases of early-stage cancer (absolute increase of 122 cases per 100 000 women) and a small decrease in cases of late-stage cancer (absolute decrease of 8 cases per 100 000 women) were observed. The trends in regional and distant late-stage breast cancer showed that the variable pattern in late-stage cancer (which includes the excess diagnoses associated with use of hormone replacement therapy in the late 1990s and early 2000s) was almost entirely attributable to changes in the incidence of regional (largely node-positive) disease. However, the incidence of distant (metastatic) disease remained unchanged (95% CI for the APC, −0.19% to 0.14%). The SEER data did not distinguish between women who were screened and those who were not screened.
Helvie et al. (2014) , similarly to Bleyer & Welch (2012) , compared the SEER breast cancer incidence and stage for the pre-mammography period (1977–1979) and the mammography screening period (2007–2009) in women older than 40 years. The authors estimated pre-screening temporal trends using several measures of APC. Stage-specific incidence values for 1977–1979 (baseline) were adjusted using APC values of 0.5%, 1.0%, 1.3%, and 2.0% and then compared with observed stage-specific incidence in 2007–2009. Pre-screening APC temporal trend estimates ranged from 0.8% to 2.3%. The joinpoint estimate of 1.3% for women older than 40 years approximated the four-decade-long APC trend of 1.2% noted in the Connecticut Tumor Registry. At an APC of 1.3%, late-stage breast cancer incidence decreased by 37% (56 cases per 100 000 women), with a reciprocal increase in early-stage rates noted from 1977–1979 to 2007–2009. The resulting late-stage breast cancer incidence decreased by 21% at an APC of 0.5% and by 48% at an APC of 2.0%. Total invasive breast cancer incidence decreased by 9% (27 cases per 100 000 women) at an APC of 1.3%. [According to the authors, a substantial reduction in late-stage breast cancer has occurred in the mammography era when appropriate adjustments are made for pre-screening temporal trends.]
Hou & Huo (2013) analysed the SEER age-standardized breast cancer incidence rates from 2000 to 2009, for 677 774 women aged 20 years and older. This study represents a descriptive analysis of population-based cancer incidence rates from 18 SEER registries with high-quality data, representing 28% of the United States population. Since 2004, incidence rates in women aged 40–49 years increased significantly for most racial/ethnic groups (overall APC, 1.1%; P = 0.001). The incidence rate of DCIS increased significantly in all racial/ethnic groups, with an APC range from 2.3% to 3.0% ( P < 0.005). The incidence rate of localized breast cancer increased significantly in non-Hispanic Black women (APC, 1.3%; P = 0.004) and Asian women (APC, 1.2%; P = 0.03). The incidence rates of regional and distant cancers decreased significantly in non-Hispanic White women from 2000 to 2004 (APC, −2.5%; P = 0.02) and in Hispanic women from 2000 to 2009 (APC, −1.1%; P = 0.006). [It is possible that the changes in incidence rates are due in part to improvements in cancer screening methods and, therefore, advances in early detection. It is unlikely that the overall trends of incidence rates are due to changes in the mammography screening rate, since mammography use did not change substantially from 2000 to 2010, although it increased by large magnitudes in small groups with growing populations, such as new immigrants and Asian-Americans.]
DeSantis et al. (2014) obtained data on incidence, probability of developing cancer, and cause-specific survival from SEER, and data on the prevalence of mammography by age from the 2010 and 2012 Behavioral Risk Factor Surveillance System, to assess the relationship between mammography screening rates in 2010 and breast cancer stage at diagnosis in 2006–2010. Among non-Hispanic White women, state-level mammography screening prevalence was positively correlated with the percentage of breast cancers diagnosed at the in situ stage (correlation coefficient, r = 0.62; P < 0.001) and negatively correlated with the percentage of breast cancers diagnosed at late stages ( r = −0.51; P < 0.001).
Given that increased mammographic breast density is associated with lower sensitivity and higher interval cancer rates ( Mandelson et al., 2000 ), its potential role as an effect modifier of mammography screening effectiveness is of interest. The effect of breast density on case fatality rate, or breast density as a modifier, has been investigated in several studies. Only one of these has examined differences in survival of women with interval cancers in those with dense versus non-dense breasts. This study in Sweden found that women with interval cancers had worse survival than women with screen-detected cancers (hazard ratio [HR], 1.69; 95% CI, 1.03–2.76, overall) and that interval-cancer survival was poorer in those with non-dense breasts (HR, 1.76; 95% CI, 1.01–3.09) than in those with dense breasts (HR, 1.26; 95% CI, 0.47–3.38) ( Eriksson et al., 2013 ). These effects were observed after adjustment for tumour size and lymph-node metastasis at diagnosis. [Before adjustment, hazard ratios were stronger.]
The remaining studies examined the impact of breast density on survival or mortality rates within populations where screening is available, but they did not differentiate between interval and screen-detected cancers. In a cohort in Denmark participating in biennial mammography at ages 50–69 years, during 1991–2001, the case fatality rate was lower in women with mixed/dense breasts than in those with fatty breasts (HR, 0.60; 95% CI, 0.43–0.84) ( Olsen et al., 2009 ). [Although the case fatality rate is lower for women with dense breasts, it should be noted that because more women with dense breasts develop breast cancer, more women with dense breasts die from breast cancer overall.] In the USA, a study using the Carolina Mammography Registry (22 597 breast cancers) showed no difference in breast cancer mortality between women with dense breasts and those with fatty breasts, after adjusting for incidence differences (HR, 0.908; P = 0.12) (stage-adjusted) ( Zhang et al., 2013 ). Similarly, the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) density score was not associated with breast cancer survival (HR for breast cancer death, 0.92; 95% CI, 0.71–1.19) in the United States Breast Cancer Surveillance Consortium ( Gierach et al., 2012 ), except for an increased risk of breast cancer death among women with low breast density (BI-RADS 1) who were obese or had tumours larger than 20 mm. The Kopparberg RCT, in Sweden, suggested that women with dense breasts have higher breast cancer incidence rates (multivariate RR, 1.57; 95% CI, 1.23–2.01) and breast cancer mortality (RR, 1.91; 95% CI, 1.26–2.91), but that there was no clear difference in survival between women with dense breasts and those with non-dense breasts (HR, 1.41; 95% CI, 0.92–2.14) (not adjusted for tumour characteristics) ( Chiu et al., 2010 ). One study found poor survival in women with dense breasts compared with those with fatty breasts in women diagnosed at the first screening round but not in those diagnosed at later rounds (rounds 5–10) ( van Gils et al., 1998 ).
[The Working Group noted that although breast cancers occurring in dense breasts are more likely to be interval cancers, there is no indication that breast cancer survival rates are poorer for these cancers (despite a shorter lead-time bias). In addition, the studies were performed with screen-film mammography, so it is difficult to extrapolate the results to digital methods.]
RCTs of mammography screening, mostly performed in the 1980s or earlier, have reported reductions in breast cancer mortality in women aged 50–69 years. However, the present-day relevance of these trials has been debated because the management and treatment of breast cancer has changed considerably in the past decades ( Gøtzsche & Nielsen, 2009 ; Kalager et al., 2010 ; Paci & EUROSCREEN Working Group, 2012 ; Marmot et al., 2013 ). Adjuvant systemic therapy has been increasingly used since the late 1980s, and its dissemination and effectiveness have progressed since then ( van de Velde et al., 2010 ). Such developments have probably affected the impact of screening, also in service screening programmes ( Berry et al., 2005 ). This section discusses studies of the effects of adjuvant systemic therapy and mammography screening in current health-care systems.
The effects of adjuvant treatment and mammography screening were calculated for the Netherlands using the Microsimulation Screening Analysis (MISCAN) model ( de Gelder et al., 2015 ). [Models can extrapolate findings from screening and adjuvant treatment trials to actual populations, can allow for comparison of intervention strategies, and can separate effects on the natural history of disease, for example screening effects and adjuvant treatment effects ( Berry et al., 2005 ; Mandelblatt et al., 2009 ) (see Section 5.1.2f ).] In the MISCAN model, the progression was modelled as a semi-Markov process through the successive preclinical invasive stages T1a, T1b, T1c, and T2+. The mean duration of the preclinical detectable phase, the probability of a transition between the stages, and the mammography sensitivity were then estimated, using detailed data from screening registries. Data on adjuvant systemic therapy were derived from comprehensive cancer centres. Cure and survival rates after screen detection were based on RCTs ( de Koning et al., 1995 ; Tabár et al., 2000 ; Nyström et al., 2002 ; Bjurstam et al., 2003 ). The risk of death from breast cancer after adjuvant treatment was modelled using the rate ratios from the meta-analysis of the Early Breast Cancer Trialists’ Collaborative Group (2005) . In 2008, adjuvant treatment was estimated to have reduced the breast cancer mortality rate in the simulated population by 13.9%, compared with a situation without treatment. Biennial screening between age 50 years and age 74 years further reduced the mortality rate by 15.7%. Extending screening to age 48 years would lower the mortality rate by 1.0% compared with screening from age 50 years; 10 additional screening rounds between age 40 years and age 49 years would reduce this rate by 5.1%. Adjuvant systemic therapy and screening reduced breast cancer mortality by similar amounts.
A previous modelling study, which included six natural history models for the population in the USA, had estimated an approximately equal contribution of adjuvant therapy and screening to the observed mortality reduction in the USA ( Berry et al., 2005 ), using very similar techniques to those described above.
These analyses have recently been updated, taking into account the receptor-specific heterogeneity of breast cancer ( Munoz et al., 2014 ), by using six established population models with ER-specific input parameters on age-specific incidence, disease natural history, mammography characteristics, and treatment effects to quantify the impact of screening and adjuvant therapy on age-adjusted breast cancer mortality in the USA by ER status from 1975 to 2000. In 2000, actual screening and adjuvant treatment were estimated to have reduced breast cancer mortality by 34.8%, compared with the situation if no screening or adjuvant treatment had been present; a reduction by 15.9% was estimated to have been a result of screening, and 23.4% as a result of treatment. For ER-positive cases, adjuvant treatment made a higher relative contribution to breast cancer mortality reduction than screening, whereas for ER-negative cases the relative contributions were similar for screening and adjuvant treatment. Although ER-negative cases were less likely to be screen-detected than ER-positive cases (35.1% vs 51.2%), when they were screen-detected, the survival gain was greater for ER-negative cases than for ER-positive cases (5-year breast cancer survival, 35.6% vs 30.7%).
5.3.1. false-positive rates.
A screening test is not diagnostic but should identify asymptomatic women who are at risk of harbouring an undiagnosed cancer. The screening episode in organized screening should end with an unequivocal diagnostic report: there is or there is not cancer ( Perry et al., 2006 ). A woman in whom an abnormality is detected by screening and whose investigations end with a negative result has a false-positive result. This result closes the screening episode.
In a recent survey of 20 population-based screening programmes in 17 European countries, the Euroscreen and EUNICE Working Group ( Hofvind et al., 2012a ) reported average recall rates varying from 9.3% at the initial screening episode (range, 2.2–15.6%) to 4.0% at subsequent screening episodes (range, 1.2–10.5%). The average rates of needle biopsy were 2.2% at the initial screening and 1.1% at subsequent screenings. The variation depends on differences between national protocols and a variety of local conditions. Over the whole diagnostic phase, the benign-to-malignant ratio ranged from 0.09 in the United Kingdom to 0.21 in Luxembourg, with an average of 0.11.
The difference in the performance of the assessment phase between opportunistic screening and service screening has been estimated by comparing screening in the USA and population-based programmes in Europe. Smith-Bindman et al. (2005) compared the performance of screening in the United Kingdom and the USA. The outcomes included (per 1000 women screened for 20 years) a detection rate of carcinoma in situ of 12.3 in the USA compared with 8.3 in the United Kingdom, a rate of non-invasive diagnostic tests for assessment of recalled women of 553 in the USA compared with 183 in the United Kingdom, and a biopsy rate of 142 in the USA compared with 85 in the United Kingdom, of which 54 and 25, respectively, were open surgical biopsies.
Hofvind et al. (2012b) compared the Norwegian mammography screening programme with screening practice in Vermont, USA (Vermont is a member of the Breast Cancer Surveillance Consortium, an initiative of the United States National Cancer Institute), showing that higher recall rates and lower specificity in the USA were not associated with higher sensitivity. These differences may be explained by professional practices, since screening centres in the USA usually have small volumes of mammography readings, and double reading is not a quality requirement in the USA as it is in Europe ( Burnside et al., 2014 ).
The cumulative risk of a false-positive recall is one of the most important harms of screening. The false-positive rate is estimated from the recall rate by subtracting the cancer detection rate in the same screening episode. The cumulative risk of a false-positive result is defined as the cumulative risk of recall for further assessment at least once during the screening period (usually 10 biennial screening episodes in organized programmes) minus the cumulative risk of cancer detection over the same period. There is a similar definition for the cumulative risk of having an invasive procedure (needle biopsy or surgical biopsy) with a benign outcome.
A systematic review has been made of publications estimating the cumulative risk of a false-positive result in European population-based mammography screening programmes ( Hofvind et al., 2012a ). Four studies were included, based on data from the 1990s and conducted in Denmark, Italy, Norway, and Spain. Results updated with a further 9 years of experience in Norway have since been published ( Román et al., 2013 ). The cumulative risk of any further assessment without cancer diagnosis varied from 8.1% to 20.4% in the most recent period (ending variously in 2001 to 2010), and the cumulative risk of assessment with an invasive procedure without cancer diagnosis varied from 1.8% to 4.1%.
The cumulative risk of false-positives is higher in opportunistic mammography screening, which is the usual modality in the USA. Elmore et al. (1998) estimated that 41% of screened women had at least one false-positive result over 10 screening episodes. Hubbard et al. (2010) applied statistical models to more recent data from the Breast Cancer Surveillance Consortium for women aged 40–59 years at entry and followed up over their screening history. The risk of a false-positive over 10 screening mammograms varied between 58% and 77%.
Román et al. (2012) assessed factors affecting the false-positive rate after any assessment, and after assessment with an invasive procedure, in a retrospective cohort in Spain. The authors reported that the false-positive risk after assessment with an invasive procedure was less for digital mammography (RR, 0.83) than for non-digital mammography, and they estimated a total cumulative risk of 20.4%, ranging from 51.4% for the highest risk profile to 7.5% for the lowest risk profile. The risk after assessment with all procedures and with invasive procedures was estimated to be higher for younger women (OR, 1.30 for age 40–44 years; OR, 1.26 for age 40–54 years; reference category, age 65–69 years).
In the USA, Kerlikowske et al. (2013) assessed the cumulative risk by breast density and risk profile. The cumulative probability of a false-positive mammography result was higher among women with extremely dense breasts who underwent annual mammography and either were aged 40–49 years (65.5%) or used combined estrogen–progestogen hormone therapy (65.8%), and was lower among women aged 50–74 years who underwent biennial or 3-yearly mammography and had scattered fibroglandular densities (30.7% and 21.9%, respectively) or fatty breasts (17.4% and 12.1%, respectively).
Indicators of the cumulative risk of false-positives are included as possible harms of screening in the balance sheet of benefits and harms. The Euroscreen mammography screening balance sheet considered 1000 women who were aged 50 or 51 years at the start of their screening regimen. The cumulative risk of false-positives was estimated to be 200 over the 10 screening rounds from age 50 years to age 69 years; 170 women were recalled for further assessment without invasive procedures, and 30 women had further assessment with invasive procedures ( Paci & EUROSCREEN Working Group, 2012 ).
The definition of overdiagnosis and estimates of overdiagnosis in randomized trials of mammography screening have been presented in Section 4.2.3c. The quantification of overdiagnosis is important in observational studies because this harm was not a primary end-point of the RCTs and estimates are influenced by local screening practice and technological innovation. Other approaches, such as radiological doubling time, have been suggested as useful indicators for the study of overdiagnosis, but in this section overdiagnosis is considered as an epidemiological construct, based on a retrospective analysis of breast cancer diagnosis in the population.
Several approaches have been proposed for estimating overdiagnosis in observational studies.
The cumulative incidence method estimates overdiagnosis by following up a cohort of women, invited and not invited to screening or screened and not screened. The ideal study would require the follow-up of pairs of birth or enrolment age cohorts in which one cohort is invited to screening and the other is not invited ( Møller et al., 2005 ; Biesheuvel et al., 2007 ). The attribution of an individual time zero to each invited woman allows for estimation of changes in incidence over the screening period in the population and monitoring of the compensatory drop phase after the end of screening ( Fig. 5.2 ).
Observed and modelled breast cancer incidence per 100 000 person–years in the presence and absence of screening in 1990–2006
The incidence-rate method compares the average annual incidence of breast cancer over a defined period of follow-up in a specified age group of women who were offered or accepted screening with an estimate of the average annual incidence of breast cancer during the same period in women who were not offered screening or were not screened. Overdiagnosis is taken to be any excess in incidence in the former over the latter once the screening lead time has been accounted for. Several methods have been suggested for the adjustment for lead time, with the aim of overcoming the frequent difficulty of too short a follow-up period for the lead time to have passed in all women under observation who had been invited to screening or were screened.
In a methodological study, Etzioni et al. (2013) contrasted an incidence excess approach with a lead-time approach. The lead-time approach uses the disease incidence under screening to make inferences about the lead time or the natural history of the disease. Using the incidence excess approach , the authors suggested that the estimate should consider the time needed for screening dissemination and the compensatory drop, as expressed by incidence rates at older ages. In the presence of a shorter follow-up time and/or unequal screening periods in the age cohorts of women, statistical adjustment for lead time is required. This can be based on estimates of lead time derived from clinical cancers (such as estimates derived from experience before the introduction of population screening programmes) or estimates from modelling studies.
Simulation, using statistical modelling, of lifetime individual histories with or without screening is often used to overcome the complexity of screening evaluation, in particular to account for lead time and to give understandable outcomes (see Section 5.1.2f ). Complex models such as these need a set of assumptions about natural history of the disease and screening performance ( Tan et al., 2006 ), which would ideally be clearly stated in reports based on the models’ use but generally are not. Importantly, too, a paucity of relevant empirical evidence means that assumptions about the proportion of preclinical cancers that are non-progressive and the range and distribution of lead time, which are critical to modelled estimates of overdiagnosis, are very uncertain.
