In the July 13, 2006 issue of NEJM, S.W. Lagakos, a statistical consultant for the journal, writes about time-to-event analysis for long term treatments in the context of the APPROVe trial. This was the trial to see if Vioxx could prevent the recurrence of colon polyps and which demonstrated an increased risk of cardiovascular events in the treated group and which in all likelihood both Merck the NEJM wishes never happened.
The article is found in the perspective section and labeled "statistics and medicine".Dr. Lagakos discussed some of the issues involved in a time-to-event analysis such as the appropriate period of follow-up and the assumptions of the proportional hazards technique and the log rank test and monotonicity.If the purpose of this article was to inform readers so that they could read studies with these tests in an at least semi-informed manner,I am not sure he succeeded. I read it twice and still do not have a good sense of the statistical issues.However,I doubt a didactic motivation was operative.
He does make it clear that-in his opinion-that one cannot conclude that rofecoxib poses no increased cardio-vascular risk in less than 18 months. This 18 month threshold has become an issue in some of the many law suits pending in the courts.
The throngs of plaintiff attorneys involved in Vioxx cases could not be more pleased.The NEJM seems to be still trying to compensate for publishing the APPROVe trial in the first place.