In their recent meta-analysis, Loke and colleagues provided evidence that long-term use of thiazolidinediones is associated with a higher risk of fractures among women but not men.1 Potentially important clinical implications of this study may have been missed because of the broadness of the inclusion criteria (treatment groups included any thiazolidinedione and control groups included any active comparator or placebo) and the absence of subgroup analyses.
When we reanalyzed the data reported by Loke and colleagues on the basis of the actual treatment used (rosiglitazone or pioglitazone), we found that patients with type 2 diabetes who received rosiglitazone had a significantly higher risk of fracture than those in the control group (fixed-effects pooled odds ratio [OR] 1.64, 95% confidence interval [CI] 1.24–2.17, p < 0.001, I2 = 21%), whereas no significant difference in fracture risk was found between patients in the pioglitazone and control groups (fixed-effects pooled OR 1.26, 95% CI 0.92–1.71, p = 0.15, I2 = 22%) (Figure 1,Figure 2). Indirect comparison between the 2 thiazolidinedione treatment groups was not feasible because no common comparator existed.2
After within-study data stratification on sex, it was possible to estimate a pooled OR only for patients who received pioglitazone (relevant data for rosiglitazone were available from only 1 study). Women receiving pioglitazone demonstrated evidence of significant fracture risk (fixed-effects pooled OR 2.14, 95% CI 1.33–3.44, p = 0.02, I2 = 13%); however, this finding did not persist in a sensitivity analysis with a random-effects model (pooled OR 2.00, 95% CI 0.86–4.66, p = 0.11) (Figure 3). As expected, men who received pioglitazone did not demonstrate evidence of a significantly higher fracture risk than those in the control group (fixed-effects pooled OR 0.84, 95% CI 0.53–1.34, p = 0.46, I2 = 0%).
In summary, pioglitazone use might not be associated with increased fracture risk in either women or men with type 2 diabetes. This finding has both clinical and research implications.
Footnotes
-
Competing interests: None declared.