Risk of Bias and Brand Explain the Observed Inconsistency in Trials on Glucosamine for Symptomatic Relief of Osteoarthritis: A Meta‐Analysis of Placebo‐Controlled Trials

Objective To determine whether study sponsor, chemical formulation, brand of glucosamine, and/or risk of bias explain observed inconsistencies in trials of glucosamine's efficacy for treating pain in osteoarthritis (OA). Methods A systematic review and stratified meta‐analysis of randomized placebo‐controlled trials was performed, and random‐effects models were applied with inconsistency (I2) and heterogeneity (tau2) estimated using Review Manager and SAS, respectively. The major outcome was reduction of pain; the standardized mean difference (SMD [95% confidence interval (95% CI)]) served as effect size. Results The inclusion criteria yielded 25 trials (3,458 patients). Glucosamine moderately reduced pain (SMD −0.51 [95% CI −0.72, −0.30]), although a high level of between‐trial inconsistency was observed (I2= 88%). The single most important explanation (i.e., covariate) was brand, reducing heterogeneity by 41% (P = 0.00032). Twelve trials (1,437 patients) using the Rottapharm/Madaus product resulted in significant pain reduction (SMD −1.07 [95% CI −1.47, −0.67]), although a sensitivity analysis of 3 low risk of bias trials using the Rottapharm/Madaus product showed less promising results (SMD −0.27 [95% CI −0.43, −0.12]), which is only a small effect size. Thirteen trials (1,963 patients) using non–Rottapharm/Madaus products consistently failed to show a reduction in pain (SMD −0.11 [95% CI −0.46, 0.24]). The second most important explanation was overall risk of bias (reducing heterogeneity by 32%). Conclusion Most of the observed heterogeneity in glucosamine trials is explained by brand. Trials using the Rottapharm/Madaus glucosamine product had a superior outcome on pain in OA compared to other preparations of glucosamine. Large inconsistency was found, however. Low risk of bias trials, using the Rottapharm/Madaus product, revealed a small effect size.