AIMS: To identify subgroups who may be more likely to respond well to a multidomain intervention among older adults with type 2 diabetes. MATERIALS AND METHODS: This study was a secondary analysis of the Japan Multimodal Intervention Trial for Prevention of Dementia. A total 531 participants aged 65-85 years with mild cognitive impairment were randomized into intervention (vascular risk management, exercise, nutritional counseling, and cognitive training) and control (health-related information) groups. The outcome was the change in average Z scores of neuropsychological tests from baseline to 18 months. Interactions between intervention and age (65-74, 75-85 years), memory impairment (amnestic, nonamnestic), HbA1c levels (within, outside target range), or genotype (0, ≥1 ε4 alleles) among participants with diabetes were evaluated using the mixed-effects model for repeated measures. RESULTS: Among 76 participants with diabetes, a significant age × intervention interaction ( = 0.007) was found, which was driven by benefits in the younger age group (Z score difference: 0.33, 95% CI: 0.09 to 0.55) that were not observed in the older age group. Intervention benefits were also detected in those with HbA1c levels outside the target range (Z score difference: 0.31, 95% CI: 0.06 to 0.56), with HbA1c levels × intervention interaction ( = 0.021). No significant interactions were detected between intervention and memory impairment or genotype. CONCLUSIONS: Multidomain interventions may benefit younger older adults or those with overly strict or lenient HbA1c control; however, these findings need confirmation in future studies.