Preposterous Universe

 
Addresssing Big Pharma publication bias

Bravo, Eliot Spitzer:

In a settlement that the New York State attorney general said would transform the drug industry, GlaxoSmithKline agreed today to post on its Web site the results of all clinical trials involving its drugs.
[...]
Mr. Spitzer filed suit in June against GlaxoSmithKline, contending that it committed fraud by publicizing the results of only one of five trials studying the effect of its huge-selling antidepressant, Paxil, in children. That single study showed mixed results. The others not only failed to show any benefit for the drug in children but demonstrated that children taking Paxil were more likely to become suicidal than those taking a placebo.[NYT permalink]

Underreporting of unfavorable trials is a threat to in clinical research. All scientists have some incentive to publish favorable results over unfavorable ones, but the potential for systemic bias is magnified when big companies insert themselves into scientific research. A single unfavorable trial of a promising new drug can send a company's shares tumbling. It's not just company scientists whose work is affected. Independent researchers who are becoming increasingly dependent upon research grants from pharmaceutical companies. These grants may restrict investigators from publishing unfavorable findings. Big pharma will spare no expense to create data on a potential blockbuster. Unlike most academics, these companies can afford to keep sponsoring trials until they get the results they want. Increased transparency is the only way to safeguard the scientific record against distortion on a major scale.

I'm glad to see Spitzer taking the lead on disclosure for clinical trials. This isn't just about consumer protection, academic integrity, or the flow of information in a free market. Selective publication is also undermining the credibility and comprehensiveness of large swathes of the biomedical literature. Over time, systematically skewed incentives distort the scientific record. The worry is not just that some bad drugs may come off looking good. The problem is far more serious. Corporate publication bias has the potential to distort the scientific record and thereby to impede scientific progress on a larger scale.

Philosopher Susan Haack has famously compared science to a partially completed crossword puzzle. Each answer constrains subsequent answers. Our willingness to entertain subsequent answers to depends, in part, on how confident we are in the answers we've pencilled in so. So, the effects of incorrect or poorly supported answers can't be encapsulated, they warp our entire strategy for solving the crossword puzzle.

The recent GSK settlement is small but important step towards scientific transparency. The current judgment applies only to GSK, but Spitzer hopes that other pharmaceutical companies will follow GSK's lead.

Many journals have already moved to safeguard the scholarly literature against various forms of bias. Mandatory disclosures of funding sources and conflicts of interests are a step in the right direction. The key journals should do even more to address publication bias, especially the corporately-mediated species. For example, journals should require all contributors to sign a declaration that they are free to publish all data resulting from their investigations according to their professional judgment. Prestigious academic journals should not publish anything from anyone who admits to publication restrictions on unfavorable results.

Here's an idea I've been toying with for a while. How about a public repository of irreproducible results? A kind of National Biomedical Data Cemetery... Here's how it might work. If a publicly funded scientist failed to get her study published, or decided not to publish her results, she could submit her data to a national archive (anonymously, if she wished). These records would be made available to scientists, historians, policy makers, and other scholars. Such an archive would be an invaluable resource. It might help answer some of the niggling questions about publication bias that plague the biomedical literature.

I'd love to hear readers' opinions on the desirability/legality/feasibility a Biomedical Data Cemetery.

[X-posted with Majikthise]

~ posted by Lindsay Beyerstein at 4:06 PM ~ link ~ ~ ~