Benefits and Harms of Mammography Screening

Magnus Løberg; Mette Lise Lousdal; Michael Bretthauer; Mette Kalager


Breast Cancer Res. 2015;17(63) 

In This Article

Breast Cancer Mortality

According to the randomized breast cancer screening trials, the relative reduction in mortality from breast cancer ranges between 15 and 25%[9,22,36,51] for women aged 50 to 69 years. The differences in these estimates are due to differences in inclusion of randomized trials in pooled estimates. For the 25% estimated reduction, mammography screening versus no-screening is compared; thus, the Canadian trial was not included because they compared physical breast examination to combined physical breast examination and annual mammography.[10,36] For the 15% estimated reduction, methodological limitations in some of the randomized trials was accounted for;[9] without this 'adjustment', a 20% reduction was found.[9,22,52] None of the randomized trials showed any effect on cancer mortality or all-cause mortality.[9] Given the number of women enrolled in the randomized trials (660,000) and a 20% reduction in breast cancer mortality, a 2% reduction in all-cause mortality should have been detectable.[52] The absence of a reduction in all-cause mortality indicates that women die of other diseases at about the same time in life with and without screening.

Study Designs

There are a number of methods to investigate the effect of mammography screening in a non-experimental setting. Cohort studies, case–control studies, and trend studies show different estimates of mortality reduction, ranging from no effect to 50% reduction in breast cancer mortality.[53,54]

Cohort Studies. The optimal non-experimental design to investigate the effect of mammography screening is a cohort study of women invited and women not-invited to mammography screening who have similar baseline risk for breast cancer and breast cancer death and similar opportunities for optimal breast cancer treatment. Only few such studies exist, and the estimated effect of mammography screening on breast cancer mortality varies from 10 to 25% reduction.[35] A pooled estimate of these trials showed a reduction in breast cancer mortality of 13 to 17%.[35]

Case–Control Studies. In case–control studies (sometimes called case-referent studies) cases are women who die of breast cancer and controls are women who are alive stratified by whether they have undergone screening mammography or not. Thus, these studies when performed in settings where mammography screening is recommended or where screening programs exist are comparisons of women who participate and who do not participate in mammography screening. The validity of these studies is low because of healthy screenee and self-selection bias, as women with breast cancer are not eligible to mammography screening or to be continued to be screened (selection of the most healthy), and women who choose to participate in mammography screening (selection) may differ with regard to risk of death from those who do not participate.[55] Attempts to adjust for these biases have been done by adjusting for the relative risk in breast cancer mortality between the non-participants and the non-invited comparison group.[7,56] The underlying assumption of these adjustments is that we do know the risk of uninvited women. In randomized trials, we can easily find the risk of breast cancer death for those not invited to mammography screening (the control group). However, in observational studies where everybody is invited or recommended to undergo mammography screening, we have to make assumptions on risk of death from breast cancer among the uninvited women. These assumptions cannot be tested and are therefore based on 'best guess' estimates. In case–control studies, a 50% reduction in mortality from breast cancer is found, and similar reductions are found in cohort studies of participants and non-participants in mammography screening.[54,57] When the randomized trial from Malmö was analyzed as a case–control study, a 58% reduction in mortality from breast cancer was found, whereas the real, observed reduction in the trial was only 4% (8% when the results were adjusted for non-compliance and contamination).[36] Thus, estimates from case–control studies systematically overestimate the effect of screening.

Trend Studies. Trend studies are studies of population-based breast cancer mortality over time in different ages (age-standardization) and geographic areas. Data on population-based breast cancer mortality are easy to retrieve, but as the yearly mortality rate is not reflective of time of diagnosis, deaths from breast cancer diagnosed before invitation influences the mortality rate some years after screening is implemented. Further, when all eligible women are invited and a screening program has been running for some time, the mortality rate is expected to reach a steady state and further reduction cannot be expected. After 7 years of follow-up in the Health Insurance Plan study, the mortality reduction was no longer apparent,[58] indicating that screening has no effect if no longer offered. For a continuing program, however, the mortality effect will not disappear, but reach a steady state. Thus, in the first years after screening has been introduced and reached full coverage in the area studied, the cause of change in trends of breast cancer mortality can be difficult to study and interpret. Most trend studies show that breast cancer mortality has declined in most European countries since the early to mid-1990s. The decline in mortality is even higher among women younger than the eligible age range for screening and for some countries a reduction is observed also for women older than the eligible age range.[59] The interpretation of these results could be that heightened awareness and improved therapy rather than mammography screening are responsible for the observed reduction.[53,59,60]

