Clinical Trials of Hyaluronic Acid Reflect Significant Publication Bias

Intra-articular injections of hyaluronic acid have been available for the treatment of knee OA since 1997. The assumed, but unproved, mechanism of action involves both augmentation of the viscous properties of the articular compartment and promotion of normal hyaluronic acid production by synovial membrane. Widespread acceptance of the compounds efficacy has been hindered by the marketing of different molecular-weight preparations, a variety of administration regimens, and underpowered clinical trials, the majority of which lack appropriate control groups. Discordance in the observed efficacy of this treatment in general clinical practice compared with published results of clinical trials, most of which involve direct industry sponsorship, bring into question the collected validity of the results. Responding to this discordance, Lo et al (JAMA 290:3115, 2003) examine the efficacy of hyaluronic acid in knee OA through meta-analysis of available clinical trial data.

Methods:

All published trials of hyaluronic acid in knee OA available on MEDLINE, the Cochrane Controlled Trials Register, manuscript bibliographies, and abstracts from scientific meetings were collected. Attempts to collect unpublished data were also made via personal contact. Requirements for inclusion in the meta-analysis were human, randomized, single- or double- blind trials using at least three weekly intra-articular hyaluronic acid injections with a minimum follow-up time of 2 months and a less than 50% dropout rate. Inclusion of an intra-articular placebo control and utilization of validated knee pain assessment scales were required. Intent-to-treat analyses (or completers-only analyses with no dropouts) were separately noted. The standardized mean difference in baseline knee pain to knee pain at 2 to 3 months after start of treatment was calculated as the measure of effect size for each study. A small effect size was defined as 0.2 to 0.5 (cited by the authors as equivalent to the difference between NSAIDs and acetaminophen in clinical trials of OA). The Cochrane Q test was used to test for heterogeneity in effect sizes. The Egger test was used to test for publication bias.

Results:

35 of 57 trials were excluded, 21 due to lack of intra-articular placebo control. The remaining 22 trials were included in the meta-analysis, 19 of which had been published in peer-reviewed journals. 17 of the 22 included studies were industry sponsored.

The combined sample sizes of all included was numbered 2949 knees randomized, and 2485 analyzed (reflecting a cumulative dropout rate of 12.4%). In all but 3 of the 22 included studies, 95% confidence intervals for effect size appropriately crossed zero (indicating no difference in efficacy between hyaluronic acid and placebo). 2 of the 3 outlying studies demonstrated large effect sizes greater than 1.5 (cited by the authors as equivalent to change in pain after total knee replacement). Both of these studies evaluated the highest molecular weight formulations of hyaluronic acid and were industry sponsored. Applying the Cochrane Q test, significance in heterogeneity among the combined studies was eliminated when the 3 outlying studies were excluded. Likewise, pooled effect size for the combined studies was reduced from 0.32 to 0.19 when the outlying studies were excluded. Since none of the studies demonstrated an effect size less than zero, applying the Egger test demonstrated significant publication bias (i.e. the likelihood of positive trials to be published over null or negative trials). Only 7 trials provided intent-to-treat (or equivalent) data suitable for meta-analysis.

In a separate analysis examining the effect of intra-articular placebo compared to hyaluronic acid injection, pooled data from the 8 studies that allowed sufficient data extraction suggested that intra-articular placebo accounted for 79% of the effect ascribed to hyaluronic acid treatment.

Conclusions:

The pooled effect size of acceptably designed available clinical trials of hyaluronic acid for the treatment of knee OA demonstrate a modest effect on knee pain at 2 to 3 months post-treatment. Clinical trials of hyaluronic acid reflect significant publication bias and most clinical trials involve industry sponsorship.

Editorial Comments:

Hyaluronic acid preparations have been approved by the FDA as medical devices. The rigorous types of clinical trials and extensive efficacy and safety analyses that we are used to seeing and are required by the FDA for approval of most drugs are not required for medical devices. Consequently, these agents were approved without clear evidence of their efficacy. Since their approval, a number of additional clinical trials have been published, most with small numbers of patients and unsatisfactory methodology. This meta-analysis is an admirable attempt to analyze the data to date with these agents and discern whether, in total, they support or refute the claim of efficacy for treatment of knee OA. The conclusions of the authors that, at least, the lower molecular weight hyaluronic acid preparations do not appear to be efficacious is in line with many practitioners impressions.

There are many limitations to meta-analysis; they cannot correct for primary study design flaws, access to the raw data is usually lacking, and combining and analyzing subjects from different studies as if they were a homogeneous group can be misleading.

Nonetheless, the vast majority of the studies analyzed show consistent and unimpressive results. Perhaps most compelling is the analysis of the intra-articular placebo controlled trials suggesting that the act of injection itself, rather than the drug delivered, is responsible for the majority of patient reported improvement in these trials.

Arthritis Center

About Arthritis Center

Founded in 1998, the Arthritis Center at Johns Hopkins is dedicated to providing quality education to patients and healthcare providers alike.