Review finds ‘publication bias’ endemic in medical research

Before a new drug is unleashed onto the market, it must first be determined that it is, broadly speaking, beneficial and safe. The drug’s sponsor, usually a drug company, submit relevant studies to an appropriate body made up of doctors, scientists, statisticians, toxicologists and the like who assess the evidence and then decide whether to grant the drug a license or not. In the USA, the organisation responsible for this process is the FDA (Food and Drug Administration).

Should it grant a license, the FDA publishes a ‘Summary Basis of Approval’, which as it names suggests, summarises the evidence which the FDA used to judge a drug as worthy of a license. However, these summaries contain only selected results from the trials submitted to the FDA, and recently scientists have questioned whether such selected publication data might give a somewhat skewed picture of the value and safety of a drug.

In order to investigate this matter three researchers from the University of California San Francisco in the USA looked at the data supporting 90 drugs licensed by the FDA between 1999 and 2000. In all, the researchers identified 909 trials that had been submitted in support of these drugs. Of these, though, a mere 43 per had been published in medical journals over the next 5 and half to 8 and a half years. Perhaps not surprisingly, those studies with statistically significant results were more than twice as likely to be published as those that did not produce statistically significant results.

Broadly, what this appears to be an example of is ‘publication bias’: a situation where (usually) drug companies publish the data which puts their product in a favourable light, but may neglect to do so for less favourable data. And the reason that this matters is because it can give an inaccurate picture of how effective a drug really is. It can ultimately lead doctors to rush to prescribe a new drug that is not materially better than an existing (and often cheaper) alternative.

Even worse, it might cause doctors to prescribe drugs that are more toxic and hazardous than existing therapies. The authors of the paper highlight a string of recent controversies in which it appears drug companies suppressed the safety risks of drugs such as rosiglitazone (for diabetes), paroxetine (and antidepressant) and rofecoxib (a painkiller).

On the plus side, a new law was passed recently in the USA that mandates that all FDA approved drugs and devises must have the basic results of all supporting trials published by the National Institutes of Health. This appears to be a step in the right direction. However, the authors of the review ask whether this law might actually increase rather than decrease publication bias. Specifically, they ask: Might sponsors feel less compelled to publish equivocal trials because the basic results will already be in the public domain? They also question whether the time pressures of publication soon after approval focus drug company efforts even more on submitting positive trials and trials of greatest interest to the journals.

The review highlights the problem of publication bias in medical research, but it also serves as some sort of benchmark against which we can judge what effect the new legislation regarding publication of trial data has. Let’s hope that the new legislation leads to more publication of trial data, not less, and as a result allows stakeholders to form truly informed opinions regarding the effectiveness and safety of newly licensed drugs.

References:

Kee K, et al. Publication of Clinical Trials Supporting Successful New Drug Applications: A Literature Analysis. PLoS Med 5(9): e191 doi:10.1371/journal.pmed.0050191

No comments yet.

Leave a Reply