Map of life expectancy at birth from Global Education Project.

Tuesday, April 29, 2008

Blood and Treasure

The takeaway lesson from this meta-analysis published on-line by JAMA is a little bit complicated, and I don't think the general media have succeeded in explaining it very well. (Thanks to C. Corax for tipping me off to the NPR report; I've also read some of the newspaper and news network web site stories.) This isn't quite like those mass murder cases where drug companies knew their products were killing people and kept it a secret. It's more of a structural problem in the way the FDA and the drug companies do research, exacerbated by greed and ass-covering, to be sure, but not primarily about that. Those are a constant that must be allowed for in the way we design our policies for studying and approving drugs and other medical products.

The story begins with concerns about the availability and safety of blood for transfusions back in the '80s, stimulated in part by the HIV epidemic and other blood-borne infections, and the difficulty of making blood available in poor countries and remote places. Various companies developed and tested blood substitutes based on hemoglobin, which as you know if you remember your high school biology is the molecule in red blood cells which carry oxygen from the lungs to the rest of the body. What you might not have learned in school is that the gaseous compound nitric oxide NO is involved in regulating the constriction of blood vessels, specifically it causes them to relax and dilate. (This has to do with the mechanism of action of those erectile dysfunction drugs, BTW.) It also inhibits blood clotting.

Hemoglobin has a strong affinity for NO as well as for oxygen, but NO doesn't cross the membrane of red blood cells, so no problem. However, free hemoglobin in the blood stream soaks up the NO, causing blood vessels to constrict, and raising the risk of blood clots. You have already figured out that this is not good -- we're talking heart attacks.

This problem was understood before anybody even tried experimenting with a hemoglobin based blood substitute. The companies tried some tricks to reduce the risk, but in almost every trial, from the very beginning, whether in trauma or surgical patients, there was an increased risk of heart attacks, and usually of death. The problem is that in any one trial, the numbers were small enough that this risk did not rise to the conventional signicance level of p < .05, in other words there was at least a 5% chance that it was due to chance. This probably doesn't seem very reassuring, and I agree that this is already a problem, but that's how we're supposed to think. If it isn't "statistically significant," it doesn't exist, even though it probably does.

Anyway, the key issue here is that if you put together the results from two or more trials, you would see that the risk was indeed significant, and quite substantial, actually. The pooled results of all the trials the reviewers were able to use show the risk of MI to be 2.77 times as high in people who got the blood substitutes as in controls, who got usual treatment (blood, plasma or saline). The risk for death was 1.27 times as high.

However, and here's the kicker, most of these trials were never published. The FDA knew about most of them, but didn't consider the metanalysis in approving new trials. The information is considered to be a trade secret, so unless the companies chose to publish the data, nobody else could find out about it. (The reviewers who published the JAMA article still couldn't get access to much of the data, or they got it in compromised form. Chances are, the situation is even worse than what they found using the data they could obtain.) This meant that Institutional Review Boards didn't have access to the information. If they had, they never would have approved the studies. So, studies continued, and subjects continued to have heart attacks and to die. There are even studies going on now. Since there is not, in fact, a shortage of blood products, there isn't even a compelling need to be met by these products.

The conclusion is that clinical trials should never be secret. If a company is trying to get a drug or medical device approved, it needs to register all of its trials with the FDA, and the methods, results and raw data need to be publicly available. Congress must change the law in order for this to happen, regardless of the new president's bowling scores.

No comments: