TZDs increase CHF events in pts with DM or with prediabetes

Clinical Question

Do thiazolidinediones increase the risk of heart failure events in patients who have diabetes or are prediabetic?

Bottom Line

Patients with diabetes or prediabetes treated with a thiazolidinediones (TZDs; rosiglitazone [Avandia] and pioglitazone [Actos]) are at higher risk of heart failure events but don't appear to be at increased risk of cardiovascular mortality. Since the TZDs improve glycemic control but have not been shown to improve clinical outcomes, they should not be used as first-line agents. (LOE = 1a)


Lago RM, Singh PP, Nesto RW. Congestive heart failure and cardiovascular death in patients with prediabetes and type 2 diabetes given thiazolidinediones: a meta-analysis of randomised clinical trials. Lancet 2007;370(9593):1129-1136.  [PMID:17905165]

Study Design

Meta-analysis (randomized controlled trials)


Self-funded or unfunded


Various (meta-analysis)


The authors systematically searched several databases looking for randomized controlled trials of TZDs that reported the frequency of congestive heart failure events (like hospitalization) and cardiovascular death. I would have preferred that they use all-cause mortality as a less biased outcome than cause-specific mortality. They only included English language studies that included men. They did not describe searching for unpublished data nor did they report duplicate independent inclusion of studies. Two investigators independently extracted data and assessed study quality. Seven studies with 20,191 patients made the final cut. The duration of the studies ranged from 12 months to 48 months. Among patients receiving a TZD, 214 (12.4% per year) experienced congestive heart failure events compared with 146 of control patients (8.4% per year). The authors report that compared with control medications, one would need to treat 107 patients per year with a TZD to cause 1 to develop congestive heart failure (95% CI, 76 - 176). These data were consistent across studies, suggesting a class effect. The authors report that there was no difference in cardiovascular death between groups, but they don't report total mortality.