Eight RCTs were included in the review (2,837patients; from 48 to 317 per trial). One RCT had a total Jadad score of 2 (low quality); other scores were 3 (three RCTs), 4 (two RCTs) and 5 (two RCTs). All were classed as high quality.
No statistically significant differences in HbA1c changes were observed in comparisons between thiazolidinediones and dipeptidyl peptidase IV inhibitors (two RCTs), thiazolidinediones and sulphonylureas (four RCTs), thiazolidinediones in comparison to each other (one RCT), and insulin and sulphonylureas (one RCT).
A statistically significantly greater reduction of fasting plasma glucose was observed with thiazolidinediones, compared with dipeptidyl peptidase inhibitors (MD -11.61mg/dL, 95% CI -17.82 to -5.39; two RCTs; Ι²=0%). Statistically significantly greater reductions in fasting plasma insulin were observed with the use of thiazolidinediones, when they were compared with dipeptidyl peptidase IV inhibitors (MD -3.50μU/mL, 95% CI -5.55 to -1.45; one RCT) and sulphonylureas (MD -5.72. 95% CI -8.21 to -3.22; four RCTs; Ι²=82%). No other statistically significant differences in changes in fasting plasma glucose or fasting plasma insulin were observed between treatment groups.
Effect estimates remained unchanged when sensitivity analyses were performed. No evidence of publication bias was found in a comparison of thiazolidinediones with sulphonylureas for change in HbA1c levels (four RCTs).