Is a Lower Dose of Rivaroxaban Required for Asians? A Systematic Review of a Population Pharmacokinetics and Pharmacodynamics Analysis of Rivaroxaban

Rivaroxaban has been widely used to prevent and treat various thromboembolic diseases for more than a decade. However, whether a lower dose of rivaroxaban is required for Asians is still debatable. This review aimed to explore the potential ethnic difference in pharmacokinetic/pharmacodynamic (PK/PD) characteristics between Asians and Caucasians. A systematic search was conducted and twenty-four studies were identified, of which 10 were conducted on Asian adults, 11 on predominantly Caucasian adults, and 3 on Caucasian pediatrics. The apparent clearance (CL/F) of rivaroxaban in Caucasian adults with non-valvular atrial fibrillation (6.45–7.64 L/h) was about 31–43% higher than that in Asians (4.46–5.98 L/h) taking 10~20 mg rivaroxaban every 24 h. Moreover, there was no obvious difference in CL/F among Japanese, Chinese, Thai, and Irani people. Regarding PK/PD relationship, prothrombin time was linked to rivaroxaban concentration in a linear or near-linear manner, and Factor Xa activity was linked with the Emax model. The exposure–response relationship was comparable between Asians and Caucasians. Renal function has a significant influence on CL/F, and no covariate was recognized for exposure–response relationship. In conclusion, a lower dose of rivaroxaban might be required for Asians, and further studies are warranted to verify this ethnic difference to facilitate optimal dosing regimens.


Introduction
Rivaroxaban is one of the most commonly used direct oral anticoagulants (DOACs) for the management of several thromboembolic disorders, such as deep vein thrombosis, pulmonary embolism, non-valvular atrial fibrillation (NVAF), and acute coronary syndrome (ACS) [1]. It has also been approved for the treatment and reduction of the recurrence risk of venous thromboembolism (VTE) in children [2,3].
Following oral administration, rivaroxaban is rapidly absorbed and reaches a peak concentration within 2-4 h [4]. The bioavailability of rivaroxaban is dose-dependent, reaching 80-100%, without being affected by food, upon the oral administration of 2.5-10 mg tablets. However, bioavailability is decreased by 34% in healthy subjects when it is administered as 20 mg tablets under fasting conditions [4,5]. The plasma protein binding for rivaroxaban is approximately 92-95%. Further, the clearance (CL) of rivaroxaban is a dual pathway: approximately 2/3 of rivaroxaban is metabolized by hepatic cytochrome P450 (CYP) enzymes 3A4/5, 2J2, and CYP-independent enzymes, while the remaining 1/3 is eliminated unchanged via the kidney, involving transporters in active renal secretion such as P-glycoprotein (P-pg) and breast cancer resistance protein (BCRP) [4,6]. The anticoagulant effect of rivaroxaban is regarded as the direct inhibition of free and clot-bound Factor Xa (FXa). Rivaroxaban could also prolong prothrombin time (PT) and activated partial thromboplastin time (aPTT) [4,7].

Comparison of Studies
The study characteristics, population PK, and population PK/PD analyses were summarized in a tabular format. The visual predictive distributions (VPDs) of the concentrationtime and PD biomarker-time profiles at steady state were generated via Monte Carlo simulations based on reported population PK/PD models from each study [17]. A total of 1000 virtual patients were simulated for each scenario. All simulations were performed using NONMEM software (version 7.5; ICON Development Solutions, Ellicott City, MD, USA).
The effects of the included covariates on the PK parameters and PD metrics were assessed using forest plots. Continuous covariates, such as age, BW, and serum creatinine (SCr) level, were scaled to the same range. In contrast, binary covariates such as sex were expressed as 0 and 1. The upper and lower limits of the parameters were estimated based on the range of the corresponding covariates. Each range was normalized to the median value in each model. Therefore, the effect of each covariate can be shown as a range of the limit to the median value, as follows (Equation (1) Data were analyzed and plotted using R software (version 4.2.1; www.r-project.org; accessed on 1 September 2022).

