Insulin Pump Therapy: a Meta-Analysis: Response to Weissberg-Benchell et al.

Weissberg-Benchell et al. (1) conclude, based on their meta-analysis on the efficacy of continuous subcutaneous insulin infusion (CSII) versus multiple daily injection (MDI) or conventional therapy (CT), that glycohemoglobin is significantly lower with CSII than with injection therapy. It is stated that “the weighted summary mean difference comparing the effect of CSII with MDI/CT was 0.95, with a 95% CI of 0.8–1.1, indicating that there was a significant difference between the two treatment approaches.” This is concluded from a meta-analysis of 11 studies with a parallel design. However, 5 of these 11 studies were not randomized, and therefore open to selection bias. Furthermore, two of these nonrandomized studies are interpreted to show the largest effect sizes, ∼2.5 and 4.5%. This will result in an overestimation of the weighted mean effect size, whereas the other studies show effect sizes between 0.25 and 0.75%. Effect sizes in this range are much more in agreement with a recent meta-analysis by Pickup et al. (2) that only included randomized studies and concluded that the effect of CSII on glycohemoglobin was 0.51%, as compared with injection therapy. Therefore, the large difference presented between insulin pump and injection therapy of 0.95% glycated hemoglobin is untenable from a methodological point of view. The mean effect size of CSII as found by Pickup et al. is consistent with a recent statement from Schade and Valentine (3): “the health care team must determine which treatment modality has the greatest potential for benefit in each diabetic patient.”