Authors'reply

摘要

We have tried to fool none of the people none of the time. However, we acknowledge the limitations of observational data, although they can have higher statistical precision and be more generalisable than data from randomised trials. The question is whetherthe systematic errors in our study misled us to believe mortality is higher after radiotherapy than after surgery for prostate cancer. Although we adjusted for almost all factors considered in the choice of treatment using different approaches that gave consistent results, the believability of our finding rests on a judgment of the size ofresiduaLonfounding. To address this, we estimated the strength of confounding needed to explain the differences that we found. Does Roach seriously believe such large effect sizes can be attributed to systematic differences in subgrades of Gleason 7 tumours or percentage positive scores between treatment groups?