Although this is a relatively better article, the way it is presented has tremendously decreased its importance. All studies with a repeated measure design (Pre-post design) should report mean change (gain score) and the standard deviation of the change, or preferably 95% confidence intervals. We are clinicians, and we should be more concerned about the clinical benefit and not about the p (or "pee") value. Thus we should try to always give 95% confidence intervals. Now that authors have not given SD of the mean change (gain score) there is no way of knowing that whether the effect of the drug on the subjects was homogenous or heterogenous, and whether there was any treatment interaction with individual subjects. Thus authors have wasted all the strengths of the pre-post design while keeping all the inherent limitations of the pre-post design. ------ Moreover, changes in lipid profile should not let one to jump to the conclusion that there will be cardioprotection. After all, hormone therapy also improves lipid profile in post-menopausal women.
Posted by: rqayyum - May 2005