Wednesday, December 9, 2009

BMJ: Medical Research is Broken

A set of articles posted yesterday on the BMJ website seems to have attracted next to no attention among the mainstream media, but raises a set of serious concerns about the integrity of medical research. Probably the most convenient entry point is via this editorial (subscription needed to access full text):

The key article in terms of the origin of the dispute is:
--which is an update of a Cochrane Collaboration review of the usefulness of oseltamivir (Tamiflu) for the prevention and treatment of influenza. The review (by Tom Jefferson and colleagues) differs from a 2005 review, that had viewed the drug as helpful for preventing serious complications of influenza. A Japanese investigator had alertly noted that the positive conclusions were apparently based on a single meta-analysis, but that the component studies on which the meta-analysis was based were not publicly available for review. So he questioned the earlier findings, and the Cochrane investigators decided that the only way they could respond to the question was by fully reconstructing the earlier meta-analysis.

Trying to summarize a very complex story (so complex that BMJ ended up teaming with Channel 4 News in the investigation), Jefferson et al. went back both to the company that sponsored all the studies (Roche, maker of Tamiflu) and to the named investigators who authored the reports of that research. The investigators did not have the original data and referred Jefferson et al. back to Roche. At least one purported investigator of a study said he never conducted that particular study. Roche first refused to provide any data, then provided partial summary data, and finally said they could not provide full data because a rival meta-analysis group was reviewing all the data. (Roche provided its side of the debate and it is printed in the same collection of BMJ articles.)

In the end, Jefferson et al. had to conclude that there are no available data to confirm the value of oseltamivir in preventing the complications of flu. Notice that governments around the world have spent billions of dollars stockpiling oseltamivir specifically under the impression that it is proven to prevent serious flu complications, and so represents a major public health response to pandemic influenza. (In yet another article, the BMJ concludes that observational studies supplied by Roche to supplement the published clinical trials also fail to demonstrate any more than a very small effect of the drug on preventing flu complications.) It now turns out that there is no well-established scientific basis for having spent all this money and having made this supposed investment in a public health response.

Fiona Godlee and Mike Clarke, the BMJ editorialists, are careful to note, "In being less then forthcoming with the raw data, Roche has done nothing wrong by current standards and even less by standards of 10 years ago. It has done exactly what the current system allows."

That leads them to the same conclusion that I proposed in HOOKED now nearly 3 years ago-- that we somehow need to unlink the machinery of medical research from commercial sponsorship of drug trials by individual drug companies, paying grants directly to individual investigators.

The BMJ series tends to support the comments of Roy Poses over at the Health Care Renewal blog (see, commenting in turn on an article published in last week's BMJ by Graf et al, proposing new standards for "good publication practice" on behalf the medical communications industry (the nice folks who bring you ghostwriting). Dr. Poses suggests that these guidelines provide a telling perspective on how the industry views research. From a typical academic-medical standpoint, a research study has a principal investigator or PI. The PI "owns" and accepts full accountability for the study. She gets the grant to fund the study; she oversees the conduct of the study and the analysis of the data; and she is chief author of the eventual publication. The buck stops with the PI. From an industry standpoint, no one owns and is accountable for the study. (The drug company may think it owns it but, as we saw with Roche, seems to run the other way when accountability is called for.) Who happens to be managing it at any point in time is a hireling who can be replaced at any instant, and the hireling who, say, authors the eventual paper may have no connection with the hireling who managed the study data collection, did the statistical analysis, or whatever. These are simply functions on an assembly line. What that means for the integrity of scientific research is something we can discuss.

Addendum 12/10: I noted in the original post that the mainstream media in the US did not seem to glom onto this story. I have since seen that AP and Yahoo did pick it up:
The news story adds the info that the WHO disputes the BMJ and says that it has a lot of faith in Tamiflu's value. There's a bit of conflict of interest here as the WHO has been one of the great promoters of the drug and has been behind a number of governments deciding to buy huge quantities for their national stockpiles. You can see how it would now be very embarrassing for them to have to admit that maybe the drug was oversold. If the WHO is so convinced that Tamiflu works, on what basis are they convinced; and why were the Cochrane reviewers unable to locate that evidence?


james gaulte said...

You are correct,the integrity of medical research is,at best,questionable.Certainly the well documented bad and at times egregious behavior of drug companies and their co-conspirators is a major element in the demise of integrity.I believe a issue of at least, parallel interest is the lack of transparency of meta-analyses.What goes on behind the curtain of a MA is a mystery to most physicians so that when the increasingly common occurrence of dueling meta-analysis emerges few know what to believe.Even in the issue of Tamiflu et al there is the issue of one MA seemingly contradicting another and still another in the wings.Besides being very suspicious about drug company sponsored studies I am becoming increasing suspicious about meta-analysis and particularly those perhaps leading to policy decisions.At best I can tell a "new" MA played some (maybe large) role in the revised guidelines for breast cancer screening from the USPSTF.I'll bet even money that we will be seeing another MA on that topic reaching a different conclusion in the not too distant future.I am not necessarily disagreeing with the USPSTF but I have no tools to readily and confidently assess a MA the belief in which for many of us may be a act of "faith-based"statistics.Dr .Steven Goodman has had some great comments on that general issue pointing out that a MA is basically an observational study with the entities being clinical trials and it is difficult to judge the rules for exclusion of trials and the outcome of the MA may hinge upon the choice of trials included for analysis.

dearieme said...

"All medical research is rubbish" is a better approximation to the truth than almost all medical research.