Twenty studies were included, of which 11 were RCTs: four (421 participants) in cognitively healthy people and seven (843 participants) in people with cognitive impairment. Eight RCTs met at least six quality criteria.
People without cognitive impairment: Meta-analysis was possible for measures of executive function. The tai chi group were significantly superior to non-active controls (SMD 0.90, 95% CI 0.03 to 1.78; four RCTs; Ι²=92%) and to active comparators (SMD 0.51, 95% CI 0.17 to 0.85; two RCTs; Ι²=0%). For the non-active comparison, removal of one RCT with high risk of bias resulted in a smaller but still statistically significant effect estimate (SMD 0.39) without heterogeneity. Effects were considered small to moderate but clinically significant and equivalent to those reported for other types of exercise and cognitive training.
People with cognitive impairment: Meta-analysis was possible for global cognition measured by the Mini-Mental State Examination. Effects were again small to moderate but considered to be clinically relevant and comparable to exercise and cognitive training (SMD 0.35, 95% CI 0.11 to 0.58 versus non-active controls and 0.30, 95% CI 0.11 to 0.49 versus active controls; both based on four RCTs).
Six RCTs reported on adverse events. Two trials had no events in either group; the remaining trials reported adverse events but none were attributable to tai chi. Results for other outcomes and from non-randomised studies were reported.