REVIEW

Intra-articular Hyaluronic Acid in Treatment of Knee Osteoarthritis A Meta-analysis Grace H. Lo, MD Michael LaValley, PhD Timothy McAlindon, MD, MPH David T. Felson, MD, MPH

H

YALURONIC ACID IS A LARGE

glycosaminoglycan composed of repeating disaccharides of glucuronic acid and N-acetylglucosamine that is naturally occurring in synovial fluid. There has been a long history of injecting hyaluronic acid intra-articularly into the joints of arthritic animals to improve their performance. This treatment has been approved by the US Food and Drug Administration as a device for the treatment of osteoarthritis (OA) in humans since 1997, with 3 to 5 weekly injections costing between $100 and $200 for each injection, not including the physician visit during which the injection is given. Despite Food and Drug Administration approval, the efficacy of intraarticular hyaluronic acid in the treatment of OA remains controversial. In 1998, a task force for the European League of Associations of Rheumatologists (EULAR) Standing Committee for Clinical Trials described intraarticular hyaluronic acid as being “probably effective in knee OA”1 in a statement issued after a systematic evaluation of the existing literature. There was insufficient evidence to allow a more definitive statement about the efficacy of hyaluronic acid in the

Context Intra-articular hyaluronic acid is a US Food and Drug Administration– approved treatment for knee osteoarthritis (OA); however, its efficacy is controversial. Objective To evaluate whether intra-articular hyaluronic acid is efficacious in treating knee OA. Data Sources We searched for human clinical trials in MEDLINE (1966 through February 2003) and the Cochrane Controlled Trials Register, using the search terms (osteoarthritis, osteoarthrosis, or degenerative arthritis) and (hyaluronic acid, Hyalgan, Synvisc, Artzal, Suplasyn, BioHy, or Orthovisc). We also hand searched manuscript bibliographies that met inclusion criteria, selected rheumatic disease journals, and abstracts from scientific meetings. Study Selection Included were published or unpublished, English and nonEnglish, single- or double-blinded, randomized controlled trials comparing intraarticular hyaluronic acid with intra-articular placebo injection for the treatment of knee OA. Trials also were required to have extractable data on pain reported by 1 of the outcome measures recommended by the Osteoarthritis Research Society. Data Extraction Two reviewers independently performed data extraction using standardized data forms. For each trial, we calculated an effect size (small-effect sizes, 0.20.5; large-effect sizes, 1.0-1.8, equivalent to a total knee replacement). We used a random-effects model to pool study results, the Cochrane Q test to evaluate heterogeneity, and a funnel plot and the Egger test to evaluate publication bias. Data Synthesis The overall dropout rate in the 22 selected trials was 12.4%. The pooled effect size for hyaluronic acid was 0.32 (95% confidence interval [CI], 0.17-0.47). There was significant heterogeneity among studies (P⬍.001). Two outlier trials, both evaluating the highest-molecular-weight hyaluronic acid, had effect sizes in excess of 1.5. However, the third trial of the same compound showed a nearly null effect. When the 3 trials of this compound were removed, heterogeneity was no longer significant (P=.58), and the pooled effect size for intra-articular hyaluronic acid decreased to 0.19 (95% CI, 0.10-0.27). There was evidence of publication bias with an asymmetric funnel plot, a positive Egger test, and identification of 2 unpublished trials whose pooled effect size was 0.07 (95% CI,−0.15 to 0.28). Conclusion Intra-articular hyaluronic acid has a small effect when compared with an intra-articular placebo. The presence of publication bias suggests even this effect may be overestimated. Compared with lower-molecular-weight hyaluronic acid, the highest-molecular-weight hyaluronic acid may be more efficacious in treating knee OA, but heterogeneity of these studies limits definitive conclusions. www.jama.com

JAMA. 2003;290:3115-3121

treatment of knee OA.1 Furthermore, a survey conducted by EULAR of a group of rheumatologists and ortho-

©2003 American Medical Association. All rights reserved.

Author Affiliations are listed at the end of this article. Corresponding Author and Reprints: Grace H. Lo, MD, 715 Albany St, A203, Boston, MA 02118 (e-mail: grace [email protected]).

(Reprinted) JAMA, December 17, 2003—Vol 290, No. 23

Downloaded from www.jama.com on March 26, 2008

3115

HYALURONIC ACID IN TREATMENT OF KNEE OSTEOARTHRITIS

Measures*

Box. Hierarchy of Pain Outcome 1. Global knee pain score (visual analog or Likert scale) 2. Knee pain on walking (visual analog or Likert scale) 3. Western Ontario and McMaster Universities (WOMAC) Osteoarthritis Index 4. Lequesne index 5. Knee pain during activities other than walking (visual analog or Likert scale) *Pain measurement scales are presented in order of preference. These are pain measure-

ment scales recommended by the task force of the Osteoarthritis Research Society.7

