The September 2006 issue of the Annals of Internal Medicine published an article (Do Oscar Winners Live Longer than less Successful Peers? A reanalysis of the evidence.Sylvestre,Huszti,and Hanley.Annals Internal Medicine,vol.145, no. 5, p. 361) that contradicted an earlier Annals'article that claimed Oscar winners live longer than their fellow actors.The abstract is available online. Why does a respected medical journal bother itself with this issue anyway?
Subscription is required for the full text of the article and also for the important letter from two of the Annals editors, Steven Goodman, an epidemiologist whose work I have found very impressive and useful and Harold Sox the editor. The original paper was published 5 years ago and found-by one analysis-a 3.9 year increase in life expectancy in Oscar winners. Here the operative words appear to be "by one analysis". The results vary significantly based on the type of analysis used and the epidemiologists-statisticians are not in agreement on how to analyse the data.
The issue here, as explained by Goodman and Sox is "when to start the clock" in a situation wherein there is a sudden change in risk due some event. Such events could include starting a treatment and in that instance there would be considerable medical interest, certaintly more than Oscar winner' longevity. Starting the clock at the wrong time can trigger something called "the immortal time bias" ( also known as "time zero" problem) something I have posted about before in the context of COPD treatment.
One study demonstrated that inhaled corticosteroids (ICS) treatment increased survival in COPD but another analysis indicated that the survival advantage was not real but an example of the immortal time bias although the original authors denied that was the case. Just as in the Oscar winner longevity controversy there was disagreement if there was or was not a bias in the ICS data due to disagreements about the clock starting issue.
The Annals editorialists state their explanatory comments are published largely because "the analytic methods at issue apply to many health care research questions ."Apparently the statistical issue of how to handle the zero time issue is not settled as Goodman and Sox invite other members of the statistical fraternity to "take up the challenge of determining the most appropriate way to measure the effect of winning an Oscar and the statistical uncertainty around the results." Of course, their concern is not really with Oscar winners's longevity but with the application of these techniques to more therapeutically relevant medical studies.
How to analyze the evidence in evidence based medicine is still to a large degree a work in progress.