Study Identification
A total of 353 publications were initially selected from the PubMed, EMBASE, and Web of Science databases, of which 167 were excluded after screening. After a full-text review, 23 studies were deemed eligible. One additional study was identified from the reference list of the included studies [18]; hence, a total of 24 articles published between 2007 and 2022 were retained. A PRISMA diagram is shown in Figure S1.

Reporting Quality
The 30-item checklist and corresponding analysis outcomes of each study are shown in Table S1. The range of compliance in the included studies was 43.3-93.3%, with a median compliance of 80.8%. The compliance in 18 of the 24 studies was greater than or equal to 80%. Moreover, there was no obvious difference in compliance between studies published before or after publishing the guidelines [15,16] (81.7% vs. 80.3%).
Zhang et al. [29] combined data from the studies by Mueck et al. [23] and Girgis et al. [27]. Willmann et al. pooled data from five previous studies by Mueck et al. [21][22][23], Xu et al. [24], and Girgis et al. [27]. The study by Speed et al. [33] applied the data obtained in study by Barsam et al. [28] and enrolled additional patients to ensure a greater sample set.
In most studies, population analysis was performed using the NONMEM software. Two used Phoenix NLME software [18,30], while one study used Monolix software [38].
PT was measured using various thromboplastin reagents in different studies. It was correlated with rivaroxaban concentration in either a linear (Equations (2) and (3)) or near-linear relationship (Equations (4) and (5)): where M represents the measurement of PT (APTT, Heptest, or PiCT), M 0 represents the baseline of PT (APTT, Heptest, or PiCT), slope represents the change in PT (APTT, Heptest, or PiCT) per unit rivaroxaban concentration change, C p represents the plasma concentration of rivaroxaban, and Hill represents the exponent of rivaroxaban concentration. FXa activity was measured directly using a two-step photometric assay [19,22,26] or expressed as a percentage change compared with the control plasma [27]. It was linked to the rivaroxaban concentration with an E max or sigmoid E max model, as shown in Equations (6)-(8): The effect of identified factor on PK, and factors influencing PK in NVAF patients where M represents the measurement of FXa activity (aPTT, or anti-Xa), M 0 represents the baseline FXa activity (aPTT, or anti-Xa), E max represents the maximum FXa activity (aPTT, or anti-Xa), EC 50 represents the rivaroxaban concentration generating 50% of the maximum FXa activity (aPTT, or anti-Xa), Hill represents the exponent of rivaroxaban concentration, and C p represents the plasma concentration of rivaroxaban. Anti-Xa activity was correlated with rivaroxaban concentration in a near-linear (Equation (5)) or E max model (Equation (8)) [20,38]. aPTT was expressed as a linear (Equation (2)), nearlinear (Equation (4)), or E max -related relationship (Equation (6)) [19,26,32,38]. Heptest was expressed as an E max -related relationship (Equations (6) and (7)) [19,27] and PiCT showed a near-linear relationship (Equation (4)) [27]. The final PD parameters of the included studies are listed in Table 3 and Table S2.

Comparison of Studies
The simulated concentration-time and PD biomarker-time profiles at the steady state of the included studies were compared. Detailed information on the simulated patient characteristics is provided in Table S3. For a comparison of PK parameters, apparent clearance (CL/F) was chosen because it is the most important PK parameter for long-term pharmacotherapy and could be compared directly among studies that employed different dosing regimens [42]. As for the PD index, FXa activity was selected for comparison because it directly reflects the inhibitive effect of rivaroxaban. PT, which was investigated in most studies, was also compared. Other PD biomarkers were not assessed because of a lack of adequate data or evidence of their clinical relevance.

