Baseline MRI markers and clinical outcomes during anti-amyloid β therapy: A real-world study in a Japanese dementia center

IMPORTANCE: Predictors of decline and of amyloid-related imaging abnormalities (ARIA) during anti-amyloid β therapy in routine care are undefined. OBJECTIVE: To identify baseline predictors of cognitive, functional, and safety outcomes. DESIGN, SETTING, AND PARTICIPANTS: Retrospective cohort of 161 consecutive patients initiating anti-amyloid β therapy at a Tokyo dementia center, 2024 to 2026 (lecanemab, 89; donanemab, 72). EXPOSURES: Lecanemab or donanemab. MAIN OUTCOMES AND MEASURES: Change in MMSE and CDR-Sum of Boxes (CDR-SB); baseline medial temporal atrophy (VSRAD Z score), amyloid PET burden, and microbleeds; ARIA. RESULTS: With lecanemab, MMSE and CDR-SB declined significantly at every assessment (CDR-SB +0.79, +1.10, +1.60 at 6, 12, 18 months); memory and orientation declined while language was preserved. Greater baseline medial temporal atrophy was associated with greater 12-month CDR-SB worsening (r = 0.39; P = .01; n = 40); patients in the most severe standard category (Z ≥ 3) worsened by 2.62 versus 0.64 points (adjusted β = +2.18; P < .001). Atrophy was associated with older age (r = 0.19; P = .03) and with baseline memory-related measures - ADAS-Cog score (r = 0.32; P < .001) and the MMSE subdomains of delayed recall (β = -0.25; P = .006) and time orientation (β = -0.30; P = .04) - but not MMSE total score. Amyloid PET burden showed no significant association at most time points. All ARIA events were asymptomatic; ARIA-H occurred in 6 of 25 with microbleeds versus 3 of 136 without (odds ratio, 14.0; P < .001). Donanemab follow-up was 6 months. CONCLUSIONS AND RELEVANCE: Baseline medial temporal atrophy was associated with functional decline and microbleeds with ARIA-H. Pretreatment MRI may provide prognostic and safety information during anti-amyloid β therapy.