Should we ignore covariate imbalance and stop presenting a stratified 'table one' for randomized trials?



I point them to the CONSORT statement, the very guidelines most of them require authors to follow!


There is much good there (although they provided an erroneous definition of the p-value) but Iā€™m sorry to see the unquestioning stance of stratifying by treatment. The main thing this accomplishes is casting doubt on perfectly good RCTs.