Twenty-nine randomised controlled trials (RCT)s were included (n=13,443). Twenty-seven were of a parallel-group design and two were of a crossover design.
Neuropsychiatric outcomes.
The data from studies using the NPI and ADAS-noncog scales were pooled separately, owing to evidence of statistical heterogeneity when they were pooled together. In studies reporting the NPI (6 RCTs, n=2,927), there was a small improvement of 1.72 points (95% CI: 0.87, 2.57) in patients receiving ChIs compared with placebo. In studies reporting the ADAS-noncog (10 RCTs, n=2,602), there was an improvement of 0.03 points (95% CI: 0.00, 0.05) in patients receiving ChIs compared with placebo. The test for statistical heterogeneity was non significant for each group of studies.
Functional outcomes.
In studies using ADL measures (14 RCTs, n=3,738), patients receiving ChIs improved by 0.10 SD (95% CI: 0.00, 0.19) compared with placebo. In studies using instrumental ADL measures (13 RCTs, n=4,176), patients receiving ChIs improved by 0.09 SD (95% CI: 0.01, 0.17) compared with placebo. The test for statistical heterogeneity was non significant for each group of studies.
Similar effect sizes were found following most of the sensitivity analyses. For the NPI, the effect size was slightly greater for the completed participants analysis than for the ITT analysis.
There was no evidence of publication bias.