Screening and Breast Cancer

Publisher Website
Google Scholar
Abstract
The article by Berry et al. (Oct. 27 issue)1 represents an impressive modeling exercise. However, the fact that all seven models of the effects of screening and adjuvant therapy on mortality from breast cancer yielded qualitatively similar results is not as remarkable as it might first appear. The reason is that the most likely source of bias is the combination of parameter estimates from different populations with different distributions of risk factors. Because all the models used the same data, any bias from the combination of different data sets would probably appear in all analyses. Another important consideration is how the models capture the effect of early detection on subsequent survival, which has a potentially large effect on the final results. A recent article reported that even after adjustment for tumor size, lymph-node status, and disease stage, there was a survival benefit for subjects whose breast cancer was detected by screening rather than clinically.2 To the credit of Berry et al., some of the models accounted for the method of detection in the estimation of survival after diagnosis, but some critical details were not provided. In future work, elucidation of this key point would be helpful.