pedic surgeons found that this treatment was regarded as not clearly being recommended.2 The American College of Rheumatology has recommended intraarticular hyaluronic acid for use in the treatment of OA in patients at increased risk for gastrointestinal tract adverse events as an alternative for oral agents.3 However, Felson and Anderson,4 in an editorial accompanying a hyaluronic acid clinical trial, pointed to an absence of data supporting treatment efficacy. Also, in a review of the evidence for treatment of OA with intraarticular hyaluronic acid, Brandt et al5 reported that many of the clinical trials of this therapy had design flaws including problems with patient selection, inadequate blinding, and an inappropriate focus on completers analyses in preference to intent-to-treat analyses. Given lingering questions about the efficacy of intra-articular hyaluronic acid in the treatment of knee OA, the aim of this meta-analysis was to perform a systematic review of randomized clinical trials that evaluated intraarticular hyaluronic acid efficacy. METHODS Study Search

We performed a search of MEDLINE (1966-February 2003) and the Cochrane Controlled Trials Register using the search terms: (osteoarthritis, osteoarthrosis, or degenerative arthritis) and (hyaluronic acid, Hyalgan, Synvisc, Artzal, Suplasyn, BioHy, or Orthovisc). We also performed hand searches of manuscript bibliographies of studies meeting inclusion criteria, rheumatic disease journals (Arthritis and Rheumatism, Osteoarthritis and Cartilage, and

Journal of Rheumatology), and abstracts of scientific meetings (American College of Rheumatology Annual Scientific Meeting and the Osteoarthritis Research Society International Meeting) from 1986 through 2002. An attempt was made to include unpublished data by contacting all authors whose studies were included, to ask if they knew of any unpublished studies evaluating intra-articular hyaluronic acid efficacy; and if so, whether results from such studies could be provided for inclusion in this meta-analysis. Inclusion Criteria

We included English and non-English, human, randomized controlled trials, reported as single- or double- blind, testing intra-articular hyaluronic acid (administered at least every week for 3 weeks as recommended by manufacturers)6 against an intra-articular placebo in the treatment of knee OA. Pain had to be assessed and reported using 1 of the outcome measures recommended by the Osteoarthritis Research Society (BOX).7 Studies included in this meta-analysis also were required to have a minimum follow-up time of 2 months and a dropout rate of less than 50%. Data Extraction

Two reviewers (G.H.L. and M.L.) independently performed data extraction using standardized data forms. We extracted data on efficacy (along with their measurements of dispersion) as measured by pain 2 to 3 months after the first intra-articular injection (however, if these data were not available, we extracted data on pain at 1 to 4 months after the first intra-articular injection), for this is the time frame in which this de-

3116 JAMA, December 17, 2003—Vol 290, No. 23 (Reprinted)

vice is claimed to be most efficacious.6 When an article reported outcomes using more than 1 pain scale, we used, for our analyses, the scale with the highest rank on the pain outcome hierarchy (Box).7 All results from the data extraction were checked for consistency between the 2 reviewers. Any identified discrepancy was discussed until the 2 reviewers reached consensus. At least 1 attempt was made to contact (by e-mail or telephone) the investigators to obtain any missing data or to clarify any abstracted data that were ambiguous based on published results (eg, we attempted to reach study investigators for exact P values when they were reported as P⬍.05). When necessary, means and measures of dispersion were estimated from figures. Each study was evaluated for the type of analysis performed (intent-to-treat vs non–intent-to-treat). An analysis was considered to be intent-to-treat if (1) it was characterized by its investigators as such and there was an attempt to analyze data from all randomized participants, or (2) there was no dropout (even if the analysis was not specifically described as intent-to-treat). In trials for which both intent-to-treat and non–intent-to-treat analyses had been performed and that included extractable data for each analysis, we used the intent-to-treat analyses for our study. Within each study, the number of participants randomized and the number analyzed were evaluated. The presence and extent of industry sponsorship were also noted. Data Analysis

Effect sizes were calculated to compare results across studies. We used the standardized mean difference as our measure of effect size for hyaluronic acid for each study (effect size = [{pain HA − pain placebo }/{pooled SD}]). When reported, we used change from baseline pain at 2 to 3 months (however, if these data were not available, we used change from baseline pain at 1 to 4 months after the first intraarticular injection). If change from baseline pain was not available, pain val-

©2003 American Medical Association. All rights reserved.

