Comparison of internal and public reports about Pfizer’s drug Neurontin reveals many discrepancies

By Rachel Ehrenberg

Web edition: January 29, 2013

Editor's note: This story was updated on January 31 with comment from Pfizer.

A rare peek into drug company documents reveals troubling differences between publicly available information and materials the company holds close to its chest. In comparing public and private descriptions of drug trials conducted by pharmaceutical giant Pfizer, researchers discovered discrepancies including changes in the number of study participants and inconsistent definitions of protocols and analyses.  

The researchers, led by Kay Dickersin, director of the Center for Clinical Trials at the Johns Hopkins Bloomberg School of Public Health, gained access to internal Pfizer reports after a lawsuit made them available. Dickersin and her colleagues compared the internal documents with 10 publications in peer-reviewed journals about randomized trials of Pfizer’s anti-epilepsy drug gabapentin (brand name Neurontin) that tested its effectiveness for treating other disorders. The results, the researchers say, suggest that the published trials were biased and misleading, even though they read as if standard protocols were followed. That lack of transparency could mean that clinicians prescribe drugs based on incomplete or incorrect information.

"We could see all of the biases right in front of us all at once,” says Dickersin, who was an expert witness in the suit, which was brought by a health insurer against Pfizer. Pfizer lost the case in 2010, and a judge ruled it should pay $142 million in damages for violating federal racketeering laws in promoting Neurontin for treating migraines and bipolar disorder.

Pfizer had in 2004 settled a case and paid $430 million in civil fines and criminal penalties for promoting Neurontin for unapproved use.

The study's results, published January 29 in PLOS Medicine, show that publications about drug trials don’t always reflect the research that was conducted, says Lisa Bero of the University of California, San Francisco, an expert in methods to assess bias in scientific publishing “We know that entire studies don’t get published and that what does get published is more likely to make a drug look favorable,” she says. “This adds another layer.”

In three of the 10 trials, the numbers of study participants in the published results didn’t match those in the internal documents. In one case, data from 40 percent of the participants were not included in the published trial. Dickersin and her colleagues also tried to directly compare several other aspects of the studies. But they found so many differences in definitions and in the analyses and protocols that the comparisons turned out to be difficult, she says.

“When we tried to draw a flow chart of who dropped out [of a trial], who stayed in — well, we couldn’t do it,” she says. “You can’t even judge if they did the right thing if you can’t figure out what they did.”

Pfizer did not immediately respond to requests for comment. The company outlined its policies for making clinical trial data public in a statement provided to Science News on January 30, concluding that the company reports on studies "in an objective, accurate, balanced and complete manner."

The Johns Hopkins analysis highlights the need for standard definitions and protocols and greater transparency in reporting clinical trials, says Bero, a longtime advocate of making raw data from clinical trials publicly available. “You’re kind of held hostage to the paper that you are reading,” she says.