GlaxoSmithKline’s Avandia (rosiglitazone) is an excellent drug. It is the only oral diabetes medication to show sustained glycemic control, it has been studied (independently as well as by GSK) in tens of thousands of patients and it has a relatively good safety profile. Yet, even if the FDA lifts restrictions, which its advisory board recommended after reviewing an independent analysis of the RECORD study in June, this medication is forever tarnished and will likely see limited expanded use. There are many lessons to be learned from the Avandia fiasco. Here are a few for the industry:
1. Attacks in the media cause lasting damage and should be countered strongly. Despite an incredible amount of data supporting Avandia’s safety, the public and physicians believed that Avandia caused heart attacks as soon as the media storm surrounding Nissen’s meta-analysis1 was unleashed in 2007. In general, the industry’s response to a media storm has been to simply release statements defending their product and relying on typical marketing techniques to reassure physicians. This approach simply does not work. Crestor (which now appears to be the safest statin) might have outsold Lipitor had its company vehemently defended their product instead of hoping safety concerns would die down.
2. Industry sponsored research is simply not trusted. At both the 2010 FDA meeting, which led to Avandia’s withdrawal, and the 2013 meeting that could bring it back, advisor’s raised concerns that because the RECORD study (which proved Avandia did not cause heart attacks) was sponsored by GSK, the evidence was tainted. This is, of course, ridiculous since most pharmaceutical research is funded by the industry. The industry must look into better ways of partnering with academia and/or the government to perform research in order to decrease the perception of bias and improve trust.
3. Meta-analyses are often wrong, and the industry should play a role in discrediting them. There has only been one actual study, Nissen’s meta-analysis that has shown Avandia increases the risk of myocardial infarction.2 Even this analysis is suspect because despite claiming a 43% relative increase in myocardial infarction, the actual event rates were about 0.6% in both groups. In addition, another meta-analysis of the same data, using slightly different methodology, showed no such increase in heart attacks. There are several other examples (hormone replacement therapy being the most significant) where a meta-analysis was later proven incorrect by a large, prospective, randomized controlled trial. While meta-analysis are very simple to perform and easy to publish, large randomized trials face many challenges. The pharmaceutical industry would be wise to include the pitfalls of meta-analyses in their communications with physicians and the public.
1. N Engl J Med. 2007; 356:2457-2471
2. Ann Intern Med. 2007; 147(8):578-581