Rosiglitazone and the Case for Safety Over Certainty

Approximately 10 years ago, the thiazolidinediones rosiglitazone and pioglitazone were introduced for the treatment of type 2 diabetes. Like their forerunner troglitazone, which was removed from the market following reports of hepatotoxicity, these drugs act on the gamma subtype of peroxisome proliferator-activated receptors (PPAR-γ) in the cell nucleus, resulting in heightened insulin sensitivity and improved glycemic control.¹ Because insulin resistance is a common feature of type 2 diabetes, the biological effects of thiazolidinediones made these drugs appealing to patients with diabetes and to their physicians who were looking for yet another way to avoid the need for insulin. Within a few years, both drugs became multibillion-dollar products despite no direct evidence that they actually prevented the complications of diabetes.

As the popularity of rosiglitazone and pioglitazone increased, reports surfaced of peripheral edema and congestive heart failure during treatment.² It quickly became apparent that these medications could cause both conditions, which are now thought to result from activation of PPAR-γ in the distal nephron, leading to increased reabsorption of sodium and water.³ The lesson here is that pharmacological tinkering with a nuclear receptor is likely to have consequences anywhere that receptor is expressed.

In May 2007, safety concerns regarding the thiazolidinediones attracted widespread attention with the publication of a meta-analysis suggesting that, compared with other treatments for diabetes, rosiglitazone was associated with a 43% higher risk of myocardial infarction (P = .03) and a 64% higher risk of cardiovascular death (P = .06).⁴ Several other meta-analyses involving rosiglitazone followed, and while their methods and conclusions varied, these reports provided a relatively consistent message that rosiglitazone might indeed increase the risk of myocardial ischemic events, albeit with an inconsistent message regarding mortality.⁵ In contrast, a meta-analysis of 19 trials involving pioglitazone suggested that even though the drug appeared to increase the risk of heart failure, it might reduce the risk of myocardial infarction, stroke, or death.⁶

Along with these meta-analyses, several large-scale pharmacoepidemiologic investigations have used health care databases to provide “real-world” data on the safety of the thiazolidinediones.⁵ While some of these studies were more rigorous than others and several did not specifically focus on the comparative safety of rosiglitazone and pioglitazone, studies in which these drugs were compared consistently showed that rosiglitazone was associated with greater risk than pioglitazone, or at best, the risks associated with these drugs were not statistically different.

In this issue of JAMA, Graham and colleagues⁷ report the results of a large cohort study examining the risk of cardiovascular events in 227 571 patients 65 years or older who were treated with rosiglitazone or pioglitazone. The authors found that, compared with pioglitazone, rosiglitazone was associated with an increased risk of adverse cardiovascular events, including heart failure and death. While the overall findings are not novel, this study is large, rigorously conducted, and exceptionally timely.

The report by Graham et al⁷ has limitations, as the authors note. Like all observational studies, because treatment assignment (in this instance, rosiglitazone or pioglitazone) was not randomized, the findings may reflect unrecognized biases or confounding—potential threats to validity that can be reliably mitigated only by randomization. However, a countervailing observation is that such studies can, when conducted carefully and with respect for their limitations, offer powerful insights into the real-world consequences of drug therapy. This is something conventional clinical trials cannot do.

With that background, several aspects of the report by Graham et al merit emphasis. First, the results are supported by a certain measure of biological plausibility. Compared with pioglitazone, rosiglitazone exhibits less favorable effects on blood lipid levels⁸ and is also a more potent PPAR-γ agonist⁹ ; consequently, the notion that rosiglitazone might impart a greater risk of adverse events than pioglitazone is not far-fetched.

Second, the findings of an increased risk of heart failure and death among patients treated with rosiglitazone are consistent with similar studies from other populations.^{10 - 11} Although some observational studies have found no difference between the 2 drugs, no study to date has suggested that rosiglitazone might actually be safer than pioglitazone.

