Medical Evidence
Glucosamine (Gl) And Chondroitin (Ch)
Treatment For Osteoarthritis (OA) Of The Knee Or Hip: Meta-Analysis
And Quality Assessment Of Clinical Trials T.E. McAlindon,
J. Gulin, D.T. Felson Arthritis Ctr., Boston Univ., MA 02118,
USA
Background: OA is a major
cause of pain and disability in the population for which effective
treatment is badly needed. GL and CH have been used for over 10
years in Europe, yet have been neglected as credible OA therapies
in the US. Because of their safety, GL and CH would be of value
even if only modestly effective. We performed a meta-analysis
and quality assessment of clinical trials to evaluate their efficacy,
and the quality of the evidence. Studies eligible for inclusion
were double-blind placebo-controlled trials of "e4 weeks duration,
testing oral or parenteral GL or CH for knee or hip OA, which
reported p values and size of treatment effect. Studies were sought
using MEDLINE, manual searches of manuscripts and journal supplements,
and by contacting authors of published manuscripts and content
experts. Study quality was scored independently by two observers
using a validated inventory with range 0-65, in which scores £
33 are considered poor, 34-45 moderate, "e46 good (JAMA 1994 272:
108). Disagreements were treated by adjudication, and by taking
averages.
Method: We decided a priori
to (i) treat the global pain score in the index joint (or, if
unavailable, the Lequesne Index) as the study 1° outcome (ii)
classify a trial as positive if the score in the treated group
at completion was "e25% lower than the placebo group and significant
[p < .05].
We then computed the probability that the observed number of
positive studies might have occurred by chance if the treatments
were in fact ineffective using the sign test, and the number of
negative studies which would be required to make this result null
(p "d .5). 13 trials met eligibility criteria: GL (4 papers, 2
abstracts), CH (5 papers, 2 abstracts). Inter-rater ICC for quality
scoring was 0.87, p = .0001. Quality scores ranged 8-36, mean
21.1, 95% CI 16.9-25.2, and were substantially lower for abstracts
compared with manuscripts (12.2 vs 25.0, p = .001). Deficiencies
related to description of randomization, blinding, and completion
rates. Only 2 included an intent-to-treat analysis. 5 of the 6
GS studies received industrial support. All studies were classified
as positive, and demonstrated large effects - mean score reduction
compared to placebo 39.5% (sd 21.9) for GS, 40.2% (6.4) for CS.
The probabilities for this outcome in the absence of any true
effect are p=0.016 (GS), p=0.008 (CS). The numbers of negative
studies required to nullify these probabilities are 45 (GS) &
62 (CS).
Results: Clinical trials
of GL and CH show substantial benefits in the treatment of OA,
but provide insufficient information about study design and conduct
to allow definitive evaluation. We conclude that further studies
are needed to test the efficacy of GS and CS.
Source: American College
of Rheumatology Annual Meeting November 10, 1998; Abstract: 994.
Poster Session D: Osteoarthritis: Clinical Trials