Effects of soy intake on glycemic control: a meta-analysis of randomized controlled trials

Evidence from animal and observational studies has supported the beneficial effects of soy intake on glycemic control, but intervention studies in humans have generated mixed results and have not been systematically examined.

We aimed to quantitatively evaluate the effects of soy intake on measures of glycemic control.

We conducted a structured electronic search of PubMed, EMBASE, the Cochrane Library, and the China National Knowledge Infrastructure (updated to March 2010) databases for randomized controlled trials that described the effectiveness of different soy regimes on measures of glycemic control [homeostatic model assessment of insulin resistance (HOMA-IR) and fasting glucose and insulin, glycated hemoglobin (Hb A(1c)), and 2-h glucose and insulin concentrations]. Data on participants, interventions, outcomes, and potential effect modifiers were extracted independently. Weighted mean effect sizes were calculated for net changes by using fixed-effects or random-effects models. We performed prespecified subgroup analyses to explore the influence of covariates on net changes of fasting glucose and insulin concentrations.

Twenty-four trials with a total of 1518 subjects were included in the meta-analysis. Soy consumption did not significantly affect measures of glycemic control. The mean (95% CI) difference was -0.69 mg/dL (-1.65, 0.27 mg/dL) for fasting glucose concentrations in the fixed-effects model (P = 0.16) and -0.18 mg/dL (-0.70, 0.34 mg/dL) for fasting insulin concentrations in the random-effects model (P = 0.50). Significant heterogeneity was noted in the results of fasting insulin concentrations and HOMA-IR.

There was not a significant overall effect of soy intake on improvements of fasting glucose and insulin concentrations; however, a favorable change in fasting glucose concentrations was observed in studies that used whole soy foods or a soy diet in the subgroup analysis. Evidence for other glycemic variables such as Hb A(1c) and 2-h postchallenge glucose and insulin concentrations was limited because of the small number of trials.