Utility of APOE testing for reducing ARIA under probabilistic stopping rates to treat with anti-amyloid therapy for ε4-homozygote patients: A simulation study.

BACKGROUND: APOE ε4/ε4 genotype increases the risk of Amyloid-Related Imaging Abnormalities (ARIA) from anti-amyloid antibody treatment (AAT). While guidelines recommend testing, its practical utility depends on the resulting probability (p) that treatment is actually withheld for ε4-homozygotes, which varies significantly across clinical settings. OBJECTIVES: To quantify the Number Needed to Test (NNT) to prevent one ARIA event as a function of p of withholding AAT in ε4/ε4 patients. DESIGN: A Bayesian simulation study using a Beta-Binomial model to analyze genotype-stratified contingency tables. SETTING: Data were derived from two published, phase 3 clinical trials: Clarity-AD (lecanemab) and TRAILBLAZER-ALZ 2 (donanemab). PARTICIPANTS: Aggregate data from source trials. INTERVENTION: Simulation of varying treatment discontinuation probability p from 0 (none) to 1 (universal for ε4-homozygotes). MEASUREMENTS: NNT to prevent one ARIA event (any ARIA-E, any ARIA-H, and symptomatic ARIA-E) and the fractional reduction in total ARIA events as a function of p. RESULTS: NNTs increased (worsened) significantly as p decreased. Under the most conservative policy (p = 1), the median NNT to prevent one any ARIA-E event was 20-30 (lecanemab) and 15-25 (donanemab), yet this only reduced total ARIA events by 10-30%. The NNT to prevent one symptomatic ARIA-E (lecanemab) was substantially higher, at 70-90 (at p = 1). CONCLUSIONS: The direct safety impact of APOE testing for ARIA mitigation is limited, even under universal discontinuation policies. Its primary value lies in supporting shared decision-making and operational planning rather than as a standalone safety lever.