Abstract Body (Do not enter title and authors here): Background:
Clinical decisions on oral anticoagulation (OAC) in atrial fibrillation (AF) become increasingly complex with age because thromboembolic and hemorrhagic risks rise together, yet evidence guiding whether to start, continue, switch, or discontinue OAC in the very elderly is scarce.
Objective:
To build a decision-support model that balances stroke/systemic embolism (SSE) against intracranial hemorrhage (ICH), thereby guides OAC strategy in patients aged ≥ 75 years with non-valvular AF.
Methods:
We analyzed 32,275 participants from the All Nippon AF in the Elderly (ANAFIE) registry (UMIN000024006). Endpoint-specific predictors identified by Shapley value (AF type, falls, smoking, blood pressure, polypharmacy, prior ablation) supplemented CHA₂DS₂-VA. OAC exposure was a five-level, time-updated variable (no therapy, continued, switch, discontinuation, initiation). For both endpoints, SSE and ICH, a five-fold cross-fitted doubly-robust learner comprising a logistic-regression propensity model and LightGBM outcome model with focal loss estimated 3-month average treatment effects (ATE). Uncertainty was quantified with ≥1,000 patient-level parametric bootstraps. Net benefit (NB) was defined as NetBenefit=−ATE_SSE−1.5×ATE_ICH, assigning 1.5-fold greater disutility to ICH.
Results:
Among 32,275 patients baseline treatment strategies comprised no therapy in 2,042 patients, continued OAC in 23,857, switching in 3,166, discontinuation in 2,852, and initiation in 358. During the predefined 3-month period, SSE occurred in 123 patients (0.38 %), and ICH in 64 (0.20 %). Compared with continued OAC, switching increased ICH by 0.25-percentage-point (pp) (95 % CI, 0.03–0.46), whereas discontinuation increased SSE by 0.35 pp (0.03–0.66). Transitioning from no therapy to initiation raised SSE by 0.12 pp (0.01–0.23) and ICH by 0.31 pp (0.23–0.38). NB was consistently negative for switching and discontinuation, and largely negative for initiation. The adverse impact of switching intensified with advancing age, whereas discontinuation harm was greatest between 75 and 79 years.
Conclusions:
In Japanese patients aged ≥ 75 years with non-valvular AF, switching or discontinuing OAC significantly worsened 3-month thromboembolic or hemorrhagic outcomes, and new initiation did not confer a net short-term benefit. These findings highlight the need for caution when altering OAC strategy in the very elderly and underscore the importance of individualized, data-driven decision support.