Duffy & Parmar (2013) , using estimates of the incidence rate in the United Kingdom and an exponential distribution of the lead time, simulated the time course of incidence rates during and after the screening period in the absence of overdiagnosis. With a 20-year period of screening (from age 50 years to age 69 years), a period of at least 10 years must elapse after the screening period (to when women are aged 79 years) for the excess incidence rate to be close to the rate observed in the absence of screening (to within 1% of excess with 30 years of follow-up from the start of screening). It is important to note that in the same simulation, 10 years of observation of a population of women screened from age 50–69 years at the start of screening will give an incidence excess of 50%. This model assumed an average lead time of 40 months. However, some estimates are much lower (see, for example, Feinleib & Zelen, 1969 ). Although there is disagreement over the average and distribution of lead time for breast cancer, the main conclusion is that an adequate correction for lead time is needed in the absence of a sufficient follow-up period to distinguish excess of incidence due to lead time from overdiagnosis .
An important factor determining the observational estimate of overdiagnosis is the estimate of the underlying incidence. In descriptive epidemiological studies, an estimate of incidence in the absence of screening is needed. In comparative studies, the reference population should be comparable to the invited population so far as is possible in terms of the background incidence rate, breast cancer risk factors, socioeconomic status, and use of health services other than for mammography. If rates from the same or another historical (pre-screening) population are used, the time trend in the underlying incidence must be estimated, a projection made to the screened population, and sensitivity analyses of the estimates made that take account of variation in the trend due to unpredicted changes in population composition or the prevalence of risk factors. Self-selection bias should also be considered and adjusted for if attenders only are evaluated.
Adjustment for lead time and estimation of the underlying incidence of breast cancer in the absence of screening (control of confounding due to differences in breast cancer risk factors between screened and unscreened women) were considered as the main problems in estimating overdiagnosis in observational studies ( Njor et al., 2013a ), but these are not the only factors to be considered. Others include ( Njor et al., 2013a ): the nature and quality of the observational data used; what estimate was actually reported as a measure of overdiagnosis (ideally classified in the terms outlined by the Independent UK Panel on Breast Cancer Screening, 2012 ), which is sometimes not clearly described, and, for the Independent United Kingdom Panel’s measure A or B, how long the period of follow-up was after screening stopped (periods beyond about 10 years from the end of screening will cause progressive “dilution” of the overdiagnosis estimate; de Gelder et al., 2011a ); whether the estimate was based on women invited to screening or women who attended screening; what the screening policies were during the period of screening to which the overdiagnosis estimate related (e.g. age at starting and at stopping screening, and screening interval); and whether the estimate is based on steady-state screening or screening that includes all or a proportion of the period after initiation of screening during which women across the whole screening age range are receiving their first invitations to screening (inclusion of this period will produce higher estimates due to greater inclusion of prevalent screens, in which the probability of overdiagnosis is higher than it is for incident screens).
Observational studies of overdiagnosis for women aged 50–69 years are summarized in Table 5.11 and Table 5.12 . Table 5.11 covers studies reviewed by the Euroscreen Working Group ( Puliti et al., 2012 ), which included all 13 observational studies conducted in Europe that were published up to February 2011. Table 5.12 covers 17 studies conducted in Europe and published from February 2011 to November 2014, when the Handbook Working Group met, or conducted outside Europe and published up to November 2014.
Studies of the estimates of overdiagnosis in Europe a .
Studies of estimates of overdiagnosis in Europe (published from February 2011 to November 2014) and in other countries (published up to November 2014) .
Estimates of the overdiagnosis risk, principally the Independent United Kingdom Panel’s measure A (the excess cancers expressed as a proportion of cancers diagnosed over the whole follow-up period in unscreened women), ranged from −0.7% to 76% for invasive cancer only and from 1% to 57% for invasive and in situ cancers together.
The Euroscreen Working Group characterized overdiagnosis estimates as made with or without correction for lead time and underlying incidence trend. The reported estimates that were considered as adequately adjusted for both biases (from 6 of the 13 studies) ranged from 1% to 10% excess over the expected incidence for all breast cancers (measure A) (1% to 10% for invasive cancer only, from 4 studies, and 1% to 7% for invasive and in situ cancers, from 4 studies). The majority of the studies used temporal trends or geographical differences in dynamic populations to adjust for the underlying incidence. Only two studies used the cohort population approach, and a few studies used statistical modelling for the estimate. The Euroscreen Working Group derived a summary estimate of overdiagnosis of 6.5% of the incidence in the absence of screening. This is the estimate of the overdiagnosis in women screened between the ages of 50 years and 69 years and followed up for 10 years after the last screening, and included carcinoma in situ ( Paci & EUROSCREEN Working Group, 2012 ), measure A as defined by the Independent UK Panel on Breast Cancer Screening (2012) .
The IARC Working Group also sought to distinguish analyses that adequately adjusted for lead time and for the underlying breast cancer incidence trend: these were the analyses of Puliti et al. (2012) , Kalager et al. (2012) , Falk et al. (2013) , Lund et al. (2013) , Njor et al. (2013a) , Heinävaara et al. (2014) (estimate A1 only), and Beckmann et al. (2015) . The range of estimates from these studies was 2% to 25% for invasive cancer only and 2% to 22% for invasive and in situ cancers together.
Over the past 50 years, breast cancer care has moved from aggressive, mutilating surgery to breast-conserving treatment ( Fisher et al., 2002 ; Veronesi et al., 2002 ). This change was the starting point for improvements in other treatment and assessment areas, such as, for example, the sentinel lymph node procedure, which has been well established in clinical practice since the early 2000s ( Veronesi et al., 2003 ). Detection of early, indolent lesions, such as carcinoma in situ ( Ernster et al., 2002 ), is a major area of concern. In a recent international survey, Lynge et al. (2014) documented the wide variability in the occurrence of in situ breast cancer across countries. In a comparison with European programmes, higher probabilities for the occurrence of carcinoma in situ were reported in the USA. This finding is associated with higher false-positive rates and biopsy rates in the diagnostic assessment phase ( Smith-Bindman et al., 2005 ).
Carcinomas in situ have high survival rates after treatment, but studies have shown that only a proportion of them, depending mainly on the pathological grade, would have progressed to invasiveness over the lifetime of the woman in the absence of early diagnosis. Overdiagnosed breast cancer cases are all overtreated. Carcinoma in situ is considered a major area of overtreatment. However, overtreatment is a harm not limited to screen-detected cases. Clinicians follow shared guidelines, primarily based on the stage at presentation of the disease. Screen-detected cases, when treated in the same cancer unit, will receive treatment by tumour characteristics. Chemotherapy and hormone therapy for breast cancer are progressively being extended to very early and less-progressive cancers ( Peto et al., 2012 ), with important implications when there is a growing proportion of early, high-survival-rate breast cancers.
An example of the relationship between overdiagnosis and overtreatment is the comparison of mastectomy rates in the screening and pre-screening epochs. In a Cochrane systematic review ( Gøtzsche & Jørgensen, 2013 ), a 31% increase in mastectomy and lumpectomy rates (20% excess of mastectomies) was estimated in the intervention group compared with the control group. This estimate considered all breast cancer cases detected in the screening period (i.e. the excess of incidence observed in the screening arm).
Zorzi et al. (2006) evaluated the use of mastectomy in Italy in the period 1997–2001, during which a large number of screening programmes were implemented, using individual data classified by stage and modality of diagnosis in relation to screening. The probability of a mastectomy increased with age and primary tumour size, and screen-detected cases were half as likely to be treated with mastectomy as non-screen-detected cases. The increasing rates of early-stage cancers (< 30 mm) and the use of breast-conserving treatment paralleled a decline in the mastectomy rate and in the incidence of advanced-stage cancers (> 30 mm), showing an appropriate use of the surgical approach.
Suhrke et al. (2011) , using population-based data in the epoch of change to a service screening programme, showed an increase in rates of breast surgery and also an increase in mastectomy rates immediately after the start of the screening programme. They described a recent decline in mastectomy rates and suggested that the change affected all age groups and that it is likely to have resulted from changes in surgical policy.
Exposure of the breast to ionizing radiation may induce breast cancers (see Section 1.3.4). The low dose of X-ray photon radiation received during mammography is thus considered as a potential adverse effect of breast cancer screening. The number of cancers caused by screening with mammography must be estimated to evaluate the balance between benefits and risks. However, due to the small number of expected cases, it is not possible to estimate such a number from epidemiological data. Thus, numerous studies have used a quantitative risk assessment approach. This approach is based on a large number of hypotheses arising from current scientific knowledge and on hypotheses about screening modalities.
(i) hypotheses about risk models.
Hypotheses about risk models come from the selection of the most reliable studies on the relationship between radiation exposure and breast cancer risk (see Section 1.3.4). Hypotheses are made about the form of this relationship, the modifying effect of time and age at exposure, the latency time between exposure and risk, and transposition from high to low dose and low exposure rate.
The most recent models for such an exercise in the general population arise from the BEIR VII models of the United States National Academy of Sciences ( National Research Council, 2006 ), with recommendations of the use of an excess absolute risk model for breast cancer risk ( National Research Council, 2006 ; ICRP, 2007 ; Wrixon, 2008 ). This model assumes no threshold, even at a very low dose, and a decreasing effect with increasing age at exposure. Coefficients are estimated from atomic bomb survivors and women medically exposed to radiation (see Section 1.3.4). Because these studies are based on a higher dose and a higher dose rate than those typically involved in mammography screening, an effort was made by some authors to produce results taking into account transposition factors from high to low dose and dose rate (dose and dose rate effectiveness factor). Values of this factor in the context of mammography generally vary between 1 and 2 ( National Research Council, 2006 ; Law et al., 2007 ; Heyes et al., 2009 ).
A hypothesis about the latency time for the induction of a breast cancer by radiation is also needed for risk assessment. A latency time of 10 years is generally used, with values varying from 5 years to 15 years.
The estimation of doses received by the glandular tissue of the breast depends on breast thickness and density. Based on an extensive literature review, a historical reconstruction of doses received during mammography shows a strong decrease over time, with an estimated mean glandular dose to the breast of 2 mGy per view since 2000 ( Thierry-Chef et al., 2012 ) (see Section 1, Fig. 1.16). Moreover, recent use of digital mammography (instead of screen-film mammography) has led to new estimates of doses received ( Hendrick et al., 2010 ; Hauge et al., 2014 ).
To fully develop the risk assessment, scenarios for the target population and screening modalities (age range, frequency, number of examinations at each screening, additional views, etc.) have been developed.
Risk assessment studies provide estimated numbers of radiation-induced breast cancer cases and/or deaths, with a range of estimates according to variations in hypotheses. Estimation of prevented deaths based on assumptions about mortality reduction by screening modalities is performed in most studies, and calculation of benefit–risk is provided. Because the risk of radiation-induced cancer applies only to women who underwent mammography, hypotheses about mortality reduction should apply only to attendees; this is not always made explicit in publications. Thus, benefit–risk estimates provided by studies should be interpreted with caution.
Risk assessment studies performed in the early 2000s or earlier used risk models that are no longer recommended by international committees ( Howe et al., 1981 ; Feig & Hendrick, 1997 ; Beemsterboer et al., 1998a ; Mattsson et al., 2000 ; Law & Faulkner, 2001 , 2002 , 2006 ; León et al., 2001 ; Berrington de González & Reeves, 2005 ; Ramos et al., 2005 ). Since 2010, all studies have used the excess absolute risk model recommended by BEIR VII and contemporary estimates of mean glandular dose to the breast from either screen-film or digital mammography ( Hendrick, 2010 ; O’Connor et al., 2010 ; de Gelder et al., 2011b ; HPA, 2011 ; Yaffe & Mainprize, 2011 ; Hauge et al., 2014 ). These recent studies are now considered to be the most relevant and are summarized below ( Table 5.13 ). In addition, one study used a biological model ( Bijwaard et al., 2010 , 2011 ).
Risk assessment studies of breast cancer induced by mammography screening a .
The Health Protection Agency estimated the number of cancer cases and cancer deaths after radiation exposure from a large number of sources, including screening mammography, in the United Kingdom population ( HPA, 2011 ). The number of radiation-induced breast cancer cases after a single two-view screen every 3 years at age 47–73 years was estimated to be 28 per 100 000 women screened, and the number of breast cancer deaths under the same conditions was estimated to be 10 per 100 000 women screened. Assuming 500 prevented deaths from screenings, the authors estimated the net benefit (deaths prevented minus deaths induced) to be 490 [ratio of prevented to induced deaths of 50].
O’Connor et al. (2010) estimated the number of breast cancer cases induced by screen-film mammography, digital mammography, and other imaging techniques in a United States setting. They estimated that 21 cancer cases would be induced by digital mammography and 27 by screen-film mammography for annual screening per 100 000 women screened at age 50–80 years, and that there would be 6 or 7 induced deaths. Using different mortality reduction hypotheses, they estimated ratios of prevented to induced deaths of 116 and 135 for screen-film and digital mammography, respectively.
In Norway, Hauge et al. (2014) estimated the number of radiation-induced breast cancer cases after a single two-view digital mammography screening every 2 years from age 50 years to age 69 years to be 10 (range, 1.4–36) per 100 000 women screened, and the number of induced deaths per 100 000 women screened to be 1 (range, 0.1–3). Assuming a 40% mortality reduction among attendees, the authors estimated that 350 lives would be saved compared with 3 or fewer deaths induced [ratio of prevented to induced deaths of at least 117].
In the Netherlands, calculations were performed for a biennial digital mammography screening between the ages of 50 years and 74 years [12 screening sessions] ( de Gelder et al., 2011b ). The authors estimated 7.7 radiation-induced breast cancer cases (range, 5.9–29.6) and 1.6 radiation-induced breast cancer deaths (range, 1.3–6.3) per 100 000 women screened, assuming a glandular dose of 1.3 mGy per view. Using a simulation model (MISCAN) to estimate deaths prevented due to screening, they estimated a ratio of prevented to induced deaths of 684. When a glandular dose of 5 mGy per view was assumed, the ratio decreased to 178 and the number of radiation-induced deaths increased to 6.3.
Bijwaard et al. (2010 , 2011 ) performed a risk assessment using a mechanistic, biologically based model that assumes a two-stage mutation for carcinogenesis. With this approach, the authors estimated that for five mammography screenings of 2 mGy starting at age 50 years [biennial screening until age 60 years], 1.3 breast cancer cases would be induced per 100 000 women screened ( Bijwaard et al., 2010 ), and 200 cases for 15 screenings of 4 mGy.
In the United Kingdom calculation ( HPA, 2011 ), the number of radiation-induced breast cancer cases after annual two-view screening at ages 40–47 years was estimated to be 61 per 100 000 women screened. Using a hypothesis about survival, the authors estimated the number of radiation-induced breast cancer deaths after annual two-view screening at ages 40–47 years to be 20 per 100 000 women screened. Assuming 100 prevented deaths from screening, they estimated the net benefit (deaths prevented minus deaths induced) to be 80 [ratio of prevented to induced deaths of 5].
In the USA, Hendrick (2010) estimated the number of deaths induced by annual mammography per 100 000 women screened at age 40–80 years to be 20 for digital mammography and 25 for screen-film mammography. In the study of O’Connor et al. (2010) , the authors estimated the number of breast cancers induced by annual mammography per 100 000 women screened at age 40–49 years to be 35 for digital mammography and 44 for screen-film mammography, and the number of radiation-induced breast cancer deaths to be 9 for digital mammography and 11 for screen-film mammography. According to a hypothesis about mortality reduction, they estimated a ratio of prevented to induced deaths of about 3 for both modalities.
In Canada, Yaffe & Mainprize (2011) estimated that mammography screening annually from age 40 years to age 55 years and biennially until age 74 years would induce 86 breast cancers cases (59 for the screening period 40–49 years) and 11 breast cancers deaths (7.6 for the screening period 40–49 years) per 100 000 women screened. Assuming a 24% reduction in mortality, they estimated a ratio of prevented to induced deaths of 46 for age 40–74 years (11.4 for age 40–49 years). The ratio of lives saved to lives lost is 78 for age 40–74 years (27 for age 40–49 years).
In the Netherlands, calculations were performed for biennial mammography screening between age 40 years and age 74 years; the authors estimated the number of breast cancer cases per 100 000 women screened to be 17.1 (range, 13.1–65.6) and the number of radiation-induced breast cancer deaths to be 3.7 (range, 2.9–14.4) ( de Gelder et al., 2011a ). Using a simulation model (MISCAN) to estimate deaths prevented due to screening, they estimated a ratio of prevented to induced deaths of 349. The study using a mechanistic model estimated 1.5 cases per 100 000 women screened for five mammography screenings of 2 mGy starting at age 40 years ( Bijwaard et al., 2010 ).
Among women at an increased risk of breast cancer, screening procedures are recommended earlier in life and at a higher frequency than in the general population (see Section 5.6 ). Due to the increased risk of radiation-induced breast cancer when exposure occurs at a younger age and because of the higher radiosensitivity of women with a familial predisposition (see Section 1.3.6), separate risk assessment must be performed for women at an increased risk.
An excess relative risk model was used to estimate the lifetime risk of radiation-induced breast cancer mortality from five annual mammography screenings in young women harbouring a BRCA mutation ( Berrington de González et al., 2009 ). They estimated the lifetime risk of radiation-induced breast cancer mortality per 10 000 women screened annually to be 26 (95% CI, 14–49) for screening at age 25–29 years, 20 (95% CI, 11–39) for screening at age 30–34 years, and 13 (95% CI, 7–23) for screening at age 35–39 years. [This calculation was based on model risk and coefficients estimated from the general population, and the higher sensitivity to radiation of these women was not taken into account.] A large European study among carriers of BRCA1/2 mutations suggested that exposure to diagnostic radiation before age 30 years for these women was associated with an increased risk of breast cancer at dose levels considerably lower than those at which increases had previously been found ( Pijpe et al., 2012 ).