Pharmacokinetic Analysis
The dose-dependent bioavailability was reported in nine studies [19][20][21][22][23][24]31,40,41]. For studies conducted on patients with NVAF, bioavailability was estimated in a dosedependent manner only by Willmann et al. [31], but it was fixed or not estimated in other studies. Therefore, the CL/F values in different dose groups were compared.   [21] also applied PT to develop PK/PD model, but did not report the parameter estimates.
Eleven studies identified renal function as significant covariates on the CL of rivaroxaban, including CrCl, eGFR, or BUN. The impact of CrCl on CL varied in these studies. Compared with patients with CrCl (or eGFR) of 80 mL/min, moderately (CrCl (or eGFR) 30-49 mL/min) or severely impaired renal function (CrCl (or eGFR) 15-29 mL/min) may lead to a decrease in the CL of approximately 4-50% and 12-72%, respectively. The change in CL/F per unit renal function was similar between Asian and Caucasian patients but had a large variability. For example, the CL/F of rivaroxaban decreased by 0.13-5.58% in the Asian population and 0.34-2.84% in the Caucasian population per 1 mL/min decrease in CrCl (or eGFR) from 15 to 120 mL/min ( Figure S4). A significant effect of hepatic function on CL was reported in four studies, which were characterized by the ALT, TBIL, and CTP levels [30,[36][37][38]. Increased TBIL (>35 μmol/L) or CTP (>6.5) levels may lead to a decrease in the CL/F of rivaroxaban by >20% (Figure 2). Nevertheless, the identified liver indices were different in each of these studies, which warrants further evaluation. Body weight (BW) was also identified as a significant covariate in three studies conducted on pediatric patients [39][40][41] (Figure 2). The body-weight-adjusted CL/F in children was higher than that in adults. For Caucasian children weighing 10, 30, and 50 kg, the median body weight-adjusted CL/F was reported to be approximately 2.05-, 1.47-, and Figure 2. Effect of covariates on the apparent clearance of rivaroxaban. The horizontal bars represent the effect of each covariate on clearance (CL) in each study. The effect of each covariate on CL was characterized as the ratio of CL in the range of each covariate to the typical CL. The gray shadow represents the range of 80-125% [18,20,25,26,28,30,31,[33][34][35][36][37][38][39][40][41].
Eleven studies identified renal function as significant covariates on the CL of rivaroxaban, including CrCl, eGFR, or BUN. The impact of CrCl on CL varied in these studies. Compared with patients with CrCl (or eGFR) of 80 mL/min, moderately (CrCl (or eGFR) 30-49 mL/min) or severely impaired renal function (CrCl (or eGFR) 15-29 mL/min) may lead to a decrease in the CL of approximately 4-50% and 12-72%, respectively. The change in CL/F per unit renal function was similar between Asian and Caucasian patients but had a large variability. For example, the CL/F of rivaroxaban decreased by 0.13-5.58% in the Asian population and 0.34-2.84% in the Caucasian population per 1 mL/min decrease in CrCl (or eGFR) from 15 to 120 mL/min ( Figure S4). A significant effect of hepatic function on CL was reported in four studies, which were characterized by the ALT, TBIL, and CTP levels [30,[36][37][38]. Increased TBIL (>35 µmol/L) or CTP (>6.5) levels may lead to a decrease in the CL/F of rivaroxaban by >20% (Figure 2). Nevertheless, the identified liver indices were different in each of these studies, which warrants further evaluation. Body weight (BW) was also identified as a significant covariate in three studies conducted on pediatric patients [39][40][41] (Figure 2). The body-weight-adjusted CL/F in children was higher than that in adults. For Caucasian children weighing 10, 30, and 50 kg, the median body weight-adjusted CL/F was reported to be approximately 2.05-, 1.47-, and 1.16-fold higher than that in adults, respectively ( Figure S5). BW was also assessed in 15 studies conducted on adult patients [20][21][22][23]25,[28][29][30][31][32][33][34][35][36][37][38], but only one study by Willmann et al. [31] found an increase in CL with a decrease in the BW of patients, which may not have clinical relevance (Figure 2).
The impact of genetic polymorphisms of ABCB1 (rs1045642, rs4148738, or rs4728709) on CL was reported, with an effect ranging from 19 to 57% among patients [34,36,37]. The study conducted by Zdovc et al. [32] identified the relative expression of the ABCB1 gene determined using the comparative Ct method as a significant covariate for CL in 17 patients [32].
Furthermore, co-medication, including CYP3A4 inhibitors, inducers, and P-gp inhibitors, was found to notably affect CL in two studies [30,31]. Co-medication with CYP3A4 inducers increased CL by approximately 30% compared to rivaroxaban monotherapy [31]. The influence of strong, moderate, and weak CYP3A4 inhibitors and P-gp inhibitors on CL was <15% in a study by Willmann et al. [31]. In contrast, it was approximately 32% in the study by Suzuki et al. [30].