Downloaded from www.jama.com on March 26, 2008

HYALURONIC ACID IN TREATMENT OF KNEE OSTEOARTHRITIS

ues at the time of interest were used (this assumed randomization was effective in creating 2 equal groups at baseline). Also, if only pain values were given without measures of dispersion, we imputed SDs using the median coefficient of variation from similar trials. As a clinical frame of reference, a small effect size is approximated as 0.2 to 0.5,8 which is equivalent to the effect of nonsteroidal anti-inflammatory drugs over that of acetaminophen in OA trials.9 A large effect size is estimated at being 1.0 or more,8 with a total knee replacement having an effect size between 1.0 and 1.8.10,11 To detect the presence of heterogeneity in effect sizes, we used the Cochrane Q test. A random effects model was used to pool study results. To test for publication bias, we used a funnel plot and the Egger test, with publication bias being the tendency for positive trials to be published and likewise the tendency for negative or null trials not to be published. Analyses were performed using SAS statistical software version 8.2 (SAS Inc, Cary, NC).

sion into the meta-analysis and for data abstraction. The trial by Russell et al14 met inclusion criteria in abstract form, but had extractable data only after details of the results were provided by the first author. One of the Pham et al32 investigators provided data from their trial report, when it had yet to be published. We computed summary effect sizes incorporating data from the latter 2 trials, but for these trials, we did not present individual effect sizes in case those investigators planned to submit their reports for publication. These nonpublished trials were excluded from our analysis evaluating publication bias. The trial by Karlsson et al31 met inclusion criteria and compared placebo with 2 formulations of hyaluronic acid, all administered intra-articularly. For our purposes, we treated this as 2 trials and compared outcomes from each formulation with those of the placebo group. Although this does not address the correlational structure of the data, the control group size was halved to avoid doubly counting any participants.

Handling of Unpublished Studies

Characteristics of Included Studies

We defined unpublished trials as (1) trials that were entirely unpublished or (2) trials that were included based on our study search but provided insufficient information to calculate an effect size until further unpublished information was provided by the investigators.

The overall dropout rate within the trials included was 12.4% with 2949 participant knees (2927 patients) randomized and 2584 participant knees included in the analysis. At least 17 of the 22 studies were industry sponsored. There was a wide range of study sizes with the smallest study having 24 and the largest, 408 participants (Table). Also, although the trial by Scale et al21 did not provide dropout information, we included it because it met all other inclusion criteria. In the Forest plot of the individual effect sizes (FIGURE 1), almost all trials had 95% confidence intervals (CIs) that included an effect size of zero, consistent with no effect. Trials by Scale et al20 and Wobig et al25 were outliers. Both trials evaluated the highest-molecularweight hyaluronic acid, and had an effect size in excess of 1.5, suggesting efficacy equivalent to that of a total knee replacement. However, another trial of the same compound by Karlsson et al31 showed a nearly null result.

RESULTS Study Search

Fifty-seven potentially relevant articles and abstracts were obtained. Of the 35 trials excluded, 21 did not have an intraarticular placebo as a comparator, 5 trials did not have an appropriate outcome measure for pain, 4 were duplicate publications, 3 did not have extractable data, 1 was a review article, and 1 did not use a sufficient number of injections. Twenty-two trials met inclusion criteria (TABLE).12-33 Of these, 19 trials were published as manuscripts in scientific journals. The trial by Cohen et al19 published only in abstract form provided sufficient information for inclu-

©2003 American Medical Association. All rights reserved.

Heterogeneity

The Cochrane Q test showed significant heterogeneity among study results (P⬍.001). This suggested that something more than random difference among the trials contributed to the variation of study effect sizes. There were, in total, 3 studies evaluating the highest-molecular-weight hyaluronic acid: Karlsson et al31 (n=105), Wobig et al25 (n=117), and Scale et al20 (n=30), 2 of which were the outliers. Heterogeneity among these 3 studies was significant (P⬍.001). When these 3 studies were removed, heterogeneity among the remaining studies was no longer significant (P=.58). Pooled Effect Sizes

With all studies included, the pooled effect size for intra-articular hyaluronic acid efficacy vs intra-articular placebo was 0.32 (95% CI, 0.17-0.47; P⬍.001), consistent with a small effect. If the trials evaluating the highestmolecular-weight formulation of hyaluronic acid were removed, the pooled effect size decreased to 0.19 (95% CI, 0.10-0.27; P⬍.001). Publication Bias Evaluation

There was marked asymmetry of the funnel plot (FIGURE 2), confirmed by the significant Egger test (P = .07; recommended level of significance is Pⱕ.10).34 Because the molecular weight of the hyaluronic acid formulation in the trial by Petrella et al30 was unknown, results from this trial were not included in the funnel plot. Also, there were no trials that had an effect size less than zero, even though the pooled effect size was small. The pooled effect size of the unpublished studies was 0.07 (95% CI,−0.15 to 0.28). These findings support the presence of publication bias. Intent-to-Treat vs Completers Analyses

Of the 9 trials that purposefully attempted an intent-to-treat analysis, only 3 studies provided intent-to-treat data. We classified an additional 3 trials that performed unspecified types of analyses as providing intent-to-treat data be-

(Reprinted) JAMA, December 17, 2003—Vol 290, No. 23

Downloaded from www.jama.com on March 26, 2008

3117

HYALURONIC ACID IN TREATMENT OF KNEE OSTEOARTHRITIS

Table. Characteristics of Included Trials

Source Dixon,12 1988

Average Molecular Hyaluronic Weight, Industry Acid kDa Sponsorship Hyalgan 615 Financial support, provided free medication

