This prospective study was approved by Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (IRB HKU/HAHKWC) (Reference number: UW 22-086). Patients with history AF receiving a stable dose of rivaroxaban were recruited from the outpatient clinics at the Department of Medicine, Queen Mary Hospital of Hong Kong. All patients provided written consent. Rivaroxaban was prescribed at dosages of 20 mg daily or 15 mg daily, according to the renal function or assessment of patients’ medical conditions by the attending physician. In general, patients with CrCl≤50 ml/min (as determined by the CKD-EPI equation) at the time of stating rivaroxaban were prescribed a reduced dose of 15 mg daily. Patients with non-Chinese ethnic origin and poor drug compliance (defined by less than 90% of compliance) were excluded from the study. At study enrollment, data on patient demographics, medical comorbidities, liver and renal biochemistry, clotting profiles were collected. The drug compliance to rivaroxaban treatment by pill-counting during outpatient clinic follow up was documented in the medical records. Patients were followed at the medical outpatient clinics every 8–12 weeks, depending on clinical situation. Dosages of rivaroxaban, concomitant medications, renal function (serum creatinine, estimated GFR by MDRD equation) and clinically significant events (i.e., thromboembolic events or bleeding complications) were documented in medical record during each follow-up. Severity of chronic kidney disease was classified according to Kidney Disease: Improving Global Outcomes (KDIGO) classification (Levey et al., 2005). As the clinicians have chosen the dosage according to the eGFR values provided by our hospital laboratory (calculated by MDRD), we have also analysed our data using CrCl estimated by the Cockroft-Gault (CG) equation to ensure consistency of our results. ISTH definition of major and minor bleeding was used. Briefly, major bleeding was defined as Hb level below 8 g/dl or drop of Hb more than 2 g/dl from baseline, limb threatening or life-threatening bleeding. All other bleeding episodes were defined as minor bleeding (Kaatz et al., 2015). Thromboembolic events were defined as clinical evidence of transient ischaemic attached (TIA) or stroke, as well as radiological evidence of new onset cerebrovascular accident (CVA) or venous thromboembolism. Clinically significant bleeding and thromboembolic complications were adjudicated independently by investigators who were unaware of the levels of rivaroxaban. Clinical data were first retrieved from the electronic health systems (ePR of the Hong Kong Hospital Authority) and verified by review of patient case records. The follow-up period of recurrent ischaemic/thromboembolic events and bleeding complications was 1 January 2018 to 31 October 2021.

To ensure a stable plasma rivaroxaban concentration, blood samples were measured in patients after receiving rivaroxaban for at least a week (more than five half-life). The peak level was collected after 2 h of rivaroxaban administration and the trough level was taken 24 h after last dose of rivaroxaban. Blood samples were taken by vacuum plastic tubes containing 3.2% trisodium citrate, followed by centrifugation at 3,750 rpm for 10 min. Platelet poor plasma were separated and stored at or below −70°C for future use.

Plasma rivaroxaban level was measured by BIOPHEN DiXal kit (Hyphen BioMed, France) in Sysmex CS5100 analyzer according to the manufacturer’s instructions. The assay utilized chromogenic anti-Xa assay specifically calibrated for rivaroxaban and plasma rivaroxaban level was expressed as unit ng/ml. Rivaroxaban level measurement was performed after the value obtained by quality control materials fell within specific ranges provided by manufacturers. Coagulation screening tests namely [prothrombin time (PT) and activated partial thromboplastin time (APTT)] were performed in Sysmex CS5100 analyzer. The reagents used for PT and APTT measurements were Thromorel S Reagent and Actin FSL Activated PTT Reagent respectively (both manufactured by Siemens).

Continuous variables were expressed as mean ± S.D. or median (range) as appropriate. The differences in continuous variables between groups were assessed by using one-way ANOVA (for parametric test) or independent-samples Kruskal–Wallis test (for non-parametric test). Categorical variables were expressed as frequencies and percentages, and they were analyzed by chi-square test or Fisher’s exact test where appropriate. Pearson correlation was used to determine the relationship between plasma rivaroxaban level with PT and APTT. Linear regression was used to examine the relationship between plasma rivaroxaban level and other clinical parameters. Logistic regression was used to determine the relationship between rate of bleeding complications, recurrent ischaemic/thromboembolic events and plasma rivaroxaban level as well as other clinical parameters including dosage, age, eGFR, presence of DM, hypertension, history of ischaemic stroke and ischaemic heart disease. Univariate regression analysis was performed first to evaluate the relationship between individual parameters, followed by multivariate analysis.

The primary outcome was defined as the relationship between plasma rivaroxaban level and renal function. The secondary outcome was the association between plasma rivaroxaban level and clinically significant events. In order to establish a correlation (r = -0.3) between plasma rivaroxaban level and eGFR, a sample size of 85 will achieve a 80% power at a 95% confidence interval (95% CI). Based on previous studies that reported a 21% event rate for all major and non-major bleeding (Wong et al., 2014), a sample size of 150 patients will achieve 80% power to determine the relationship between plasma rivaroxaban level and bleeding events by univariate logistic regression at a 95% CI.

All statistical analysis was performed by using IBM SPSS software version 27 and p-values of less than 0.05 were considered statistically significant.