Pharmacodynamics Analysis: PT
Most studies displayed similar VPDs in the PT-time profiles ( Figure S6), except for the study by Zdovc et al. [32], which may be due to the imprecise estimation of the PK and PD parameters in that study (ω ka : 794%; ω CL : 81%; ω slope : 109%). Therefore, that study was excluded from further analyses.
The relationship between rivaroxaban concentration and PT according to the type of PT reagent used is shown in Figure 3. As illustrated in Figure 3a (Neoplastin Plus), no obvious difference between Asian and Caucasian populations was observed when the same bioassay for PT was used. Moreover, the estimated baseline PT (PT 0 ) was similar between Asian (11.4-14 s) and Caucasian (11.4-13.9 s) populations, as listed in Table 3. The estimated change in PT per unit rivaroxaban concentration change (slope) was also similar between Asians [0.0018-0.0467 s/(µg/L)] [20,25,26,30,34,38] and Caucasians [0.032-0.0458 s/(µg/L)] [19,[22][23][24]27,32], but with large variability, which may be due to the different PT reagents employed in these studies.

Pharmacodynamics Analysis: FXa Activity
The relationship between FXa activity and rivaroxaban concentration in Asians was similar to that in Caucasians (Figure 4). The estimates of parameters related to FXa activity in the Asian population (FXa 0 : 0.803 U/mL; E max : 0.928 U/mL; EC 50 : 221 µg/L) [26] were comparable to those in Caucasian populations (FXa 0 : 0.87-1.0 U/mL; E max : 0.86-0.942 U/mL; EC 50 : 172-296 µg/L) using the same bioassay for FXa activity determination as listed in Table 3.
The VPDs of FXa activity-time profiles at steady state in all included studies are shown in Figure S8. Few covariates have been reported for FXa activity. Only the study by Girgis et al. [27] found that age had an impact on baseline FXa activity, but no details were provided. Because only one study was conducted on Japanese patients and two studies were conducted in a non-Japanese population, further investigation may need to be

