Physicians are informed by studies which examine group data but deal with individual patients. How to apply the group data in clinical setting is not as easy as it might appear at first glance.
I blogged about this general topic several years ago.See here.
The term "Heterogeneity of treatment effects" (HET) is the translation into the jargon of the statistician of the basic fact that everyone does not respond the same to a particular treatment. Can the patient in the doctor's office be assumed to have the average response to a given treatment reported in a medical journal article? In a given group treated with a certain medication some subjects will fare better than average along some parameter of interest while others respond not at all and some in either group may have adverse effects,some serious some minor.You cannot expect every patient receiving a given treatment to do well let alone better than average which only occurs in the statistically impossible world of the children highlighted by Garrison Keillor.
RL Kravitz,N Duran and J Braslow authored the classic article on the issue of HET. See here for full text of the article which should be part of every medical student's education.
Dr. Michel Accad in his Blog Alert and Oriented discusses a recent paper that offers suggestions for ways to tame the problem of HET. The suggestions are aimed as those who carry out the clinical trials . See here for Accad's discussion entitled "Dealing with variable risk" and see here for a link to the full text of the article by Kent et al that he references.