Do collagen supplements reduce symptoms of osteoarthritis? Meta-analytic results do not support strong conclusions

Abstract

To the editor: Consumers purchased $1.8 billion in collagen supplements during 2019 [1], yet evidence for collagen’s health effects is limited. In 2019, International Orthopaedics published a meta-analysis on the effects of oral collagen supplements on symptoms of osteoarthritis (OA) [4]. Summarizing five randomized placebo-controlled trials, the authors concluded that “collagen is effective in improving OA symptoms by the decrease of both WOMAC [Western Ontario and McMaster Universities Osteoarthritis] index and VAS [Visual Analog Scale pain] score.” More caution is warranted in drawing conclusions about collagen’s effects. Figure 1 replicates the authors’ random effects meta-analysis, appending information from the metaanalyzed studies and other tables in the authors’ article. The information suggests several reasons for caution. All five of the meta-analyzed articles disclosed financial relationships with companies that make or sell collagen (Nitta Gelatin [6], InterHealth Nutraceuticals [7], Protein SA[2], Gelita AG [8], and BioCell Technology [9]). As detailed in Fig. 1, in every study, collagen companies provided funding, employed authors, or paid an author as a consultant. Three studies enrolled fewer than 30 patients each, and, with one exception, smaller studies reported larger effects (a pattern that Fig. 1 highlights by sorting the results by sample size). This pattern is a common sign of publication bias—the tendency to publish significant but not null effects, especially if the sample is small. Although other explanations are possible, and formal tests for publication bias require at least ten studies [10], the possibility of publication bias coupled with conflicts of interest raises concerns that additional small studies may have been conducted but not submitted or published because they did not find an effect. The results were very heterogeneous (I2 = 78% for WOMAC and 91% for VAS), indicating low agreement across studies. Results can vary for different reasons, including dose and duration, but the published effects of collagen did not increase with dose or duration. Two studies with doses of 2 to 10 g/day [6, 9] reported larger effects than three studies with doses of 10 to 40 g/day [2, 7, 8]. Studies that lasted 10 to 13 weeks [6, 9] reported larger effects than studies that lasted 24 to 48 weeks [7, 8]. These patterns offer little guidance for how large dose patients should take or how soon they should expect relief. It is important not just to test whether effects differed significantly from zero, but to ask whether they exceeded the minimum clinically important difference, which is approximately 10 for VAS [5] and 16.5 for the authors’ Likert-scaled version of the WOMAC [3]. By these standards, the meta-analytic estimates, exceeded clinical importance for VAS (ES − 16.57, 95% CI − 26.24 to − 6.89) but not for WOMAC (ES − 8.00, 95% CI − 13.04 to − 2.95). Among the individual articles, only one showed a clinically important effect on VAS [6], and two reported clinically important effects on WOMAC [6, 9]. Conclusions about clinical importance are sensitive to which studies are included. For example, excluding the article by Kumar (2014)—which reported the largest effects while having small samples and four authors working for a collagen company—shrinks the meta-analytic estimates to clinical unimportance for both WOMAC (ES − 4.96, 95% CI − 6.59 to − 3.32) and VAS (ES − 5.00, 95% CI − 6.88 to − 3.13). Unbiased estimation of collagen’s effect would require a large, preregistered trial that is run independently of the collagen industry. Until such a trial occurs, the meta-analyzed trials provide little assurance that collagen supplements have reliable, clinically important effects on OA symptoms. * Paul T. von Hippel [email protected]