How the FDA Can Make Better Decisions on Drug Safety

Narrator: In May of 2007, the diabetes drug rosiglitazone, commonly known as Avandia, was issued a strict warning by the US Food and Drug Administration after initial studies indicated it could increase risk of heart problems. This in turn dropped use of the drug by 78 percent over 15 months and plummeted sales. According to research by Chicago Booth’s John R. Birge and his coauthors, had the FDA used a more data-driven approach to its postmarket drug surveillance, they would have found a different verdict on this drug.

John R. Birge: In this research, we were looking at what was the overall effect of this drug, this rosiglitazone, and what we were trying to address is that there had been a meta-analysis about this, which is a study that puts together a number of other studies, and in those other studies, there was some indication that the drug actually had caused increased number of heart attacks and deaths from heart attacks and strokes. What we wanted to do was to use a different set of data to address independently whether that drug, rosiglitazone, was actually associated with these coronary events. And then we decided that we would also like to look at this as a template for how postmarket surveillance should be done, that we should be looking at the overall range of events that might occur and using these independent data sources in order to be able to determine whether the drug is actually associated with these adverse kinds of events.

Narrator: The new approach proposed by the researchers makes use of several reliable and longitudinal databases alongside established econometrics methods to analyze the link between approved drugs and potentially related adverse health events. For the specific case of rosiglitazone, they compiled Veterans Health Administration data for more than 320,000 diabetes patients over an eight-year period.

John R. Birge: The data showed us that rosiglitazone was not associated with major coronary events—stroke, heart disease—and in fact, rosiglitazone, when we did it on a matching analysis, trying to make sure that we matched exactly the patients who were taking rosiglitazone with patients who weren’t taking rosiglitazone, when we did it by that matching analysis, rosiglitazone was actually associated with fewer deaths, with lower mortality, and not associated with any kind of adverse coronary events. And so by using this independent data set, we were able to come up with a different decision, or a different outcome, from what was observed in this first study, this meta-analysis, and what it actually prompted the FDA to issue a black box warning on rosiglitazone, which made rosiglitazone prescription numbers go down precipitously.

Narrator: The researchers further demonstrated a generalizability of their new approach by retroactively analyzing two additional FDA postmarket drug warnings, one for statins and the other for atenolol, and determined that both warnings were, in fact, warranted. Further, they indicate that the current postmarket drug analysis is incomplete and prone to errors. These shortcomings can produce disastrous financial effects for providers and firms, as well as serious consequences to patient health.

John R. Birge: Today, it’s somewhat passive in how they’re collecting these adverse drug-event datas. It’s not as much a comprehensive approach as what we would ask for and what we would promote in our paper. Instead of having this comprehensive approach, the FDA has somewhat a passive approach. It collects information that it gathers about adverse drug events, but a lot of that reporting is just voluntary. In fact, there’s some studies that say that 94 percent of the adverse events are not actually reported. So many of these events are not being reported, and so we’re not seeing as many adverse events as probably are occurring, and it highlights even more those adverse events that are reported. So if we see maybe a few rare events that are reported, the FDA may think that those are more common than they actually are and may react to them more severely than they would otherwise. But I think that the overall effect, maybe, of having so much industry support is that we’re not seeing as much of a comprehensive approach to actually trying to get all of the reports of the adverse drug events that there might be.

Narrator: The researchers suggest that if the FDA were to use a more data-driven approach in its postmarket drug analysis, there would be less cases where safe drugs are wrongly penalized when they could actually be helping people.