Benefit–risk estimates for women at an increased risk need to consider: the age-dependent higher risk of radiation in younger women and in women with specific gene mutations; their age-dependent overall measured breast cancer risk; and the contribution of mammography to early detection, which itself may depend on patient age, the type of genetic mutation ( BRCA1 vs BRCA2 ), and the availability of magnetic resonance imaging (MRI).
Participation in breast cancer screening can have psychological or psychosocial consequences for women. Section 3.1.4 summarizes the psychological impacts of an invitation to screening, of a negative result, of a diagnosis of breast cancer, and of interval cancer, as well as the impact of a false-positive result on further participation. This section presents the studies reviewed for the evaluation of the psychological consequences of a false-positive result and of DCIS.
Several reviews have focused on the long-term psychological implications of a false-positive result ( Rimer & Bluman, 1997 ; Steggles et al., 1998 ; Brodersen et al., 2004 ; Brett et al., 2005 ; Brewer et al., 2007 ; Hafslund & Nortvedt, 2009 ; Salz et al., 2010 ; Bond et al., 2013a , b ). The two reviews by Bond et al. (2013a , b ) evaluate the same set of studies, so one has been excluded. The review by Rimer & Bluman (1997) has also been excluded, due to its lack of relevance. In this section, the outcomes of the informative reviews ( Table 5.14 ) and results from more recent individual studies are presented.
Measures used in 70 studies of psychological consequences of a false-positive result of mammography screening.
Negative outcomes were reported from studies using validated measures during the period between receiving a recall letter and the recall appointment ( Sutton et al., 1995 ; Chen et al., 1996 ; Lowe et al., 1999 ; Lampic et al., 2001 ; Sandin et al., 2002 ), at the recall appointment ( Ellman et al., 1989 ; Cockburn et al., 1992 ; Swanson et al., 1996 ; Lowe et al., 1999 ; Ekeberg et al., 2001 ; Meystre-Agustoni et al., 2001 ), or immediately after receiving a recall letter ( Cockburn et al., 1994 ; Lidbrink et al., 1995 ; Olsson et al., 1999 ; Lindfors et al., 2001 ).
The main psychological consequences of a false-positive result were psychological distress, somatization, depression, fear, anxiety, worry, an increase in women’s perceived likelihood of developing breast cancer, a decrease in the perceived benefits of mammography, and an increase in the frequency of breast self-examination (BSE) ( Salz et al., 2010 ). [These outcomes may be contextualized as symptoms, but it is unclear how they would affect women in their everyday lives.]
Salz et al. (2010) performed a meta-analysis of the effect of false-positive mammograms on generic and specific psychosocial outcomes. From 17 studies presented in 21 articles, they found that across six generic outcomes, the only consistent effect was generalized anxiety ( Ellman et al., 1989 ; Gram et al., 1990 ; Bull & Campbell, 1991 ; Lerman et al., 1991a , 1993 ; Cockburn et al., 1994 ; Ong et al., 1997 ; Scaf-Klomp et al., 1997 ; Brett et al., 1998 ; Pisano et al., 1998 ; Olsson et al., 1999 ; Aro et al., 2000 ; Lipkus et al., 2000 ; Brett & Austoker, 2001 ; Lampic et al., 2001 , 2003 ; Meystre-Agustoni et al., 2001 ; Sandin et al., 2002 ; Barton et al., 2004 ; Jatoi et al., 2006 ; Tyndel et al., 2007 ).
All reviews concluded that there are short-term psychological consequences (up to 3 months) from having a recall. In one review ( Brodersen et al., 2004 ), all 22 studies that investigated short-term consequences reported adverse short-term consequences. In a review based on 54 articles, Brett et al. (2005) concluded that the negative psychological impact was significantly higher for women who had a recall than for women who received a clear negative result after participation in mammography screening, although three studies reported no difference in the psychological impact of mammography screening between women who received a clear negative result and those who had a false-positive result ( Bull & Campbell, 1991 ; Lightfoot et al., 1994 ; Aro et al., 2000 ). Other negative consequences reported in women who had a false-positive result were more intrusive thoughts, worry about breast cancer, greater requirements for social support, being more busy than usual to keep their thoughts away from the clinical visit, or difficulties sleeping ( Bull & Campbell, 1991 ; Lightfoot et al., 1994 ; Scaf-Klomp et al., 1997 ; Gilbert et al., 1998 ; Aro et al., 2000 ). Two studies reported that 30% ( Austoker & Ong, 1994 ) and 40% ( Scaf-Klomp et al., 1997 ) of women felt very anxious when they received a recall letter. One study that looked at how having a false-positive result influences quality of life found a marked decrease in quality of life for recalled women ( Lowe et al., 1999 ).
Based on the available reviews, results about long-term consequences are more ambiguous and inconsistent ( Brodersen et al., 2004 ; Brett et al., 2005 ; Brewer et al., 2007 ). Several studies did not find increases in long-term levels of anxiety among women who had a false-positive result ( Gram et al., 1990 ; Cockburn et al., 1994 ; Lidbrink et al., 1995 ; Gilbert et al., 1998 ; Lowe et al., 1999 ; Ekeberg et al., 2001 ; Lampic et al., 2001 ; Sandin et al., 2002 ), and two studies were inconclusive ( Scaf-Klomp et al., 1997 ; Aro et al., 2000 ). Other studies reported that the anxiety experienced was greater among women who had a false-positive result than among women who received a clear negative result, at 4–6 months after recall ( Ellman et al., 1989 ; Brett et al., 1998 ; Olsson et al., 1999 ; Lampic et al., 2001 ), 6–12 months after recall ( Lampic et al., 2001 ; Hislop et al., 2002 ), and 24 months after recall ( Lipkus et al., 2000 ). One review found no long-term symptoms of depression among women who received a false-positive result ( Brewer et al., 2007 ).
One review investigated the effects on health-care use and symptoms ( Brewer et al., 2007 ). The findings suggested that having a false-positive result increases anxiety related to breast cancer-specific measures ( Brewer et al., 2007 ). Three studies found that women who received a false-positive result reported conducting BSE statistically significantly more frequently ( Bull & Campbell, 1991 ; Aro et al., 2000 ; Lampic et al., 2001 ). Women who had a false-positive result also reported higher levels of worry and increased concern about breast cancer ( Lerman et al., 1991a , b ; Scaf-Klomp et al., 1997 ; Brett et al., 1998 ; Aro et al., 2000 ; Lipkus et al., 2000 ; Sandin et al., 2002 ; Absetz et al., 2003 ). In their meta-analysis, Salz et al. (2010) found statistically significant effects on all eight breast cancer-specific outcomes: distress about breast cancer, somatization or symptoms in the breast, fear of developing breast cancer, anxiety about breast cancer, worry about breast cancer, perceived likelihood of breast cancer, perceived benefits of mammography, and frequency of BSE. The largest effect was for anxiety about breast cancer ( r = 0.22) and the smallest was for fear ( r = 0.08); all eight pooled effect sizes were statistically significant.
Screening factors associated with greater adverse psychological effects were: previous false-positive results ( Brett & Austoker, 2001 ; Haas et al., 2001 ; Lampic et al., 2001 ), pain at previous mammography screening ( Ong & Austoker, 1997 ; Drossaert et al., 2002 ), dissatisfaction with information and communication during screening ( Austoker & Ong, 1994 ; Brett et al., 1998 ; Brett & Austoker, 2001 ; Dolan et al., 2001 ), and waiting time between recall letter and assessment appointment ( Gram et al., 1990 ; Thorne et al., 1999 ; Brett & Austoker, 2001 ; Lindfors et al., 2001 ).
Elements of the structure of the screening programme were also found to be important. The extent of further investigation seemed to determine the extent of negative psychological outcomes. Women who underwent a surgical biopsy before receiving a clear result experienced the greatest anxiety ( Ellman et al., 1989 ; Lerman et al., 1991b ; Ong & Austoker, 1997 ; Brett et al., 1998 ; Lampic et al., 2001 ), as did those asked to come back for further tests after 6 months or 1 year ( Ong et al., 1997 ; Brett et al., 1998 ; Brett & Austoker, 2001 ). On-site evaluation was shown to reduce the stress of having a false-positive result ( Lindfors et al., 2001 ). Biopsy-specific events appeared to be more distressing than follow-up mammography, and distress risk factors included younger age, less education, and no family history of breast cancer ( Steffens et al., 2011 ).
Reported sociodemographic factors often associated with greater adverse psychological outcomes were younger age, less education, living in an urban area, having one child or no children, and manual occupation ( Brett et al., 2005 ). Other studies found no impact of age ( Brett et al., 1998 ; Brett & Austoker, 2001 ; Lampic et al., 2001 ) or employment ( Olsson et al., 1999 ). One study with 910 participants in California, USA, found that Asian ethnicity, annual income greater than US$ 10 000, and weekly attendance of religious services were significantly associated with decreased depressive symptoms ( Alderete et al., 2006 ).
More recent studies, not included in the reviews, have used the Hospital Anxiety and Depression Scale, the Psychological Consequences Questionnaire, and the Consequences of Screening in Breast Cancer questionnaire to study psychological consequences of mammography screening ( Table 5.14 ). Consistent with findings from a study conducted in 1996–1997 ( Ekeberg et al., 2001 ), Schou Bredal et al. (2013) found that recall after mammography among women with a false-positive result was associated with transiently increased anxiety and a slight increase in depression. At 4 weeks after screening, the level of anxiety was the same and depression was lower compared with the general female Norwegian population ( Schou Bredal et al., 2013 ).
In a study in Spain, participants were found to worry little until they underwent mammography, but levels of worry increased when the women were notified by telephone call of the need for further testing ( Espasa et al., 2012 ). A substantial proportion of women requiring further assessment reported that they were at least somewhat worried about having breast cancer throughout the screening process, but levels of anxiety and depression, measured by the Hospital Anxiety and Depression Scale, showed no statistically significant differences among women who had invasive complementary tests, non-invasive tests, and negative screening results ( Espasa et al., 2012 ).
In a longitudinal study in Denmark, psychological effects of false-positive results were assessed with the Consequences of Screening in Breast Cancer questionnaire. At 6 months after the final diagnosis, women with a false-positive finding reported changes in existential values and inner calmness as great as those reported by women with a diagnosis of breast cancer; 3 years after the final diagnosis, women who had a false-positive result consistently reported greater negative psychosocial consequences in all 12 psychosocial outcomes compared with women who had a normal finding ( Brodersen & Siersma, 2013 ). However, after 5 years, there was no statistically significant difference between the two groups in reported psychosocial aspects ( Osterø et al., 2014 ).
When women who were first-time participants in mammography screening were compared with women with repeated screening experience, women in both groups reported experiencing high levels of anxiety before the diagnosis was known, and no differences were found in anxiety, depressive symptoms, or quality of life ( Keyzer-Dekker et al., 2012 ).
In a study in 98 women, women reported a significant increase in anxiety after being notified of the need to return for follow-up testing, and significant positive associations were found between anxiety and behavioural approach, behavioural avoidance, cognitive approach, and cognitive avoidance coping in cross-sectional analyses ( Heckman et al., 2004 ). Moreover, cognitive avoidance coping was a strong predictor of final levels of state anxiety in these women ( Heckman et al., 2004 ).
These findings are consistent with qualitative studies in Scandinavia and North America. Norwegian women expressed mixed emotions over being recalled; information about recall rates and breast cancer risk was seen as alarming, and the short time between recall and examination was seen as reassuring but was also perceived as an indication of malignancy ( Solbjør et al., 2011 ). Swedish women who were recalled described the recall process as “a roller coaster of emotions” ( Bolejko et al., 2013 ). Qualitative studies from North America have described the psychological effects of the waiting process experienced by women, their unmet informational and psychosocial needs ( Doré et al., 2013 ), anxieties generated by waiting and wondering, and fears of iatrogenic effects of follow-up tests such as biopsies and repeat mammograms ( Padgett et al., 2001 ).
Psychological consequences of DCIS are included in this section because increasing participation in mammography screening means an increasing number of DCIS detections among women, but the effect of DCIS on psychological issues has been little explored. Women may not be aware of having DCIS, because surgeons might differ in how they inform women about this condition. Potentially, some women with DCIS are informed that they have breast cancer while others are informed that they do not have breast cancer. A study with semi-structured interviews of women previously diagnosed with and treated for DCIS identified six key themes: (i) invisibility of DCIS, (ii) uncertainty, (iii) perceptions of DCIS, (iv) acceptance of treatment, (v) social support, and (vi) moving on, which highlight the substantial challenges faced by women diagnosed with DCIS ( Kennedy et al., 2008 ).
No articles focused on non-invasive breast cancer or DCIS before 1997 ( Webb & Koch, 1997 ). A review of quality-of-life issues among patients with DCIS ( Ganz, 2010 ) found that women with DCIS experience psychological consequences to a lesser extent than women with breast cancer, but few studies have compared these women with healthy women. Of greater concern, women with DCIS demonstrate severe misconceptions about their risk of invasive breast cancer ( Ganz, 2010 ).
One study of 10 women with DCIS found that they would have liked to have received more information about DCIS when they were invited to routine screening ( Prinjha et al., 2006 ). In another study, 45 women took part in an initial interview after a diagnosis of DCIS, and 27 took part in a follow-up interview 9–13 months later ( Kennedy et al., 2012 ). Women’s early perceptions of DCIS merged with and sometimes conflicted with their beliefs about breast cancer, and their perceptions and experiences of the condition shifted over time.
A study in Australia also found misunderstanding and confusion among women diagnosed with DCIS and a desire for more information about their breast disease ( De Morgan et al., 2011 ). Approximately half of the participants worried about their breast disease metastasizing, approximately half expressed high decisional conflict, 12% were anxious, and 2% were depressed. Logistic regression analysis demonstrated that worry about dying from the breast disease was significantly associated with not knowing that DCIS could not metastasize ( De Morgan et al., 2011 ). In five focus group interviews involving 26 women diagnosed with DCIS, women were confused about whether or not they had cancer that could result in death, and this confusion was compounded by the use of the term “carcinoma” and by the recommendation of treatments such as mastectomy ( De Morgan et al., 2002 ).
In a study of 487 women who were newly diagnosed with DCIS, financial status was inversely associated with anxiety and depression at the 9-month follow-up, and women with medium or low socioeconomic status were vulnerable to escalating anxiety and depression after a DCIS diagnosis ( de Moor et al., 2010 ). A study in the USA of approximately 800 Latina and Euro-American women with DCIS found that younger age, not having a partner, and lower income were related to lower quality of life in various domains ( Bloom et al., 2013 ).
Decisions about implementation of health-care interventions are based primarily on benefits and a favourable harm–benefit ratio, but – to use limited resources efficiently – are also often based on cost–effectiveness analyses. A cost–effectiveness analysis compares different policies, including the current one, with no intervention (average cost–effectiveness) or compares a more-intensive programme with a less-intensive one (incremental cost–effectiveness). Often, the incremental cost–effectiveness ratio (ICER) is estimated for each policy, expressed as the ratio of the change in costs to the change in effects compared with a less-intensive alternative or the current policy. In a cost–effectiveness analysis, future costs and effects are taken into account and both are discounted at a chosen annual discount rate, to account for time preference. A new strategy is considered cost-effective if it results in an additional effect (compared with a baseline) at acceptable additional costs (or even savings). One should stress the fact that the change in effects is as important as, and in the practice of policy-making even more important than, the change in costs: how much will the population benefit from the resources invested? Effects are often defined as disease-specific deaths prevented and life years gained but are ideally adjusted for quality of life, resulting in quality-adjusted life years (QALYs) ( Weinstein & Stason, 1977 ). For breast cancer screening, factors that could negatively affect quality of life are, among others, the screening examination, false-positive referrals, earlier and often more intensive treatment, overdiagnosis, and simply the earlier knowledge of cancer ( Korfage et al., 2006 ). All such harms are included when adjusting the life years gained for negative quality-of-life effects. Positive side-effects, such as a reduced need for expensive palliative treatments because fewer women are dying of breast cancer, can and should also be incorporated into such cost–effectiveness analyses.
To determine whether an intervention produces reasonable amounts of benefits and limited harms for the resources invested, the cost–effectiveness ratios are usually compared with cost–effectiveness thresholds. A frequently used cost–effectiveness threshold is £30 000 per QALY gained ( NICE, 2014 ). In the USA, interventions below the threshold of US$ 50 000 per QALY are generally considered cost-effective, interventions between US$ 50 000 per QALY and US$ 100 000 per QALY are considered moderately or borderline cost-effective, and those that exceed US$ 100 000 per QALY are generally not considered cost-effective ( Grosse, 2008 ). It has recently been recommended that a threshold of US$ 200 000 per QALY should be used for the USA ( Neumann et al., 2014 ). The relatively high threshold of US$ 200 000 per QALY relates to the fact that health-care costs in the USA are generally considerably higher than those in Europe. Looking more globally, the World Health Organization (WHO) has suggested a cost–effectiveness threshold of 3 times the national gross domestic product per capita ( WHO, 2014 ). Practically, for low-income regions the maximal values for being cost-effective are about US$ 5000 ( WHO, 2001 ). [A clear distinction has to be made for cost-efficacy estimates of trials, which often relate to the limited time frame of an RCT, in which not all benefits have accrued yet but where it is likely that cost and harms have already been prominent.]
Costs that should be considered in a cost–effectiveness analysis of breast cancer screening are costs associated with the organization of the programme (e.g. cost of invitations, screening costs), costs related to the diagnostic workup of both true-positives and false-positives, and additional treatment costs (e.g. due to more and earlier treatments). A few years after implementation, screening will lead to cost savings in treatment due to a decrease in the number of cases of advanced disease needing treatment ( de Koning et al., 1992 ). The cost savings depend mostly on the cost for advanced disease and the magnitude of the effectiveness of the screening programme. In a full cost–effectiveness analysis, direct medical costs, direct non-medical costs (travel and time), and indirect costs (e.g. due to sick leave) must be considered.