Discussion
Rivaroxaban is a DOAC that was first approved in 2008 and is currently widely used globally, for the treatment and prevention of thromboembolic disorders. However, whether a lower rivaroxaban dose is required for Asian people is still debatable. To our knowledge, this is the first systematic review to summarize the knowledge regarding such potential ethnic differences from the perspective of the population PK and PK/PD profiles of rivaroxaban.
Population PK studies conducted in real-world patients showed that the median CL/F of rivaroxaban in adult Caucasian patients was approximately 31-43% higher than in Asian patients with normal renal and hepatic function. This finding is also supported by phase II and III clinical studies conducted on Japanese patients with NVAF [25,26]. An approximate one-third decrease in the CL/F in Asians might lead to the requirement of one-third reduction in the rivaroxaban dose, as compared to that for Caucasians. The effects of BW, age, renal function, and hepatic function could not explain the difference.
The difference in CL/F may be partially attributed to the genetic polymorphism of ABCB1. A population PK study conducted by Zhang et al. [37] identified that patients carrying ABCB1 (rs 4728709) AA/GA had a 47.6% higher CL/F than those with GG. Moreover, Zhang et al. [36] identified that patients carrying ABCB1 (rs1045642) CT/TT had a 25.7% higher CL/F than those with a CC polymorphism [36]. The frequency of ABCB1 (rs 4728709) AA/GA in American and African-American populations was determined to be approximately 2.1-fold higher than that in Asians. Moreover, it was also reported that the trough rivaroxaban concentration between different genotypes of ABCB1 at the rs128503 locus was significantly different [43]. However, all of these studies were performed in a Chinese population, and further studies in other ethnicities may be considered to explore the impact of genetic polymorphisms.
Moreover, food intake may also play a role in the differences in CL/F because the bioavailability of rivaroxaban is food-dependent upon administration at doses between 15 and 20 mg. Differences in dietary habits and diet content among ethnicities may also explain the differences observed. However, information about food, such as meal timing and dietary content, was not recorded in most of the studies and may deserve further investigation.
According to European Medicines Agency, no clinically relevant inter-ethnic differences are observed among Caucasian, African-American, Hispanic, Japanese, or Chinese patients regarding rivaroxaban PK and PD characteristics [44]. However, the statement was different from the Food and Drug Administration (FDA), which reported that healthy Japanese subjects have 20-40% higher exposures on average, compared to those in other ethnicities including Chinese, and these differences in exposure are reduced when corrected for body weight [45]. The reason for this different statement is unclear.
The PK and PD characteristics in Chinese healthy subjects (n = 8) taking a single dose was reported to be in line with that in Caucasian subjects when not corrected by body weight, even though a 47% lower area under the concentration-time curve (AUC) per body weight was observed in Chinese individuals [46]. However, AUC at steady state in Caucasian healthy subjects taking multiple doses of rivaroxaban were 1.32-to 1.51-fold of that in Chinese subjects [46,47]. Data from healthy subjects might not fully represent patients in the real world. Population analyses of real-world data could thus contribute to comprehensively delineating PK and PD characteristics.
Previous exposure-response analyses based on patients with NVAF showed that the risk of ischemic stroke, non-central nervous system systematic embolism, or all-cause death in Asians was not significantly different from that in individuals from other regions. However, Asians had a statistically significantly higher risk of major or non-major clinically relevant bleeding compared to that in West European and Latin American individuals [48]. The different risks of bleeding between geographic regions were also reported in other types of patients [49,50]. This could be partially explained by the other confounding factors, such as history of gastrointestinal bleeding, baseline use of non-steroidal anti-inflammatory drugs or aspirin, and age [48]. Moreover, the higher risk of bleeding in Asians might be attributed to higher rivaroxaban exposure compared to that in other ethnicities with the same dose regimen.
Renal function has a significant impact on the CL/F of rivaroxaban because 1/3 of the drug is eliminated unchanged in the urine, and about half of the metabolites are eliminated by the kidney. The change in CL/F by unit renal function has large variability among studies, ranging from 0.13 to 5.58%, but no obvious difference was observed between Asian and Caucasian populations.
Although approximately 2/3 of rivaroxaban is metabolized by hepatic enzymes [4], only 4 of 24 studies have reported the effect of the hepatic function index on the CL/F of rivaroxaban. The identified indices were ALT, TBIL, and CTP; none were well-recognized in previous studies [30,[36][37][38]. This may be because none of these covariates could fully reflect hepatic function.
The effect of age on CL/F was ≤30% and may not have significant clinical relevance as the drug label recommended by the FDA [45]. However, it should be noted that elderly patients in the real world have more concomitant diseases and are prone to have an unstable morbid state [51]. Besides the risk of increased PK exposure, elderly patients also have a higher risk of thrombotic and bleeding events. Therefore, further studies may need to be performed to assess the effects of rivaroxaban treatment on the elderly population.
It has been reported that Caucasian children weighing ≤ 30 kg had at least 47% higher CL/F per body weight (kg) than Caucasian adults. This can be explained by the fact that the ratio of liver to total body mass is larger in children than in adults, resulting in increased blood flow. Stronger hepatic metabolic enzyme activity may also partially contribute to this [52]. Owing to the lack of population PK studies in Asian children, further studies need to be conducted.
The exposure-PT relationship was described as linear or near-linear in Asian and Caucasian patients. The baseline PT and change in PT induced by rivaroxaban concentration in Asians were not significantly different from that in Caucasian populations, suggesting no distinct difference between Asian and Caucasian populations. However, it should be noted that PT may not be a specific biomarker for rivaroxaban and could be influenced by the bioassay method and test reagents used.
In addition, the relationship between exposure and FXa activity in the Asian population was similar to that in the Caucasian population, as described by the E max model, revealing comparable estimates in baseline FXa, EC 50 , and E max . However, only three studies (one in Japanese patients and two in non-Japanese patients) were conducted, which may require further investigation. Moreover, further studies are needed to explore the relationship between FXa activity and clinical outcomes, such as the rate of bleeding and thrombotic events.
Considering the lower CL/F of rivaroxaban in Asians than in Caucasians and the similar PK/PD relationship among ethnicities, it could be inferred that a lower dose of rivaroxaban is required for Asians. This finding is supported by studies based on other Asian populations, except for Japanese individuals [10][11][12]. Meanwhile, the efficacy of a lower rivaroxaban dose was also confirmed with Japanese patients in real world clinical settings [53]. However, owing to inconsistent findings [54][55][56], further studies with larger sample sizes are warranted for verification.
Our study had several limitations. First, because only literature published in English was included in our analysis, studies published in other languages were omitted. However, this may not essentially impact our conclusion because studies published in other languages may be limited by their small sample size. Second, we did not assign weights to studies with different sample sizes and reporting quality when comparing PK/PD characteristics. However, approximately 3/4 of the included studies had a sample size greater than 100 and a compliance rate ≥ 80%, which means that they should be recognized as well-reported.

