**Background:** Cognitive impairment and dementia affect approximately one-fifth of the elderly population globally, with an estimated 46.8 million people aged ≥60 years living with dementia in 2015, projected to reach 131.5 million by 2050. Modifiable lifestyle risk factors include physical inactivity, social isolation, and cognitive inactivity. While multidomain interventions targeting multiple risk factors have shown promise, methodological issues such as low adherence and insufficient intensity have limited previous trials. This study aimed to evaluate a community-based multidomain intervention combining dual-task exercise and social engagement over 10 months on cognitive and physical function in older adults with mild to moderate cognitive decline.
**Methods:** Participants were recruited from a sub-cohort of the National Center for Geriatrics and Gerontology-Japan Study of Geriatric Syndrome (NCGG-SGS) in Midori Ward, Nagoya, Japan. From 24,271 older adults invited by mail, 5,257 participated in screening. After applying inclusion criteria (MMSE score 21–24, no gait dysfunction, no serious health problems) and exclusions, 280 participants were randomized 1:1 to intervention or control groups. The intervention group (n=140) attended once-weekly 90-minute exercise sessions (40 sessions total) including warm-up, 30 minutes of aerobic exercise at 60% maximum heart rate, 30 minutes of dual-task training (cognitive tasks during exercise, termed 'COGNICISE'), and cool-down. They also participated in twice-monthly mentally stimulating social activities (14 sessions) in groups of 4–5 participants. The control group (n=140) received three 1-hour educational lessons on nutrition, oral care, and healthy longevity. Assessments were conducted at baseline and 10 months by blinded assessors. Primary outcomes were cognitive function measured by the NCGG-FAT (word list memory, logical memory, Trail Making Test A and B, Digit Symbol Substitution Test). Secondary outcomes included physical function (grip strength, 5-times chair stand, gait speed, 6-minute walk test), social engagement (conversation time, Lubben Social Network Scale), and accelerometer-measured physical activity (daily steps, MVPA). Statistical analysis used repeated-measures ANCOVA adjusting for age, gender, education, and medication, with intention-to-treat analysis using last observation carried forward for missing data.
**Key Results:** The 10-month retention rate was 90.0% in the intervention group and 91.4% in the control group. Baseline characteristics were similar between groups (mean age 76.4 years, SD=4.1; 39.6% women; mean education 11.9 years, SD=2.6). For primary outcomes, logical memory recognition scores showed a significant time×group interaction (F=5.04, p=0.026), with the intervention group improving from 6.4 to 6.9 (p<0.001) while the control group showed no significant change (6.5 to 6.6). Both groups improved in immediate logical memory (intervention: 5.4 to 5.8, p<0.01; control: 5.4 to 5.8, p<0.05) and recall logical memory (intervention: 5.3 to 5.7, p<0.05; control: 5.1 to 5.6, p<0.001), with no significant between-group differences. No significant group differences were found for TMT-A, TMT-B, digit symbol substitution, or word list memory tasks. For secondary outcomes, significant time×group interactions were observed for 6-minute walk test (F=12.48, p<0.001; intervention: 453.5 to 463.3 m, p<0.001; control: 450.0 to 447.5 m, non-significant), daily steps (F=7.99, p=0.005; intervention: 6916.7 to 7150.4, p<0.01; control: 6478.8 to 6243.2, p<0.01), and MVPA (F=4.62, p=0.033). The 5-times chair stand improved in the intervention group (7.6 to 7.2 sec, p<0.05) but not in controls. Grip strength significantly decreased in the control group (28.4 to 27.9 kg, p<0.01) but not in the intervention group. Time spent discussing health increased in the intervention group (2.2 to 3.2, p<0.001) with a trend toward group interaction (F=3.74, p=0.054). Social network scores showed a significant interaction in per-protocol analysis (F=4.13, p<0.05) but not in ITT analysis.
**Clinical Implications:** This study demonstrates that a 10-month community-based multidomain intervention combining dual-task exercise and social activity can produce modest improvements in logical memory recognition, aerobic fitness (6-minute walk distance), and daily physical activity (steps and MVPA) in older adults with mild to moderate cognitive decline. The intervention group maintained grip strength while the control group declined. The findings support the potential of non-pharmacological multidomain programs as a preventive strategy for cognitive decline, though effects were limited to specific cognitive domains (memory) and did not extend to attention, executive function, or processing speed. The high adherence rate (81.4% exercise attendance; average 33 of 40 sessions) suggests the program was feasible and acceptable. Limitations include the relatively short 10-month follow-up, lack of genetic outcome data (e.g., APOE ε4), and the possibility that the intervention intensity was insufficient to produce broader cognitive effects. Longer-term follow-up on dementia onset is ongoing.