Performance of anticoagulation risk scores for predicting ischaemic stroke in dialysis patients with atrial fibrillation

Type of funding sources: None.

Risk models such as the CHADS₂ and revised CHA₂DS₂-VASc scores are recommended for anticoagulation decision-making in atrial fibrillation (AF) patients. Despite this, their prognostic utilities in dialysis patients with AF are not well established. We evaluated the performance of conventional anticoagulation risk scores at predicting adverse outcomes in dialysis patients with AF.

Pubmed database was searched until 3/20/2021 for studies reporting the area under receiver-operative characteristics curve (AUC) and/or hazards ratios (HR) of anticoagulation scores for adverse outcomes in dialysis patients with AF. Data was pooled using the random-effects DerSimonian-Laird method and OpenMetaAnalyst software.

Amongst 260 studies screened, 34 full-text articles were reviewed and 7 eligible studies totalling 13,693 patients were analysed. Pooled AUCs (95% confidence intervals) for discriminating ischemic stroke were 0.705 (0.574-0.836) in 4 studies for the CHADS₂ score, and 0.719 (0.606-0.832) in 3 studies for the CHA₂DS₂-VASc score. Pooled HRs (95% confidence intervals) for stroke were 1.66 (1.26-2.18) in 4 studies for the CHADS₂ score, and 1.40 (1.17-1.67) in 3 studies for the CHA₂DS₂-VASc score. There were significant heterogeneity in all analyses (I2 inconsistency statistic 80.5-93.1%, P < 0.001-0.006). Only one study reported either AUC or HR for predicting other outcomes such as bleeding and deaths in these patients, insufficient for pooled analyses.

The CHADS₂ and CHA₂DS₂-VASc scores had moderate discriminative ability for ischemic stroke during follow-up, comparing similarly to general AF populations. Further studies are required to assess if these scores can also guide selection for anticoagulation in this challenging to manage high-risk cohort.