Six randomised placebo controlled trials were included in the meta-analyses (n=1,032). The sample sizes ranged from 107 to 226. Drop out rates varied from 14% to 63% in the lithium arm and 13% to 65% in the placebo arm.
An overall medium effect size of 0.40 (95% CI: 0.28 to 0.53) was determined. There was no evidence of statistically significant heterogeneity. The overall difference in response rate between lithium and placebo was 17% (95% CI 0.08 to 0.26) and there was no evidence of statistically significant heterogeneity. The number of patients needed to be treated for one patient to benefit was approximately six (95% CI 3.8 to 12.5). The statistical significance of these findings was unclear.
When the one study that showed a lack of assay sensitivity was removed from the analysis, the overall difference between lithium and placebo became 20% and the number needed to treat became five. There was a large variation in the improvement observed in the placebo group.