Ideally, all possible screening policies that are relevant are compared in a cost–effectiveness analysis. However, it is not feasible to compare all scenarios of interest in an RCT or an observational study. In addition, trials deliver (at best) costs per case detected. This is not an appropriate measure for cost–effectiveness because it lacks information about the effectiveness of screening (in terms of life years gained or breast cancer deaths averted). Furthermore, the aim of a cost–effectiveness analysis on breast cancer screening is to assess the effectiveness of a screening programme in an actual population rather than in a controlled setting. By the use of mathematical models, findings from RCTs and observational studies can be extrapolated to simulated populations ( Berry et al., 2005 ). Models are especially suitable for a cost–effectiveness analysis because the key elements of screening, including the screening strategy (starting age, stopping age, and screening interval), the target population (e.g. at average or increased risk), and the time point of the analysis, can be altered and/or compared. Furthermore, long-term lifetime effects can be predicted, and life years gained or QALYs can be calculated ( Groenewoud et al., 2007 ) (see Section 5.1.2f for further details).
Under the assumption that mammography screening programmes are effective in reducing breast cancer mortality in women at average risk of breast cancer, numerous cost–effectiveness analyses have shown that organized mammography screening can be cost-effective ( van Ineveld et al., 1993 ; Leivo et al., 1999 ; Stout et al., 2006 ; Groenewoud et al., 2007 ; Carles et al., 2011 ; Pataky et al., 2014 ).
Most population-based screening programmes screen women at biennial intervals ( Giordano et al., 2012 ). Annual screening strategies may improve the detection of rapidly growing tumours. However, despite the greater effectiveness, screening strategies that consist of annual screening are often found to be less efficient and less cost-effective, due to a disproportionate increase in costs or due to diminishing returns; about 80% of the effect of annual screening is retained when screening is performed every 2 years ( Mandelblatt et al., 2009 ; Stout et al., 2014 ). Schousboe et al. (2011) demonstrated that, in the United States setting, even if annual mammography is restricted to certain risk groups, based on age or breast density, the costs exceed US$ 100 000 per QALY gained. In contrast, Carles et al. (2011) reported several cost-effective annual screening strategies in Spain. However, ICERs increased markedly when comparing annual screening with biennial screening, as reported in other studies.
Organized mammography screening has been shown to be more cost-effective than opportunistic mammography screening ( Bulliard et al., 2009 ; de Gelder et al., 2009 ). In Switzerland, the costs per life year gained of opportunistic screening were twice those of organized screening ( de Gelder et al., 2009 ). This difference was caused predominantly by the higher costs of mammography for opportunistic screening and the more frequent use of additional imaging in combination with opportunistic screening.
Cost–effectiveness ratios obtained from studies of screening programmes in different countries are not easily comparable, due to differences in assumptions about effects and costs, time horizon, discount rate, and calculation methods ( Brown & Fintor, 1993 ; de Koning, 2000 ). Furthermore, epidemiological factors (background risk of breast cancer), the performance of the screening test, and the organization of the national screening programme and the health-care system all influence cost–effectiveness. The cost–effectiveness of a screening programme also depends on its characteristics, including attendance rate, screening interval, and age group targeted for screening.
A detailed cost–effectiveness analysis on breast cancer screening in India has been reported, in which the natural history of breast cancer was calibrated against available data on breast cancer incidence, stage distribution, and mortality in India ( Okonkwo et al., 2008 ). The model was used to estimate the costs of breast cancer screening in India, its effects on mortality, and its cost–effectiveness (i.e. costs of screening per life year gained or per life saved). Screening using CBE or mammography among different age groups and at various frequencies was analysed. Stage-dependent sensitivities of CBE in this study were based on data from the Canadian National Breast Screening Study (CNBSS) ( Rijnsburger et al., 2004 ). Alternative (lower) estimates of stage-dependent sensitivities of CBE were based on data from 752 000 CBEs delivered to low-income women in the USA in 1995–1998 through the National Breast and Cervical Cancer Early Detection Program of the United States Centers for Disease Control and Prevention ( Bobo et al., 2000 ).
Okonkwo et al. (2008) expressed costs in international dollars (Int.$), the currency used by WHO; an international dollar has the same purchasing power in a particular country as a United States dollar has in the USA. Under the assumption that such screening programmes are as effective as is seen in mammography trials, the estimated mortality reduction was the greatest for programmes targeting women between age 40 years and age 60 years. Using a 3% discount rate, a single CBE at age 50 years had an estimated cost–effectiveness ratio of Int.$ 793 per life year gained and resulted in a reduction in breast cancer mortality of 2%. The cost–effectiveness ratio increased to Int.$ 1135 per life year gained for every 5-yearly CBE (age 40–60 years) and to Int.$ 1341 for biennial CBE (age 40–60 years); the corresponding reductions in breast cancer mortality were 8.2% and 16.3%, respectively. CBE performed annually from age 40 years to age 60 years was predicted to be nearly as efficacious as biennial mammography screening for reducing breast cancer mortality, while incurring only half the net costs.
The main factors affecting cost–effectiveness were breast cancer incidence, stage distribution, and cost savings on palliative care averted ( Okonkwo et al., 2008 ). The estimated cost–effectiveness of CBE screening for breast cancer in India compares favourably with that of mammography in developed countries. [The study relied on an assumption about the efficacy of CBE in reducing breast cancer mortality in India, which has not been verified in randomized trials comparing CBE with no screening but was based on the CNBSS 2 trial, assuming that the effect of stage shift from mammography trials can be extrapolated.]
More recently, several studies have investigated the expected cost–effectiveness of different strategies in Costa Rica and Mexico ( Niëns et al., 2014 ), Ghana ( Zelle et al., 2012 ), and Peru ( Zelle et al., 2013 ). In Costa Rica, the current strategy of treating breast cancer at stages I to IV at a geographical coverage level of 80% seems to be the most cost-effective, with an ICER of US$ 4739 per disability-adjusted life year (DALY) averted. At a coverage level of 95%, biennial CBE screening could double life years gained and can still be considered very cost-effective (ICER, US$ 5964 per DALY averted). For Mexico, the results indicate that at a coverage level of 95%, a mass media awareness-raising programme could be the most cost-effective (ICER, US$ 5021 per DALY averted). If more resources are available in Mexico, biennial mammography screening for women aged 50–70 years (ICER, US$ 12 718 per DALY averted), adding trastuzumab (ICER, US$ 13 994 per DALY averted), or screening women aged 40–70 years biennially plus trastuzumab (ICER, US$ 17 115 per DALY averted) are less cost-effective options ( Niëns et al., 2014 ). Breast cancer in Ghana is characterized by low awareness, late-stage treatment, and poor survival. Biennial screening with CBE of women aged 40–69 years, in combination with treatment of all stages, seems the most cost-effective intervention (ICER, US$ 1299 per DALY averted). Mass media awareness-raising is the second-best option (ICER, US$ 1364 per DALY averted) ( Zelle et al., 2013 ). The current breast cancer programme in Peru (US$ 8426 per DALY averted) could be improved by implementing 3-yearly or biennial screening strategies. These strategies seem the most cost-effective in Peru, particularly when mobile mammography is applied (from US$ 4125 per DALY averted) or when CBE screening and mammography screening are combined (from US$ 4239 per DALY averted).
The impact of the various screening interventions on stage distribution was estimated on the basis of a model using proportional detection rates ( Duffy & Gabe, 2005 ). The authors applied a stage shift from developing countries to the Dutch screening programme and corrected this shift for locally relevant attendance rates and the epidemiology and demography. The age-specific sensitivity of tests and the sojourn times (CBE sojourn times are two thirds those of mammography) were based on the literature ( Duffy & Gabe, 2005 ; NETB, 2014 ). The effectiveness of the awareness-raising interventions is based on a study in Malaysia ( Devi et al., 2007 ), where a 2-fold reduction in advanced breast cancer was observed when a mass media campaign was applied. However, evidence on the effectiveness of awareness-raising, CBE, and mammography screening is absent in many countries. Also, these programmes require substantial organizational, budgetary, and human resources, and the accessibility of diagnostic, referral, treatment, and palliative care facilities for breast cancer should simultaneously be improved.
As already pointed out, the expected effects – both benefits and harms – and the cost of an intervention are context-specific. In public health, medicine, and any other field, inferences and extrapolations to other populations and individuals are needed. The average estimates for relative benefits, observed in IBM, nested case–control cohort, and case–control studies, in which biases have been minimized as much as possible, need to be extrapolated, as well as the estimates for overdiagnosis, false-positives, and radiation risk. To incorporate all of these and to estimate values as specifically as possible for different populations with different age structures, life expectancies, incidence, mortality, and treatment levels, statistical models are used.
The harm–benefit ratio has been calculated for different settings. The Independent United Kingdom Panel estimated that the United Kingdom screening programmes currently prevent 1300 deaths from breast cancer per year, equivalent to about 22 000 years of life being saved. Per 10 000 women invited to screening, it is estimated that 43 deaths from breast cancer are prevented and 129 cases of breast cancer represent overdiagnosis ( Marmot et al., 2013 ). The Euroscreen Working Group estimated that for every 10 000 women screened biennially from age 50 or 51 years until age 68 or 69 years, about 80 deaths from breast cancer are prevented, versus about 40 cases overdiagnosed ( Paci & EUROSCREEN Working Group, 2012 ). In the Netherlands, it has been estimated that each year 775 breast cancer deaths are prevented, versus 300 overdiagnosed cases (1 million invitations per year) ( NETB, 2014 ).
Women younger than 50 years may benefit less from mammography screening, due to a lower breast cancer incidence, a lower sensitivity of mammography due to denser breast tissue, a lower PPV, higher false-positive rates, and possibly more aggressive tumour growth ( Carney et al., 2003 ; Buist et al., 2004 ). Therefore, the cost–effectiveness ratio is less favourable for younger women than for older women. For instance, a recent analysis showed that for Canada the most cost-effective strategies were biennial screening from age 50 years to age 69 years (ICER, US$ 28 921 per QALY), followed by biennial screening from age 40 years to age 69 years (ICER, US$ 86 029 per QALY) ( Pataky et al., 2014 ).
In addition, the efficacy or effectiveness of screening, in terms of breast cancer mortality reduction, in women screened from age 40 years ( Alexander et al., 1999 ; Smith et al., 2004 ; Moss et al., 2006 ; Hellquist et al., 2011 ) is less precisely estimated, due to small numbers of breast cancer deaths, than that in women screened from age 50 years, and may therefore be underestimated or overestimated in cost–effectiveness analyses. It could even be more cost-effective to screen women aged 50–69 years more frequently than to include women younger than 50 years ( de Koning et al., 1991 ).
A study in which the Dutch MISCAN model was used to assess the cost–effectiveness of different policies for breast cancer screening in Catalonia, Spain (using Dutch data on costs) demonstrated that it is comparably cost-effective to extend screening from age 50 years to age 45 years and to extend screening from age 64 years to age 69 years ( Beemsterboer et al., 1998b ). The researchers emphasized that extending the upper age limit would result in a greater reduction in breast cancer mortality, whereas extending screening to younger women could lead to more life years gained. A more recently performed cost–effectiveness analysis, also focusing on screening in Catalonia, showed that biennial screening from age 45 years (to age 69 years or 74 years), annual screening from age 40 years (to age 69 years or 74 years), and annual screening from age 45 years (to age 69 years) (ranked in order of effectiveness) are all cost-effective strategies, with incremental costs per QALY gained of less than €30 000 ( Carles et al., 2011 ).
A study based on data from the USA demonstrated that biennial mammography screening from age 40 years to age 49 years is cost-effective only for women with BI-RADS 3 or 4 breast density, women with both a previous breast biopsy and a family history of breast cancer, and women with BI-RADS 3 or 4 breast density and either a previous breast biopsy or a family history of breast cancer, assuming a cost–effectiveness threshold of US$ 100 000 per QALY gained ( Schousboe et al., 2011 ). In contrast, another study, using five independent models of digital mammography screening in the USA, found that extending biennial screening from women aged 50–74 years to those aged 40–49 years would lead to incremental costs of US$ 55 100 per QALY gained, which was considered to be cost-effective ( Stout et al., 2014 ). Annual mammography, which may improve detection of rapidly growing tumours that may be more common among younger women, was considered not cost-effective in both studies. As mentioned previously, age considerations may be different for developing countries.
Breast cancer incidence and breast cancer detection rates are higher in women aged 70 years and older, which may increase the effect of screening. However, compared with younger women, older women are more subject to numerous illnesses and conditions that negatively affect life expectancy, thereby limiting the beneficial effect of screening on life expectancy and potentially increasing costs of screening. Furthermore, attendance rates may be lower among older women, which would also negatively affect the cost–effectiveness ratio.
Women older than 74 years were not included in any breast cancer screening trial (see Section 4.2). Model simulations demonstrated that screening women aged 50–75 years and screening women with high bone mineral density up to age 79 years are both cost-effective strategies ( Boer et al., 1995 ; Kerlikowske et al., 1999 ). Correspondingly, two systematic reviews showed that ceasing screening at age 75 years or 79 years instead of at age 65 years or 69 years is cost-effective, even for women who are not screened regularly before age 65 years ( Barratt et al., 2002 ; Mandelblatt et al., 2003 ).
In several countries, digital mammography has practically replaced film mammography ( NHS, 2005 ; NETB, 2014 ). The sensitivity of digital mammography may be higher than that of film mammography for women younger than 50 years and for women with dense breasts ( Pisano et al., 2008 ). However, the specificity of digital mammography may be slightly lower than that of film mammography ( Skaane, 2009 ; Kerlikowske et al., 2011 ). Referral rates are likely to increase with digital mammography, depending on the baseline situation of referrals, but this is especially pertinent in the implementation phase. Because of the differences in test characteristics and in costs of mammography, cost–effectiveness ratios are likely to differ as well. A modelling study that used data from the DMIST trial found that, compared with film mammography, digital mammography is not cost-effective (US$ 331 000 per QALY gained), except when limited to women aged 40–49 years ( Tosteson et al., 2008 ). However, digital mammography targeted to younger ages combined with film mammography from age 50 years is usually not a feasible strategy because film mammography has practically been replaced by digital mammography. Another study showed that digital mammography increases the number of false-positive findings by 220 per 1000 women compared with film mammography, leading to additional costs of US$ 350 000 per 1000 women, whereas the gain in benefits relative to film mammography is small ( Stout et al., 2014 ).
In most countries, organized mammography screening applies to all women in a targeted age group (usually 50–69 years or 50–74 years) with a relatively low (average) risk of breast cancer. Because breast cancer risk is associated with risk factors including age, reproductive history, a previous breast biopsy, and a family history of breast cancer (see Section 1.3), costs and benefits of screening may be affected by a woman’s individual risk of breast cancer. More personalized mammography screening, by selecting the starting and stopping ages and the screening interval based on a woman’s breast cancer risk profile, is therefore being considered in several research projects.
A cost–effectiveness study based on data from women in the USA showed that biennial mammography from age 40 years is cost-effective for women with high breast density (BI-RADS 3 or 4) and either a family history of breast cancer or a previous breast biopsy (< US$ 50 000 per QALY gained), and moderately cost-effective for women with high breast density only or both a previous breast biopsy and a family history of breast cancer (< US$ 100 000 per QALY gained) ( Schousboe et al., 2011 ). Annual mammography was estimated to cost more than US$ 100 000 per QALY gained for any group at an increased risk, and was therefore not considered cost-effective.
Another study based on population data from the USA, using five independent models, showed that annual digital mammography screening for women aged 40–74 years with high breast density (BI-RADS 3 or 4) resulted in 3-fold higher incremental costs per additional QALY gained relative to biennial screening for all women aged 40–74 years ( Stout et al., 2014 ). The incremental benefits of annually screening women aged 40–49 years with (extremely) dense breasts were small, predominantly accounting for the increase in ICERs.
Women with heterogeneously or extremely dense breasts and a negative screening mammogram may be considered for supplemental screening. The most readily available supplemental screening modality is ultrasonography, but little is known about its effectiveness when performed after negative screening mammography (see Section 5.5.1a ). Sprague et al. (2015) used three independent simulation models to assess the lifetime benefits, harms, and cost–effectiveness from the payer perspective of supplemental ultrasonography screening for women with dense breasts compared with screening with digital mammography alone. They found that supplemental ultrasonography screening for women with dense breasts undergoing routine digital mammography screening would substantially increase costs while producing relatively small benefits in breast cancer deaths averted and QALYs gained. The cost–effectiveness ratio was US$ 325 000 per QALY gained (range, US$ 112 000–766 000). Restricting supplemental ultrasonography screening to women with extremely dense breasts would cost US$ 246 000 per QALY gained (range, US$ 74 000–535 000) relative to biennial mammography alone for women aged 50–74 years.
A Dutch analysis of cost–effectiveness and quality of life conducted in 1991 included estimates on 15 phases induced and/or prevented by the screening programme ( de Koning et al., 1991 ). It appeared that 85% of the decrements in quality of life due to screening were due to the additional years in follow-up after diagnosis (of which about half were due to earlier detection and about half due to life years gained). False-positives comprised only a small component, as did the initial years of overdiagnosed cases. However, about 66% of the decrements were counterbalanced by gains; 70% of these gains imply reductions in palliative treatments for women with advanced disease. It was estimated that correcting the life years gained for quality of life would imply a 3% difference, that is, 3% fewer life years gained when adjusted for quality of life. The most unfavourable sensitivity analysis estimated a 19.7% decrease.
