Volume and mortality in coronary artery bypass grafting

Further reanalysis of existing trials is unlikely to yield new insights, which will require new studies.Anderson and Tomenson recommend investigation of antidepressants prescribed in clinical practice rather than in clinical trials.Such observational studies are subject to many potential biases, which reduce the validity of comparisons.'What is needed are large, well designed pragmatic randomised controlled trials in primary care settings; such trials will allow both reliable and generalisable estimation of the relative effectiveness and cost effectiveness of different classes of antidepressants, as used in routine clinical practice and over a longer period.Future randomised controlled trials comparing classes of antidepress- ants should also use more relevant outcomes to measure quality of life in those who continue to use treatments and those who drop out.