In Reply: Our meta-analysis indicated that baseline HbA1c level and exercise frequency explained nearly 70% of the heterogeneity in the effect of structured exercise. Therefore, we agree with Dr Armstrong and colleagues that the inclusion of more aerobic exercise–only studies with high baseline HbA1c level and greater exercise frequency might have resulted in a larger effect size for aerobic-only training.
We reanalyzed the meta-analysis removing the studies that used intention-to-treat, and the results were not significantly affected. However, as recently stated by White et al,1 it is unclear how to apply the intention-to-treat principle when some patients depart from randomized treatment in a trial, a common challenge in studies of lifestyle interventions. For instance, in the 2 largest trials of structured exercise,2- 3 linear mixed-effects models for repeated measures were used so that patients who dropped out could be analyzed up to the time of drop-out or study completion.4 The trial by Church et al,2 which showed 1 of the smallest effects on HbA1c reduction, presented both intention-to-treat and per-protocol analyses.