Seven RCTs (833 patients) were included in the meta-analysis. Four RCTs were journal publications and 4 RCTs were internal company reports (Dupont Merck).
The populations differed between the studies in terms of the age of the patients, the number of years of drinking, the percentage employed, and the percentage in a stable relationship.
The quality scores ranged from 10 to 11 (maximum 12).
Relapse rates.
Naltrexone was significantly more effective in preventing relapse into heavy drinking than placebo. Using a random-effects model, the RD was -14% (95% CI: -23, -5) and the pooled RR was 0.72 (95% CI: 0.55, 0.94). Significant heterogeneity was detected (chi-squared 11.9, d.f.=6, p=0.06)
Abstinence.
Naltrexone was associated with significantly higher abstinence rates than placebo. Using the fixed-effect model, the RD was 10% (95% CI: 3.5, 16.3) and the pooled RR was 1.28 (95% CI: 1.08, 1.52). No significant heterogeneity was detected (chi-squared 9.7, d.f.=6, p=0.14).
Alcohol consumption outcomes.
Naltrexone-treated patients reported significantly fewer drinking days than placebo-treated patients. Using a fixed-effect model, the pooled WMD was -3% (95% CI: -5.4, -0.5). No significant heterogeneity was detected (chi-squared 5.75, d.f.=4, p=0.22). Naltrexone-treated patients reported significantly fewer drinks per drinking day per patient than placebo-treated patients. Using a fixed-effect model, the pooled WMD was -1.04% (95% CI: -2.0, -0.1). No significant heterogeneity was detected (chi-squared 2.97, d.f.=3, p=0.40).
Safety.
There was no statistically-significant difference between naltrexone and placebo in the number of patients reporting at least one adverse event (RD 1%, 95% CI: -6, 8) or the number of patients who discontinued the study due to an adverse event (RD 2%, 95% CI: -1, 5). No significant heterogeneity was detected for either outcome (p=0.30 and p=0.65, respectively).