Authors' reply to: "Questionable method for estimating the influence of mammography screening on breast cancer mortality in the Netherlands".

Abstract

We appreciate the comments of Autier and Boniol on our study “The effect of population-based mammography screening in Dutch municipalities on breast cancer mortality: 20 years of follow-up.” In their letter, they raise several questions concerning the methodology of this paper and the interpretation of results. Our study clearly shows that the implementation of screening played a role in the decline in breast cancer mortality. In our discussion, we argue extensively that the decline in breast cancer mortality is caused by the effect of both the introduction of adjuvant treatment and screening but that it is difficult to quantify the treatment and screening effect separately. We refer to modeling studies1, 2 in which the separate effects of treatment and screening on breast cancer mortality were estimated. Therefore, we do not deny the role of (adjuvant) treatment in the reduction in breast cancer mortality over the years; on the contrary, we think that the introduction and increasing use of adjuvant treatment has been of great influence on breast cancer mortality. However, we do believe that our results indicate that, after its introduction, screening also played an important role in the reducing mortality. We agree that our study has limitations with respect to quantifying the effect of screening. However, as these limitations are inherent in trend analyses, the authors' re-analysis with jointpoint may not solve this issue. In their re-analysis the authors do not take into account the time of screening implementation and therefore also neglect the gradual aspect of this implementation over a period of 7–10 years. An important disadvantage using this approach is the inclusion of calendar years directly after the introduction of screening, which is likely to dilute the effect per year in terms of mortality reduction. This has been emphasized in a systematic review of all significant trend studies on breast cancer mortality in Europe.3 The reason why we transformed calendar years to “years since introduction of screening” and estimated the annual percentage change in mortality from the turning point in the trend is to account for this drawback. Despite the limitations of their jointpoint analysis, the authors' findings are largely in line with ours, with a jointpoint in 1993 in the age group 55–74 and a jointpoint in 1999 in the age group 75–79. This finding is in disagreement with their reasoning that the decline is in all age groups equivalent, and that greater provision in adjuvant therapy is most likely responsible for the change in trends. Rather, these results support that the later downturn in the trend in the age group 75–79 (vs. 55–74) is related to the fact that women between the ages 70 and 74 years were later invited to screening than women aged 50–69 years. Concerning the comparison between early and late starters (Fig. 3, article), municipalities were grouped in a way to obtain three groups of comparable size (similar number of municipalities in all groups). The reason why the early group reflects a greater duration in terms of calendar years (6 years vs. 2 and 3 years) is that during the major part of this period (1987–1990) screening was implemented in very few municipalities. Therefore, the later years of the “early group” (1991–1992) can still be considered as early years for implementation of screening. Finally, we did perform our analysis for all age groups, equal to the re-analysis of the authors, including the groups 40–54 and 75–79 (Table 1; results are also reported in the Supporting Information). Our results may be more accurate as our analyses were conducted at the municipality level, using mortality data of each municipality separately. Our analysis, thus, takes into account differences in calendar year of introduction of screening between municipalities, which is not the case in the jointpoint analysis of Autier and Boniol. We also used the age group 40–54 as a reference group for the group invited to screening (55–74) and found that a turning point similar to the one observed in women aged 55–74 (shortly after implementation of screening; Year 2) was absent in this reference group. The presence of a turning point related to the time of implementation of screening in the group invited to screening alone indicates that introduction of screening also affected the mortality trend in this group, rather than the increase in the use of adjuvant therapy alone. To conclude, our previous analysis4 as well as the supplemental jointpoint analysis of Autier and Boniol shows substantial declines in breast cancer mortality in the Netherlands with changes in trends shortly after the introduction of screening in the age group invited to screening, supporting our previous conclusion that the implementation of screening played an important role in the decline in breast cancer mortality in the Netherlands.