Baseline characteristics

A total of 10,594 patients with AF (age 70.35±9.92 years, 55% males, 56% paroxysmal AF) were included and divided into five groups by CrCl: ≥ 95, 80 to < 95, 50 to < 80, 30 to < 50 and < 30 mL/min (Table 1). Overall, 2,797 (26%) patients received apixaban, 2,273 (21%) dabigatran, and 2,485 (23%) rivaroxaban.

Table 1 Baseline characteristic of patients with atrial fibrillation in different levels of creatinine clearance

Patients with CrCl 30 to < 50 mL/min were the oldest (79.22 ± 6.21 years) while patients with CrCl > 95 mL/min had the highest mean BMI (33.12 ± 6.60 kg/m2). For patients with CrCl < 30 mL/min, 51% received VKA and 28% received apixaban, with the latter having the highest CHA2DS2-VASc score (4.33 ± 1.46).

3-year cumulative incidence of clinical adverse results

After three years of follow-up, NOAC users had lower cumulative incidence rates of all-cause death (6.48% vs.10.35%, P < 0.01), composite outcomes (11.33% vs. 15.36%, P < 0.001), CV death (2.82% vs. 4.55%, P < 0.001), and major bleeding (3.70% vs. 5.07%, P = 0.001) compared with VKA therapy (Table 2).

Table 2 Three-year cumulative incidence rate of different clinical adverse events in two different OACs cohorts (VKA or NOAC therapy) and 5 levels of renal function

For patients with CrCl > 95 mL/min, participants receiving NOAC had a lower cumulative incidence of all-cause death (3.40% vs. 5.79%, P = 0.004), composite outcomes (7.11% vs. 9.93%, P = 0.013) and major bleeding (2.54% vs. 4.28%, P = 0.016). For patients with CrCl < 30 mL/min, patients prescribed NOAC had a lower cumulative incidence rate of all-cause death (20.37% vs. 39.47%, P = 0.002), composite outcomes (27.78% vs. 44.74%, P = 0.009) and CV death (10.19% vs. 22.81%, P = 0.012).

Multivariable nonlinear analysis and RCS curves for renal function

After adjustment, the estimated RCS curve suggested nonlinear associations in outcomes of all-cause death, composite outcomes, and CV death in VKA and NOAC groups with an ‘L-shaped’ curve. For patients with CrCl < 80 mL/min, the risk of all-cause death, composite outcomes and CV death declined with increasing CrCl in patients with VKA/NOAC therapy. For patients with CrCl > 80 mL/min, those prescribed NOAC had lower risk of major bleeding compared to those prescribed a VKA (Fig. 1).

Fig. 1

Restrictive cubic spline curve. VKA, vitamin K anticoagulant; Ref., reference; NOACs, Non-vitamin K antagonist oral anticoagulants

Fig. 2

Forest plot of analysis of risk of clinical end events in AF patients of different creatinine clearance by cox regression comparing NOAC vs. VKA (reference) treatment. Models were adjusted by age, sex, race (Asian /not Asian), BMI, smoking and drinking status, history of hypertension, cardiovascular disease, congestive heart failure, history of thromboembolism, chronic obstructive pulmonary disease, peripheral artery disease, diabetes, previous bleeding, and any antiplatelet drug use. CrCl (mL/min): creatinine clearance; No., number of events; HR: hazard ratio; CI, confidence intervals; VKA, vitamin K anticoagulant; Ref., reference; NOACs, Non-vitamin K antagonist oral anticoagulants; CVD: cardiovascular death; MI: myocardial infarction; TE, thromboembolism events

Univariable and multivariable analysis for OAC type

Figure 2 shows the forest plot of the Cox regression analysis. Considering all AF patients, those prescribed NOAC had lower risk of all-cause death (HR 0.62, 95% CI 0.54–0.71), composite outcomes (HR 0.72, 95% CI 0.65–0.81), CV death (HR 0.62, 95% CI 0.50–0.76) and major bleeding (HR 0.72, 95% CI 0.59–0.88) compared to those prescribed VKA.

Patients prescribed with NOAC consistently had lower risk of all-cause death at every CrCl groups. For those with CrCl < 30 mL/min, NOAC prescription was associated with lower risk of all-cause death (HR: 0.48, 95% CI: 0.29–0.80), composite outcomes (HR 0.58, 95% CI 0.37–0.90), CV death (HR: 0.42, 95% CI: 0.21–0.85). After adjustment for age, sex, BMI, smoking/drinking status, comorbidities and pharmacotherapies, similar results were found in patients with CrCl > 95 mL/min and CrCl < 30 mL/min.

Table 3 Analysis of risk of clinical end events in AF patients of different creatinine clearance by cox regression comparing individual NOACs vs. VKA (reference) treatmentIndividual NOACs versus VKA

Comparisons of individual OAC drugs are shown in Table 3. Given the small number of patients prescribed edoxaban (CrCl at all levels) and dabigatran (CrCl < 30 mL/min), these data were not analysed. Compared to a VKA, apixaban and dabigatran were associated with a decreased risk of all-cause death (HR: 0.69, 95% CI: 0.57–0.82) (HR: 0.63, 95% CI: 0.51–0.78) and composite outcomes (HR: 0.79, 95% CI: 0.68–0.91) (HR: 0.70, 95%CI: 0.60–0.83).

Compared to a VKA, apixaban was consistently associated with a decreased risk of all-cause death in patients with CrCl = 30–50 and < 30 mL/min. Dabigatran was associated with a lower risk of all cause death in patients with CrCl > 95 and 30–50 mL/min. Rivaroxaban was associated with decreased risk of all-cause death, composite outcomes, and major bleeding in patients with CrCl > 95mL/min.

Age subgroups

The baseline table suggested that patients with lower renal function were older, and were more likely to receive VKA. Therefore, we explored the impact of age on outcomes with NOAC versus VKA. Supplementary Fig. 1 shows the age subgroup of association between outcomes and NOAC versus VKA. Patients receiving NOAC were associated with a lower risk of major bleeding in patients with age ≥ 75 (HR: 0.60, 95% CI: 0.46–0.80, Pinteraction = 0.042). No significant interaction was noted between age groups and OAC use in other outcomes (All Pinteraction > 0.05).

Association between Asian/Non-Asian individuals and clinical events

Previous studies have shown that Asian patients are prescribed NOAC less frequently [15] and have a higher risk of bleeding [16]. In our baseline, we observed significant heterogeneity in renal function among Asian patients. Therefore, we performed the analysis to assess the outcome risks in Asian patients with varying renal functions (Supplementary Fig. 2). Non-Asian individuals were associated with higher risk of all-cause death (HR: 1.51, 95% CI: 1.17–1.93), composite outcomes (HR: 1.38, 95% CI: 1.15–1.66), and major bleeding (HR: 1.46, 95% CI: 1.05–2.03). No significant interaction was noted between ethnic groups and renal function (All Pinteraction >0.05).