Applying group data to individual patients-the problem revisited and perhaps a solution sugested

How does one apply the results of even the very best designed and executed randomized clinical trials (RCT) to the individual patient? I guess the answer is very carefully and not too literally.

I have written about this issue here in discussing the classic,everyone-must-read, article from the 2004 Milbank Quarterly by Kravitz, Duan and Braslow entitled "Evidence-based Medicine;Heterogeneity of treatment effect and the trouble with averages". It can be found here.

The basic fact is that everyone does not react the same to a given medication.The summary statistic of a clinical trial does not reveal that there is a mixture of

"substantial benefit for some, little benefit for many and harm for a few"

In other words you cannot expect the average effect to occur in your patient.

In the Sept 12, 2007 issue of JAMA, David Kent and Rodney Hayward revisit this issue and offer what they believe to be a solution to the problem. ( "Limitations of Applying summary results of clinical trials to individual patients-The need for risk
stratification",JAMA vol 298, no 10, pg 1209.

They propose that the answer is risk-stratified analysis which will greatly increase the power of the detecting the critical differences in treatment effect but ,the authors suggest is rarely used. I am not sure I have ever seen it used in a major clinical trial publication.This risk stratification is supposed to be a definite improvement over the simple sub-group analysis which typically takes one characteristics( e.g. gender or some co-mobidity) at a time,the problems of which are well known including lack of power and increased likelihood of false positive associations.

Had the authors treated us to a few good examples it would have been more convincing;the one graphical illustration of how this method improves things probably will not turn the heads of many readers.But it is good that folks who know more about statistics and analysis than I do are thinking seriously about improving the way we use group data to manage the patients who have a pernicious tendency to be individuals.