By December 31, 2017, a search of a database of the New Territories West Cluster of the Hospital Authority of Hong Kong, China identified patients with non-valvular atrial fibrillation taking 1 of the 3 NOACs. Three cohorts (rivaroxaban vs. dabigatran, apixaban vs. dabigatran, and apixaban vs. rivaroxaban) were set up for comparison, and inverse probability of treatment weighting was used to balance the baseline characteristics. Patients were tracked until drug discontinuation or June 30, 2019 (study end date), whichever came first. The primary endpoint included a combination of a greater than 30% decline in the estimated glomerular filtration rate, acute kidney injury, doubling of serum creatinine, and renal failure. The Cox proportional hazards regression model was used for survival analysis.

Results: A total of 1,153 patients were included in the analyses. Rivaroxaban was associated with a higher incidence of the primary endpoint compared to dabigatran (hazard ratio: 1.43, 95% confidence interval: 1.20–1.70, P < 0.001). Apixaban was associated with a lower risk compared to rivaroxaban in the analysis of the subgroup of patients with an estimated glomerular filtration rate ≥ 60 mL/(min·1.73 m2) (hazard ratio: 0.62, 95% confidence interval: 0.31–0.94, P = 0.010). No significant differences were observed between dabigatran and apixaban.

Conclusions: Rivaroxaban was associated with worse renal outcomes compared to dabigatran in the main cohort and com- pared to apixaban in the subgroup analysis. Variations in renal outcomes exist among the NOACs.