Biomarkers for the prediction of cognitive decline in atrial fibrillation patients

Abstract

Abstract Background Patients with atrial fibrillation (AF) are at increased risk of cognitive decline. Biomarkers might help to identify the highest risk patients. Purpose To investigate associations of a broad panel of biomarkers with cognitive decline in a large population of AF patients. Methods We enrolled 1440 AF patients with available baseline biomarkers and cognitive testing by the Montreal Cognitive assessment (MoCA) score at inclusion and at ≥2 yearly follow-ups over three years. Cognition over time was assessed by modelling of a slope over three years. Cognitive decline was defined as a decreasing slope based on 1 standard deviation of the baseline MoCa score. We investigated the associations of biomarkers and age with cognitive decline in univariate logistic regression models by AUCs. We used Lasso modelling to build combined prediction models with biomarkers, clinical variables and both. Results Mean age was 72 years, 75% were male, 47% had paroxysmal AF and 90% were anticoagulated. During a follow-up of 3 years, cognitive decline occurred in 93 patients (6.5%). Patients with cognitive decline were older (77 vs 72 years, p=0.001), more often had paroxysmal AF (50 vs 47%, p=0.007) or a history of stroke (24 vs 11%, p=0.001), but there were no differences in anticoagulation (94 vs 90%, p=0.289). The three biomarkers with the highest univariate AUC for cognitive decline were Growth Differentiation Factor (GDF)-15 (0.67 [0.62-0.72]), Cystatin C (0.67 [0.61-0.72]) and high-sensitivity Troponin T (0.65 [0.60-0.70]), while age had the highest overall AUC (0.70 [0.64-0.75]) (Figure). The combined prediction model with the highest AUC of 0.69 (0.67-0.72) included GDF-15, age and baseline cognition. Conclusion Over 3 years, 6.5% of AF patients had cognitive decline despite a high rate of anticoagulation. Besides age, GDF-15, Cystatin C and high-sensitivity Troponin T had the highest predictive value for cognitive decline.Univariate AUCs for biomarker/ageFigure legend