You can tell a big thing from a little thing but a litle thing from nothing at all is really hard

I have obsessed about the issue of small relative risks before and quoted some folks who know a lot more about it that I do. One of those folks' quote is paraphrased in the title to this piece. Michale Thun,VP of epidemiology and Surveillance at the American Cancer Society said:

With epidemiology you can tell a little thing from a big thing.What's very hard to do it to tell a little thing from nothing at all.

If you don't think it is hard, plow through the material presented to the FDA panel and their considerations in regard to the issue of Avandia cardiovascular safety.

Robert Temple ( of the FDA-at least at the time he was quoted) said in regard to the topic of "small" relative risks,

My basic rule is if the relative risk isn't at least 3 or 4 , forget it.

The relative risk in the Avandia, NEJM Nissen article was 1.43. That was not ignored and not forgotten. So is this purported increase risk in cardiovascular events that is attributed to Avandia something or nothing at all ? The experts at the recent FDA hearing couldn't really agree on that. Some thought the evidence indicated an increased risk really exists while others believed that the data was not sufficient to reach that conclusion.

The fact that something is a close call does not mean we can suspend action. Even if something is close sometimes someone has to make the call. In the course of human events this happens all the time. Human brains seemed to have evolved to make decisions with inadequate data. Issues cannot be debated ad infinitum, at some point a call is made and we move on. The panel made a call: add a warning label about c-v risk and do not remove it from the market. The action of many individual docs and many individual patients will likely to a large degree remove it anyway. I can think of no reason for a physician to add Avandia to a diabetic patient's treatment program in light of recent events even though we may never know with apodictic certainty if there is an increased risk of heart attack or not. I wouldn't prescribe it because people are worried about it and there is a drug in the same family that could be used not because I know that the Peto method-used by Nissen-was the statistically correct thing to do and his analysis lead to the "correct" answer. ( Heck, I have no clue as to what the Peto method really is.)

The Nissen relative risk may not have been "big" in the sense of a big RR such as 7 or 8 designating the risk of cigarette smoking and lung cancer and we may never be able to determine what the risk situation really is but it was big enough to kick off a argumentative storm that likely will lead to the demise of Avandia with the statistical arguments put forth by the statistical team at Glaxo falling on the deaf ears of many doctors and patients.