Rivaroxaban plasma levels in patients admitted for bleeding events: insights from a prospective study

This analysis was part of a prospective observational cohort study, conducted in two teaching hospitals in Belgium to assess the preventability of serious adverse drug reactions related to the use of oral anticoagulants [21]. We collected plasma samples from rivaroxaban patients admitted to the ED for a bleeding event, between July 2015 and January 2016.

The following routine clotting assay was performed: prothrombin time (PT) using RecombiPlasTin2G® (Werfen, Lexington, USA; normal range 12.4–14.5 s). Rivaroxaban plasma concentrations were estimated with the Biophen® Direct Factor Xa Inhibitors (Biophen®DiXaI, Hyphen BioMed, Neuville-Sur-Oise, France), a calibrated chromogenic anti-Xa assay. The procedure was performed on a STA-R® Evolution analyzer according to the manufacturer recommendations, using calibrators from Hyphen BioMed. Commercial rivaroxaban anti-Xa assays have previously demonstrated good accuracy (bias below 8%) and acceptable precision (inter-laboratory coefficients of variation (CV) 6–25%) [22]. A procedure dedicated to low concentrations was applied to rivaroxaban plasma concentrations < 50 ng/ml, where plasma samples were diluted 1:8 in buffer and low concentrations standards were used [23].

We extrapolated rivaroxaban plasma concentrations at trough (i.e. 12 h and 24 h after the last drug intake for twice daily and once daily regimen, respectively). To this end, we used a Pop PK model that was previously developed to predict adequate discontinuation intervals in a cohort of rivaroxaban patients scheduled for cardiac catheterization [24]. Residual rivaroxaban concentrations were described by a 2-compartment model with first-order absorption and elimination. Results were then compared to the expected on-therapy ranges at trough (5th–95th percentiles): 12–137 ng/ml (rivaroxaban 20 mg) and 18–136 ng/ml (rivaroxaban 15 mg) for stroke prevention in non-valvular atrial fibrillation (NVAF), and 6–87 ng/ml for treatment and secondary prevention of venous thromboembolism (VTE) [25].

A comprehensive medication history was performed with patients and/or relatives to collect sociodemographic, clinical and medication data (including dose regimen, indication and timing of the last rivaroxaban intake). Patient renal function was estimated based on serum creatinine on admission using Cockroft-Gault equation. We reviewed electronic medical records and contacted GPs, community pharmacists or relatives when necessary. Bleeding severity was assessed using the International Society on Thrombosis and Haemostasis definition. Patient follow-up was undertaken to assess 3-month mortality.

Appropriateness of rivaroxaban prescribing was analyzed according to the Medication Appropriateness Index (MAI) [26]. We applied a DOAC-specific form, previously developed in a pilot study [27]. We documented the concomitant use of P-glycoprotein (P-gp) and CYP3A4 inhibitors or inducers. Regarding potential pharmacodynamics interactions, we considered other anticoagulants, antiplatelet therapy, non-steroidal anti-inflammatory drugs (NSAID), selective serotonin reuptake inhibitors (SSRI) and serotonin-norepinephrine reuptake inhibitors (SNRI). Patient adherence to anticoagulant therapy was evaluated during medication history using an adapted version of the Morisky Medication Adherence Scale [28].

For all patients except one (because of misunderstanding), an additional blood sample was drawn in EDTA tube and frozen at − 20 °C without delay until experiments. Genomic DNA was extracted from whole blood, using QIAamp DNA Blood Mini kit (Qiagen, CA, USA). The following single nucleotide polymorphisms (SNP) of ABCB1 were detected: rs1128503 (1236C > T, Gly412Gly), rs2032582 (2677G > T, Ala893Ser), rs1045642 (3435 C > T, Ile1145Ile) and rs4148738. Allelic discrimination was performed by TaqMan® (Applied Biosystems, CA, USA) Drug Metabolism Genotyping Assays (C___7586662_10, C_11711720C_30, C__15951365_20, C___1253813_10). Statistical analyses were performed using JMP Pro 13.2.0 (SAS Institute, Cary, NC, USA). Kruskal-Wallis tests were carried out to compare estimated trough concentrations between different genotypes. A p-value of less than 0.05 was considered statistically significant.