Russell,13 1992

1000

Financial support

No. of Patients

Effect Size (95% Confidence Interval)

Randomized Analyzed 63 53 0.35 (−0.19 to 0.90)

Intent-to-Treat Analysis Status

Patients Who Dropped Out, % 15.9

139

117

Omitted because data Performed and discussed, are currently but extractable data unpublished not provided

15.8

110

95

0.40 (−0.01 to 0.81) Performed and discussed, but extractable data not provided

13.6

209

195

Dougados,14 1993

Hyalgan

615

Puhl,15 1993

Artzal

900

Financial support, at least 1 author employed by sponsor

Dahlberg,16 1994

Artzal

900

Financial support

52

Creamer,17 1994

Hyalgan

615

Financial support

Henderson,18 1994

Hyalgan

615

0.29 (0.01 to 0.57)

6.7

52

0.21 (−0.35 to 0.77) Provided extractable data for the meta-analysis

0

24

24

0 (−0.80 to 0.80) Provided extractable data for the meta-analysis

0

91

84

0 (−0.43 to 0.43)

7.7

Cohen,19 1994

Hyalgan

615

Financial support

39

37

0.46 (−0.19 to 1.11)

Scale,20 1994

Synvisc

6000

Financial support

30

30

1.76 (0.92 to 2.60)

Unknown

Corrado,21 1995

Hyalgan

615

40

35

0.58 (−0.10 to 1.26)

12.5

Sala,22 1995

Hyalgan

615

40

40

0.39 (−0.24 to 1.01) Provided extractable data for the meta-analysis

0

Carrabba,23 1995

Hyalgan

615

Sponsor employee assisted with statistics for paper preparation

60

60

0.98 (0.42 to 1.55)

0

Lohmander,24 1996

Artzal

900

Financial support, at least 1 author employed by sponsor

240

189

0.07 (−0.21 to 0.36) Performed and discussed, but extractable data not provided

Wobig,25 1998

Synvisc

6000

Financial support

117

117

1.55 (1.14 to 1.97)

Altman,26 1998

Hyalgan

615

Financial support, at least 1 author employed by sponsor

330

244

0.04 (−0.21 to 0.29) Performed and discussed, but extractable data not provided

26.1

Huskisson,27 1999

Hyalgan

615

100

81-100†

0.35 (−0.05 to 0.74) Performed and discussed, but extractable data not provided

0-19.0‡

Brandt,28 2001

Orthovisc

2000

Financial support, at least 1 author employed by sponsor

226

135

0.25 (−0.09 to 0.59) Performed and discussed, but extractable data not provided

40.3

Tamir,29 2001

BioHy

3000

At least 1 author employed by sponsor

49

36

0.16 (−0.49 to 0.82)

26.5

Petrella,30 2002

Suplasyn

Financial support, at least 1 author employed by sponsor

120

108

0.20 (−0.18 to 0.57)

10

Karlsson,31 2002

Artzal

900

Financial support

125

104

0.11 (−0.33 to 0.54)

16.8

Karlsson,31 2002

Synvisc

6000

Financial support*

121

105

0.10 (−0.33 to 0.54)

13.2

Pham,32 2003

Hyalart

615

Financial support

216

216

Omitted because data Provided extractable data are currently for the meta-analysis unpublished

0

Jubb,33 2003

Hyalgan

615

At least 1 author employed by sponsor

408

408

0.15 (−0.04 to 0.35) Provided extractable data for the meta-analysis

0

5.1

Provided extractable data for the meta-analysis

Provided extractable data for the meta-analysis

21.3

0

*Financial support provided by a company other than the one manufacturing the treatment used in the study. †The number of patients analyzed is somewhere between 81 and 100, so exact dropout rate could not be calculated.

3118 JAMA, December 17, 2003—Vol 290, No. 23 (Reprinted)

©2003 American Medical Association. All rights reserved.

Downloaded from www.jama.com on March 26, 2008

HYALURONIC ACID IN TREATMENT OF KNEE OSTEOARTHRITIS

cause the investigators reported that there were no dropouts. The last trial contributing to intent-to-treat data was the unpublished study by Pham et al,32 in which the study investigators shared, on request, the raw data on all participants. In total, only 7 trials provided intent-totreat data for this meta-analysis. COMMENT Based on the findings of this metaanalysis, intra-articular hyaluronic acid has, at best, modest efficacy in the treatment of knee OA. This effect is equivalent to the effect of nonsteroidal antiinflammatory drugs over that of acetaminophen, an effect that itself remains controversial.9 We also found evidence of publication bias, so even this estimate of efficacy may be inflated. Our findings suggest the controversy surrounding the efficacy of intra-articular hyaluronic acid is justified and that the best available data does not support its efficacy. We evaluated efficacy of once weekly intra-articular hyaluronic acid injection for at least 3 injections, at 2 to 3 months after the first injection because this is the regimen and time frame in which it is believed to have its greatest efficacy.6 That the 2 outlier trials had effect sizes comparable with an effect size of a total knee replacement does not seem realistic. Furthermore, the effect size of most trials had 95% CIs that included an effect size of zero, consistent with no effect, again supportive of a small effect of this treatment or perhaps no effect at all. We recognize that a meta-analysis cannot address the problem of design flaws in the original studies. We designed our inclusion criteria to identify trials with a high quality of design, but even so, design and analysis flaws can get incorporated into meta-analytic summaries. Of note, in our meta-analysis, multiple trials used completers analyses and had substantial dropout rates. Appropriate randomization of patients in clinical trials creates equivalent groups of patients that vary only in their experimental treatment assignments. This allows investigators to attribute differences observed between the groups to their treatment as-