Vilaprinyo et al. (2014) estimated QALYs for the different breast cancer disease states. They used the health-related quality of life measures obtained from the EuroQol EQ-5D self-classifier in the study of Lidgren et al. (2007) , which provided health-related quality of life measures for the first year after primary breast cancer (EQ-5D = 0.696), the second and following years after primary breast cancer or recurrence (EQ-5D = 0.779), and the metastatic breast cancer state (EQ-5D = 0.685). For false-positive mammograms, the authors assumed an average annualized loss of quality of life of 0.013. To obtain the value of 0.013, they assumed that 50% of women with a false-positive result would experience anxiety sufficient to increase the mood subscale of the EuroQol instrument from 0 to 1, lasting a total of 2 months. According to the United States EQ-5D tariffs, such a change for an entire year represents a decrease in the QALY value of 0.156. In the sensitivity analysis, the authors assessed the impact of changing the disutility by false-positives to 0 and to 0.026.
This section reports evidence on the efficacy or effectiveness of imaging modalities other than screen-film mammography or standard digital mammography, where applied for population screening of asymptomatic women of about average (population) risk. Studies that included women at above average risk were considered, but not those in which study subjects were restricted to classifications of increased risk. Studies of cohorts of women defined by dense breast tissue on mammography (but not restricted to women at an increased risk) were also reviewed.
The following imaging technologies were reviewed: breast ultrasonography, digital breast tomosynthesis, MRI (other than screening of women at increased risk), electrical impedance technology for breast imaging, scintimammography, and positron emission mammography. No RCTs examining the efficacy of these imaging technologies for population breast screening were available to the Working Group.
For two imaging technologies (ultrasonography in dense breasts and digital breast tomosynthesis in population screening), there was evidence from non-randomized studies of incremental (additional) cancer detection when applied as adjunct screening to mammography. The evidence for the preventive effects, adverse effects, and cost–effectiveness of these two technologies is presented in Sections 5.5.1–5.5.3 , respectively. Other imaging technologies, for which there was very little or no data on efficacy or effectiveness, or for which population screening studies have not been conducted, are briefly outlined in Section 5.5.4 .
(a) breast ultrasonography.
Ultrasonography has had a role in diagnosis of breast disease for approximately 30 years and has been used for the workup of screen-detected abnormalities and for image-guided needle biopsy (see Section 2.2.1 for technical details). Because dense breast tissue is a risk factor for breast cancer ( McCormack & dos Santos Silva, 2006 ) and reduces the sensitivity of mammography, and hence is associated with a greater likelihood of an interval cancer in mammography screening ( Ciatto et al., 2004a ), evaluations of breast ultrasonography screening have often focused on populations defined by mammographic density ( Buchberger et al., 2000 ; Houssami et al., 2009 ; Corsetti et al., 2011 ; Houssami & Ciatto, 2011 ; Venturini et al., 2013 ).
No RCTs examining the efficacy of screening by ultrasonography or of adjunct ultrasonography in women with dense breast tissue on mammography (i.e. mammography alone vs mammography plus ultrasonography) were identified by the Working Group. A recent Cochrane systematic review ( Gartlehner et al., 2013 ) evaluated the literature to assess the effectiveness of ultrasonography screening as adjunct to mammography in women at average risk of breast cancer. None of the studies identified (no randomized, prospective, or controlled studies) reported sound evidence supporting ultrasonography as adjunct to mammography in population breast screening. An RCT on the efficacy of adjunct ultrasonography for breast cancer screening, called the Japan Strategic Anti-Cancer Randomized Trial, was noted ( Ishida et al., 2014 ). This trial aimed to recruit 100 000 women aged 40–49 years and has recently closed to recruitment; its results have not yet been reported.
Several studies of breast ultrasonography screening, all non-randomized and without a comparison or control group, have examined the incremental cancer detection of breast ultrasonography in women with dense breast tissue and negative mammography . Table 5.15 presents the studies that have reported data for both true-positive detection and false-positives (or additional recall) attributed to ultrasonography screening. Studies that recruited women with dense breast tissue conditional to also being classified as at an increased risk were not considered (e.g. Berg et al., 2008 ). However, studies that defined subjects on the basis of dense breast tissue but also included some women or subgroups with additional risk factors were included and reviewed.
Studies of adjunct ultrasonography in screening asymptomatic women with mammography-negative dense breast tissue.
The majority of the studies were retrospective, and all were designed to assess incremental cancer detection (as an indicator of potential effectiveness) within screened subjects; none of these studies were designed to assess screening benefit in terms of mortality reduction or using a surrogate for effectiveness of screening, such as a reduction in interval cancer rates. Incremental detection of breast cancer by ultrasonography was in the range of 0.19% to 0.52% of all screens. The highest estimate ( Kelly et al., 2010 ) included women at an increased risk, including some women with a history of breast cancer, and reported a modest cancer detection rate for mammography. Therefore, the incremental detection of breast cancer by ultrasonography was substantial but heterogeneous, representing approximately 14% to 48% of the detected cancers ( Corsetti et al., 2008 ; Venturini et al., 2013 ). [These data should be interpreted taking into account that several studies included, among women with dense breasts, subgroups of women at increased risk due to other risk factors (i.e. dense breasts plus other risk factors), and many studies included young women, and therefore the evidence may not be generalizable to population screening of women with dense breasts.] The two prospective studies reported the lowest incremental detection rates for ultrasonography, of 0.19% ( Brem et al., 2014 ) and 0.24% ( Venturini et al., 2013 ) of screens. Ultrasonography-only detected cancers were frequently early-stage cancers, generally at a comparable or earlier stage than cancers detected with mammography, although comparative data on cancer characteristics were not comprehensively reported.
Giuliano & Giuliano (2013) examined detection measures for automated breast ultrasonography screening in women with dense (density > 50%) breast tissue (test group) and used a different cohort of women with dense breasts from an earlier time frame as a control group for mammography screening. [This study is limited by the comparison of two cohorts with different underlying breast cancer prevalence (test group, 1.25%; control group, 0.60%).] For the test group ( n = 3418; median age, 57 years) screened with mammography and ultrasonography, the screening sensitivity was 97.7%, the specificity was 99.7%, the cancer detection rate was 12.3 per 1000 screens, and the mean tumour size of detected cancers was 14.3 mm. For the control group ( n = 4076; median age, 54 years) screened with digital mammography alone, the screening sensitivity was 76.0%, the specificity was 98.2%, the cancer detection rate was 4.6 per 1000 screens, and the mean tumour size of detected cancers was 21.3 mm. [This mean size is larger than expected for a screened population. The inferred 2.6-fold increase in the cancer detection rate, which represents one additional detection in approximately 0.70% of screens, was attributed to ultrasonography. This is well above estimates from all the other reviewed studies and is probably due to the comparison of cohorts with different underlying breast cancer risk. In addition, the relatively high specificity in the test group, based on the combined screening approach, is unusual and is inconsistent with all the other studies . Because of these limitations, this study was considered uninformative.]
One prospective screening study of ultrasonography in a multimodality setting (CBE, mammography, and ultrasonography) included 3028 Chinese women aged 25 years and older ( Huang et al., 2012 ), not restricted to women with dense breasts. The sensitivity was higher for mammography (84.8%) than for ultrasonography (72.7%); however, ultrasonography detected 3 cancers not detected with mammography (all were in women with dense breasts). Ultrasonography yielded an incremental cancer detection rate of [0.99 per 1000] screens of all screening participants. Mammography-detected cancers were more frequently smaller than 20 mm and node-negative than those detected with ultrasonography or CBE.
Two non-randomized studies of adjunct ultrasonography for screening dense breasts reported data on interval cancers ( Kelly et al., 2010 ; Weigert & Steenbergen, 2012 ). [Given that these studies did not have a comparison estimate and had a relatively short follow-up period (12 months), it is difficult to interpret the estimated interval cancer rates.] Corsetti et al. (2008 , 2011 ) reported indirect comparisons based on follow-up for first-year interval cancers in a cohort of self-referring women attending a breast service in Italy. The estimated first-year interval cancer rate was 1.1 per 1000 screens (from 7172 negative screens with follow-up) in women who underwent adjunct ultrasonography and had dense breasts, compared with 1.0 per 1000 screens (from 12 438 negative screens with follow-up) in women who received mammography only and did not have dense breasts.
Digital breast tomosynthesis is a derivative of digital mammography that produces quasi three-dimensional images, which reduces the effect of tissue superimposition and can therefore improve mammography interpretation (see Section 2.1.4 for details). A recent systematic review ( Houssami & Skaane, 2013 ) examined the available evidence on the accuracy of digital breast tomosynthesis. The studies identified were relatively small ( n = 14), comprised mostly test-set observer (reader) studies or clinical series that included symptomatic and screen-recalled cases, and were generally enriched with breast cancer cases. Taking into consideration the limitations of the studies, the evidence can be summed up as follows ( Houssami & Skaane, 2013 ): (i) two-view digital breast tomosynthesis has accuracy that is equal to or better than that of standard two-view mammography; (ii) one-view digital breast tomosynthesis does not have better accuracy than two-view mammography; (iii) the addition of digital breast tomosynthesis to digital mammography increases interpretive accuracy; (iv) improved accuracy from using digital breast tomosynthesis (relative to, or added to, digital mammography) was the result of increased cancer detection or reduced false-positive recalls, or both; and (v) subjective interpretation of cancer conspicuity consistently found that cancers were equally or more conspicuous on digital breast tomosynthesis relative to digital mammography.
A review of the literature did not identify any RCTs examining the efficacy of digital breast tomosynthesis in population breast screening; however, digital breast tomosynthesis was the only other imaging technology investigated in population-based screening programmes in women at average (population) risk ( Ciatto et al., 2013 ; Haas et al., 2013 ; Rose et al., 2013 ; Skaane et al., 2013a , b , 2014 ; Friedewald et al., 2014 ; Houssami et al., 2014a ; Table 5.16 ). All these studies investigated digital mammography with tomosynthesis (also referred to as integrated two-dimensional/three-dimensional [2D/3D] mammography), using various methodologies (different design and reading/recall protocols). None were designed with the aim of assessing screening benefit in terms of mortality reduction or using a surrogate for effectiveness of screening, such as a reduction in interval cancer rates. Also, none of the studies reported estimates of overdiagnosis. Two studies were prospective population-based trials embedded within organized screening programmes in Europe: the Screening with Tomosynthesis or Standard Mammography (STORM) trial in Italy ( Ciatto et al., 2013 ) and the Oslo trial in Norway ( Skaane et al., 2013a , b , 2014 ). Both studies used double reading according to European standards, but they used different recall protocols. Both studies performed digital mammography with tomosynthesis in all participants, and hence they reported paired data for screened women (within screening participant comparison).
Studies evaluating tomosynthesis for population breast cancer screening: three-dimensional mammography as adjunct to digital mammography.
The STORM trial ( Ciatto et al., 2013 ; Houssami et al., 2014a ) compared sequential screen-readings by the same readers for the same women: digital mammography alone and integrated 2D/3D mammography. The study reported a significant incremental cancer detection rate of 2.7 per 1000 screens for integrated 2D/3D mammography versus digital mammography ( P < 0.001). The Oslo trial ( Skaane et al., 2013a , b ) randomized readers to four screen-reading strategies that used digital mammography or integrated 2D/3D mammography, allowing assessment of reconstructed 2D mammography in one of the study arms ( Skaane et al., 2014 ). The study showed a significant incremental cancer detection rate of 1.9 per 1000 screens for integrated 2D/3D mammography versus digital mammography in a reader-adjusted analysis ( P = 0.001) ( Skaane et al., 2013a ) and of 2.3 per 1000 screens for double reading of integrated 2D/3D mammography versus digital mammography ( P < 0.001) ( Skaane et al., 2013b ). A further analysis ( Skaane et al., 2014 ) found that integrated 2D/3D mammography yielded a similar incremental cancer detection rate compared with digital mammography whether by dual acquisition of digital mammography with tomosynthesis (acquired 2D and 3D images) or by tomosynthesis acquisition with synthetic 2D mammography (3D acquisition only, and 2D images reconstructed from the 3D data).
A third prospective screening trial, also conducted within a population-based programme, was in progress in Malmö, Sweden, at the time of the Handbook Working Group Meeting, in November 2014. This trial differs from the other screening studies of this technology in that it compares screen-reading using digital mammography alone (two views) with screen-reading using tomosynthesis alone (one 3D mammography view); hence, it is the only population-based breast screening study reporting detection estimates for tomosynthesis alone. [Note added after the Meeting: The results of the trial have been published ( Lång et al., 2015 ). The incremental cancer detection rate was 2.6 per 1000 screens using tomosynthesis alone versus digital mammography ( P < 0.0001).]
Three retrospective studies have also examined digital mammography with tomosynthesis for population screening ( Haas et al., 2013 ; Rose et al., 2013 ; Friedewald et al., 2014 ); all three studies were conducted in the USA and hence used single reading as practised in the USA. Two studies ( Rose et al., 2013 ; Friedewald et al., 2014 ) used a before–after methodology, comparing detection measures before and after the introduction of integrated 2D/3D mammography, whereas one study ( Haas et al., 2013 ) compared services using digital mammography with services using integrated 2D/3D mammography within the same time frame. The largest retrospective study ( Friedewald et al., 2014 ) was a comparison of 281 187 versus 173 663 screens before and after the introduction of tomosynthesis as adjunct to digital mammography screening in 13 radiology services, and reported a significant incremental cancer detection rate of 1.2 per 1000 screens. Overall, the three studies showed a modest incremental detection rate with the use of adjunct tomosynthesis (range, 0.5–1.4 per 1000 screens) relative to the prospective trials; however, the direction of the estimated increased cancer detection is consistent across all studies.
Four out of five studies provided limited data on the characteristics of the cancers detected with integrated 2D/3D mammography compared with digital mammography. [Studies were generally not powered for such analyses.] Two studies indicated that the increased cancer detection achieved by digital mammography with tomosynthesis was mostly of invasive disease ( Rose et al., 2013 ; Friedewald et al., 2014 ), whereas two studies showed incremental detection of both invasive and in situ disease ( Ciatto et al., 2013 ; Skaane et al., 2013b ).
Data on interval cancer rates for this technology are limited to the follow-up report from the STORM trial; the estimated interval cancer rate based on only 12 months of follow-up is 0.82 per 1000 (95% CI, 0.30–1.79) ( Houssami et al., 2014a ).
Several studies reported on the use of integrated 2D/3D mammography screening in reducing false-positive recalls ( Table 5.16 ). The reduction in false-positive recalls is most marked in the retrospective studies reported from the USA (absolute decreases in false-positive results range from 1.6% to 3.6%), where the baseline false-positive recall rates for digital mammography alone are relatively high (range, 8.7–12.0%). The estimated reduction in false-positive recalls in the prospective studies, which were conducted in European population screening programmes and had relatively low recall rates, was modest (0.8% and 2%), and the latter was an estimate conditional to 3D mammography positivity. Furthermore, one of the studies ( Skaane et al., 2013b ) showed that for double reading, digital mammography with tomosynthesis reduced false-positive recalls compared with mammography alone, but increased overall recall (see Table 5.16 ). [It is likely that the potential for digital mammography with tomosynthesis to reduce false-positive recalls will depend on both the false-positive recall rates at digital mammography and the recall rules, which vary according to the screening programme.]
The adverse effects of breast ultrasonography screening have been examined in non-randomized retrospective and prospective studies in women with dense breast tissue ( Buchberger et al., 2000 ; Kaplan, 2001 ; Kolb et al., 2002 ; Corsetti et al., 2008 , 2011 ; Kelly et al., 2010 ; Hooley et al., 2012 ; Weigert & Steenbergen, 2012 ; Venturini et al., 2013 ; Brem et al., 2014 ). The main adverse effect is additional false-positive intervention. Ultrasonography caused additional testing (needle biopsy or imaging follow-up) in 1.2–6.3%, and also surgical biopsy (although some studies included non-surgical biopsy in this percentage) in 0.9–2.7% due to false-positives ( Table 5.15 ). The study of Kelly et al. (2010) , which included some women at an increased risk, reported an overall recall rate [not distinctly false-positive recall] of 7.2% for ultrasonography (vs 4.2% for mammography; P < 0.01), and the combined strategy had an overall recall rate of 9.6% in that study. Venturini et al. (2013) reported a false-positive biopsy rate for ultrasonography of 0.9% (vs 0.1% for mammography) in a cohort of young women (aged 40–49 years) with dense breast tissue and intermediate lifetime risk. Brem et al. (2014) reported an overall recall rate of 28.5% for adjunct ultrasonography with mammography (vs 15% for mammography alone; P < 0.001).
Given that there is substantial increased detection of breast cancer using adjunct ultrasonography in women with mammography-negative dense breasts, it seems possible that overdiagnosis could occur in this context. However, overdiagnosis has not been reported in any of the studies reviewed ( Buchberger et al., 2000 ; Kaplan, 2001 ; Kolb et al., 2002 ; Corsetti et al., 2008 , 2011 ; Kelly et al., 2010 ; Hooley et al., 2012 ; Weigert & Steenbergen, 2012 ; Venturini et al., 2013 ; Brem et al., 2014 ). [It would be difficult to attempt to estimate overdiagnosis based on the available data, due to (but not limited to) the lack of a control or comparison cohort and the heterogeneity of the screened populations, including variable underlying risk profiles.]
All studies reviewed reported a reduction in false-positive recalls using integrated 2D/3D mammography ( Table 5.16 ). Therefore, this does not seem to be an adverse effect of this technology. [The same may not apply for 3D screening alone.]
Given that there is increased detection of breast cancer using digital mammography with tomosynthesis, it seems possible that overdiagnosis could occur in this context. Several studies ( Rose et al., 2013 ; Skaane et al., 2013a ; Friedewald et al., 2014 ) have suggested that digital breast tomosynthesis mostly increases detection of invasive cancers. However, none of the studies have reported on overdiagnosis. [The currently available data do not allow inferences relating to overdiagnosis from the increased cancer detection attributed to tomosynthesis.]