Plasma samples from 10 rivaroxaban patients were included in this analysis. Median age of patients was 75 years, and 80% were female. Five patients had moderate renal impairment on admission (creatinine clearance < 60 ml/min). Indications of rivaroxaban therapy were stroke prevention in NVAF (6/10) and treatment and secondary prevention of VTE (4/10). Clinical characteristics are detailed in Table 1 for each patient. The most frequent adverse events were gastrointestinal bleeding (4/10). Half of bleeding events were considered major and half occurred spontaneously. Red blood cells transfusion and prothrombin complex concentrates administration were operated in 4 and 1 patients, respectively. All patients were alive at 3-month follow-up. Details of bleeding events are presented in Table 2.

Measured rivaroxaban levels varied from 5 to 358 ng/ml, with a delay since the last drug intake ranging from 9 to 38 h. PT was above the normal range in all patients with rivaroxaban concentrations above 30 ng/ml (Table 3). When using PopPK model, estimated plasma concentrations at trough were between 12 and 251 ng/ml (median value 94 ng/ml). Elimination clearance (CL/F) was between 2.0 and 7.9 L/h (median value 3.3 L/h). Measured and extrapolated rivaroxaban concentrations are described in Table 3 for each patient, as well as the elimination clearance. Four patients had rivaroxaban levels at trough higher than the expected on-therapy range (No 2, 3, 5, 9).

The analysis of rivaroxaban prescriptions revealed that at least one criterion of the MAI was inappropriate in 8 of 10 patients. Three patients received an inadequate dosing regimen: 20 mg OD instead of 15 mg OD (No 3, 7), and 15 mg BID instead of 20 mg OD (No 9). Two of them had estimated trough concentration above the on-therapy range. Three patients had an excessive alcohol consumption (≥ 2 units/day; No 1, 4, 7). Moreover, in two patients treated for VTE prevention, treatment indication had not been reassessed for years (No 3, 10).

Half of patients were taking at least 1 drug with potential pharmacokinetic interactions (Table 3). Three of them had higher-than-expected rivaroxaban plasma concentrations at trough. The most frequent interacting drugs were amiodarone and diltiazem. Moreover, half of patients were taking at least 1 drug increasing the risk of bleeding (Table 3). We mainly observed the concomitant use of SSRI (n = 3) and NSAID (n = 2). Regarding adherence to anticoagulant therapy, three patients reported forgetting to take their medicine in rare circumstances (No 1, 4, 7). One patient mentioned a variable time of intake (No 7).

ABCB1 genetic polymorphisms were analyzed in 9 of 10 patients. Genotyping data are presented in Table 4 for the ABCB1 1236C > T, 2677G > T, 3435 C > T and rs4148738 SNP. Of the four patients with higher-than-expected rivaroxaban trough levels, three carried at least one variant allele for each of these SNP (No 2, 3, 5 – blood sample not available for patient No 9). However, no clear association between ABCB1 genotype and estimated trough concentrations was observed (p > 0.05).

The management of DOAC-related bleeding includes the temporary discontinuation of the oral anticoagulant, supportive measures and the administration of reversal agents, depending on the severity of the event [34, 35]. Andexanet alpha has recently been approved by the FDA to reverse the anticoagulant effect of rivaroxaban, while prothrombin complex concentrates were also suggested as an alternative [36‐38]. In this context, rapid laboratory measurement may provide valuable assistance to warrant and monitor antidote administration, or manage urgent interventions [7].

In the present study, we observed that PT was prolonged in all patients with clinically relevant concentrations (> 30 ng/ml) of rivaroxaban. However, we employed a sensitive thromboplastin reagent (RecombiPlasTin 2G®) for the assay [39]. A nationwide Belgian survey has previously highlighted a wide variation in response to rivaroxaban according to the reagent used [40]. Commercial specific assays are currently available for all DOACs [41]. They are more accurate but require calibrators and controls. Turnaround times around 30 min have nevertheless been reported in a daily practice context [42, 43].

