Viscosupplementation for Osteoarthritis of the Knee: A Systematic Review of the Evidence

BACKGROUND: The purpose of this analysis was to determine the clinical significance of injectable hyaluronic acid (HA) in the treatment of knee osteoarthritis, and to assess which trial-level factors influence the overall treatment effect of HA on pain (as measured by a VAS [visual analog scale] or the WOMAC [Western Ontario and McMaster Universities Osteoarthritis Index]) and the WOMAC function and WOMAC stiffness subscales.

METHODS: A comprehensive literature search of PubMed, EMBASE, the Physiotherapy Evidence Database, and the Cochrane Central Register of Controlled Trials was done to locate randomized controlled trials that compared HA with control treatment and had a minimum of thirty patients per subgroup. To be considered for inclusion, each article had to include VAS or WOMAC pain, WOMAC function, and/or WOMAC stiffness as outcomes because the minimal important difference (MID) has been established for these instruments. A "best-evidence" systematic review and meta-analysis of nineteen trials was performed; because of high heterogeneity among the trials, meta-regression analyses were conducted to determine the influence of trial characteristics on overall HA treatment effects for pain, function, and stiffness.

RESULTS: The most consistent finding was that double-blinded, sham-controlled trials had much smaller treatment effects than trials that were not sufficiently blinded (p < 0.05). For double-blinded trials, the overall treatment effect was less than half of the MID for pain, function, and stiffness. Other significant associations were found for cross-linked HAs and follow-up duration. However, the effect sizes among double-blinded trials of cross-linked HAs were still less than half of the MIDs for pain and stiffness. The statistically significant effect of follow-up duration disappeared when the open-label trials were removed from the analysis.

CONCLUSIONS: Meta-analysis of only the double-blinded, sham-controlled trials with at least sixty patients did not show clinically important differences of HA treatment over placebo. When all literature was added to the analysis, the overall effect was greater but was biased toward stronger treatment effects because of the influence of nonblinded or improperly blinded trials.