A new analysis of a dozen studies in The New England Journal of Medicine examined possible new uses for the epilepsy drug and found the published results were often fudged, indicating the pill worked better than internal company documents showed. In essence, when the primary finding of a company-funded study wasn’t favorable, the study authors usually buried the result and found something else positive to highlight, without disclosing the change, the Associated Press notes.

The authors identified 20 trials for which internal documents were available from Pfizer and Parke-Davis, and of these trials, 12 were published. For eight of them, the primary outcome defined in the publication differed from what was described in the protocol. The primary outcome was changed in five of eight published trials for which statistically significant differences favoring Neurontin were reported. Of 21 primary outcomes described in the protocols of the published trials, six weren’t reported at all and four were reported as secondary outcomes. Of 28 primary outcomes described in the published reports, 12 were newly introduced (here is the abstract).

The documents used in the review were obtained by lawyers suing Pfizer for refunds on prescriptions paid for by insurers and consumers, the AP notes. The lawyers claim Pfizer concealed evidence that Neurontin didn’t work for those unapproved uses, including nerve pain, migraines and bipolar disorder. One author, Kay Dickersin, is an expert witness for the plaintiffs, while another, Swaroop Vedule, has received fees from the lawyers.

Pfizer disputes the report’s conclusions, saying it never “attempted to mislead the medical community about the effectiveness” of the drug for certain uses. “We believe the review suffers from significant bias, insufficient data, poor methodology, and cannot pass the threshold of credible scientific research,” the drugmaker tells the AP. The analysis comes two months after Pfizer was fined a record $2.3 billion, including an unprecedented $1.2 billion criminal fine, for illegally marketing other drugs.

Sid Wolfe, who heads health research at Public Citizen and is a memberof the FDA’s drug safety advisory committee, tells the AP that the analysis offers the first comprehensive look “at studies in which a company and people working for it so maliciously manipulated the data to make a drug look more effective than it actually was. In every instance, the published article made the drug look better than it would have. This results in harm.”

The authors cited limitations, such as not knowing who made changes in reported outcomes. “We cannot be certain that selective reporting was a decision made by employees of Pfizer and Parke-Davis, since the authors of the published reports included nonemployees,” the researchers wrote.

Art Caplan, director of the University of Pennsylvania’s Center for Bioethics, tells the AP that the report is “one of the most ethically disturbing papers I’ve read in some time” and “an indication that people have been playing fast and loose with studies,” particularly industry studies.

The AP reminds us that some experts believe most Neurontin sales were for off-label uses, such as those in the reviewed studies. Sales peaked at $2.7 billion in 2004, when Pfizer paid $430 million in government fines to settle allegations it improperly marketed the epilepsy drug for unapproved uses. By last year, Neurontin sales fell to $387 million due to cheaper generic versions.