Several risk factors for bleeding events were highlighted in our 10 rivaroxaban patients. First, half of them were older than 75 years, and half of them had moderate renal impairment. Older age and renal insufficiency were previously demonstrated as contributing factors to major bleeding in rivaroxaban patients [14]. Recently, glomerular filtration rates below 60 ml/min have been independently associated with higher-than-expected residual rivaroxaban levels in a perioperative setting [44]. This is not surprising since one third of the rivaroxaban dose is eliminated unchanged by the kidneys [25]. However, in our analysis, only two patients with moderate renal impairment had trough concentrations above the on-therapy range. Second, despite the highest convenience and fixed-dose regimen of DOAC, some of our patients did not receive the appropriate dose of rivaroxaban. In particular, one patient (No 9) was still taking rivaroxaban 15 mg BID while the 21-day treatment phase of VTE had been completed. Yao and colleagues highlighted that patients with indication for dose reduction were often potentially overdosed, leading to an increased risk of major bleeding [19].

Drugs interactions with P-gp or CYP3A4 inhibitors were frequent in our patients, as previously reported [32, 45]. One patient (No 2) was taking diltiazem and clarithromycin, two inhibitors of both P-gp and CYP3A4. He had an estimated rivaroxaban level of 181 ng/ml at trough, exceeding the 95th percentile of the on-therapy range (136 ng/ml). In healthy volunteers, clarithromycin has been shown to promote a 2-fold increase in rivaroxaban exposure [46]. Similarly, we extrapolated a rivaroxaban trough concentration of 125 ng/ml in a patient taking simvastatin (No 3). This P-gp inhibitor has been associated with a higher risk of major bleeding in patients taking dabigatran [47]. Pharmacodynamic interactions were also widely observed. Aspirin has often no valid indication in anticoagulated patients, while combination therapy has been shown to increase the risk of major bleeding [16, 48]. A similar increase was associated with the addition of SSRIs, the most prescribed class of antidepressants [49].

An original aspect of this work was the investigation of several ABCB1 polymorphisms. All 3 patients with higher-than-expected rivaroxaban levels and available genotyping data were at least heterozygous mutated for the 1236C > T, 2677G > T, 3435C > T and rs4148738 SNP. Previous experiments have shown that the 1236 T–2677 T-3435 T variant haplotype decreased ABCB1 expression or transport towards several drugs such as anticancer agents [50‐52]. These results support our observations, as a decreased efflux of rivaroxaban may have led to drug accumulation. However, in a recent study conducted in 60 healthy volunteers, the 1236 T–2677 T-3435 T haplotype was not a significant determinant of rivaroxaban pharmacokinetics [46]. This strengthens the need for additional studies to clarify the impact of ABCB1 polymorphisms on rivaroxaban transport. Indeed, our findings might also be due to the high frequency of the variant genotype 1236 T–2677 T-3435 T (up to 50% of the Caucasian population is expected to heterozygous mutated). Furthermore, these three patients were also receiving concomitant interacting medications, as previously discussed.

The study presents several limitations. First, statistical analysis was limited by the small number of plasma samples collected. However, this allowed an in-depth analysis of rivaroxaban measurements, including clinical, medication and genetic characteristics at the individual level. Second, three patients (No 2–4) were transfused with red blood cells (RBC) before sampling. For these patients, we cannot exclude that rivaroxaban measurements were not influenced by fluid volume or factor Xa content of packed RBC. However, Biophen®DiXaI was designed for minimizing the interference of plasma factors. Third, estimation of renal function was only based on serum creatinine on admission, as previous laboratory results performed in primary care were not available. Finally, the Pop PK model we used assumed no variability in the volume of distribution, while inter-individual variability in V/F was 18% in another Pop PK model from AF patients [31]. However, simulations were previously performed and showed the limited influence of this variability on predicted exposure estimates.