Number needed to harm in the post-marketing safety evaluation: results for rosiglitazone and pioglitazone.

Abstract

Our aim is to investigate the usefulness of metric indices in post-marketing safety evaluations by estimating number needed to harm (NNH) values for cardiovascular (CV) adverse outcomes for rosiglitazone and pioglitazone. Reports from regulatory authorities (RAs) were consulted, and Medline searches were performed to identify studies assessing CV risks [all-cause death, CV death, myocardial infarction (MI), stroke, or congestive heart failure (CHF)] for thiazolidinediones. Meta-analyses were performed to pool evidence from randomized controlled trials (RCTs) and observational studies (OS). NNHs [with 95% confidence intervals (CI)] per year were estimated for CV adverse events. Reports from RAs included two meta-analyses of short-term RCTs, two long-term RCTs (RECORD and PROACTIVE), and a systematic review of OS (n = 29). The Medline search identified six additional OS. Statistically significant NNH values were obtained for the following: (i) rosiglitazone versus control on MI and CHF in the meta-analysis of RCTs (NNH = 16; 95%CI = 10-255; and NNH = 7; 95%CI = 5-16, respectively) and meta-analysis of OS (NNH = 12; 95%CI = 9-20; and NNH = 5; 95%CI = 32-131, respectively) and on CHF in the RECORD (NNH = 6; 95%CI = 4-14); (ii) pioglitazone versus control on CHF (NNH = 11; 95%CI = 6, 403) in the meta-analysis of RCTs and PROACTIVE (NNH = 12; 95%CI = 8-43); and (iii) rosiglitazone versus pioglitazone on MI (NNH = 69; 95%CI = 32-379), stroke (NNH = 36; 95%CI = 20-225), CHF (NNH = 33; 95%CI = 19-47), and all-cause death (NNH = 63; 95%CI = 49-100) in the meta-analysis of OS. The NNH values suggested an increased CV risk with rosiglitazone versus pioglitazone across several sources of information. The inclusion of objective metrics in post-marketing drug's benefit-risk assessments could be of increased value and help RAs to make consistent decisions on drug safety.