Are Existing Risk Scores for Nonvalvular Atrial Fibrillation Useful for Prediction or Risk Adjustment in Patients With Chronic Kidney Disease?

BACKGROUND: Comparative effectiveness studies are common in patients with nonvalvular atrial fibrillation (NVAF) and chronic kidney disease (CKD), but the accuracy of current thromboembolic (n = 4) and bleeding (n = 3) prediction scores used for risk adjustment are uncertain in these patients because previous studies have included few CKD patients. METHODS: This was a retrospective cohort study, using Cox models adjusted for time-varying coefficients, of nonanticoagulated adults with incident NVAF and kidney function (defined into Kidney Disease: Improving Global Outcomes [KDIGO] CKD categories) between 2002 and 2013. RESULTS: Of 58,451 patients (mean age 66 years, 31.3% with CKD) followed for a median of 31 months, 21.3% died, 12.6% had a thromboembolic event (4.2 per 100 patient-years), and 7.8% had a major bleed (2.6 per 100 patient-years). There were graded associations between kidney function and all-cause mortality (adjusted hazard ratio [aHR], 1.88 [95% confidence interval (CI), 1.79-1.98] for very high vs low risk KDIGO category), major bleeding (aHR, 1.61 [95% CI, 1.47-1.76]), and thromboembolic events (aHR, 1.13 [95% CI, 1.04-1.23]). All 7 prediction scores had significantly poorer c statistics in patients with CKD: 0.50-0.59; all P < 0.0001 compared with those with normal kidney function (c statistics 0.69-0.70 for the 4 thromboembolic risk scores and 0.60-0.68 for the 3 bleeding risk scores). Inclusion of KDIGO category did not improve calibration or discrimination statistics for current prediction scores. CONCLUSIONS: Existing NVAF risk scores exhibit poor discrimination in patients with CKD, limiting their utility for clinical decision-making or for risk adjustment in comparative effectiveness studies. Although CKD is an independent risk factor for adverse events, adding KDIGO class to current risk scores did not improve their performance.