signments. If substantial dropout occurs in a randomized clinical trial and completers analyses are performed, these assumptions no longer hold and the validity of treatment comparisons is threatened. In our meta-analysis, 2 of the largest trials (Brandt et al28 and Altman and Moskowitz26) had dropout rates of 40.3% and 26.1%, respectively, and provided only completers analyses, raising concerns about the accuracy of the effects reported. In both of these trial publications, their intent-to-treat analyses were described as being nonsignificant but no details were provided.26,28 The trial by Scale et al20 did not report dropout data and its investigators did not respond to our request for information. Strictly speaking, this trial should have been excluded from our meta-analysis. However, in an effort to

allow every opportunity for this metaanalysis to show a positive effect of intraarticular hyaluronic acid, and given that this was 1 of the pivotal trials used for Food and Drug Administration approval for the highest-molecularweight formulation of hyaluronic acid in the United States, we allowed this trial into our meta-analysis. Notably, this is 1 of the outlier trials that had an effect size comparable with that of a total knee replacement. Were we to remove this trial from the meta-analysis, the pooled effect size for intra-articular hyaluronic acid would have fallen slightly from 0.32 to 0.29 (95% CI, 0.14-0.43); the Cochrane Q test for heterogeneity would have still been significant (P⬍.001). In our meta-analysis, at least 17 of the 22 trials included were industrysponsored, likely reflecting the increas-

Figure 1. Forest Plot of Trial Effect Sizes Source

Hyaluronic Acid Molecular Weight, kDa

Petrella,30 2002∗ Dixon,12 1988 Dougados,14 1999 Creamer,17 1994 Altman,26 1998 Henderson,18 1994 Huskisson,27 1999 Carrabba,23 1999 Cohen,19 1994 Jubb,33 2003 Sala,22 1995 Corrado,21 1995 Pham,32 2003 Lohmander,24 1996 Dahlberg,16 1994 Puhl,15 1993 Karlsson,31 2002 Russell,13 1992 Brandt,28 2001 Tamir,29 2001 Wobig,25 1998 Scale,20 1994 Karlsson,31 2002

615 615 615 615 615 615 615 615 615 615 615 615 900 900 900 900 1000 2000 3000 6000 6000 6000

Pooled Without HighestMolecular-Weight Hyaluronic Acid All Studies Pooled –1

0

1

2

3

Effect Size (95% Confidence Interval)

Error bars represent 95% confidence intervals. There were 2 outlier trials20,25 both evaluating the highestmolecular-weight hyaluronic acid. Both were industry sponsored and both showed an effect roughly equal to that of a total knee replacement. *We were unable to find the molecular weight of the hyaluronic acid formulation used in the study by Petrella et al.

©2003 American Medical Association. All rights reserved.

(Reprinted) JAMA, December 17, 2003—Vol 290, No. 23

Downloaded from www.jama.com on March 26, 2008

3119

HYALURONIC ACID IN TREATMENT OF KNEE OSTEOARTHRITIS

Figure 2. Funnel Plot of Hyaluronic Acid Efficacy in Treating Knee Osteoarthritis Hyaluronic Acid Compound Highest Molecular Weight

Other Molecular Weight

No. of Participants in Analyses

450 400 350 Egger Test P = .07

300 250 200 150 100 50 0 –1

0

1

2

Effect Size

The zero line represents an effect size of zero. The recommended level of significance is Pⱕ.10. Only published articles are represented. The trial Petrella et al30 was not included.

