Fourteen controlled clinical trials (CCTs) were included (504 patients), of which 11 were RCTs (412 patients) and 3 were non-RCTs (92 patients).
Study quality.
The studies were of moderate to low quality (mean Jadad score: 1.6).
HbA1c.
Exercise versus no exercise: 9 CCTs (8 RCTs and 1 non-RCT; 11 comparisons; 310 patients) were included. Exercise significantly lowered HbA1c compared with no exercise. HbAlc was 7.65% in the exercise group and 8.31% in the control group; the WMD was -0.66% (95% CI: -0.98, -0.34, P<0.001). No significant heterogeneity was detected (P=0.46).
Exercise plus diet versus no exercise plus no diet: 2 RCTs (3 comparisons; 142 patients) were included. Exercise plus diet significantly lowered HbA1c compared with no exercise plus no diet; the WMD was -0.76% (95% CI: -1.32,-0.20, P=0.008). No significant heterogeneity was detected (P=0.69).
The sensitivity analysis found similar results for randomised compared with non-randomised studies and for aerobic exercise compared with resistance training. The results were similar after excluding one study that enrolled only older patients.
Body mass.
Exercise and no exercise: 11 CCTs (8 RCTs and 3 non-RCTs; 13 comparisons; 368 patients) were included. There was no significant difference in body mass between exercise and no exercise; the WMD was 0.06% (95% CI: -0.15, +0.26, P=0.60). No significant heterogeneity was detected (P=0.95).
Exercise plus diet versus no exercise plus no diet: 2 RCTs (3 comparisons; 142 patients) were included. There was no significant difference in body mass between exercise plus diet and no exercise or diet; the WMD was -0.20% (95% CI: -0.54, +0.14, P=0.24). No significant heterogeneity was detected (P=0.94).
Differences in HbA1c between treatment groups were not significantly related to weight loss, exercise intensity, or exercise volume.