Discordance in estimated kidney function by different equations and outcomes in patients with atrial fibrillation on oral anticoagulation: a COMBINE-AF analysis
P Krisai, A Benz, Q Yi, T F Chao, R P Giugliano, C Granger, J Healey, L Johnson, F Mahfoud, M Patel, L Wallentin, J Eikelboom, Z HijaziAbstract
Background
In the pivotal non-Vitamin K oral anticoagulant (NOAC) trials in patients with atrial fibrillation (AF), creatinine clearance (CrCl) was used to estimate kidney function, whereas CKD-EPI equations are mainly used in clinical practice.
Aim
To investigate differences in estimated kidney function by CKD-EPI equations instead of CrCl and its association with thromboembolic and bleeding events in patients with AF.
Methods
Individual patient data from the pivotal NOAC trials in AF were pooled in the COMBINE-AF dataset. All patients randomized to standard dose of NOAC or warfarin with available creatinine measurements at baseline were included. Next to CrCl (mL/min), GFR (mL/min/1.73m2) was estimated based on creatinine (eGFRcrea), cystatin C (eGFRcys) and both biomarkers (eGFRcrea-cys). Reclassification across a 50 mL/min cut-off, and its association with stroke or SE and major bleeding was assessed.
Results
58,373 patients were included. Cystatin C was available in 24,690 patients. Median CrCl was 70.1, eGFRcrea 66.4, eGFRcys 70.6 and eGFRcrea-cys 68.9 mL/min/1.73m2. Of 11,180 patients classified ≤50 by CrCl, 38.9% were reclassified to >50 by eGFRcrea. Similar reclassification proportions were observed when using each CKD-EPI equation instead of CrCl. (Fig. 1) For stroke or SE, the highest incidence rates of 2.91 per 100 patient years (PY) were in patients reclassified from CrCl ≤50 mL/min to >50 mL/min/1.73 m2 with eGFRcrea. Patients that were classified ≤50 mL/min/1.73 m2 by both equations had the second highest incidence rates of 2.46/100 PY. The two groups with >50 mL/min by CrCl had the lowest incidence rates of 1.61 and 1.51 /100 PY for ≤50 and <50 by eGFRcrea. This pattern was consistent for eGFRcrea-cys but not observed when using eGFRcys. (Fig. 2) For major bleeding, patients classified as having ≤50 mL/min/1.73 m2 by both CrCl and the respective CKD-EPI equation had the highest incidence rate. This was followed by the two groups with discordant classification, and finally by patients consistently classified as >50 mL/min/1.73 m2 by both methods, who had the lowest bleeding rates.
Conclusion
A large proportion of patients were reclassified when using CKD-EPI equations instead of CrCl. Reclassified patients from ≤50 by CrCl to >50 by CKD-EPI equations experienced the highest rates of thromboembolic events.Sankey diagrams showing reclassificationIncidence curves stroke/SE & maj. bleed