Safety and efficacy of S-adenosylmethionine (SAMe) for osteoarthritis

For pain and functional impairment we computed the difference in the average response between treatment groups and control groups, standardized to account for differences in the measurement scale across studies. The result is a difference effect size (ES) with a positive ES favoring SAMe. We also applied a correction factor¹⁰ that adjusts for the positive bias in the ES estimate for small samples. For the binary outcome of adverse effects, we computed the odds ratio (OR) for the individual trials.¹¹ An OR of less than 1 indicated that treatment with SAMe was more effective than the control.

Heterogeneity in the strategy to measure pain was expected. Either individual studies pooled several pain items (eg, day pain and resting pain) that were rated using a 4- or 5-point rating scale or Visual Analog Scale (VAS), or studies used a single-item VAS. Functional limitation reflects stiffness, swelling, and joint mobility as rated by the physician according to the degree of joint movement (eg, flexion, extension, abduction, adduction, and rotation). In some studies, this score also included a pain item. Adverse effects refer to patient reports of nonspecific gastrointestinal complaints, mucocutaneous symptoms, and central nervous systems disturbances. Finally, a pooled dropout rate because of side effects was computed across studies as a measure of the tolerability of SAMe.

Statistical analysis

Outcomes for each subject measured at multiple time points tend to be correlated, which introduces dependency between corresponding ESs. To avoid this dependency, we computed the ES for the end-of-treatment only, rather than for all time points. Although dependency is also a concern when results are reported for more than one outcome within a study,^12-14 we did not control for this. Following the test for homogeneity or consistency within the set of ESs using the Q statistic with α = .10,11we computed the weighted mean ES with 95% confidence intervals (CI) across studies for each outcome, weighting for sample size (the inverse of the variance). The choice of a fixed-effects model was dependent on the finding of homogeneity of results.

To assess sensitivity of the results, we examined the relationship of the ES to the dosage of SAMe, length of treatment, and study quality rating. Subgroup analyses examined differences related to the location of the OA to estimate the robustness of results. Finally, we assessed potential publication bias informally by using the funnel plot of ES by precision, and statistically through the rank correlation between the standardized ES and standardized study variance.¹⁵

Results

Description of studies

Twenty studies were identified through our search and 11 of them^16-26 met the inclusion criteria (Table). We excluded one duplicate study₂₇and one study whose sample included persons with rheumatoid arthritis.²⁸ Other excluded studies compared the routes of administration of SAMe,²⁹ compared SAMe plus ketoprofen with ketoprofen alone,³⁰ or were not randomized controlled trials.^31-34 Four of the included studies^18,20,21,25 were published in Italian; the others were published in English. The majority of studies (7 of 11) were conducted in Italy.

Quality assessment

Percent agreement between raters for the items on the Jadad scale averaged 87.5%. Following discussion, the raters reached consensus for all items. Using Jadad’s criteria, all studies were rated of high quality (score 3), although only 2 studies^16,23 included a description of the method of randomization. None of the studies addressed allocation concealment.

Study characteristics

Ten of the 11 studies used a parallel groups design including one with 3 arms¹⁹; the 11th one²⁵ used a crossover design (Table W1).* The SAMe dosage in 6 studies was 1200 mg per day orally^{18,19,22-24,26}; 3 studies used 600 mg per day orally^17,21,25; and one used 400 mg per day intravenously.²⁰ In one study¹⁶ the dosage varied. Duration of treatment ranged from 10 days to 84 days; a duration of 28 or 30 days was used in 8 of the studies. A variety of NSAIDs served as active comparators and 2 studies^16,19 used placebo. The studies involved 1442 subjects with a mean age of 60.3 years, of whom 70.1% were women. Mean duration of OA was 5.7 years, ranging from 2.6 years to 9.1 years. In 5 studies, the majority of subjects had OA of the knee; across all studies 54.2% of the subjects had OA of the knee.

TABLE
Characteristics of studies included in meta-analysis