Friday, May 27, 2005

statins and colon cancer-still another case-control study will not answer the question

Here is a case control study that I really would like to believe but I am skeptical. In the May 26, 2005 issue of the NEJM a 47% reduction in colon cancer is reported by JN Poynter et al from Israel.I have been taken Pravachol for 5+ years and I have a strong family history of colon cancer so I would really like it if my pill were multi-tasking prevention wise. The related editorial give us some context. There have been other case control studies- which no one should be surprised- have given somewhat conflicting results. (The case control studies regarding the effect on coronary artery events and hormone replacement therapy were also some what conflicted and we know how that worked out)
We are also told that secondary analysis from of three large meta-analyses of prospective statin trials designed to detect reduction in coronary events failed to show any colon cancer risk reduction. And then a comment is about how those studies might have not been adequately powered to detect a cancer risk difference. A Canadian case control showed a 28% reduction in colon cancer risk (surprisingly in only 2.7 years of follow-up causing one to wonder if some patients already had early colon cancer when the drugs were started) while another study showed an increase in colon as well as prostate and bladder cancer.
This exemplifies the sort of thing you see all the time when you look a set of articles that use coarse grain instruments (such as case controls and meta-analysis) to determine associations and causal links. This is what you have here. Some studies show decreased risks, some increased risk and some no difference. The authors of the article in their introduction say " to clarify the association " we evaluated data in... a case-control study. In light of the context of multiple conflicting case-control studies, how reasonable is it to expect that another case-control will ever clarify anything? Case control studies and multiple logistic regression analysis, according to Harvard epidemiologist, Kenneth J. Rothman, are two of the most important developments in epidemiology in the last fifty years. Case control studies are said to be efficacious and efficient and he characterizes it as the central tool of modern epidemiology. How many times have we heard the phrase " hypothesis generating" in regard to case control studies? Perhaps someone more skilled in epidemiology than I can explain what is the utility of doing another case control in a situation in which several earlier case control studies have resulted in contradictory findings in regard to a particular association. It reminds me of what one of our endocrinology attendings was fond of saying" if one endocrine test is equivocal, all the other tests you do to clarify it will also be equivocal." The editorialist opines that we won't really know more until a RCT is done, but that comment was just as cogent before this publication appeared.

No comments: