Does effect size in naltrexone trials for alcohol dependence differ for single-site vs. multi-center studies?

Background: The opioid antagonist naltrexone was first shown in single-site trials to be efficacious in the treatment of alcohol dependence. Recent clinical trials of the medication have used multi-center designs, which permit greater generalization and increased statistical power. We compared effect sizes for these two kinds of trial design on the hypothesis that multi-center trials introduce sources of variation that reduce the observed effect size. Methods: A meta-analysis of data from 19 placebo-controlled trials of the efficacy of naltrexone (7 multi-center and 12 single-site studies) was performed. Effect size estimates for these two study designs were compared using two outcomes: percentage of days drinking and percentage of subjects relapsing to heavy drinking. Results: Compared with single-site studies, multi-center studies were estimated to yield a nonsignificantly smaller effect on the percentage of days drinking (Cohen's d = 0.20 vs. 0.33, respectively) and a significantly smaller effect on the percentage of subjects relapsing to heavy drinking (Cohen's d = 0.17 vs. 0.41, respectively; p = 0.014). Earlier studies showed a larger effect size than later studies. Conclusion: The smaller effect size seen with multi-center studies may reflect random error due to heterogeneity among the sites. However, because multi-center studies were, in general, conducted more recently than single-site studies, it was not possible conclusively to disentangle the moderating impact of study type and year of publication on effect size. Further research on factors that moderate effect size can contribute to the development of medications to treat alcohol dependence.