The main potential adverse effect of digital mammography with tomosynthesis relates to the radiation dose to the breast if dual acquisition is used. Digital breast tomosynthesis is reported to deliver on average similar doses to digital mammography ( Feng & Sechopoulos, 2012 ; Houssami & Skaane, 2013 ). Thus, using dual acquisition by digital mammography with tomosynthesis approximately doubles the radiation dose. In the two population screening studies, the mean glandular dose per view was 1.58 mGy for digital mammography and 1.95 mGy for digital breast tomosynthesis in the Oslo study ( Skaane et al., 2013a ) and 1.22 mGy for digital mammography and 2.99 mGy (1.22 + 1.77 mGy) for integrated 2D/3D mammography in the STORM study ( Bernardi et al., 2014 ). Recent tomosynthesis technology allows reconstruction of the 2D images from the data obtained from the tomosynthesis acquisition (also referred to as synthetic 2D mammography), eliminating the need for dual acquisition. Reconstruction of the 2D images from the tomosynthesis acquisition decreases the radiation dose by 45% compared with the dual acquisition ( Skaane et al., 2014 ) and performs similarly to digital mammography with tomosynthesis from dual acquisition (see Section 5.5.1 and Table 5.16 ).
There were no studies of breast ultrasonography for population breast screening that reported on cost per life year gained or QALY saved. Cost analyses were reported by four of the studies that investigated ultrasonography in women with dense breasts. Studies conducted in the USA ( Hooley et al., 2012 ; Weigert & Steenbergen, 2012 ) reported relatively higher costs than those conducted in Europe ( Corsetti et al., 2008 ; Venturini et al., 2013 ). Hooley et al. (2012) estimated the cost of adjunct ultrasonography, factoring in the costs of ultrasonography and related biopsy and short-interval imaging follow-up (using the Medicare reimbursement rate), to be $US 60 267 per detected breast cancer. Weigert & Steenbergen (2012) , using the average reimbursement rate for ultrasonography and related biopsy, estimated the cost of adjunct ultrasonography screening to be $US 110 241 per detected breast cancer.
In the European setting, Corsetti et al. (2008) estimated the cost of adjunct ultrasonography, factoring in the costs of ultrasonography and related testing and any form of biopsy, to be in the range of €14 618–15 234 per detected breast cancer. Venturini et al. (2013) reported the cost of screening young women with dense breasts; mammography was estimated to cost €6377 per detected breast cancer, whereas adjunct ultrasonography in the same programme was estimated to cost €19 158 per detected breast cancer.
There were no studies available of the cost–effectiveness, or any cost analyses, of digital mammography with tomosynthesis in population breast screening. Digital breast tomosynthesis is more expensive than digital mammography and requires more imaging storage and display infrastructure, all of which increase the costs and the resources needed for screening implementation. Digital mammography with tomosynthesis also increases screen-reading time, resulting in an approximate doubling ( Houssami & Skaane, 2013 ); based on the Oslo trial ( Skaane et al., 2013a ), the mean interpretation time was 91 seconds for integrated 2D/3D mammography versus 45 seconds for digital mammography ( P < 0.001).
(a) magnetic resonance imaging.
Breast MRI has been shown to have superior screening sensitivity to mammography in women at an increased risk of developing breast cancer (see Section 5.6 ). Searches of the literature did not identify any studies of MRI for screening of women considered at average (population) risk. One recent study ( Kuhl et al., 2014 ) of an abbreviated (fast) MRI protocol screened 443 women “referred to MRI screening on clinical grounds”; 82% of the women were considered to be at mildly or moderately increased risk, because of either dense breast tissue or a mild or moderate family history of breast cancer. The 146 women with a personal history of breast cancer were having imaging of the contralateral breast. In this selected subject group, reportedly “pre-screened” with digital mammography and ultrasonography [data not reported for either], MRI yielded an incremental cancer detection rate of 18 per 1000 screens. False-positive rates varied by the applied MRI protocol and were in the range of 5.6–29%. [The findings from this “proof-of-concept” reader study are early and do not represent population screening.]
The literature search did not identify any RCTs or population-based studies of electrical impedance scanning for breast screening. Studies of electrical impedance technologies for imaging of the breast have used various devices and instrumentation, operated at various frequencies and interpreted using variable methods (e.g. visual, computer algorithms, or other methods) ( Malich et al., 2001 ; Martín et al., 2002 ; Wersebe et al., 2002 ; Diebold et al., 2005 ; Fuchsjaeger et al., 2005 ; Zheng et al., 2008 , 2011 ; Wang et al., 2010 ; Lederman et al., 2011 ).
All these studies were relatively small clinical series or diagnostic studies of women who had suspicious or equivocal (mammography or other image-detected) findings and included both symptomatic and asymptomatic women; these studies were based on women who were undergoing biopsy (surgical or core needle biopsy), and hence the studies were highly enriched with breast cancer cases (prevalence in the range of 5–60%).
One relatively large study assessed electrical impedance imaging for “risk-stratification” and screening of asymptomatic young women (aged 30–45 years) ( Stojadinovic et al., 2005 , 2008 ). [One limitation of this study is that the study participants included women with mammographic findings or clinical abnormalities who were scheduled to undergo biopsy.] The study reported an extremely low sensitivity for screening of 26.4%, and specificity of 94.7%.
The literature search did not identify any studies evaluating the efficacy or effectiveness of this technology for breast screening of women at average (population) risk.
Scintimammography has been used and evaluated in various clinical applications for breast imaging, predominantly in small and/or highly selected clinical series and diagnostic studies highly enriched with breast cancer cases (19–100%), including, but not limited to: diagnostic workup of suspicious or indeterminate mammography-detected (or other image-detected) findings; breast assessment in women scheduled for biopsy on the basis of clinical or mammographic abnormalities; staging of a known cancerous breast lesion; monitoring response to treatment; and detecting breast cancer recurrence ( Bekiş et al., 2004 ; Rhodes et al., 2005 ; Adedapo & Choudhury, 2007 ; Duarte et al., 2007 ; Gommans et al., 2007 ; O’Connor et al., 2007 ; Spanu et al., 2007 , 2008 , 2009 ; Hruska et al., 2008 ; Kim et al., 2009 ; Sharma et al., 2009 ; Xu et al., 2011 ; Lee et al., 2012 ; Spanu et al., 2012 ; Weigert et al., 2012 ; BlueCross BlueShield Association, 2013 ). A meta-analysis ( Xu et al., 2011 ) of 45 extremely heterogeneous diagnostic accuracy studies of scintimammography reported meta-estimates of 83% for sensitivity and 85% for specificity; in the subgroup of subjects without a palpable mass, meta-estimates were 59% for sensitivity and 89% for specificity.
Three studies reported screening of defined asymptomatic populations, which included women at an increased risk. Brem et al. (2005) screened with scintimammography 94 women at an increased risk who had normal mammograms and CBE. They detected 2 additional invasive (9 mm) cancers (+2%); however, this was at the trade-off of 14 additional false-positives (+15%). Rhodes et al. (2011) screened 936 women (aged 25–89 years) with dense breasts and at an increased risk (personal history of breast cancer or lobular carcinoma in situ [LCIS] or atypical proliferations, or BRCA mutations) using dedicated dual-head gamma imaging (with the radiotracer 99m Tc-sestamibi). The detection yield was 3.2 per 1000 screens for mammography and 9.6 per 1000 screens for scintimammography (incremental cancer detection rate, 7.5 per 1000 screens). Most of the cancers detected on scintimammography only were node-negative invasive cancers (median size, 11 mm). [The sensitivity of mammography was extremely low (27%).] False-positive recall rates (9% for mammography, 8% for scintimammography) and specificity (91% for mammography, 93% for scintimammography) were similar for the two tests. Finally, Hruska et al. (2012) reported a study of molecular breast imaging with 99m Tc-sestamibi in 306 asymptomatic women (aged 37–88 years), including some women at an increased risk, such as those with a personal history of breast cancer, who were undergoing myocardial perfusion imaging. Scintimammography had an incremental cancer detection yield of 13 per 1000 screens (4 cancers) relative to mammography in the previous 12 months, and caused additional false-positives in approximately [6%] of subjects.
The radiation dose to the whole body from this technology (see Section 2.2.4 for details) is reported to be 15–30 times the radiation dose from digital mammography ( BlueCross BlueShield Association, 2013 ).
Literature searches did not identify any population breast screening studies of positron emission mammography. This technology has been evaluated in very specific and limited clinical applications of breast imaging, predominantly for staging of a lesion; for preoperative assessment of disease extent (generally in comparison with MRI); for “screening” of the contralateral breast in preoperative staging; for response monitoring, in very small series of women with a biopsy of suspicious findings; or in phantom studies ( Raylman et al., 2000 ; Levine et al., 2003 ; Tafra et al., 2005 ; Berg et al., 2011 , 2012a ; Schilling et al., 2011 ; Schilling, 2012 ; Shkumat et al., 2011 ; Eo et al., 2012 ; Kalles et al., 2013 ). Positron emission mammography involves much higher doses of radiation (whole-body radiation) and a much longer acquisition time (for two views of both breasts) than mammography (see Section 2.2.3).
Few studies have measured psychosocial harm from imaging techniques other than mammography. One study found that MRI screening was more distressing than X-ray mammography both shortly after and 6 weeks after the screening procedure ( Hutton et al., 2011 ), whereas another study found no difference between MRI and mammography screening in psychological outcomes ( Brédart et al., 2012 ). As with other screening processes, psychological harm may depend on the conduct of the technology, such as the number of false-positive and false-negative screens and the waiting time from examination to result (see also Sections 3.1.4 and 5.3.5 ).
In some women, the risk of developing breast cancer during their lifetime is increased compared with that of women in the general population, and usually with an earlier expected age of onset. This increased risk may be attributed to the presence of a genetic or familial predisposition to breast cancer, to a personal history of invasive breast cancer or DCIS, or to the presence of lobular neoplasia or atypical proliferations. It should be noted that a familial predisposition, if not assessed by a specialized genetic centre, should not be used as an indication for screening outside the scope of the population breast cancer screening programme.
In general, it is preferable that women at an increased risk be screened outside the scope of a population breast cancer screening programme, for two reasons. First, regular population screening programmes with mammography might be insufficient, due to the earlier age of onset of breast cancer in these women and due to the reduced sensitivity of mammography in these women. In addition, women with a BRCA1/2 mutation are more susceptible to radiation risk. Second, these women often require additional care, assessment, counselling, and information relevant to primary prevention and risk-reduction strategies (as might be provided, for example, through specialized genetics teams/units) that are generally well outside of the health-care brief of mammography screening programmes.
Evidence on the outcomes of screening for breast cancer in the several subgroups of women at an increased risk is summarized and discussed here.
This section reports evidence on the effectiveness of screening with MRI alone, adjunct MRI, adjunct ultrasonography, or adjunct CBE as compared with mammography alone in women with a high familial risk, with or without a BRCA1 or BRCA2 mutation. Table 5.17 presents individual prospective studies, and Table 5.18 summarizes pooled and meta-analyses, and systematic reviews. The included studies are those that were performed prospectively, in which MRI and mammography were performed in the same screening round, and in which the review of the diagnostic test was performed blinded for the outcome of the other test. Studies that were performed retrospectively or unblinded, or in which MRI, ultrasonography, or mammography were not performed in parallel were excluded.
Prospective studies in women with a BRCA1/2 mutation or a familial breast cancer risk screened with magnetic resonance imaging, mammography, ultrasonography, or clinical breast examination.
Systematic reviews, pooled analysis, and meta-analyses of women at an increased risk of breast cancer screened with adjunct magnetic resonance imaging compared with mammography alone, with or without ultrasonography.
In addition, three reports reviewing the evidence of the effectiveness of adjunct MRI in the screening of women at an increased risk of breast cancer were identified ( Table 5.18 ). One is a systematic review of the literature ( Lord et al., 2007 ), one is a systematic review and meta-analysis at the level of published studies ( Warner et al., 2008 ), and one is a pooled analysis of individual patient data ( Phi et al., 2014 ).
(i) sensitivity and specificity in women with a brca1/2 mutation.
Several studies focused on the added value of MRI compared with mammography and/or ultrasonography in the screening of women with a BRCA1 or BRCA2 mutation ( Table 5.17 and Table 5.18 ). In the meta-analysis ( Warner et al., 2008 ) and the pooled analysis ( Phi et al., 2014 ), the estimates of the sensitivity of mammography were comparable, at about 40%, and increased with mammography combined with MRI similarly in both studies, to 94% (95% CI, 90–97%) in Warner et al. (2008) and 93.4% (95% CI, 80.2–98.0%) in Phi et al. (2014) . The specificity of adjunct MRI was also similar in the two analyses, to 77.2% (95% CI, 74.7–79.7%) in Warner et al. (2008) and 80.3% (95% CI, 72.5–86.2%) in Phi et al. (2014) . Thus, adding MRI to mammography in the screening of women with a BRCA1/2 mutation leads to a statistically significant increase in sensitivity of the screening strategy, accompanied by a decrease in specificity that was also statistically significant (see Table 5.18 ).
In the pooled analysis using individual data in women with BRCA1/2 mutations, for the screening of women aged 50 years and older, the highest sensitivity was reported for adjunct MRI (94.1%; 95% CI, 77.7–98.7%) compared with mammography alone (38.1%; 95% CI, 22.4–56.7%) and compared with MRI alone (84.4%; 95% CI, 61.8–94.8%) ( Phi et al., 2014 ); the specificity was lowest for adjunct MRI.
Only two informative studies assessed the sensitivity and specificity of mammography and MRI separately for women with a familial risk without a known BRCA1 or BRCA2 mutation ( Kuhl et al., 2005 ; Rijnsburger et al., 2010 ). Two other studies were considered uninformative due to the small number of breast cancers in that category ( Lehman et al., 2007 ; Trop et al., 2010 ; see Table 5.17 ). For mammography, the reported estimates for the sensitivity were 25–46% and for the specificity were 95–97%. For MRI, the reported estimates for the sensitivity were 73–100% and for the specificity were 89–98%. [All estimates reported by the earlier study ( Kuhl et al., 2005 ) are outside the confidence intervals of the two published meta-analyses ( Warner et al., 2008 ; Phi et al., 2014 ). Given the lower expected incidence of breast cancer among women without a BRCA1 or BRCA2 mutation, the PPV of screening with MRI will be much lower than that among women with a BRCA1 or BRCA2 mutation.]
There are no randomized trials assessing the efficacy of adjunct MRI in terms of mortality reduction in women at an increased risk with or without a BRCA gene mutation ( Nelson et al., 2013 ). Several prospective observational studies with long-term follow-up reported on stage distribution and mortality reduction by annual MRI plus mammography screening compared with women without intensified screening.
Three studies analysed the stage distribution of cancers detected in follow-up rounds of intensified screening programmes ( Schmutzler et al., 2006 ; Rijnsburger et al., 2010 ; Passaperuma et al., 2012 ). In two of the studies ( Schmutzler et al., 2006 ; Rijnsburger et al., 2010 ), an increase of N0 stages was reported (N0 stages of 67% vs 52% and 83% vs 56%, respectively). In the third study ( Passaperuma et al., 2012 ), a significant reduction of late stages from 6.6% to 1.9% with intensified screening was observed.
Prospective studies assessing the effectiveness of adjunct MRI in terms of mortality reduction are summarized in Table 5.19 . In a four-country study (England, the Netherlands, Norway, and Scotland), the 5-year survival was assessed for 249 women (205 non- BRCA1/2 mutation carriers with a family history of breast cancer, 36 BRCA1 mutation carriers, and 8 BRCA2 mutation carriers) prospectively diagnosed with breast cancer during screening ( Møller et al., 2002 ). All women were under breast cancer surveillance at a dedicated clinic, including annual mammography and CBE, and were diagnosed with breast cancer in this setting. The 5-year survival was 63% for women with a BRCA1 mutation compared with 91% in the women with a family history of breast cancer and without a known BRCA1/2 mutation.
Prospective studies of 5-year and 10-year survival of women with a BRCA1/2 mutation screened with mammography and/or magnetic resonance imaging.
In 2001, as part of a national initiative, women in Norway with a BRCA1 mutation were offered annual breast screening with MRI in addition to mammography. The observed 5-year breast cancer-specific survival for breast cancer patients with a BRCA1 mutation was 75% (95% CI, 56–86%) and the 10-year survival was 69% (95% CI, 48–83%) ( Møller et al., 2013 ). These results are in contrast with those of two other recent studies ( Rijnsburger et al., 2010 ; Passaperuma et al., 2012 ). In one study ( Rijnsburger et al., 2010 ), the estimated overall survival at 6 years in BRCA1/2 mutation carriers was 92.7% (95% CI, 79.0–97.6%). In the other study ( Passaperuma et al., 2012 ), out of 28 previously unaffected women with a BRCA1 mutation diagnosed with invasive breast cancer, only 1 died after relapse. [The Working Group noted that the study of Møller et al. (2013) included only women with a BRCA1 mutation, whereas the other two studies also included women with BRCA2 mutations, which could explain the difference in outcome.]
In a recent publication ( Evans et al., 2014 ), a survival analysis was conducted between BRCA1/2 mutation carriers screened with MRI plus mammography and unscreened BRCA1/2 mutation carriers ( Table 5.19 ). There were no differences in 10-year survival between the groups screened with MRI plus mammography and with mammography only, but survival was significantly higher in the group screened with MRI plus mammography (95.3%) compared with the unscreened cohort (73.7%; P = 0.002). After adjustment for age at diagnosis, this difference was still statistically significant (HR, 0.13; 95% CI, 0.032–0.53). [In this study, there were no deaths among the 21 BRCA2 carriers who received adjunct MRI, indicating that there might be differences in growth time between BRCA1 and BRCA2 tumours.]
The low specificity linked to screening with mammography plus MRI implies that after several screening rounds a significant percentage of screenees will have experienced either a recall or an image-guided (often MRI-guided) biopsy or will have undergone short-term follow-up ( Hoogerbrugge et al., 2008 ). In one systematic review on the adverse effects of adjunct MRI in the screening of women at an increased risk of breast cancer ( Lord et al., 2007 ), there was a 3–5-fold higher risk of patient recall for investigation of false-positive results compared with that of mammography alone.
