Letter to the Editor: “An analysis of 11.3 million screening tests examining the association between recall and cancer detection rates in the English NHS breast cancer screening programme”

by Dr Sian Taylor-Phillips, Olive Kearins, and the Public Health England Breast Screening Clinical Advisory Group (S.Taylor-Phillips@warwick.ac.uk)

An analysis of 11.3 million screening tests examining the association between recall and cancer detection rates in the English NHS breast cancer screening programme

Dear Editor,

We read with great interest the article entitled “An analysis of 11.3 million screening tests examining the association between recall and cancer detection rates in the English NHS breast cancer screening programme” by Roger Blanks and colleagues [1]. We would like to congratulate the authors on an excellent paper offering some fascinating insights from a very large cohort of women attending breast screening in the UK.

The authors suggest that their data is used to underpin UK screening standards. Such analysis of existing data is an essential tool in defining standards in a population screening programme, but further analysis and clarity is required to enable standard setting. There are two areas in the paper where further description would aid interpretation for policy-makers and researchers alike: showing the data itself in addition to the models, and information about model fit.

Data points are provided on Figure 1, whereas the other figures only provide the details of the models. The data points show the range where the models are driven by data, and the ranges where there are no data and the model is based purely on extrapolation. For example, in Figure 1, there are 5 data points ranging between 4% and 10% recall rate. The shape of the curve below 4% is based on extrapolation, so for example, it would be unwise to define a lower bound of acceptable recall rates based on the shape of this curve below 4%. Such consideration cannot be applied to the other figures as the data points are not shown. Providing the data points and confidence intervals would allow the reader to examine the heterogeneity between centres while also assuring policy makers that recall rates could be changed. It would also be interesting to see the Dutch and English data separated: if the Dutch recall rates are systematically lower, then other differences between the two countries could influence the shape of the curve.

The finding that rates of DCIS detection increase in a linear fashion with increasing recall may be of particular significance considering the debate around overdiagnosis. However, again without the accompanying data points or model fit, drawing any conclusions around this may be premature.

It would also be fascinating to extend the analysis to interval cancers to ascertain whether increases in cancer detection rates at screening are accompanied by reductions in interval cancers, or reductions in later stage cancer diagnosis. Furthermore, an examination of how the characteristics of detected cancers varies with recall threshold would be of value.

We have additional research planned to extend these analyses, particularly to investigate the impact on interval cancers and beyond. The intention is that these extensive, additional analysis together with a review of the existing published evidence will then inform any modifications of the recall standards in the English NHS Breast Cancer Screening Programme.

We would like to thank the authors for such an interesting and thought provoking article, a very enjoyable read.