New use of rosiglitazone decreased following publication of a meta-analysis suggesting harm

Aims — It is uncertain whether meta-analyses lead to changes in prescribing practices. We studied trends in the prescribing of glucose-lowering therapy before and after the publication of a meta-analysis suggesting harm from rosiglitazone.

Methods — We examined the prescription records of all residents of Ontario, Canada, aged > or = 66 years. For each week between January and December 2007, we identified new users of five categories of glucose-lowering medications: rosiglitazone, pioglitazone, metformin, glibenclamide (glyburide) and insulin. The effect of the meta-analysis was assessed using interventional autoregressive integrated moving-average models.

Results — Following the release of the meta-analysis, there was a sudden decline in new users of rosiglitazone (P = 0.01), mirrored by a nearly identical but transient increase in new users of pioglitazone (P < 0.001). There was also a net decline in new users of thiazolidinediones as a class (P < 0.001). The number of new users of other glucose-lowering medications did not change.

Conclusions — A highly-publicized meta-analysis regarding rosiglitazone's potential harms led to an abrupt decline in new users of the drug, as well as a transient surge in new use of pioglitazone.