Overall, the sensitivity of ultrasonography for the screening of women at an increased risk of breast cancer is comparable to or lower than that of mammography, and it is always lower than that of MRI ( Warner et al., 2004 ; Kuhl et al., 2005 , 2010 ; Cortesi et al., 2006 ; Lehman et al., 2007 ; Riedl et al., 2007 ; Weinstein et al., 2009 ; Trop et al., 2010 ; Sardanelli et al., 2011 ; Berg et al., 2012b ; Table 5.17 ).
As part of the screening programme offered to women at an increased risk of breast cancer with and without a BRCA1 or BRCA2 mutation, CBE is offered in some settings in addition to mammography and/or MRI. The evidence on the topic was recently reviewed ( Roeke et al., 2014 ), including seven studies ( Tilanus-Linthorst et al., 2000 ; Warner et al., 2001 , 2004 ; Kuhl et al., 2010 ; Rijnsburger et al., 2010 ; Trop et al., 2010 ; Sardanelli et al., 2011 ). The percentage of breast tumours detected by CBE varies from 0 out of 120 (0%) ( Warner et al., 2001 , 2004 ; Kuhl et al., 2010 ; Trop et al., 2010 ; Sardanelli et al., 2011 ) to 1 out of 260 (0.04%) ( Tilanus-Linthorst et al., 2000 ) and 3 out of 97 (3.1%) ( Rijnsburger et al., 2010 ) screen-detected cancers. [These latter two studies reported lower screen detection by mammography and/or MRI compared with studies in which no additional cases were detected by CBE. Furthermore, it is not clear whether CBE was performed blinded for the other tests, or whether these cases were detected during the screening or between the screening rounds, as most studies had annual screening with MRI plus mammography (with or without ultrasonography) and biannual screening with CBE.]
Women with a personal history of invasive breast cancer or DCIS are at an increased risk of developing breast cancer. This section reviews the evidence on the performance of screening with mammography and on whether adjunct ultrasonography or MRI improves screening performance in these women ( Table 5.20 ).
Studies of the effects of screening in women with at least one risk factor for breast cancer.
Women with a personal history of breast cancer are at an increased risk of ipsilateral or contralateral breast recurrence, or of a second primary breast cancer. Several studies have shown that a follow-up surveillance programme, including annual mammography, may be considered beneficial to these patients ( Ciatto et al., 2004b ; Lash et al., 2007 ; Lu et al., 2009 ). Only studies that included a comparison group were considered by the Working Group.
One large multicentre cohort study affiliated with the Breast Cancer Surveillance Consortium assessed the accuracy and outcomes of mammography screening in women with a personal history of breast cancer compared with those without such a history ( Houssami et al., 2011 ; Table 5.20 ). Mammography data of women with a personal history of early-stage breast cancer (58 870 mammograms in 19 078 women) were matched on age, breast density, and year of screening to women without a personal history of breast cancer (58 870 mammograms in 55 315 women). Mammography screening in women with a personal history of breast cancer had lower sensitivity and specificity and a higher interval cancer rate, but a similar proportion of detected early-stage disease, compared with that in women without such a history ( Houssami et al., 2011 ).
In a large study on the detection of breast cancer with the addition of annual screening with ultrasonography or a single screening with MRI to mammography in women at an increased risk, about 50% of the women had a personal history of breast cancer, and at baseline, about 55% of the women had a visually estimated breast density at scan of more than 60% ( Berg et al., 2012b ; Table 5.20 ). In this study, 111 cancers were detected: 33 with mammography only, 32 with ultrasonography only, and 26 by the combination of mammography and ultrasonography. In a substudy, after three rounds of mammography and ultrasonography, 9 additional cancers were detected with MRI. Overall, adding ultrasonography to mammography gave a statistically significant increase in sensitivity of the screening (first round, 55.6% vs 94.4%; subsequent rounds, 52% vs 76%) as well as a statistically significant increase in the recall rate (first round, 11.5% vs 26.6%; subsequent rounds, 9.4% vs 16.8%) ( Berg et al., 2012b ). When women with a personal history of breast cancer were compared with those without such a history, there were no statistically significant differences in yield between the two groups. However, the increase in the recall rate due to adjunct ultrasonography was statistically significantly smaller in the group of women with a personal history of breast cancer compared with those without such a history.
In a substudy in which MRI was added to the combination of mammography and ultrasonography, the sensitivity increased from 43.8% to 68.8%, whereas the recall rate increased from 16.3% to 36.3% ( Berg et al., 2012b ; Table 5.20 ). [The low sensitivity of the combined mammography and ultrasonography screening compared with the whole study might indicate an overselection of women with dense breast tissue in this substudy. The change in the recall rate due to supplementary MRI was statistically significantly higher in the group of women with a personal history of breast cancer compared with those without such a history. In this study, at baseline, about 55% of the women had a visually estimated breast density at scan of more than 60%.]
Women with lobular neoplasia or atypical proliferations are estimated to be at an increased risk of developing breast cancer ( Collins et al., 2007 ; Tice et al., 2013 ). One large study affiliated with the Breast Cancer Surveillance Consortium assessed the accuracy and outcomes of screening women with LCIS, atypical lobular hyperplasia, atypical ductal hyperplasia, or atypical hyperplasia compared with those without such lesions ( Houssami et al., 2014b ; Table 5.20 ). The cancer rates in the cohorts of women with LCIS or with atypical lobular hyperplasia were 2–3 times that in the reference cohort, and the cancer rate in the cohort of women with atypical ductal hyperplasia was 3–4 times that in the reference cohort. There were no statistically significant differences in sensitivity between the four cohorts. However, mammography screening of women with LCIS, atypical lobular hyperplasia, atypical ductal hyperplasia, or atypical hyperplasia resulted in lower specificities and higher interval cancers rates compared with their referent population. [The higher interval cancer rates partly reflect the higher underlying breast cancer risk.]
A few studies have examined the sensitivity of MRI in screening women with LCIS ( Friedlander et al., 2011 ; Sung et al., 2011 ; King et al., 2013 ) and those with LCIS or atypical hyperplasia ( Port et al., 2007 ). In the two studies that did not have a comparison group, high sensitivities were reported for MRI screening in women with LCIS ( Friedlander et al., 2011 ; Sung et al., 2011 ). [The Working Group noted that in the study of Sung et al. (2011) , only 80% of the screens were routine screens; the remaining 20% had non-specified indications, and the indications for the routine screens were not specified. Similarly, the study of Friedlander et al. (2011) reported only results from routine breast MRI screens, but the indications for the routine screens were not specified. The estimated sensitivities are thus likely to be biased in both studies.]
In the other two studies ( Port et al., 2007 ; King et al., 2013 ), women with high-risk lesions (LCIS and/or atypical hyperplasia) screened annually with mammography plus MRI were compared with women with high-risk lesions screened with annual mammography only. [In both studies, women with high-risk lesions selected to undergo adjunct MRI screening were younger and had stronger family histories of breast cancer compared with those screened by mammography only.] In both studies, adjunct MRI screening generated more follow-up biopsies compared with mammography alone.
5.7.1. preventive effects of clinical breast examination.
Randomized trials of CBE versus no screening have shown a significant shift from late-stage (T3/T4) to early-stage (T1/T2) breast cancers in the intervention arm ( Pisani et al., 2006 ; Mittra et al., 2010 ; Sankaranarayanan et al., 2011 ; see Section 4.3). Compliance with screening is one of the factors that determine effectiveness. In all three trials of CBE, the compliance with screening was high (> 85%), indicating acceptance of the procedure and ease of administering CBE. Access to care after recall and diagnosis is of paramount importance in the success of any screening trial, as is evident in the two randomized trials in India of CBE versus no screening ( Mittra et al., 2010 ; Sankaranarayanan et al., 2011 ). This was the major reason that the study in the Philippines was discontinued ( Pisani et al., 2006 ). The active intervention was stopped after the first screening round due to poor compliance (35% of screen-positive women) of participants with clinical follow-up for confirmation of diagnosis and treatment.
In the Mumbai study, the recall rate after CBE was 0.71%. Out of 153 130 screens by CBE, 1539 women were recalled for diagnostic investigations and 81 were confirmed to have invasive cancers ( Mittra et al., 2010 ).
Some harm of CBE may be attributed to pain or discomfort. Baines et al. (1990) carried out a survey of women who participated in the CNBSS to document women’s attitudes to screening by CBE and mammography. Of those who underwent CBE, 8.4% reported moderate discomfort and 2.1% extreme discomfort, whereas of those who underwent mammography, 36.2% reported moderate discomfort and 8.7% extreme discomfort.
Determining the cost–effectiveness of CBE alone is difficult because no trial has reported independent efficacy of CBE versus no screening. There have been many reports of cost–effectiveness analyses ( Okonkwo et al., 2008 ; Ahern & Shen, 2009 ) on screening with reference to CBE. [The Working Group noted that most reports made assumptions about mortality reductions to simulate or estimate cost–effectiveness that were not realistic. It may be appropriate to look at cost analysis instead.] The cost of delivering breast cancer screening by CBE is less than one third that of mammography ( Sarvazyan et al., 2008 ).
5.8.1. preventive effects of teaching breast self-examination.
Randomized trials and multiple observational studies have generally shown little or no reduction in mortality from breast cancer in women who practised BSE (see Section 4.4). If BSE is to have an effect on breast cancer mortality, it will have to be practised competently, and more frequently than in the Shanghai trial (see Section 4.4). Table 5.21 shows results of 11 surveys on BSE practice, based on self-reports, conducted primarily in countries with limited resources. Proficiency of BSE practice was not assessed in any of the studies. [It is unlikely that the proportion of women who reported practising BSE in any of the studies was sufficiently high to result in a meaningful reduction in breast cancer mortality rates in the populations surveyed.]
Percentage of women who reported practising breast self-examination in surveys conducted in selected countries.
Results of two studies of BSE practice before and after BSE instruction have been reported. Approximately 1000 women aged 30–50 years in Madhya Pradesh, India, attended BSE instruction sessions in which a film was shown, reinforced by a lecture with flip charts showing proper technique, and including a question-and-answer period ( Gupta et al., 2009 ). None of the women were practising BSE before the instruction. Two months after the instruction, 53% reported practising BSE regularly. [It is uncertain what regular practice means in just 2 months of alleged practice.] In Lower Saxony, Germany, women invited to instruction sessions received a lecture on BSE techniques followed by individual BSE training by a gynaecologist ( Funke et al., 2008 ). The self-reported prevalence of monthly BSE practice was 21% before the instruction and 62% 1 year after the instruction. Proficiency of BSE practice was not assessed in either of these studies. [It is therefore unclear whether a sufficient number of women in either study practised BSE with sufficient competence and frequency to result in a reduction in mortality from breast cancer.]
In three studies, BSE practice after BSE instruction was compared with BSE practice in a control group that did not receive instruction. In a study in rural women in the Republic of Korea ( Lee et al., 2003 ), women were given BSE instruction after appraisal of their individual risk on the basis of a questionnaire. Three months after the instruction, 30.5% of the women reported practising BSE regularly, compared with 10.2% in a control group. In a study of Latinas in the USA ( Jandorf et al., 2008 ), women were randomized to a group receiving information on BSE and CBE or to a control group. Telephone interviews 2 months after the instruction revealed that 45% of the women in the instruction group practised BSE compared with 27% in the control group. [Proficiency was not assessed in either of these studies.] In a BSE instruction programme in Ribe County, Denmark, up to 20 women at a time attended an intensive BSE training session lasting up to 2 hours that included videos as well as individual instruction on breast models and on the women’s own breasts ( Sørensen et al., 2005 ). An unreported number of years later (< 5 years), a questionnaire was mailed to the women who had participated and to a sample of women in the county who had not participated; 485 (77%) and 313 (53%) responded, respectively. Women were asked about frequency of BSE practice and whether they practised the various components of the BSE technique that was taught (positioning, use of mirror, and palpation pattern). On the basis of their answers, women were classified as performing BSE correctly, nearly correctly, or partly correctly. A higher percentage of women in the intervention group than in the control group practised BSE monthly (30.7% vs 21.1%) and practised it correctly or nearly correctly (27.6% vs 10.2%).
[The level of BSE practice in women taught BSE in all five of the evaluations of BSE instruction summarized in this section was lower than that in the trial in Shanghai, which showed no reduction in breast cancer mortality from BSE instruction. It is therefore reasonable to conclude that the level of BSE activity that was probably achieved in these studies was insufficient to have a meaningful impact on breast cancer mortality rates in the populations in which they were conducted. All of these studies except one were conducted in developed countries in which women, like the women in the Shanghai trial, had reasonable access to care, and in which women would be expected to seek medical attention for breast symptoms suggestive of breast cancer early in the course of the disease. The study in India may be an exception. In that country, many women with breast cancer typically present with advanced disease. It is unknown whether breast cancer mortality would be reduced if women in that country could be motivated to practise BSE on a regular basis, as was reported in the study by Gupta et al. (2009 ), and to do so competently.]
In both randomized trials of BSE, more women in the instruction group than in the control group found breast lumps that required further evaluation and that were subsequently confirmed as not being breast cancer (Section 4.4). In the trial in St Petersburg ( Semiglazov et al., 2003 ), nearly twice as many women were referred for further evaluation in the instruction group than in the control group; in the Shanghai trial, 80% more women in the intervention group than in the control group were found to have a histologically confirmed benign lesion ( Thomas et al., 2002 ). Such false-positives on screening can produce considerable anxiety, and the further evaluation of suspicious findings is not a trivial expense. Given that there is no proven benefit of BSE in reducing mortality from breast cancer, the risk–benefit ratio is very high.
Given that there is no good evidence that BSE, as it has been reported to be practised in studies to date, contributes to a reduction in mortality from breast cancer, there can be no estimate of the cost per life year gained by practising BSE. Based on data from the study in Ribe County, Denmark, Sørensen & Hertz (2003) estimated the cost per avoided cancer with spread to lymph nodes to be €15 410 and the cost of avoiding a cancerous tumour larger than 20 mm to be €16 318. [In their model, they assumed that there was considerable shift to a lower stage as a result of BSE practice, but as discussed in Section 4.4, the evidence for this is questionable and inconsistent, and the results of their estimates are highly dependent on the assumptions that they made as to the magnitude of the stage shift. They used only the cost of the BSE programme in their model. Their estimates did not take into account the costs of diagnostic confirmation or of changes in treatment if there is a stage shift at the time of diagnosis by BSE practice. If there truly is a stage shift, then this could result in less aggressive and less costly treatment, which would be a benefit even in the absence of a reduction in mortality. However, given the uncertainties as to any beneficial effects of BSE, no meaningful cost–effectiveness estimates are possible.]
Your browsing activity is empty.
Activity recording is turned off.
Turn recording back on
Connect with NLM
National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894
Web Policies FOIA HHS Vulnerability Disclosure
Help Accessibility Careers
See Advances in Breast Cancer Research for an overview of recent findings and progress, plus ongoing projects supported by NCI.
Drs. Ruth Pfeiffer and Peter Kraft of NCI’s Division of Cancer Epidemiology and Genetics discuss how breast cancer risk assessment tools are created and how people can use them to understand and manage their risk.
Some people with no evidence of cancer in nearby lymph nodes after presurgical chemotherapy can skip radiation to that area without increasing the risk of the cancer returning, a clinical trial found. But some experts caution that more details are needed.
For women in their 70s and older, the risk of overdiagnosis with routine screening mammography is substantial, a new study suggests. The findings highlight the need for conversations between older women and their health care providers about the potential benefits and harms of continuing screening mammography.
Many young women who are diagnosed with early-stage breast cancer want to become pregnant in the future. New research suggests that these women may be able to pause their hormone therapy for up to 2 years as they try to get pregnant without raising the risk of a recurrence in the short term.
For younger women with advanced breast cancer, the combination of ribociclib (Kisqali) and hormone therapy was much better at shrinking metastatic tumors than standard chemotherapy treatments, results from an NCI-funded clinical trial show.
In a large clinical trial, a condensed course of radiation therapy was as effective and safe as a longer standard course for those with higher-risk early-stage breast cancer who had a lumpectomy. This shorter radiation course makes treatment less of a burden for patients.
Adding the immunotherapy drug pembrolizumab (Keytruda) to chemotherapy can help some patients with advanced triple-negative breast cancer live longer. In the KEYNOTE-355 trial, overall survival improved among patients whose tumors had high levels of the PD-L1 protein.
People with metastatic breast cancer whose tumors had low levels of HER2 protein lived longer after treatment with trastuzumab deruxtecan (Enhertu) than those treated with standard chemotherapy, results of the DESTINY-Breast04 clinical trial show.
NCI researchers have shown that an experimental form of immunotherapy that uses an individual’s own tumor-fighting immune cells could potentially be used to treat people with metastatic breast cancer who have exhausted all other treatment options.
Most breast cancer risk tools were developed with data mainly from White women and don’t work as well for Black women. A new tool that estimates risk for Black women may help identify those who might benefit from earlier screening, enabling earlier diagnosis and treatment.
In people with metastatic HER2-positive breast cancer, the targeted drug trastuzumab deruxtecan (Enhertu) markedly lengthened progression-free survival compared with trastuzumab emtansine (Kadcycla), new study results show.
In a large clinical trial, women with HR-positive, HER2-negative metastatic breast cancer treated with ribociclib (Kisqali) and letrozole (Femara) as their initial treatment lived approximately 1 year longer than women treated with letrozole only.
Women with early-stage breast cancer who had one or both breasts surgically removed (a unilateral or bilateral mastectomy) had lower scores on a quality-of-life survey than women who had breast-conserving surgery, a new study has found.
For women undergoing chemotherapy for breast cancer, meeting the national physical activity guidelines may help alleviate cognitive issues, a new study suggests. The benefits may be even greater for patients who were physically active before treatment.
Sacituzumab govitecan (Trodelvy) now has regular FDA approval for people with locally advanced or metastatic triple-negative breast cancer (TNBC). The update follows last year’s accelerated approval of the drug for people with TNBC.
For some people with ER-positive breast cancer, a new imaging test may help guide decisions about receiving hormone therapy, according to a new study. The test can show whether estrogen receptors in tumors are active and responsive to estrogen.
