Search

Objective Recent meta-analyses on dose-response relationships of SSRIs are largely based on indirect evidence. We analyzed RCTs directly comparing different SSRI doses. Method Systematic literature search for RCTs. Two raters independently screened articles and extracted data. Across SSRIs, doses defined as low, medium, and high doses, based on drug manufacturers' product monographs, were analyzed in pairwise random-effects meta-analyses and in a sensitivity network meta-analysis with regard to differences in antidepressive efficacy (primary outcome). We also analyzed all direct comparisons of different dosages of specific SSRIs. (Prospero CRD42018081031). Results Out of 5333 articles screened, we included 33. Comparisons of dosage groups (low, medium, and high) resulted in only small and clinically non-significant differences for SSRIs as a group, the strongest relating to medium vs low doses (SMD: -0.15 [95%-CI: -0.28; -0.01) and not sustained in a sensitivity analysis. Among different doses of specific SSRIs, no statistically significant trend emerged for efficacy at higher doses, but 60 mg/day fluoxetine are statistically significantly inferior to 20 mg/day. Paroxetine results are inconclusive: 10 mg/day are inferior to higher doses, but 30 and 40 mg/day are inferior to 20 mg/day. Meaningful effects cannot be ruled out for certain drugs and dosages, often investigated in only one trial. Dropout rates increase with dose-particularly due to side effects. Network meta-analyses supported our findings. Conclusions There is no conclusive level I or level II evidence of a clinically meaningful dose-response relationship of SSRIs as a group or of single substances. High SSRI doses are not recommended as routine treatment.

ObjectiveTo evaluate the efficacy and tolerability of baclofen vs. placebo for long-term treatment of alcohol use disorder. MethodSystematic review and meta-analysis following methods of the Cochrane Collaboration Handbook (PROSPERO registration: CRD42017073663). Primary outcome was the random-effects summary estimate of all standardized mean differences (SMDs), as calculated from the primary outcomes of each study. ResultsFourteen double-blind RCTs (1522 patients) were included. Heterogeneity was substantial for most analyses (I-2 about 75%). Baclofen showed a small, but not statistically significant superiority over placebo: SMD=0.22 ([95% CI: -0.03; 0.47], P=0.09). This result was supported by a leave-one-out-analysis, and Orwin's fail-safe N, by predefined secondary analyses (on abstinence rates and amount of drinking), and by a post hoc-analysis of high-dose studies (>80mg/day). An analysis of low risk of bias studies (SMD=0.10 [-0.20; 0.41], P=0.51, I-2=43.3%) found no effect. Exclusion of four studies focusing on patients with comorbidity yielded a small positive effect. Drop-out rates were similar. ConclusionOur results question baclofen's utility in the long-term treatment of alcohol use disorder at both normal and high doses. While the confidence intervals indicate that marginally harmful or moderately beneficial effects of baclofen remain a possibility, the most likely effect size is slightly above placebo effects.