Conclusions
In this review, the potential ethnic difference in the PK/PD of rivaroxaban was evaluated from the perspective of a population analysis. Approximately 31-43% lower CL/F of rivaroxaban was observed in Asians than Caucasians. However, the relationship between rivaroxaban concentration and PT or FXa activity was similar between the two ethnicities. Renal function was identified as a significant covariate of the CL/F of rivaroxaban, and no well-recognized covariates significantly affected PT or FXa activity. A lower dose of rivaroxaban might be required for Asians, and further studies are needed to explain the difference in the CL/F of rivaroxaban between Asian and Caucasian populations, which is essential for optimal patient dose regimens.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/pharmaceutics15020588/s1, Figure S1: PRISMA flow diagram for identifying population pharmacokinetic and population pharmacokinetic-pharmacodynamic studies of rivaroxaban; Figure S2: Concentration-time profiles of rivaroxaban at steady state; Figure S3: Distribution of apparent clearance of rivaroxaban in Japanese patients, Chinese patients, and other Asian patients with non-valvular atrial fibrillation; Figure S4: The apparent clearance of rivaroxaban versus renal function as reported in population pharmacokinetic models; Figure S5: The apparent clearance per body weight in pediatric and adult patients; Figure S6: Prothrombin time-time profiles of rivaroxaban at steady state for (a) Asian patients with NVAF and (b) Caucasian patients with or without NVAF; Figure S7: Effect of covariates on baseline PT (a) and slope (b); Figure S8: FXa activity-time profiles of rivaroxaban at steady state; Table S1: Checklist for literature quality when reporting a clinical pharmacokinetic study; Table S2: Final population pharmacodynamic parameters of the included studies; Table S3: Demographic information of the simulated patients.