Factors associated with the effect-size of thiazolidinedione (TZD) therapy on HbA(1c): a meta-analysis of published randomized clinical trials

Abstract

Objective: To examine factors affecting the size of the HbA(1c) response to thiazolidinedione (TZD) therapy.

Research design and methods: Meta-analysis of randomized TZD controlled trials which were identified using PubMed, EBSCO and Sci-lit databases and were published in English. Sociodemographic and clinical data were extracted from each trial. HbA(1c) effect size was defined as either a placebo-subtracted change in HbA(1c) or a change in HbA(1c) from baseline. Weighted multivariable regression was used to examine factors associated with changes in HbA(1c). Bootstrapped smearing estimates were computed to obtain reliable estimates of HbA(1c) effect size.

Results: Forty-two trials yielded 60 trial arms which represented 8322 patients treated with thiazolidinediones. Weighted placebo-subtracted change in HbA(1c) was -0.99% +/- 0.02% with an average baseline HbA(1c) of 9.1% +/- 1.0%. Weighted bootstrapped smearing estimate of the placebo-subtracted change in HbA(1c) was -1.02% +/- 0.004%. After controlling for other variables, the baseline HbA(1c) level had a significant negative association with placebo-subtracted HbA(1c) change (p = 0.004) and also with change in HbA(1c) from baseline (p = 0.002). Longer trial duration was associated with greater placebo-subtracted HbA(1c) change (p = 0.01) but not with the change in HbA(1c) from baseline. The multivariable models explained 72% of the variation in placebo-subtracted HbA(1c) change. It was not possible to estimate effects of the run-in period and obesity on TZD effect size.

Conclusion: Baseline HbA(1c) and trial duration significantly impacted the effect size of TZD therapy on HbA(1c). Age, gender, duration of diabetes and prior use of anti-diabetic therapy were not associated with the TZD effect size.