GlaxoSmithKline (NYSE: GSK) believes that conclusions drawn from the most recent meta-analyses published by Drs. Nissen et. al. and Furberg et. al. in the Journal of the American Medical Association (JAMA) do not confirm a difference in the safety profile of Avandia (rosiglitazone) and Actos (pioglitazone).
These analyses reflect limitations that are common to all meta-analyses, by the authors’ own admission. These analyses do not yield data robust enough to guide doctors in selecting appropriate diabetes treatments for their patients. Comparisons between different meta-analyses with different endpoints and patient populations are even more unreliable.
The pioglitazone meta-analysis is based on a small number of studies (19) provided directly by Takeda, and is heavily biased by data from the PROactive study (5,238 patients) which contributed 80 percent of the endpoint data. The patient population in the PROactive study was at high risk of cardiovascular disease.
PROactive compared diabetic patients on pioglitazone to those on placebo. Patients taking pioglitazone to control blood sugar might be expected to have fewer cardiovascular events than those who were not controlled on medication, consistent with the primary results of PROactive (Hazard Ratio 0.90 p=0.095). However, as described by FDA in the recent Advisory Committee, at six months PROactive actually showed an increased risk of heart attack (HR 1.2).
Applying the endpoint of CV death, myocardial infarction and stroke to the data on rosiglitazone across long-term clinical trials to enable a closer like-for-like comparison with the pioglitazone meta-analysis shows no statistical difference between rosiglitazone and comparators (a hazard ratio of 1.03). In RECORD, a study specifically designed to look at cardiovascular events, no real difference was seen between rosiglitazone and comparators (HR 0.96). In all of these analyses, the rosiglitazone and pioglitazone hazard ratios (HR 0.82 for pioglitazone in Nissen meta-analysis) do not suggest an increased risk versus the comparators.
No long-term, head-to-head clinical trial data specifically evaluates cardiovascular risk between rosiglitazone and pioglitazone; however, the head-to-head data that does exist, and the overwhelming majority of comparative observational data5 show no significant differences in CV events.
The JAMA article on rosiglitazone is yet another iteration of previously analyzed data, and offers no new information on the safety of rosiglitazone. The suggested increase in heart attack cited comes from a selective re-analyses of previously published and highly selective data from only 4 of 116 available studies, and reflects a difference of only 11 events in 14,291 patients between rosiglitazone and control. In this limited meta-analysis, in the context of all the other evidence, we believe it is inappropriate for the author to advise doctors to disregard the FDA’s advice which is to keep patients who are effectively controlling their diabetes on rosiglitazone. These data have been presented to an expert advisory panel of the FDA, which voted to keep rosiglitazone available to patients – a vote that reflects the role of this medicine as an important treatment option to help diabetes patients control their blood sugar.
The conclusions of these meta-analyses conflict with the wealth of accumulated data on rosiglitazone – including 116 clinical trials in more than 52,000 patients and epidemiological studies of databases in over one million patients. Analyzed studies show no difference in the ischemic cardiovascular effects of rosiglitazone versus other oral anti-diabetic medicines, including pioglitazone.