Objective: To test the hypothesis that a long-term, structured cognitive leisure activity program is more effective than a health education program at reducing the risk of further cognitive decline in older adults with mild cognitive impairment syndrome (MCI), a high risk for dementia.

Design: A 3-arm, single-blind randomized controlled trial.

Setting: Community.

Participants: A total of 201 Japanese adults with MCI (mean age: 76.0 years, 52% women).

Interventions: Participants were randomized into 1 of 2 cognitive leisure activity programs (60 minutes weekly for 40 weeks): dance (n = 67) and playing musical instruments (n = 67), or a health education control group (n = 67).

Measurements: Primary outcomes were memory function changes at 40 weeks. Secondary outcomes included changes in Mini-Mental State Examination and nonmemory domain (Trail Making Tests A and B) scores.

Results: At 40 weeks, the dance group showed improved memory recall scores compared with controls [mean change (SD): dance group 0.73 (1.9) vs controls 0.01 (1.9); P = .011], whereas the music group did not show an improvement compared with controls (P = .123). Both dance [mean change (SD): 0.29 (2.6); P = .026] and music groups [mean change (SD): 0.46 (2.1); P = .008] showed improved Mini-Mental State Examination scores compared with controls [mean change (SD): -0.36 (2.3)]. No difference in the nonmemory cognitive tests was observed.

Conclusions: Long-term cognitive leisure activity programs involving dance or playing musical instruments resulted in improvements in memory and general cognitive function compared with a health education program in older adults with MCI.

Trial registration: UMIN-CTR UMIN000014261.