Warfarin anticoagulation control and risks of cerebral small vessel disease and cognitive impairment among elderly patients with ischemic stroke and atrial fibrillation: a retrospective cohort study

Objective This study applied individual exposure metrics derived from population pharmacokinetic-pharmacodynamic (PopPK/PD) modelling to investigate the associations between warfarin anticoagulation quality, cerebral small vessel disease (CSVD) burden, and cognitive outcomes among older adults with ischemic stroke and non-valvular atrial fibrillation (NVAF). Methods We included 300 patients aged ≥65 years with ischemic stroke and AF who had received warfarin for ≥3 months between January and December 2025. A nonlinear mixed-effects model was used to estimate individualized pharmacokinetic parameters and steady-state exposure metrics. Multivariable regression was performed to assess the associations of these metrics with CSVD burden, cognitive impairment, and the composite endpoint; the ability to identify high CSVD burden was compared between model-derived exposure metrics and conventional anticoagulation quality metrics, including time in therapeutic range (TTR) and international normalized ratio (INR) variability. Results Steady-state average concentration (Cavg ss ) was significantly positively associated with CSVD burden (per 100 ng/mL increment: OR = 1.32, p < 0.001), whereas TTR was not associated with burden ( p = 0.444). High CSVD burden was significantly associated with cognitive impairment (Cox HR = 2.04, p = 0.007). The incidence of cognitive impairment increased stepwise across exposure tertiles (42.2%, 54.0%, and 75.4%, p < 0.001). The composite endpoint occurred significantly more often in patients with INR standard deviation (INR_SD) > 0.5 ( p = 0.009). For identifying high CSVD burden, PopPK/PD-derived exposure parameters demonstrated differing within-cohort discrimination for identifying high CSVD burden: steady-state peak concentration (Cmaxss), Cavgss, and degree of fluctuation (DF) (AUC = 0.695, 0.679, and 0.651, respectively), with better discrimination compared with INR_SD and TTR (AUC = 0.557 and 0.493, respectively; p < 0.05). Conclusion Within this cohort, PopPK/PD-derived individualized exposure parameters showed better discrimination for identifying patients with high CSVD burden compared with traditional TTR and INR variability, although discriminative ability remained modest for all markers, including those whose anticoagulation quality appeared adequate by conventional metrics. These findings provide a pharmacological foundation for refined anticoagulation quality assessment in this high-risk elderly population.