Tumor Stage

Another benefit of mammography screening could be that breast cancers detected at screening are smaller and thus less advanced than those detected clinically. In general, smaller tumors are more likely to be resected by lumpectomy, and with less node-positive disease, less adjuvant therapy is needed. Based on the randomized mammography screening trials, however, this is not the case; screening was associated with an increase in the number of mastectomies of about 20%.[9] The reason is that mammography increased both the number of women diagnosed with invasive breast cancer and the number found to have multiple microscopic cancers distributed throughout the breast, for which mastectomy is recommended. Further, in the National Health Service breast screening program in the UK, 30% of DCIS and 24% of invasive breast cancers were treated with mastectomy, so earlier detection does not necessarily mean less aggressive treatment.[61] As mentioned above, another benefit of mammography screening could be less aggressive adjuvant therapy, due to smaller and less aggressive tumors. As seen in the stage distribution in screening and non-screening groups in Norway,[41] screening led to the diagnosis of 58% more stage I (localized cancer) and 22% more stage II (regional cancer or cancer involving the lymph nodes) cancers, without any reduction in advanced stage disease (stages III and IV). Since all these patients receive surgery (either mastectomy or breast-conserving surgery with radiation) and most stage II patients are recommended to receive adjuvant chemotherapy, screening may have led to 58% more women undergoing breast surgery and 22% more women undergoing adjuvant chemotherapy.[41] Thus, screening mammography does not seem to reduce the burden of receiving more aggressive treatment.

Cause of Death

The number of women saved from breast cancer death might be outweighed by death from other causes due to harms of treatment; however, due to uncertainty about the overall number of women saved, we present different estimates of women saved from breast cancer in different meta-analyses of randomized and observational studies of breast cancer[19,22,32,34,35] (Figure 3). The number needed to be invited to mammography to save or harm women is highly dependent on the underlying risk of breast cancer or death from breast cancer (Figures 4 and 5, showing risk of breast cancer and death from breast cancer in the US and UK[49,62]). In the estimates shown in Figure 1, we use UK data from 2007 for mortality from breast cancer in women aged 55 to 74 years,[34] and the relative reduction of 13 to 17% in breast cancer mortality based on a meta-analysis of observational studies.[35] For 1,000 women invited to mammography screening every second year for 20 twenty years from age 50, 2 to 3 women are prevented from dying from breast cancer (Figure 1).

Figure 4.

Benefit and harm with screening mammography and use of aspirin over 10 years [62] . Shown are the 10-year risk of death from breast cancer (bars above 0) and the 10-year risk of the diagnosis of breast cancer (bars below 0) among women aged 40 years and 50 years, with and without mammography screening. Also shown are the 10-year risk of death from cancer (bar above 0) and the 10-year risk of major extracranial bleeding, defined as bleeding necessitating transfusion or resulting in death (bar below 0), associated with the use or non-use of aspirin as a primary preventive measure (on the basis of findings from randomized trials). In each pair (no screening versus screening and no aspirin versus aspirin), the difference between the percentages represented by the bars shows the absolute benefit or harm associated with screening mammography or the use of aspirin. Background data are derived from the literature.

Figure 5.

Twenty year risk for diagnosis of, and death from, breast and prostate cancer with and without screening in the United Kingdom [49] . Displayed are 20-year absolute risks for incidence (including overdiagnosis) and mortality with and without screening. Overdiagnosis is set to 45% for prostate cancer and 22% for breast cancer, respectively (age 50 to 69 years). Mortality reduction is set to be 20% for both cancers (age 55 and 74 years). For prostate cancer, the estimates are based on the observed incidence and mortality in 1998 (before any widespread use of prostate-specific antigen (PSA)) and for breast cancer in 2007 (latest data available).