The test, which helps guide treatment decisions, was not as good at predicting the risk of death from breast cancer for Black patients as for White patients, a new study has found. The findings highlight the need for greater racial diversity in research studies.
The drug abemaciclib (Verzenio) may be a new treatment option for people with the most common type of breast cancer, with new study findings suggesting that it can reduce the risk of the cancer returning.
Fertility preservation for young women with breast cancer doesn’t increase their risk of dying in the ensuing decades, a new study affirmed. Experts said the findings support routinely offering fertility preservation to patients who want it.
Some postmenopausal women with HR-positive, HER2-negative breast cancer may not benefit from chemotherapy and can safely forgo the treatment, according to clinical trial results presented at the San Antonio Breast Cancer Symposium.
A heart-related event, like a heart attack, may make breast cancer grow faster, a new study suggests. In mice, heart attacks accelerated breast tumor growth and human studies linked cardiac events with breast cancer recurrence, researchers reported.
FDA has approved sacituzumab govitecan (Trodelvy) for the treatment of triple-negative breast cancer that has spread to other parts of the body. Under the approval, patients must have already undergone at least two prior treatment regimens.
Women with high-risk breast cancer who engaged in regular exercise before their cancer diagnosis and after treatment were less likely to have their cancer return or to die compared with women who were inactive, a recent study found.
Researchers have developed a “microscaled” approach to analyze the proteins and genetic changes (proteogenomics) of a tumor that uses tissue from a core needle biopsy. The analyses can provide important information that may help guide treatment.
Tucatinib improved survival for women in the HER2CLIMB trial, including some whose cancer had spread to the brain. Trastuzumab deruxtecan improved survival and shrank many tumors in the DESTINY-Breast01 trial, which led to its accelerated approval.
A TAILORx analysis shows women with early-stage breast cancer and high recurrence scores on the Oncotype DX who received chemotherapy with hormone therapy had better long-term outcomes than what would be expected from hormone therapy alone.
Men with breast cancer may be more likely to die of the disease than women, particularly during the first 5 years after diagnosis, a new study suggests. The higher likelihood of death was linked in part to undertreatment and later diagnosis.
In a survey of nearly 600 breast cancer survivors, researchers found that the cost of care factored into the decisions the women made about what type of surgery to get. Many women also reported never discussing costs with their physicians.
FDA has expanded the approved use of the drug ado-trastuzumab emtansine (Kadcyla), also called T-DM1, to include adjuvant treatment in some women with early-stage HER2-positive breast cancer.
Many women diagnosed with ovarian and breast cancer are not undergoing tests for inherited genetic mutations that can provide important information to help guide decisions about treatment and longer-term cancer screening, a new study has found.
FDA has approved atezolizumab (Tecentriq) in combination with chemotherapy for the treatment of some women with advanced triple-negative breast cancer. This is the first FDA-approved regimen for breast cancer to include immunotherapy.
The build-up of connective tissue around some types of cancer can act as a barrier to immunotherapy. A new study uses a bone marrow transplant drug, plerixafor, to break down this barrier and improve the efficacy of immune checkpoint inhibitors in animal models of breast cancer.
A new study in mice shows that disrupting the relationship between breast cancer cells that spread to bone and normal cells surrounding them makes the cancer cells sensitive to treatment.
In women with early-stage breast cancer, two clinical trials have shown that both whole- and partial-breast radiation therapy are effective at preventing the cancer from returning after breast-conserving surgery.
Researchers are testing a topical-gel form of the drug tamoxifen to see if it can help prevent breast cancer as effectively as the oral form of the drug but with fewer side effects.
Findings from a clinical study and a mouse study may shed light on genetic risk factors for developing cancer-related cognitive problems in older breast cancer survivors. The results suggest a gene associated with Alzheimer’s disease may play a role.
Arsenic trioxide and retinoic acid work together to target the master regulator protein Pin1, a new study shows. In cancer cell lines and mice, the drug combination slowed the growth of triple-negative breast cancer tumors.
FDA has expanded the approved uses of ribociclib (Kisqali) for women with advanced breast cancer, including new uses in pre- and postmenopausal women. It’s the first approval under a new FDA program to speed the review of cancer drugs.
Using a liquid biopsy to test for tumor cells circulating in blood, researchers found that, in women with breast cancer, the presence of these cells could identify women at risk of their cancer returning years later.
Findings from the TAILORx clinical trial show chemotherapy does not benefit most women with early breast cancer. The new data, released at the 2018 ASCO annual meeting, will help inform treatment decisions for many women with early-stage breast cancer.
Do cancer study participants want to receive their genetic test results? A recent study involving women with a history of breast cancer tested an approach for returning genetic research results and evaluated the impact those results had on the women.
Researchers compared the risk of death for women with breast cancer who had low skeletal muscle mass, or sarcopenia, at the time of their cancer diagnosis and women who had adequate muscle mass.
Some people who have been treated for breast cancer or lymphoma have a higher risk of developing congestive heart failure than people who haven’t had cancer, results from a new study show.
FDA has approved the CDK4/6 inhibitor abemaciclib (Verzenio) as a first-line treatment in some women with advanced or metastatic breast cancer. Under the approval, the drug must be used in combination with an aromatase inhibitor.
A new study in mice raises the possibility that using microscopic, oxygen-carrying bubbles may improve the effectiveness of radiation therapy in the treatment of breast cancer.
The drug olaparib (Lynparza®) is the first treatment approved by the Food and Drug Administration for patients with metastatic breast cancer who have inherited mutations in the BRCA1 or BRCA2 genes.
Joint pain caused by aromatase inhibitors in postmenopausal women with breast cancer can cause some women to stop taking the drugs. Reducing their symptoms may translate into better adherence to therapy.
Kamala Thiagarajan
Meenakshi Gupta, who is blind, works as a "medical tactile examiner" to identify breast tumors. The mannequin is used in the training program for would-be examiners. The strips enable the examiners to identify and carefully examine each zone of the breast. Smita Sharma for NPR hide caption
Meenakshi Gupta has been blind since birth. And yet she can identify what many patients and medical specialists miss: the tiniest lumps in a woman’s breast that could be malignant.
One of 30 blind women from India trained as part of a global project called “ Discovering Hands, ” Gupta, 31, has been working for the past two years as a medical tactile examiner at Medanta Hospital in the northern Indian city of Gurgaon.
Introduced in India in 2017, the program is now part of major hospitals in key Indian cities: Bengaluru, Varanasi, Gurugram and Delhi. In a country where the equipment to perform mammograms is in short supply, the expertise of these examiners is crucial.
The Discovering Hands project itself first evolved in Germany a decade earlier.
From left : Neetu Garg, Neha Suri, Meenakshi Gupta, Neha Singh and Nisha Nishkam. The group of five women work as Medical Tactile Examiners (MTE) at the National Association of the Blind India Centre for Blind Women and Disability Studies in Delhi. Smita Sharma for NPR hide caption
Dr. Frank Hoffman, a gynecologist and founder of the program, says he was appalled by the sheer numbers of cases of early-stage breast cancer that were being missed, not just in Germany but around the world.
“It occurred to me that we could improve the outcome of the breast examination if we were to specially train others to do it as support staff for doctors,” he says. He decided to focus on training people who were blind; studies that have shown that the brains of blind people can develop a heightened sense of touch.
Like all the examiners, Gupta was rigorously trained by four sighted trainers who taught her about the female body, in particular the anatomy of the breast. Her training lasted 9 months: a 6-month study course and a 3-month internship. She was given mobility training as well — she uses a white cane to make her way independently to the hospital and asks for help to navigate traffic-ridden roads if she needs it. The effort is worth it.
After being trained as a breast examiner, Meenakshi Gupta can identify lumps at initial stages, even before they show up on imaging scans. Smita Sharma for NPR hide caption
“They were so successful that they were 30% better at detecting tissue changes than doctors,” Hoffman says of the trainees. “The MTEs can identify lumps at the very initial stages, even before they show up on imaging scans.”
Over the years, several independent studies have verified this. A pilot study conducted in 2023 by the department of obstetrics and gynecology at the Erlangen University Hospital in Germany involved 104 patients and concluded that clinical breast examinations by MTEs who were visually impaired were as accurate in identifying potentially cancerous tumors as doctors trained in this procedure.
In many cases, a lump in the breast can be noncancerous. Tests are needed to rule out cancer, and the earlier this is done, the better, says Dr. D Pooja, gynecologist and CEO of Apar health, who is not affiliated with the Discovering Hands program. "A MTE's work is very empowering, especially in a low-tech setting when not everyone can have access to mammograms. It eases the burden on doctors who deal with over-crowded waiting rooms too," she says.
Examiners like Gupta are proving to be a powerful force for identifying breast cancer and allowing for effective treatment , says Dr. Pooja, who adds: "However, we need more clinical studies to establish how their work adds value to the health-care system."
As an arts graduate who had studied science only in high school, Gupta says it was challenging to learn about breast anatomy and conduct clinical exams.
A model of skin used for medical training for breast exams at the National Association of the Blind India Centre for Blind Women and Disability Studies in Delhi. Smita Sharma for NPR hide caption
In their training sessions, the MTEs practice on plastic models with silicone breasts. “One of the first things we learned was how to map the breast,” she says. Using skin-friendly tape, they divide each breast into four zones. Probing gently with fingertips and using varying pressures, they closely examine every centimeter of the breast. The process takes up to an hour for both breasts. If the MTEs locate a lump, mapping the breast this way helps the doctor locate it quickly and precisely for further examination.
MTEs document their findings to share with doctors, Gupta says. “We examine the consistency of each area of the breast: Is it hard or soft? If we find any lump, we make a note of its location, depth, size and shape. Our duties end with the examination. We’re not authorized to say whether it could be cancerous or not."
After her training, Gupta began working as an intern at Medanta Hospital.
“To build our understanding and confidence at first, the doctor would ask us to identify the nature of the lump in patients who had already been diagnosed with breast cancer,” she says. They spent months noting down the feel, size, shape and consistency of these lumps. Soon, they moved on to examining patients who came in for checkups.
Meenakshi Gupta pastes tape on a dummy — a practice that helps divide the body into zones for breast exams. Smita Sharma for NPR hide caption
“I remember being so nervous that my hands shook when I examined my first patient,” she says. The responsibility weighed on her: “This is a living person. What if I miss anything? I was worried.” That anxiety eased within a couple of months, as her experience and confidence grew.
She now sees 5 to 8 patients a day, spending roughly an hour with each. On average, she says, one or two tend to have abnormalities. She records her findings on a laptop and flags the cases for follow-up with the doctor.
Two women have been instrumental in establishing the Discovering Hands program in India and training MTEs.
One of them is Shalini Khanna Sodhi, Founding Director and Secretary of the National Association for the Blind, India’s Centre for Blind Women and Disability Studies in New Delhi. . “Blind women especially were a very forgotten lot,” Sodhi says. “In addition to providing diagnostic support for doctors who are facing severe overcrowding in waiting rooms, these roles give visually impaired women who often struggle to find jobs, dignity and purpose,” she says.
Sodhi’s efforts were supported by Dr. Kanchan Kaur, a surgeon who reconstructs breasts after a mastectomy and who is practicing in Medanta Hospital, where Gupta now works. Both women traveled to Germany to observe and learn the technique.
Kaur says that another issue is the lack of ‘breast awareness’ among Indian women. Not every woman has access to annual mammograms, a screening procedure that is considered routine in other countries. According to India’s Ministry of Health and Family Welfare, breast cancer is one of the most common cancers among women, with roughly 75,000 deaths every year. One of the biggest reasons is the lack of access to mammography equipment, especially in rural areas where mammograms are not a part of routine care.
“I’m very aware that I treat a disease that’s potentially curable if women come in time,” Kaur says. “But in India, this is a huge problem. Women may sense that they have a lump, but because of the stigma of having their breasts examined, many delay getting treatment until it’s too late.” In India, the biggest issue is that women don't go to the doctor on time. That's what causes the delay in diagnosis --there is still stigma involved and in such a conservative society, women are hesitant to expose their breasts to doctors for medical check ups. Nearly half of the cases Kaur sees are women who seek treatment at an advanced stage, “when the cancer is very aggressive and the mortality rate is high,” she says.
According to a report by the Indian Council of Medical Research (ICMR), only one of every two women diagnosed with breast cancer in India survives. Studies have found that the disease occurs at a younger age for women in India , (between the ages of 45-49) compared to the West, and because of this, survival rates are poor. That’s why it’s critical to catch the disease early, says Kaur.
The fact that breast examinations are being done by blind women has helped ease some of the stigma around going for a check-up, she says. Because the MTEs are blind, the women who come in for an exam do not feel embarrassed or ashamed about exposing their breasts. This was something that health-care professionals hadn’t quite thought about when they began the program in Germany, Hoffman says.
If more MTEs were trained and breast examinations became part of a routine checkup in hospitals across the country, it could save a lot more lives, Sodhi says. However, that's not a goal that can be easily met.
Currently, the intensive training for each MTE costs around 2 lakh rupees — about $2,500. “We are funded by private donors and the number of MTEs we train every year would depend on this funding,” Sodhi says. “But we’re hopeful we can expand soon. If so, Discovering Hands can save many more lives. It’s heartening when you think of how these women who cannot see themselves are showing us the way.”
Kamala Thiagarajan is a freelance journalist based in Madurai, Southern India. She reports on global health, science and development and has been published in The New York Times, The British Medical Journal , the BBC, The Guardian and other outlets. You can find her on X @kamal_t
IMAGES
COMMENTS
Global guidelines for breast cancer screening: A systematic ...
The introduction of screening mammography in the United States has been associated with a doubling in the number of cases of early-stage breast cancer that are detected each year, from 112 to 234 ...
Update from the ACR and Society of Breast Imaging
Screening for Breast Cancer: Evidence Report and ...
Breast Cancer Screening (PDQ®) - NCI
ACS Breast Cancer Screening Guidelines
Advances in Breast Cancer Research - NCI
Breast cancer screening is used to identify women with asymptomatic cancer with the goal of enabling women to undergo less invasive treatments that lead to better outcomes, ideally at earlier stages before the cancer progresses. ... effect of mutation status on cancer incidence. Breast Cancer Research and Treatment. 2009 Dec; 118 (3):539-546 ...
TMIST (Tomosynthesis Mammographic Imaging Screening Trial) is a research study that will help researchers learn about the best way to find breast cancer in women who have no symptoms. It is a randomized breast cancer screening study that compares two types of Food and Drug Administration (FDA)-approved digital mammograms for their ability to ...
Recommendation: Breast Cancer: Screening
Breast cancer is the second most common cause of cancer death for women in the US. Each year, there are about 240 000 cases diagnosed and nearly 43 000 women die from breast cancer. Notably, Black women are 40% more likely to die from breast cancer than White women, even though they get breast cancer at a roughly similar rate.
Breast-Cancer Screening — Viewpoint of the IARC ...
Concluding Remarks. This research underscores the potential impact of mammography frequency on breast cancer mortality rates, accentuating a lower risk with annual screenings. The data suggests that tailored approaches to mammography, especially for certain demographics, may improve breast cancer outcomes.
breast cancer screening in 2018, compared to 64% of insured women. • The prevalence of up-to-date breast cancer screening was 70% or higher among lesbian women, college graduates, and those ages 55-74 years. • In 2016, by state, the prevalence of up-to-date mammography among women ages 45 and older
New ACR Breast Cancer Screening Guidelines call for ...
Breast Cancer Screening: Common Questions and Answers
Screening for Breast Cancer
Clinical trials are research studies that involve people. The clinical trials on this list are for breast cancer screening. All trials on the list are NCI-supported clinical trials, which are sponsored or otherwise financially supported by NCI. NCI's basic information about clinical trials explains the types and phases of trials and how they ...
Breast screening
In most developed countries, both organized screening (OrgS) and opportunistic screening (OppS) coexist. The literature has extensively covered the impact of organized screening on women's survival after breast cancer. However, the impact of opportunistic screening has been less frequently described …
To systematically update the 2009 USPSTF review on screening for breast cancer in average risk women age 40 years and older. ... This report is based on research conducted by the Pacific Northwest Evidence-based Practice Center (EPC) under contract to the Agency for Healthcare Research and Quality (AHRQ), Rockville, MD (Contract No. HHSA-290 ...
Armed with information about the complexities of breast density and its impact on breast cancer screening, women can discuss their personal risk with their providers and evaluate the options for ...
About 240,000 cases are diagnosed annually and nearly 43,000 women die from breast cancer. The nudge toward earlier screening is meant to address two vexing issues: the increasing incidence of ...
Mindpeak: Algorithms for breast biomarkers and pan tumour PD-L1 for lung, gastric, esophageal, bladder and breast cancers; Owkin: Algorithm for the screening of microsatellite stability in colorectal cancer; Qritive: Algorithms for screening and grading of prostate cancer, analysing lymph nodes for metastasis, and screening for colon cancer
Breast Cancer Screening - NCI
Description. This ESMO Deep Dive Breast Cancer Webinar focused on HER2 positive metastatic breast cancer. This webinar was developed with the aim to provide a second, deeper layer of educational experience in emerging data related to: molecular biology and classification, translational research and biomarkers for precision medicine, and unknowns, hypotheses, and ongoing research in areas of ...
This section considers measures of screening quality and major beneficial and harmful outcomes. Beneficial outcomes include reductions in deaths from breast cancer and in advanced-stage disease, and the main example of a harmful outcome is overdiagnosis of breast cancer. The absolute reduction in breast cancer mortality achieved by a particular screening programme is the most crucial indicator ...
Despite an overall decline in deaths from breast cancer for women, mortality rates for Black women remain 40% higher compared to white women, according to research from the American Cancer Society ...
Find research articles on breast cancer, which may include news stories, clinical trials, blog posts, and descriptions of active studies. ... For women in their 70s and older, the risk of overdiagnosis with routine screening mammography is substantial, a new study suggests. The findings highlight the need for conversations between older women ...
Dr. Frank Hoffman was appalled by the sheer numbers of cases of early-stage breast cancer that were being missed. Then he had an idea: What if "we were to specially train others to do it."