Observations Open Data

Rosiglitazone: a case of regulatory hubris

BMJ 2013; 347 doi: http://dx.doi.org/10.1136/bmj.f7428 (Published 11 December 2013) Cite this as: BMJ 2013;347:f7428
  1. Steven E Nissen
  1. 1chairman of cardiovascular medicine, Cleveland Clinic, Cleveland, Ohio
  1. nissens{at}ccf.org

The FDA’s defensiveness over its decisions means further drug safety disasters may occur

The recent announcement by the US Food and Drug Administration (FDA) lifting restrictions on the prescription of rosiglitazone is the latest twist in a tortured saga dating back more than a decade.1 Rosiglitazone was approved in 1999 to treat type 2 diabetes, on the basis of results of small and limited duration studies designed primarily to show that this drug lowered blood glucose concentrations.2 At this time the regulatory requirements for approval of antidiabetes drugs emphasized reduction in glycemic indices as the principal basis for approval, not improvement in health outcomes.

Although the pre-approval studies were small, they showed two disturbing findings: that rosiglitazone raised concentrations of low density lipoprotein cholesterol by a mean of 18.6%; and that there was an excess number of ischemic cardiovascular events, approaching significance (hazard ratio 1.8 (95% confidence interval 0.9 to 3.6)). The US regulators approved the drug, but European authorities insisted on a post-approval clinical outcome trial, known as the RECORD (rosiglitazone evaluated for cardiovascular outcomes and regulation of glycemia in diabetes) trial.

After approval, the drug, marketed as Avandia, rapidly became a huge money maker for GlaxoSmithKline, eventually becoming the largest selling antidiabetes drug in the world. But something was clearly wrong. In 2005, at the insistence of the World Health Organization, GSK performed a meta-analysis of all 37 trials involving use of rosiglitazone, finding a hazard ratio of 1.29 (0.99 to 1.89). In 2006 the company updated the analysis, now including 42 trials and showing a hazard ratio of 1.31 (1.01 to 1.70). GSK secretly provided these analyses to the FDA, but neither the company nor the agency warned prescribers or patients of the hazard. There was complete silence.

Finally, in May 2007, suspicious about the safety of this drug, we acquired the study level data for all rosiglitazone trials. Access was incredibly fortuitous and due solely to a successful lawsuit by the state of New York after revelations that GSK had suppressed trial results showing increased suicidality among children and adolescents taking the antidepressant paroxetine. The provisions of the legal settlement required GSK to disclose the main results of all its clinical trials.

We published a meta-analysis showing a hazard ratio of 1.43 (1.03 to 1.98) for myocardial infarction and 1.64 (0.98 to 2.74) for death.3 The FDA’s reaction was surprising and disturbing. FDA officials were infuriated with me for challenging the drug’s safety, interpreting my challenge as an attack on their competence and commitment to public health. Several years later in a private meeting one of the highest ranking FDA officials, red faced with anger, said, “Why did you have to publish your findings? You could have come to us privately and expressed your concerns. Why did you deliberately embarrass us?”

The FDA’s embarrassment and hubris have been the driving force behind its subsequent regulatory approach to the drug. In 2009 the RECORD trial was published, showing no significant increase in cardiovascular events. In 2010 the FDA assembled an advisory panel to exonerate the drug, on the basis of the RECORD results, but the meeting backfired. The FDA’s deputy commissioner, a man of integrity recently appointed by the Obama administration, insisted on fair hearings. The FDA’s cardiovascular reviewer of the RECORD trial wrote a scathing analysis of the trial, pointing out that it was extraordinarily unblinded.4 Treating physicians, patients, the contract research organization, and even GSK itself were all aware which patients were receiving rosiglitazone. Even more disturbing, the referral of cases for adjudication of cardiovascular events was highly biased in favor of rosiglitazone. I was permitted to present an independent view to the panel. After two days of hearings the panel overwhelmingly supported either the removal of rosiglitazone from the market or severe restrictions on prescribing. In response, European regulators banned the drug entirely, while the FDA restricted access to rosiglitazone only as a drug of last resort.

Before the 2010 panel meeting the US Senate issued an investigation based on documents subpoenaed from GSK.5 It was revealed that GSK had obtained a copy of our meta-analysis manuscript while it was under review at the New England Journal of Medicine. Within hours of submission a reviewer at the journal had faxed a copy to GSK, which the company exploited, circulating the manuscript to more than 40 company executives. GSK statisticians prepared a comprehensive analysis of the meta-analysis that concluded, “There is no statistical reason for disregarding the findings.” In an email the head of research at GSK stated, “FDA, Nissen and GSK all come to comparable conclusions regarding increased risk for ischemic events, ranging from 30 percent to 43 percent!” Nevertheless, forewarned about publication, the company planned a public relations offensive designed to discredit the findings.

In addition to the Senate report, the New York Times obtained company documents showing that GSK had conducted a secret study in 1999 confirming that a competitor’s drug, pioglitazone, was safer.6 In an email a GSK executive wrote, “This was done for the US business, way under the radar. Per Sr. Mgmt. request, these data should never see the light of day to anyone outside of GSK.” Eventually, in 2012, the company paid out a record $3bn civil and criminal settlement to the US Department of Justice for unlawful promotion of several drugs, including rosiglitazone.

But the FDA never gave up the fight to resurrect the drug and thereby clear the agency of any failure of regulatory oversight. It invited the company to submit the RECORD trial for “readjudication” by physicians at Duke University. The very idea of reanalyzing an open label trial begun 10 years earlier was ludicrous, but it served the agency’s purpose. Even more bizarrely, the FDA allowed GSK to submit for adjudication exactly the same carefully selected dossiers used for the original RECORD analysis. Of course, the result was very similar to the original RECORD publication.

Armed with a new version of the RECORD trial, the FDA convened a new advisory panel meeting.7 But the agency did not ask some of the most critical members of the 2010 panel; nor did I have the opportunity to present alternative analyses. In the end, the FDA got exactly what it sought: a vote to ease restrictions on prescribing of rosiglitazone. Luckily, only about 3000 patients currently take the drug, and the FDA’s decision will have little practical effect.

Nevertheless, the precedent is disturbing. When challenged on issues of drug safety the FDA reacts with hostility, defending its bureaucracy at all costs from any implication of poor decision making. During the most progressive administration in decades the FDA operates with little or no oversight, more concerned about its image than the health of the public. Thus rosiglitazone may not be the last drug safety disaster. Stay tuned.


Cite this as: BMJ 2013;347:f7428


  • bmj.com Feature (with timeline): Rosiglitazone: what went wrong? (BMJ 2010;341:c4848, doi:10.1136/bmj.c4848)