The mean difference in pain scores between treatment and placebo at week 1 was 4.4 (95% CI, +1.1, +7.2) and –1.0 (95% CI, -3.2, +1.2) for analysis restricted to the 7 good quality trials. The mean difference in pain scores at 5 to 7 weeks was 17.6 (95% CI, +7.5, +28.0) and 7.2 (95% CI, +2.4, +12.0) for the analysis restricted to the 2 good quality studies. At weeks 8 to 12 the mean difference in VAS between treatment and control was 18.1 (95% CI, +6.3, +29.9), and 7.1 (95% CI, +3.0, +11.3) in the analysis restricted to good quality trials. At weeks 15 to 22, the mean difference was 4.4 (95% CI, –15.3, +24.1). The Egger test was not statistically significant (2.3; P=.096; 95% CI, –0.5, +5.2) suggesting that there is no publication bias.
High heterogeneity was observed at all time intervals except 1 week (FIGURE 1). Of the 5 trials outside the confidence bounds (positioned 2 units above and below the regression line), 4 were poor-quality studies.
TABLE 3 shows the random-effect regression models we used to test the influence on the outcome of type of pain measured (pain with activity or pain at rest), type of medication (hyaluronan or hylan G-F 20), and study quality (good or poor). No significant association between treatment efficacy and type of pain used as outcome variable was observed. Clinical trials using hylan GF-20 showed statistically significant better results than those using hyaluronan at weeks 5 to 7 and 8 to 12. Poor-quality studies showed a larger treatment effect, but the difference was statistically significant only at week 1.
TABLE 3
Regression models to assess the sources of heterogeneity in the meta-analysis
| WEEK 1 | WEEKS 5–7 | WEEKS 8–12 | ||||
|---|---|---|---|---|---|---|
| Coef. (SE) [95% CI] | P value | Coef. (SE) [95% CI] | P value | Coef. (SE) [95% CI] | P value | |
| Pain | ||||||
| With activity | 1.7 (3.8) | 1.8 (4.3) | –0.5 (4.6) | |||
| At rest | [–5.8, +9.2] | 0.657 | [–6.6, +10.2] | 0.671 | [–9.5, +8.6] | 0.916 |
| Medication | ||||||
| Hyaluronan** | –3.4 (6.8) | 0.614 | 17.5 (4.9) | <0.001 | 14.8 (6.1) | 0.016 |
| Hylan G-F 20 | –16.7, +9.9] | [+7.8, +27.1] | [+2.8, +26.8] | |||
| Quality* | ||||||
| Poor (<0.75) | –19.9 (5.7) | 0.001 | –7.4 (4.9) | 0.131 | –11.7 (7.0) | 0.092 |
| Good (≥0.75) | [–31.1, –8.7] | [–17.0, +2.2] | [–25.3, +1.9] | |||
| Constant | 18.4 (5.6) | 0.001 | 13.5 (4.5) | 0.003 | 19.2 (5.8) | 0.001 |
| [+7.5, +29.3] | [+4.6, +22.3] | [+7.9, +30.5] | ||||
| *<.75 quality score: Grecomoro .439, Scale .570, Wobig .731, Huskisson .718. | ||||||
| ** 9 trials for hyaluronan (3 for Hyalgan®) and 2 trials for hylan G-F20 (Synvisc®). | ||||||
Discussion
This meta-analysis synthesized data from 9 randomized, double-blinded, placebo-controlled trials that evaluated the efficacy of intra-articular hyaluronic acid. Our findings show significantly decreased pain as measured by VAS at 5 to 7 weeks and at 8 to 12 weeks after the last injection. Intra-articular hyaluronic acid was not more effective than placebo in relieving pain at 1 week or at 15 to 22 weeks after the last injection. Because only 3 of the trials assessed patients after 12 weeks, however, the sample size is too small to definitively rule out a significant therapeutic effect after 12 weeks.
Reasons for the differences in efficacy among trials of hyaluronic acid in the treatment of knee osteoarthritis include dose, type, and frequency of administration, genetic or age differences among the study subjects, severity of osteoarthritis, time of follow-up, and quality of the studies. We confirmed that the treatment effect is time dependent. Although our meta-regression analysis (TABLE 2) suggests that hylan GF-20 is more effective than hyaluronan at 5 to 12 weeks, the number of clinical trials is relatively small and both of the hylan G-F20 studies were of poor quality. Therefore, we cannot say with confidence that one form is better than the other. Data in these trials were insufficient to assess the impact of body mass index, genetics, or severity of osteoarthritis.
We did not evaluate functional improvement in this meta-analysis because functional status was not measured in some trials and the assessment methods were too variable in the trials that did assess functional status. Publication bias, or the possibility that unpublished data would contradict the results of published studies, is always a potential source of bias in meta-analysis. However, the Egger test was not statistically significant (6.5; 95% CI, –0.5, +13.5) suggesting that there is no publication bias.25
Finally, the presence of heterogeneity of results indicates there were important differences among the studies. Exclusion of clinical trials considered of poor quality diminished this heterogeneity substantially. Subanalysis restricted to good-quality studies supports the efficacy of intra-articular hyaluronic acid in the treatment of knee osteoarthritis pain, although the effect size is smaller when one considers only the good quality studies.