Inflated effect estimates for vitamin D supplementation are driven by common meta-analytical errors.

Abstract

Han et al. (J Int Soc Sports Nutr 16:55, 2019) sought to quantify the effects of vitamin D supplementation on strength outcomes among athletes in a meta-analysis. The authors reported a pooled effect size (standardized mean difference; SMD) of -0.75 (95% CI: -1.82 to 0.32, p = 0.17) in favor of supplementation, but the analytical approach was not appropriate for a pooled analysis of randomized controlled trials and the effect sizes were calculated incorrectly. This letter discusses how these issues impact the results and interpretation of the paper, then provides an update on the estimated average effect of vitamin D on strength outcomes in athletes. Identified errors included the use of within-group rather than between-group effect size metrics, the use of standard error values in place of standard deviations, and failure to account for correlated observations within the model. The data were reanalyzed after correcting for these common meta-analytic errors. The results of this reanalysis reflect a dramatically smaller and statistically nonsignificant pooled effect estimate of SMD = 0.16 (-0.24 to 0.56, p = 0.43) in favor of supplementation. Further, the model from this reanalysis has more favorable statistical characteristics than the original analysis, as evidenced by a fairly symmetrical funnel plot and a nonsignificant result for Cochrane's Q test (Q = 5.02, p = 0.41). In order to disseminate robust information to sports nutrition practitioners and researchers, it is critically important for meta-analyses to produce valid effect estimates that are appropriate for the underlying study designs and calculated without error. This letter highlights common errors to inform the calculation and interpretation of future meta-analyses in sports nutrition.