About cancer screenings and saving lives: measuring the effects of cancer screening programs through meta-analyses—A comment to the meta-analysis “Estimated Lifetime Gained With Cancer Screening Tests” by Bretthauer et al. (2023)

A meta-analysis of randomized clinical trials conducted by Bretthauer et al. to evaluate the advantage of cancer screening, recently published by Jama Internal Medicine, concluded that “common cancer screenings do not save lives with the possible exception of sigmoidoscopy screening” (1). The Authors derive their conclusion from estimates of lifetime gained with screening by “comparing all-cause mortality in people who underwent screening with those who did not.” They used the relative risk of death from any cause measured from randomized trials of cancer screenings and the average follow-up time of the unscreened group to obtain estimates of lifetime gained with screening. Both Bretthauer et al. in their meta-analysis and a comment paper appeared in the same number of JAMA Internal Medicine express the view that only randomized controlled trials can provide evidence of (cancer) screening efficacy and that a reduction of all-cause mortality is the measure of choice to evaluate efficacy (instead of the commonly used cancer-specific mortality) (1, 2). The reason for their choice is that a reduced risk of cancer specific deaths, if it is not associated with a reduced risk of all- cause mortality, can be considered the consequence of deaths associated with harmful effects of screening counterbalancing the screening benefit or of substitution of cancer specific deaths with death from competing causes. Nevertheless, we contend that the use of too stringent criteria led to an underestimation of the influence of screening on all-cause mortality in the meta-analysis authored by Bretthauer et al. and that the use of all-cause mortality implies small and unreliable estimates of screening efficacy (1). We believe that small estimates of relative risk for all-cause mortality should not be interpreted as minor effect of a cancer screening but indicate the opportunity to investigate the presence of bias in cause of death assignment and eventual harm of screening. With respect to the results published by Bretthauer et al., we also remark that 10–15 years of follow-up are insufficient to fully evaluate the impact of screening. Furthermore, low adherence to screening and uptake of screening in the control arm led to underestimation of screening efficacy in some randomized trials. Finally, evidence from observational studies should not be completely ignored, particularly for cancer screening that reduces incidence of infiltrative cancers.