Screening for breast cancer with mammography
Peter C Gotzsche,Karsten Juhl Jorgensen
DOI: https://doi.org/10.1101/2024.06.06.24308542
2024-06-06
Abstract:Background
A variety of estimates of the benefits and harms of mammographic screening for breast cancer have been
published and national policies vary. This is an update of a review previously updated 2013 and originally
published 2001.
Objectives
To assess the effect of screening for breast cancer with mammography on mortality and morbidity.
Search methods
For this 2023 update, we searched PubMed, CENTRAL, the Cochrane Breast Cancer Group Specialised
Register, the World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) and
ClinicalTrials.gov up to 28 February 2023.
Selection criteria
Randomised clinical trials (RCTs) comparing mammographic screening with no mammographic screening.
Data collection and analysis
Two authors independently extracted data. Study authors were contacted for additional information. Our main
outcomes of interest were deaths due to breast cancer, any cancer, and due to any cause, and harms measured
as overdiagnosis, number of mastectomies, lumpectomies, use of radiotherapy and of chemotherapy. Certainty
of evidence was assessed with GRADE.
Main results
Eight eligible trials from Europe and North America that compared women offered screening mammography with
women not offered screening were included. We excluded a trial because the randomisation failed to produce
comparable groups. The eligible trials included 600,000 women in the age range 39 to 74 years.
The trials with adequate randomisation did not show a benefit in terms of a reduction in breast cancer mortality at
13 years (risk ratio (RR) 0.90, 95% confidence interval (CI) 0.79 to 1.02; 33 vs 30 deaths from breast cancer per
10,000 women; 3 RCTs; 292,153 participants). The findings at 24 years were similar to those at 13 years. Our
certainty in both estimates was downgraded 1 level to 'low' due to changes in technology and treatment
(indirectness) and due to imprecision. The trials with suboptimal randomisation showed a reduction in breast
cancer mortality at 13 years with an RR of 0.75 (95% CI 0.67 to 0.83; 4 RCTs; 306,937 participants; very low
certainty evidence).
In women below age 50 years, the results from adequately randomised trials did not show a reduction in breast
cancer mortality at 13 years of follow-up (RR 0.87, CI 0.73 to 1.03; 28 vs 24 deaths from breast cancer per 10,000
women; 3 RCTs; 218,697 participants, low certainty evidence), nor for women at least 50 years (RR 0.94, CI 0.77
to 1.15; 53 vs 50 deaths from breast cancer per 10,000 women; 2 RCTs; 74,261 participants, low certainty
evidence). Only one trial included women aged 70 years and above and could not provide a reliable effect
estimate.We found that breast cancer mortality was an unreliable outcome that was biased in favour of screening, mainly
because of the risk of differential misclassification of cause of death. The trials with adequate randomisation did
not find an effect of screening on total cancer mortality, including breast cancer, (RR 1.00, 95% CI 0.96 to 1.04;
288 vs 288 cancer deaths per 10,000 women; 3 RCTs; 292,954 participants; moderate certainty evidence; the
follow-up was 10.5 years for Canada, 9 years for Malmo and 23 years for the UK age trial). All-cause mortality
was not reduced (RR 0.98, 95% CI 0.94 to 1.03 after 7 years; RR 0.99, 95% CI 0.95 to 1.03 after 13 years; 324 vs
328 deaths per 10,000 women; and RR 1.01, 95% CI 0.99 to 1.04 after 24 years; 773 vs 765 deaths per 10,000
women; 2 RCTs; 250,671 participants; moderate certainty evidence) in the adequately randomised trials.
There were more lumpectomies and mastectomies combined in the screened groups, likely reflecting
overtreatment (RR 1.31, 95% CI 1.22 to 1.42; 164 vs 214 operations per 10,000 women; 2 RCTs; 132,321
participants; moderate certainty evidence), as were the number of mastectomies alone (RR 1.20, 95% CI 1.08 to
1.32; 122 vs 102 per 10,000 women; 2 RCTs; 132,321; moderate certainty evidence). The use of radiotherapy
was similarly increased whereas there was no difference in the use of chemotherapy (data for each outcome
available from only one adequately randomised trial; low certainty evidence). Breast screening increased the
number of breast cancer diagnoses (overdiagnosis)(RR 1.25, CI 1.18 to 1.34, 142 vs 113 diagnoses at 7 to 9
years of follow-up; 3 RCTs, 292,979 participants; moderate certainty evidence) in trials that did not screen the
control group after the intervention phase.
Authors' conclusions
Because of substantial changes in screening technology, treatment, and breast cancer awareness since the
trials were done, the estimates from the trials are uncertain in today's setting. As breast cancer mortality is an
unreliable outcome that is biased in favour of screening, it is noteworthy that screening did not reduce total
cancer mortality or total mortality. Breast screening does not meet the criteria that population screening should be
based on rigorously performed randomised trials that show that the benefits outweigh the harms. No studies
have been completed in low income countries and one small study from Colombia has yet to provide data on long
term outcomes. Women, clinicians and policy makers should consider the trade-offs and the uncertainties
carefully when they decide whether or not to attend or support breast screening programmes.