CardioExchange Archive

We welcome Sanjay Kaul, MD, lead author of a recent American Heart Association/American College of Cardiology science advisory about the cardiovascular risks of thiazolidinedione drugs, to answer our questions about rosiglitazone. We encourage you to ask yours.

You and your coauthors call the evidence on the cardiovascular risks of rosiglitazone “inconclusive.” When you prescribe a thiazolidinedione, which agent do you choose and why?

For diabetic patients with known ischemic heart disease — particularly older, high-risk patients who take nitrates, ACE inhibitors, or insulin — I prefer pioglitazone over rosiglitazone. Indeed, the FDA has warned about the risk for myocardial ischemia from rosiglitazone (but not pioglitazone) in some of these high-risk patients. That said, for a diabetic patient without high-risk characteristics and with well-controlled blood sugar on rosiglitazone, I see no compelling reason to switch to pioglitazone. However, I don’t object if such a patient wants to switch to pioglitazone or any other antidiabetic drug. 
 
Notably, both rosiglitazone and pioglitazone are contraindicated in patients with NYHA class III or IV congestive heart failure (CHF). Given that the drugs are associated with weight gain and fluid retention, caution is warranted in all CHF patients and in patients with signs and symptoms of CHF.
 
What is your threshold for accepting that the cardiovascular risk is real — for example, if an estimate showed a clinically important risk increase of 20% to 30%?
 
Let me answer this important question using rosiglitazone.
 
First, we have the RECORD trial, an open-label, industry-funded study in which about 4500 patients with inadequate glucose control while taking metformin or a sulfonylurea were randomized to receive either both drugs (controls) or add-on rosiglitazone. Compared with controls, rosiglitazone recipients showed no significant difference in the primary endpoint of time to first cardiovascular hospitalization or cardiovascular death (ranging from a 15% risk reduction to a 16% risk increase) or for myocardial infarction (ranging from a 20% risk reduction to a 63% risk increase). In a prespecified subgroup of patients with preexisting ischemic heart disease, the lack of a primary-endpoint difference between rosiglitazone and the control regimen persisted, although subgroup analyses are inherently limited. Indeed, this entire trial has well-known limitations, so its findings are inconclusive.
 
Next, we have evidence from an FDA meta-analysis that focused only on diabetes trials; assessed robust outcomes of cardiovascular death, MI, or stroke; and used stringent analytical and statistical methods. With regard to the risk for CV death, MI, or stroke, rosiglitazone had a nonsignificant odds ratio of 1.15 (ranging from a 20% risk reduction to a 60% risk increase, compared with control). Thus, a 15% elevated risk is the most plausible point estimate we have.
 
If we use a probabilistic approach based on Bayes’ theorem, the likelihood of a 20% to 30% clinically important risk associated with rosiglitazone is less than 50% — lower than the “preponderance of the evidence” threshold used in civil court trials and nowhere near the 95% level of certainty that one would want. Ultimately, risk-benefit assessments are something of an art practiced by a clinician as he or she carefully considers the scientific evidence along with the clinical profile and the wishes of an individual patient.
 
What should we do when years have passed since FDA approval of a drug but we still don’t know whether it is safe?

In the face of insufficient evidence, patients, providers, and insurers often find themselves at odds with one another, frequently confused, and sometimes distrustful despite the presumed good intentions of industry sponsors, investigators, and regulators. As Jerry Avorn wrote in the NEJM in 2007, “The approval, prescribing, and safety surveillance of prescription drugs involve a complicated mix of science, regulatory law, clinical judgment, business, and politics.” I hope that the quiet, dispassionate voice of science will be allowed to render verdicts through its natural but rigorous course of refutation or confirmation. Meanwhile, the Hippocratic Oath compels us as clinicians to avoid (even potentially) harmful interventions, especially if we have alternatives.

The story of rosiglitazone, like that of ezetimibe, is a cautionary tale that highlights key shortcomings inherent in drug/device development, evaluation, and approval. The FDA approved rosiglitazone more than a decade ago, and we still lack conclusive evidence about its effects on cardiovascular outcomes. I believe that all the major stakeholders bear responsibility for this evidence vacuum: drug developers, regulators, investigators, physicians, payors, patients, and the U.S. Congress.

A potential remedy is the so-called “lifecycle evaluation” of a drug. Ideally, the FDA should grant only “provisional approval” when surrogate endpoints are used (rosiglitazone was approved on the basis of glycemic control, as other diabetes drugs have been). The conditions for full approval should be timely demonstration of efficacy and safety in large, well-designed clinical-outcomes trials that examine cardiovascular endpoints. Furthermore, Congress should empower the FDA to enforce postmarketing commitments. Only then can we hope to avoid controversies like those rosiglitazone has brought us.