The effect of the Flemish breast cancer screening program on breast cancer-specific mortality: A case-referent study

1 Introduction

Currently, breast cancer is the most common cancer in women in Europe, both in terms of incidence and mortality [1] . Over the last decades, screening programs were introduced across Europe with the primary aim of reducing mortality from breast cancer. This followed the results of randomized control trials performed in the 1980s that showed breast cancer screening by mammography to be efficacious in reducing breast cancer mortality [23] . Since then, observational studies have been performed to estimate the effectiveness of mammographic screening on breast cancer-specific mortality in a real-world setting. Although most of these studies report a reduction in breast cancer-specific mortality for screened women, the magnitude of the estimates varies. For example, estimates from studies in Western Europe range from 12 % to 58 % reduction in breast cancer-specific mortality [4] . This wide range of estimates is a consequence of large differences in design, participation rate and quality of the studies [45] . Moreover, screening programs and therapy have improved over time [6] , making it valuable to monitor effectiveness over time. In addition, most observational studies were performed in countries with low screening rates outside the screening program (opportunistic screening).

The impact of the population-based mammography screening program (PMSP) on breast cancer-specific mortality has never been studied in Belgium. Therefore, the aim of this study was to assess the effectiveness of PMSP in Flanders (the Northern region of Belgium) in reducing breast cancer-specific mortality. In Flanders, opportunistic screening remained common after the introduction of PMSP in 2001: its coverage gradually declined from slightly over 20 % to just below 15 % since 2015 [7] . To this end, we investigated the association between participation in PMSP and breast cancer-specific mortality in all women eligible for screening and living in Flanders. The outcome is interpreted as the additional impact of the organised program on top of ‘routine care’ (i.e. the availability of opportunistic screening and clinical diagnosis). We also performed several sensitivity analyses. First, we investigated whether the association differed across provinces. Second, we explored the effect of adjustment for additional potential confounders. Third, we postponed the pseudodiagnosis date with one year to explore exposure opportunity bias. Fourth, women who were at risk of hereditary breast cancer were excluded because these women might be more frequently screened opportunistically and thus might bias the estimates. Fifth, women who were screened exclusively opportunistically were excluded to compare the effect of participation in PMSP with ‘occasional opportunistic screening and clinical diagnosis’ rather than ‘opportunistic screening and clinical diagnosis’. Last, we corrected our unadjusted effect estimates for self-selection bias.