Boston, MA - Older diabetic patients treated with rosiglitazone (Avandia, GlaxoSmithKline) are at higher risk of dying or developing heart failure than patients treated with pioglitazone (Actos, Takeda Pharmaceuticals), a new analysis of Medicare beneficiaries suggests [1]. The observational study—the largest of its kind to date—appears in the November 24, 2008 issue of the Archives of Internal Medicine and is likely to fan the ongoing controversy over the adverse-effect profile of rosiglitazone as compared with other drugs in the thiazolidinedione (TZD) class.

Dr Wolfgang C Winkelmayer (Brigham and Women's Hospital, Boston, MA), lead author on the study, told heartwire that while the analysis was observational and retrospective, patients and physicians should nonetheless consider the findings on top of previous research in making decisions about treatment. As previously reported by heartwire, a number of meta-analyses published in the past year and a half have pointed to an increased risk of MI and/or cardiovascular death with rosiglitazone.

For their study, Winkelmayer and colleagues looked at more than 28 000 diabetic patients over age 65 and followed them for a total of 29 060 person-years. The analysis was restricted to patients who stayed on the drug they'd initially been prescribed, without switching over to another agent. Over the study period—January 1, 2000 to December 31, 2005—1869 patients died.

In the study's primary analysis, which assumed that patients were exposed to the drug for just 60 days after the date of their most recently filled prescription (and adjusted for patient characteristics), diabetic patients initially treated with rosiglitazone had a 15% higher mortality rate than patients on pioglitazone. Rates of first hospitalization for congestive heart failure—a known side effect of TZDs—were 13% higher in the rosiglitazone group than in the pioglitazone group. In contrast to recent studies pointing to a risk of ischemic events with rosiglitazone, rates of MI and stroke were no different between the groups.

In a secondary analysis, which assumed patients to be exposed on a constant basis to either drug, the mortality findings for rosiglitazone were, in the authors' words, "less pronounced" for both all-cause mortality (8% higher), and time to first hospitalization for heart failure (11% higher).

In an interview with heartwire, Winkelmayer explained that the secondary analysis (the constant-exposure analysis) most closely approximates an "intention-to-treat" analysis in a randomized controlled trial, but the primary analysis in this study (the as-treated analysis) is closer to that of drug-safety research. "In drug-safety research, you actually want to know whether a patient continues to take the drug and whether they cross over, because you want to be more certain that you can attribute an outcome to the correct exposure," he explained. "You're more likely to pick up any safety problems."

But more important, he notes, "The flavor of the finding is the same whether you use one approach or the other approach. In both cases, mortality was elevated, and, in both cases, hospitalization for heart failure was elevated. So there's no discussion of whether there is a risk or not, it was just the magnitude that was different."

Winkelmayer added that he was surprised not to see an increased risk of stroke or MI in the rosiglitazone group—an increased risk of ischemic events was seen in last year's meta-analyses of randomized clinical trials. But he pointed out that, in this study, average patient age was 76, much older than that of patients included in the randomized trials. As such, patients in this Medicare cohort may have actually died from MI or stroke, meaning that their events would not have been recorded in the Medicare records, and any difference between the two drug groups may not have been adequately captured.

Winkelmayer would not comment on whether he thought the totality of data supports much stricter warnings on rosiglitazone's labeling or even its withdrawal from the market. Such decisions are up to policy makers, he emphasized.

"All I can say, very carefully, cautiously, is that our study provides yet another piece of evidence from another angle that is very much compatible with the increased risk that was found with rosiglitazone in the very high-quality meta-analyses of last year and which was absent in another high-quality meta-analysis for pioglitazone," he said. "This provides another piece of evidence: after this we can see things a little bit more clearly. But it is up to policy makers to decide when the picture is clear enough to draw any conclusions with regard to policy action."

But Winkelmayer believes the results, despite being nonrandomized, are powerful enough to be considered by doctors and their patients. Indeed, he points out, patient characteristics in the study were so closely matched at baseline that they actually resemble those of a randomized clinical trial—selection bias was likely limited, he suggested, since both drugs appeared on the market within months of each other and at the time were viewed as more or less equivalent. "The main driver of whether a patient was put on one drug or the other was probably practitioner preference, likely based on which drug rep came through the door last," he said.

As such, he concluded to heartwire, "Patients and their providers can make their own decisions. Some evidence was available before now that there might be a difference between these drugs; this study reinforces that impression. People who are considering glitazone therapy should have a good look both at the meta-analyses, as well as this current paper, and use it as the basis for the treatment decision."

Dr Steven Nissen (Cleveland Clinic, OH), a coauthor on the first meta-analysis to really galvanize the rosiglitazone safety debate, agreed to comment on the current study for heartwire.

"These findings are consistent with other observational studies and our two meta-analyses of cardiovascular outcomes, one with pioglitazone and the other with rosiglitazone," Nissen said. "There is no reason to expect that these differences are limited to older patients. . . . There is clearly a consensus that, if a TZD is prescribed, pioglitazone is strongly preferred over rosiglitazone."

Nissen also pointed to a "consensus algorithm" from the American Diabetes Association and the European Association for the Study of Diabetes on the initiation and adjustment of therapy for hyperglycemia [2]. In it, consensus group members "unanimously advised against using rosiglitazone," the document reads.

The consensus document was released last month.

Also contacted by heartwire, a spokesperson for GlaxoSmithKline emphasized that the latest study was nonrandomized and observational, and not the "gold-standard" randomized clinical trial. In a statement provided to heartwire, spokesperson Jeff McLaughlin said that the company "strongly supports the safety and efficacy of Avandia based on extensive clinical-trial experience and widespread postmarketing use. This new study is inconsistent with evidence from randomized clinical trials and has significant limitations."

Importantly, he noted, the primary outcome in this observational analysis is all-cause mortality, yet the ongoing long-term randomized clinical trial RECORD has shown no statistically significant differences between the rosiglitazone-treated patients and the control group regarding death from cardiovascular causes or any cause. "The results of this observational study for both all-cause mortality and cardiovascular outcomes may be biased due to imbalances in comorbid conditions—cardiovascular disease, chronic obstructive pulmonary disease, and malignancies—between the two treatment groups," the statement reads. "Final long-term data from ongoing cardiovascular outcome studies [including RECORD and BARI2D] should be available within a year and should be taken into consideration before final conclusions are reached about safety and effectiveness of Avandia."