ing number of clinical trials funded by industry overall in the current research environment.35 Findings from a meta-analysis of trials comparing studies that were industry-sponsored with those that were otherwise funded showed that research funded by pharmaceutical companies was less likely to be published and was more likely to have outcomes favoring the sponsor.35 We found strong evidence of publication bias. There was asymmetry of the funnel plot, confirmed by a significant Egger test (P=.07). Also, because the pooled effect size was relatively close to zero, the lack of any trials with a negative effect size also supported the presence of publication bias. In a review of intra-articular hyaluronic acid efficacy Kirwan36 suggested that there is likely a large placebo response related to aspiration of the knee that may inflate results in uncontrolled evaluations of intra-articular therapies. Of the 57 trials evaluated for inclusion into this meta-analysis, 21 trials (37%) were excluded because they did not use intra-articular placebo for comparisons. We made a specific point of choosing trials that evaluated intra-articular hyaluronic acid against an intra-articular placebo to account for the possibility that any type of intra-articular injection might

be efficacious in relieving pain related to OA. In fact, all 22 trials included in our meta-analysis reported improvement of pain in the intraarticular placebo groups. To our knowledge, there are no headto-head trials comparing an intraarticular placebo with an oral placebo to evaluate greater efficacy of intraarticular placebo. The trial by Altman et al26 had a treatment arm with daily naproxen (and a sham injection) in addition to an intra-articular placebo group. Comparison of these 2 groups showed roughly equivalent efficacy for treating pain, supportive of the hypothesis that an intra-articular placebo injection has efficacy for treating knee OA. To explore the possibility that most of the benefit derived from intraarticular hyaluronic acid injection is related to the injection itself and not to the content of the injection, we pooled the change from baseline in the hyaluronic acid treatment groups separately from the change from baseline in injection placebo groups for the 8 trials that reported change from baseline and provided SDs for the change.13-16,19,24,31,32 The pooled change in the hyaluronic acid arms was 0.82 SD units and the pooled change in the placebo groups arms was 0.65 SD units, suggesting that the intraarticular placebo effect accounted for 79% (0.65 SD units/0.82 SD units) of the efficacy of intra-articular hyaluronic acid. This supports our hypothesis that the majority of the effect of intra-articular hyaluronic acid is an intra-articular placebo effect. We believe that this explains the dramatic relief of joint pain after intra-articular hyaluronic acid injection reported by some patients. Although an alternative explanation could be that there is a unique effect of intraarticular hyaluronic acid injection in a particular subgroup, to our knowledge, none has been consistently identified across multiple trials. Without individual trials in our meta-analysis providing detailed information about specific subgroups, addressing this question is beyond the scope of our review. Compared with patients without arthritis, the synovial fluid of patients with OA has lower viscoelasticity and a re-

3120 JAMA, December 17, 2003—Vol 290, No. 23 (Reprinted)

duced concentration and molecular weight of hyaluronic acid.5 The perception that high-molecular-weight hyaluronic acid is superior to lowermolecular-weight preparations is based on the following claims: high-molecularweight hyaluronic acid normalizes synovial fluid, induces the synovial membrane to synthesize normal hyaluronic acid, restores normal joint function, augments flow of the joint fluid, inhibits hyaluronic acid degradation, and/or results in effective joint lubrication.5,37 Furthermore, the highest-molecularweight formulations of hyaluronic acid, hylans, are chemically cross-linked forms of hyaluronic acid and said to have a longer residence time within the joint.37 However, comprehensive reviews of existing literature evaluating such theories5,37 suggest that there is little evidence to support these theories. Also, the trial by Karlsson et al,31 a head-to-head trial of the highest-molecular-weight hyaluronic acid vs a lower-molecularweight formulation, showed no difference between the 2 treatments. In a separate trial evaluating the highestmolecular-weight hyaluronic acid against the denatured form of the same compound, there was equivalent efficacy.38 Denaturing the compound eliminated the viscoelastic properties of the highmolecular-weight hyaluronic acid, suggesting that viscosupplementation does not afford intra-articular hyaluronic acid injections their efficacy.38 In total, 3 trials included in this metaanalysis evaluated the highest-molecularweight hyaluronic acid, the 2 outlier trials of this meta-analysis (Scale et al 20 [n=30]), Wobig et al25 [n=117]), and the trial by Karlsson et al31 (n=105). The 2 outlier trials, both industry sponsored, showed effect sizes in excess of 1.5 (comparable with a total knee replacement). The trial conducted by Karlsson et al was also industry sponsored, but unlike the others, it was sponsored by the manufacturer of a competing formulation of hyaluronic acid; it showed an effect size of 0.1, smaller than our mean effect (Figure 2). With so few studies evaluating the higher-molecular-weight treatment and the presence of heterogeneity among

©2003 American Medical Association. All rights reserved.

