Kidney disease increases the risk of cardiovascular events in patients with device-detected atrial fibrillation: NOAH-AFNET 6

Patients with device-detected atrial fibrillation (DDAF) without ECG-documented atrial fibrillation (AF) have a lower rate of stroke than patients with ECG-documented paroxysmal AF,1,2 possibly due to their relatively low AF burden.3 Reduced kidney function is associated with higher rates of stroke and cardiovascular events in patients with AF.4–7 Kidney function is not included in the CHA2DS2-VA score or any other risk score in patients with DDAF. It is not known whether chronic kidney disease (CKD) affects the cardiovascular treatment effects of anticoagulation in patients with DDAF. This prespecified analysis of the NOAH-AFNET 6 (n = 2534 patients) trial compared anticoagulation with edoxaban to no anticoagulation in patients with DDAF at various stages of CKD. The double-blind, double-dummy design of NOAH-AFNET 6 ensured that all patients randomized to edoxaban1 received the approved dose. Fifty-five per cent of patients randomized to no anticoagulation received low-dose aspirin as part of the randomized treatment. Chronic kidney disease was defined at baseline according to the 2024 Kidney Disease Improving Global Outcomes (KDIGO) criteria8 as no kidney disease or CKD stage KDIGO G1-G4. Chronic kidney disease Stage G5 patients were excluded from the trial, following the label of edoxaban and other anticoagulants. Exploratory analyses estimated the effect of kidney function as a continuous parameter on cardiovascular outcomes. The efficacy outcome of this analysis was the primary outcome of NOAH-AFNET 6, a composite of stroke, systemic embolism, or cardiovascular death. The safety outcome was the safety outcome of NOAH-AFNET 6, major bleeding according to ISTH criteria or death.1 The outcomes are reported descriptively as subgroup-specific event numbers, event rates per 100 patient-years and as adjusted cause-specific hazard ratios (HRs) with two-sided 95% confidence intervals (CIs) and corresponding P-values. The treatment effects of creatinine on the outcome are presented using a locally estimated scatterplot smoothing (LOESS). The LOESS analysis allows graphic inspection but is merely hypothesis-generating. The impact of CKD on outcome was assessed with cause-specific Cox proportional hazard models with fixed effects for randomized group, sex, age, congestive heart failure history, hypertension history, diabetes mellitus, prior stroke/transient ischaemic attack/systemic thromboembolism, vascular disease history, aspirin intake at baseline, kidney function by estimated glomerular filtration rate (eGFR), and a frailty term for study site. All analyses are exploratory, reflecting the limited power of subgroup analyses, and thus, no adjustment for multiple testing was performed. Results are therefore hypothesis-generating. Primary outcome events increased with increasing severity of CKD—Stage G1, 9 events/265 patient-years (3.4%); G2, 72/2854 (2.5%); G3a, 59/1245 (4.7%); G3b, 28/540 (5.2%); and G4, 16/147 (10.9%)—which was driven by higher rates of cardiovascular death and systemic embolic events. Of note, total stroke rate was low across all CKD stages (0.7–0.9%/year): G1, 2/265 (0.8%); G2, 25/2854 (0.9%); G3a, 15/1245 (1.3%); G3b, 5/540 (0.9%); and G4, 1/147 (0.7%, Figure 1B, left panel). Safety events also increased with increasing severity of CKD, with more major bleeding than death: G1 12/258 (4.6%), G2 110/2870 (3.8%), G3a 78/1232 (6.3%), G3b 46/534 (8.6%), and G4 17/146 (11.6%, Figure 1B, right panel). Anticoagulation with edoxaban did not interact with eGFR or CKD groups (pinteraction = 0.50, and 0.20). A LOESS plot suggests that anticoagulation could be more effective in patients with advanced CKD: event rates increased more in patients randomized to no anticoagulation than in patients randomized to edoxaban than in patients randomized to no anticoagulation (Figure 1C, left panel). A LOESS plot of the primary safety outcome suggested an increase in event rates with reduced kidney function without differences between treatment groups (Figure 1C, right panel). A multivariable analysis showed that age [HR 1.057, 95% CI (1.032–1.083); P < 0.001] and kidney function [HR 0.985, 95% CI (0.976–0.994); P < 0.001] were the strongest predictors of the primary outcome. (A) CONSORT flow chart of this prespecified secondary analysis of the NOAH-AFNET 6 trial. Displayed are distributions of kidney function and baseline characteristics in each chronic kidney disease (CKD) group [according to KDIGO (Kidney Disease Improving Global Outcomes) definition], including sex, age, CHA2DS2-VASc-Score, previous stroke, median duration of device-detected atrial high-rate episodes (AHRE), and heart failure. (B) Stacked column plot of efficacy outcomes by CKD stage and randomized group in NOAH-AFNET 6. Efficacy outcomes (left side) [cardiovascular death (black), stroke (dark red), systemic embolism (lighter red)] by CKD stage and by randomized treatment and safety outcomes [right side, all-cause death (black) and major bleeding (blue)] by CKD stage and by randomized treatment. (C) LOESS (locally estimated scatterplot smoothing) plot of the hazard for a primary outcome event (left side) and for a safety outcome (right side) by estimated glomerular filtration rate (eGFR). Blue solid lines show the estimated event rate for patients randomized to placebo, while red solid lines the estimated event rate for patients randomized to edoxaban. Dotted lines indicate 95% confidence intervals. In patients with reduced kidney function (eGFR