A total of 92 Chinese patients (CKD stage 1 n = 10, stage 2 n = 53 and stage 3 n = 29) were recruited into the study (Table 1). The primary indication of rivaroxaban was AF. Sixty-eight patients (73.9%) received a dose of 20 mg daily and 24 patients (26.1%) received a dose of 15 mg daily. All recruited patients showed good drug compliance (i.e., > 90% compliance by pill-counting). The time taken from starting rivaroxaban to blood taking for plasma rivaroxaban level was at least 1 month apart, with a median time period of 13 months (range: 1–32 months). Patients receiving rivaroxaban 15 mg daily were older and showed higher CHADS₂ score, lower eGFR and higher incidence of hypertension compared with those taking rivaroxaban 20 mg daily (p < 0.05 for all). Twenty-four patients received rivaroxaban 15 mg daily. Dosage in these patients was reduced because of patient’s clinical condition (18 related to impaired renal function and six because of mild bleeding symptoms at the time of rivaroxaban initiation). When CrCl were estimated by the CG equation, 22 out of 68 patients (32.4%) who received rivaroxaban 20 mg daily should in fact receive 15 mg daily instead. Diabetes nephropathy and hypertensive nephropathy were the commonest causes of renal impairment for our patients (79.3% in total).

Despite receiving a lower dose of rivaroxaban, patients receiving 15 mg daily showed a significantly higher plasma trough level when compared with those patients taking 20 mg daily (60.4 ± 35.8 ng/ml vs. 41.8 ± 29.6 ng/ml, p = 0.014) (Table 2). The peak rivaroxaban level was also numerically higher in those receiving 15 mg daily compared with those taking 20 mg daily, although statistical significance was not reached (318.1 ± 127.1 ng/ml vs. 272.6 ± 119.9 ng/ml, p = 0.126).

For patients taking rivaroxaban 20 mg daily, the plasma trough level was inversely correlated with eGFR (r = 0.391, p = 0.001) (Figure 1A). Plasma trough levels in patients with stage 3 CKD were significantly higher than those with stage 2 and 1 CKD (66.0 ± 34.9 ng/ml vs. 35.7 ± 24.7 ng/ml vs. 34.7 ± 26.2 ng/ml, respectively, p = 0.005) (Figure 1B). However, there was no significant correlation between plasma peak level and eGFR (r = -0.087, p = 0.485) (Figure 2A). The plasma peak level was numerically higher in those patients with CKD stage 3 when compared with stage 2 and stage 1, although not reaching statistical significance (330.8 ± 116.1 ng/ml vs. 250.5 ± 118.1 ng/ml and 291.8 ± 114.7 ng/ml in patients with stage 3, 2 and 1 CKD respectively, p = 0.098) (Figure 2B).

For patients receiving rivaroxaban 15 mg daily, there was no correlation between the plasma peak level and eGFR (r = -0.014, p > 0.1) (Figure 3A), and the plasma peak levels also showed no difference between patients with different stages of CKD (Figure 3B). The plasma trough level showed a trend of inverse correlation with eGFR though statistically not significant (r = -0.402, p = 0.051) (Figure 4A), and the trough plasma rivaroxaban levels were also numerically higher in patients with CKD stage 3 when compared with stage 2 (68.4 ± 38.5 ng/ml vs. 47.0 ± 27.6 ng/ml, p > 0.05) (Figure 4B).

The plasma trough rivaroxaban level showed a positive correlation with PT (r = 0.646, p < 0.001) and APTT (r = 0.437, p < 0.001) (Figures 5A,B), while the peak level showed no relationship with both PT (r = 0.111, p = 0.301) and APTT (r = 0.028, p = 0.795) (Figures 6A,B).

The median follow-up period was 36.1 months (range: 1–45 months). There was a total of 27 episodes of bleeding complications (annual rate of 7.7%) and five episodes of recurrent ischaemic/thromboembolic events (annual rate of 1.3%) during the entire follow-up period (Table 3). The median time from blood sample for plasma rivaroxaban level to occurrence of clinical events was 399.5 days (range: 24-950 days). The plasma trough rivaroxaban level showed a positive correlation with the rate of bleeding complications [odds ratio (OR): 1.018, 95% CI 1.004–1.033, p = 0.014] but not the rate of recurrent ischaemia/thromboembolic event (OR 0.976, 95% CI 0.935–1.018, p = 0.257) upon univariate regression analysis. Multi-variate regression analysis further demonstrated that plasma trough rivaroxaban level was positively associated with the rate of bleeding complications after adjustment for eGFR, age, dosage, history of DM, hypertension, ischaemic heart disease and ischaemic stroke (OR: 1.020, 95% CI 1.002–1.038, p = 0.028). Patients who developed bleeding complications also showed to have a significantly higher plasma trough rivaroxaban level than those who did not (59.9 ± 35.6 ng/ml vs. 41.1 ± 29.2 ng/ml, p = 0.011) (Table 4).

The peak plasma rivaroxaban level did not show any relationship with the rate of recurrent ischaemic/thromboembolic events and bleeding complications (p > 0.05 for both). Nevertheless, patients with recurrent ischaemic/thromboembolic events showed lower peak (230.0 ± 120.0 ng/ml vs. 287.6 ± 123.2 ng/ml, p = 0.336) and trough plasma rivaroxaban levels (30.3 ± 14.8 ng/ml vs. 47.6 ± 32.7 ng/ml, p = 0.348) compared with those who did not. The plasma peak rivaroxaban levels were similar in patients with or without bleeding complications (278.2 ± 122.8 ng/ml vs. 286.9 ± 123.6 ng/ml respectively, p = 1.00) (Table 4).