Downloaded from www.jama.com on March 26, 2008

HYALURONIC ACID IN TREATMENT OF KNEE OSTEOARTHRITIS

these studies, the hypothesis that highermolecular-weight hyaluronic acid injections have greater efficacy than those of lower-molecular-weight can be neither confirmed nor refuted based on this meta-analysis. In summary, we have found that intra-articular hyaluronic acid, at best, has a small effect in the treatment of knee OA compared with an intra-articular placebo. The likely presence of publication bias suggests that even this effect may be an overestimate. Approximately 80% of the treatment effect of intra-articular hyaluronic acid was accounted for by the placebo effect of an intra-articular injection. Compared with lower-molecular-weight hyaluronic acid, the highest-molecular-weight formulation may have greater effects, but heterogeneity of these studies precludes definitive conclusions. Recommendations for the use of these compounds should be reevaluated based on the findings of this meta-analysis. Further independent trials using intentto-treat analyses are needed to establish whether higher-molecular-weight hyaluronic acid has greater efficacy and to identify a subgroup of patients who have a greater response to this treatment, if such a subgroup exists. Author Affiliations: The Clinical Epidemiology Research and Training Unit (Drs Lo and Felson) and the Arthritis Center (Drs Lo, McAlindon, and Felson), Boston University School of Medicine and Department of Biostatistics (Dr LaValley), Boston University School of Public Health, Boston, Mass. Dr McAlindon is now with New England Medical Center, Boston, Mass. Funding/Support: This study was supported by grant AR 47785 from the National Institutes of Health. Acknowledgment: We thank Camlin Tierney, PhD, Biostatistics Department, Harvard School of Public Health, for her help in translating a trial that was published in German. REFERENCES 1. Mazieres B, Bannwarth B, Dougados M, Lequesne M. EULAR recommendations for the management of knee osteoarthritis: report of a taskforce of the Standing Committee for International Clinical Studies Including Therapeutic Trials. Joint Bone Spine. 2001; 68:231-240. 2. Pendleton A, Arden N, Dougados M, et al. EULAR recommendations for the management of knee osteoarthritis: report of a taskforce of the Standing Committee for International Clinical Studies Including Therapeutic Trials (ESCISIT). Ann Rheum Dis. 2000; 59:936-944. 3. Recommendations for the medical management of

osteoarthritis of the hip and knee: 2000 update. American College of Rheumatology Subcommittee on Osteoarthritis Guidelines. Arthritis Rheum. 2000;43: 1905-1915. 4. Felson DT, Anderson JJ. Hyaluronate sodium injections for osteoarthritis: hope, hype, and hard truths. Arch Intern Med. 2002;162:245-247. 5. Brandt KD, Smith GN Jr, Simon LS. Intraarticular injection of hyaluronan as treatment for knee osteoarthritis: what is the evidence? Arthritis Rheum. 2000; 43:1192-1203. 6. Synvisc-Hylan G-F 20 [package insert]. Ridgefield, NJ: Biometrics Inc; 2000. 7. Altman R, Brandt K, Hochberg M, et al. Design and conduct of clinical trials in patients with osteoarthritis: recommendations from a taskforce of the Osteoarthritis Research Society: results from a workshop. Osteoarthritis Cartilage. 1996;4:217-243. 8. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Hillsdale, NJ: L Erlbaum Associates; 1988:xxi, 567. 9. Felson DT. The verdict favors nonsteroidal antiinflammatory drugs for treatment of osteoarthritis and a plea for more evidence on other treatments. Arthritis Rheum. 2001;44:1477-1180. 10. Liang MH, Larson MG, Cullen KE, Schwartz JA. Comparative measurement efficiency and sensitivity of five health status instruments for arthritis research. Arthritis Rheum. 1985;28:542-547. 11. Roos EM, Nilsdotter AK, Toksvig-Larsen S. Patient expectations suggest additional outcomes in total knee replacement [abstract]. Arthritis Rheum. 2002; 46(suppl):9. 12. Dixon AS, Jacoby RK, Berry H, Hamilton EB. Clinical trial of intra-articular injection of sodium hyaluronate in patients with osteoarthritis of the knee. Curr Med Res Opin. 1988;11:205-213. 13. Russell IJ, Michalek JE, Lawrence VA, Lessard JA, Briggs BT, May GS. A randomized, placebo and nointervention controlled, trial of intra-articular 1% sodium hyaluronate in the treatment of knee osteoarthritis [abstract]. Arthritis Rheum. 1992;35(suppl):9. 14. Dougados M, Nguyen M, Listrat V, Amor B. High molecular weight sodium hyaluronate (hyalectin) in osteoarthritis of the knee: a 1 year placebo-controlled trial. Osteoarthritis Cartilage. 1993;1:97-103. 15. Puhl W, Bernau A, Greiling H. Intra-articular sodium hyaluronate in osteoarthritis of the knee: a multicenter, double-blind study. Osteoarthritis Cartilage. 1993;1:233-241. 16. Dahlberg L, Lohmander LS, Ryd L. Intraarticular injections of hyaluronan in patients with cartilage abnormalities and knee pain: a one-year double-blind, placebocontrolled study. Arthritis Rheum. 1994;37:521-528. 17. Creamer P, Sharif M, George E, et al. Intraarticular hyaluronic acid in osteoarthritis of the knee: an investigation into mechanisms of action. Osteoarthritis Cartilage. 1994;2:133-140. 18. Henderson EB, Smith EC, Pegley F, Blake Dr. Intraarticular injections of 750 kD hyaluronan in the treatment of osteoarthritis: a randomised single centre double-blind placebo-controlled trial of 91 patients demonstrating lack of efficacy. Ann Rheum Dis. 1994; 53:529-534. 19. Cohen MA, Shiroky JB, Ballechey ML, Neville C, Esdaile JM. Double-blind randomized trial of Intraarticular Hyaluronate in the treatment of osteoarthritis of the knee. Arthritis Rheum. 1994;34(suppl):6. 20. Scale D, Wobig M, Wolpert W. Viscosupplementation of osteoarthritic knees with hylan: a treatment schedule study. Curr Therapeut Res. 1994;55:220232. 21. Corrado E, Peluso GF, Gigliotti S. The effects of intra-articular administration of hyaluronic acid on osteoarthritis of the knee: a clinical study with immunological and biochemical evaluations. Eur J Rheumatol Inflamm. 1995;15:47-56.