Third, rosiglitazone and pioglitazone did not differ with respect to risk of myocardial infarction, another observation made previously.^{10 - 12} The significance of this seemingly unexpected finding is easy to miss. If patients treated with rosiglitazone were systematically “sicker” than those treated with pioglitazone (the description of patients suggests otherwise), recipients of poorer care, or destined for some other reason to have worse outcomes than those treated with pioglitazone, they also should have had an increased risk of myocardial infarction. However, that they did not (adjusted hazard ratio, 1.06 [95% confidence interval, 0.95-1.17]) provides considerable reassurance regarding concerns of bias and confounding for the study en bloc. To be clear, this finding should not be construed to mean that rosiglitazone does not increase the risk of myocardial infarction—only that it is associated with no greater risk than pioglitazone.

Meta-analyses and observational studies are rarely definitive. Clinical trials sometimes are, but the only major trial examining “hard” outcomes with rosiglitazone failed to show a benefit over conventional treatment,¹³ whereas the major trial of pioglitazone showed benefit only in a secondary analysis.¹⁴ In an effort to definitively address some of the uncertainties surrounding the safety of these thiazolidinediones, the US Food and Drug Administration (FDA) compelled GlaxoSmithKline (GSK), the manufacturer of rosiglitazone (Avandia), to undertake a large, multicenter controlled clinical trial. The Thiazolidinedione Intervention With Vitamin D Evaluation (TIDE) trial commenced in May 2009, with an anticipated enrollment of 16 000 patients and completion targeted for 2015. A major objective of this trial is to compare rosiglitazone and pioglitazone with regard to cardiovascular death, myocardial infarction, or stroke. The trial has been criticized by some who perceive it to lack equipoise and view it largely as a test of the dangers of one drug over another.^{15 - 16} Moreover, negative media accounts regarding rosiglitazone seem to have hampered recruitment.¹⁷ The number of investigational sites has nearly tripled in the past 3 months, now standing at 247 sites, with the bulk of the expansion occurring in South America, India, Pakistan, and Eastern Europe.¹⁸ Nearly half of participating sites are not yet recruiting patients, in some instances because investigators are awaiting the result of an impending review of rosiglitazone's safety by a panel convened by the FDA.

The report by Graham et al⁷ will undoubtedly contribute to the FDA deliberations, which are likely to conclude with one of two possible courses of action concerning rosiglitazone. One option is to recommend removal of rosiglitazone from the US market and the termination of the TIDE trial. Under this scenario, GSK loses, but clinicians and patients still have pioglitazone as an option, and public safety is prioritized over the desire for certainty about the safety of rosiglitazone relative to pioglitazone. A second option is to do nothing for now other than await the results of the TIDE trial. If the FDA elects this course of action, regulators around the world are likely to follow their lead, and millions of patients will continue to receive rosiglitazone (assuming GSK continues to market this drug). Under this scenario, the desire for certainty trumps safety, patients may lose, and an ethically questionable trial will continue to seek participants who, it seems, may not fully appreciate the potential risks of participation.

The epilogue of the rosiglitazone story has yet to be written, but a few observations can now be made with confidence. First, there is no direct evidence that rosiglitazone prevents vascular events in patients with diabetes. Second, converging lines of evidence suggest that rosiglitazone is less safe than pioglitazone, whereas no data suggest that the converse might be true. Third, because the evidence to date is not conclusive, differing views have emerged on how to proceed in the face of uncertainty. A consensus panel of the American Heart Association and the American College of Cardiology Foundation has called for more controlled clinical trials,⁵ whereas the American Diabetes Association and its European counterpart have advised against the use of rosiglitazone.¹⁹ The latter view incorporates a simple fact that has frequently gone overlooked: rosiglitazone confers no therapeutic advantage over pioglitazone. Whether rosiglitazone and pioglitazone really do have different cardiovascular safety profiles is an intriguing question but one with a misplaced focus. Accumulating concerns about rosiglitazone make it difficult to advance a cogent argument regarding why, exactly, a patient might want to receive the drug or why a physician would choose to prescribe it when there is an available and quite possibly safer alternative.