Patients asking their doctors if a new drug is right for them would do well to also ask for supporting evidence. Conclusions about drug safety and effectiveness in reports submitted to the FDA are sometimes changed to favor the drug in the medical literature, a new analysis finds. And nearly a quarter of submitted drug trials were never published at all, researchers report in the Nov. 25 PLoS Medicine.
Information published in journals is the most accessible to health care professionals and also drives marketing of new drugs. The new study suggests that this information is incomplete and biased, says health policy expert Lisa Bero of the University of California, San Francisco, who led the study.
An-Wen Chan, who wrote an accompanying commentary but was not involved with the work, says he does not think health care providers will be surprised to learn of suppression and inaccurate reporting of new drug information.
“These new findings confirm our previous suspicions that this is happening on a much broader systemic level. It shows that information is unavailable to those who really need it the most — the clinicians and the researchers,” says Chan, of the Mayo Clinic in Rochester, Minn. “If we take the view that research on humans is ethical, is allowed based on an assumption of public good, then all clinical trial information should be publicly available.”
Drug manufacturers are required to submit all their studies to the U.S. Food and Drug Administration as part of new drug applications. That’s the last step in drug development, following testing on animals, trials with healthy people, and larger trials with sick and healthy people. Ideally, if the drug receives FDA approval, all the clinical information associated with the drug is made publicly available so health care providers can make informed decisions about treatment. This is typically done by publishing in the scientific literature.
The new analysis examined 164 trials for 33 new drugs that were approved by the FDA from January of 2001 to December 2002. By June 2007, 22 percent of the trials were either published only in a partial form — as an abstract, or part of a pooled publication — or were not published at all. The unpublished trials were predominantly those with unfavorable results, the researchers report.
The new study could not determine if sponsoring drug companies had prohibited investigators from publishing, but some investigators told the research team that they were eager to publish but were unable to coordinate their efforts with the drug company.
Development of one drug can require several trials. But among the drugs for which findings were published in the scientific literature, only 52 percent disclosed results from every trial.
Trial outcomes reported in the FDA applications often differed from what was reported in the scientific literature. Outcomes are predefined measures that indicate, for example, whether the new drug is more effective than existing treatments, or whether a drug has other effects on a patients’ health. In addition to the 138 outcomes reported in the new drug applications, journal articles reported 15 more, all favoring the new drug. Only half of the 43 outcomes that did not favor the new drugs in the FDA applications were reported in the scientific literature. Nine conclusions were actually changed to favor new drugs, the researchers report.
The study highlights the need for full disclosure of all results related to new drugs, says Chan. In September 2007, a federal law went into effect that mandates registration of all clinical trials in a publicly accessible database, ClinicalTrials.gov, run by the National Institutes of Health. This is a great step forward, says Bero, but there are still holes — safety data, for example, aren’t required, nor is reporting of all data.
Others have suggested that the review boards, required for institutions that conduct drug research with federal funding, could insist that results be published in full, says John Scoggins of the Fred Hutchinson Cancer Research Center in Seattle.
“The novelty of this article isn’t that it reveals publication bias — it’s just been hard to find the data to prove it,” Scoggins says. “The evidence is just now trickling in of just how bad it is.”