©2003 American Medical Association. All rights reserved.

22. Sala SF, Miguel RE. Intra-articular hyaluronic acid in the treatment of osteoarthritis of the knee: a shortterm study. Eur J Rheumatol Inflamm. 1995;15:3338. 23. Carrabba M, Paresce E, Angelini M, Re KA, Torchiana EEM, Perbellini A. The safety and efficacy of different dose schedules of hyaluronic acid in the treatment of painful osteoarthritis of the knee with joint effusion. Eur J Rheumatol Inflamm. 1995;15:25-31. 24. Lohmander LS, Dalen N, Englund G, et al, for the Hyaluronan Multicentre Trial Group. Intra-articular hyaluronan injections in the treatment of osteoarthritis of the knee: a randomised, double blind, placebo controlled multicentre trial. Ann Rheum Dis. 1996;55: 424-431. 25. Wobig M, Dickhut A, Maier R, Vetter G, et al. Viscosupplementation with hylan G-F 20: a 26-week controlled trial of efficacy and safety in the osteoarthritic knee. Clin Ther. 1998;20:410-423. 26. Altman RD, Moskowitz R, for the Hyalgan Study Group. Intraarticular sodium hyaluronate (Hyalgan) in the treatment of patients with osteoarthritis of the knee: a randomized clinical trial. J Rheumatol. 1998;25: 2203-2212. 27. Huskisson EC, Donnelly S. Hyaluronic acid in the treatment of osteoarthritis of the knee. Rheumatology (Oxford). 1999;38:602-607. 28. Brandt KD, Block JA, Michalski JP, Moreland LW, Caldwell JR, Lavin PT, for the ORTHOVISC Study Group. Efficacy and safety of intraarticular sodium hyaluronate in knee osteoarthritis. Clin Orthop. 2001; 385:130-143. 29. Tamir E, Robinson D, Koren R, Agar G, Halperin N. Intra-articular hyaluronan injections for the treatment of osteoarthritis of the knee: a randomized, double-blind placebo controlled study. Clin Exp Rheumatol. 2001;19:265-270. 30. Petrella RJ, DiSilvestro MD, Hildebrand C. Effects of hyaluronate sodium on pain and physical functioning in osteoarthritis of the knee: a randomized, double-blind, placebo-controlled clinical trial. Arch Intern Med. 2002;162:292-298. 31. Karlsson J, Sjogren LS, Lohmander LS. Comparison of two hyaluronan drugs and placebo in patients with knee osteoarthritis: a controlled, randomized, double-blind, parallel-design multicentre study. Rheumatology (Oxford). 2002;41:1240-1248. 32. Pham T, Le Hananff A, Ravaud P. Lack of symptomatic and structural efficacy of a new hyaluroninc acid compound, (NRD101), when compared to diacerein and placebo in a one year controlled study in symptomatic knee osteoarthritis [abstract]. Artritis Rheum. 2003;48(suppl):9. 33. Jubb R, Piva S, Beinat L, Dacre I, Gishen P. A oneyear, randomised, placebo (saline) controlled clinical trial of 500-730 kDa sodium hyaluronate (Hyalgan) on the radiological change in osteoarthritis of the knee. Int J Clin Pract. 2003;57:467-474. 34. Sterne JA, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol. 2000; 53:1119-1129. 35. Lexchin J, Bero LA, Djulbegovic B, Clark O. Pharmaceutical industry sponsorship and research outcome and quality: systematic review. BMJ. 2003;326: 1167-1170. 36. Kirwan J. Is there a place for intra-articular hyaluronate in osteoarthritis of the knee? Knee. 2001;8: 93-101. 37. Aviad AD, Houpt JB. The molecular weight of therapeutic hyaluronan (sodium hyaluronate): how significant is it? J Rheumatol. 1994;21:297-301. 38. Allard S, O’Regan M. The role of elastoviscosity in the efficacy of viscosupplementation for osteoarthritis of the knee: a comparison of hylan G-F 20 and a lower-molecular-weight hyaluronan. Clin Ther. 2000; 22:792-795.

(Reprinted) JAMA, December 17, 2003—Vol 290, No. 23

Downloaded from www.jama.com on March 26, 2008

3121