Predictive thresholds of peak and trough anti-Xa levels for bleeding risk in rivaroxaban-treated nonvalvular atrial fibrillation

Abstract

Background

See More

While anti-Xa assays are increasingly used to monitor rivaroxaban therapy, evidence supporting specific thresholds for bleeding risk remains limited.

Objective

To determine predictive thresholds of peak and trough anti-Xa levels for bleeding complications in patients with nonvalvular atrial fibrillation (NVAF) receiving rivaroxaban.

Methods

In this prospective multicenter cohort study conducted at four tertiary care hospitals in Damascus, Syria, we enrolled 70 NVAF patients receiving rivaroxaban (20 mg/day; 15 mg/day if eGFR < 50 mL/min). Anti-Xa levels were measured at peak (1–3 h post-dose) and trough (pre-dose). Patients were followed for 6 months for bleeding events.

Results

Twenty-five patients (35.7%) experienced bleeding (8 major, 17 minor). Bleeding patients demonstrated significantly higher anti-Xa levels (peak: 399 ± 78 vs. 206 ± 67 ng/mL, p < 0.0001; trough: 41 ± 14 vs. 20 ± 10 ng/mL, p < 0.0001). ROC analysis identified optimal predictive thresholds of 298 ng/mL for peak levels (AUC = 0.985, sensitivity 89.5%, specificity 93.3%) and 27.5 ng/mL for trough levels (AUC = 0.887, sensitivity 86.4%, specificity 76.3%).

Conclusion

Anti-Xa levels strongly predict bleeding risk in rivaroxaban-treated NVAF patients. The identified thresholds may guide clinical decision-making regarding dose adjustment, particularly in high-risk patients. However, it is important to acknowledge limitations, including the modest sample size and the exclusion of patients on antiplatelet therapy, which may affect generalizability.

Introduction

Direct oral anticoagulants (DOACs), including rivaroxaban, have replaced vitamin K antagonists in many clinical settings due to predictable pharmacokinetics and reduced monitoring needs [1]. Some of the drugs in this group are inhibitors of active factor X such as rivaroxaban, apixaban, and edoxaban, and some are direct thrombin inhibitors such as dabigatran [1]. Unlike vitamin K antagonists, they are characterized by fewer interactions with other drugs, and their pharmacokinetics are more predictable, which has made it possible to apply them in fixed doses, limiting the need for routine monitoring of coagulation function [2]. Furthermore, in some cases, determining the anticoagulant effectiveness of DOACs is necessary for making a clinical decision, such as bleeding or thromboembolic events, preparation for surgery, or suspected excessive or insufficient anticoagulant action (e.g., renal or hepatic impairment, severe weight gain or loss, co-administration with drugs that have interfering effects, or alteration of the gastrointestinal tract anatomy) [3].

However, routine coagulation tests such as prothrombin time (PT) and activated partial thromboplastin time (aPTT) are not well suited for monitoring DOACs [18]. Their values do not reliably correlate with DOAC plasma levels, and prolonged times are not always consistently associated with the actual anticoagulant effect [18, 19]. Standard tests often fall short due to complex interactions between the drugs and the assays, and their sensitivity and specificity are insufficient for accurately reflecting plasma concentrations [18, 20]. This lack of reliable correlation makes them inadequate for assessing individual drug exposure or guiding clinical decisions [18].

Anti-factor Xa activity assay has been used (uncommonly) to measure the severity of anticoagulation function and to adjust the dose for patients taking DOACs [8,9,10]. It has been used previously to measure plasma levels of heparin and low molecular weight heparins. While anti-Xa assays correlate linearly with rivaroxaban plasma levels measured by high-performance liquid chromatography (HPLC), the gold standard for drug quantification [11,12,13,14,15,16,17]. Few studies have shown whether patients benefit fromlaboratory assessment of DOAC levels by anti-factor X activity assay, and in some studies,a correlation has been demonstrated between DOAC levels measured by anti-factor Xactivity assay and bleeding complications [7, 15,16,17]. In light of the limited data related tomonitoring DOAC levels in clinical practice, this study aims to evaluate the importance ofmonitoring rivaroxaban levels by anti-factor X activity assay in a group of patientsprescribed rivaroxaban (patients with nonvalvular atrial fibrillation) in several centers andhospitals Damascus and to study the correlation between the results of this assay andclinical data related to bleeding. Evidence supporting their clinical utility remains limited.This study evaluates anti-Xa assay thresholds for predicting bleeding complications inNVAF patients.

Methods

Study design and participants

This multi-center prospective cohort study was conducted across four major hospitals in Damascus, Syria: Al-Mowasat University Hospital, National Hospital, Damascus Hospital, and Red Crescent Hospital. The study enrolled 70 patients with non-valvular atrial fibrillation (NVAF), which was confirmed by electrocardiography (ECG), and who had been on rivaroxaban therapy for at least one month.

Inclusion and exclusion criteria

Adults aged 18 years or older with NVAF receiving guideline-directed rivaroxaban dosing (20 mg/day or 15 mg/day if eGFR was less than 50 mL/min) were included in the study. Patients were excluded if they had hepatic failure (Child-Pugh B/C), pre-existing bleeding disorders such as hemophilia or von Willebrand disease, concurrent antiplatelet therapy like aspirin or clopidogrel, or active peptic ulcer disease. Additionally, patients on medications known to significantly increase bleeding risk, such as nonsteroidal anti-inflammatory drugs (NSAIDs) or selective serotonin reuptake inhibitors (SSRIs), were also excluded to minimize confounding factors related to bleeding events not directly associated with rivaroxaban’s anticoagulant effect.

Data collection and procedures

Baseline assessments

Sociodemographic and clinical data, including age, sex, weight, and CHADS-VASc score (assessing stroke risk for congestive heart failure, hypertension, age 75 or older, diabetes, stroke/TIA history, vascular disease, and sex category), were collected at baseline.

Blood sampling and anti-xa measurement

Blood samples were collected at both peak (1–3 h post-dose) and trough (pre-dose) times. For sample handling, citrate tubes containing 3.2% sodium citrate were centrifuged at 2500 RCF for 15 min to obtain platelet-poor plasma, while EDTA tubes were processed within 6 h at room temperature for a complete blood count (CBC). Citrate plasma was stable for 2 h at room temperature or 1 month at − 20 °C and was thawed at 37 °C for 15 min before testing. Anti-Xa activity was measured using a chromogenic assay (Technochrom^® Anti-Xa, Austria) on an automated analyzer (SAT 450, AMS, Italy) with rivaroxaban-specific calibrators. The CBC was analyzed using a hematology analyzer (Medonic M20, Boule, Sweden).

Follow-up and outcomes

Patients were monitored for 6 months for bleeding events, including major or clinically relevant non-major events, and thrombotic events, such as stroke or systemic embolism. Major bleeding was defined according to the International Society on Thrombosis and Haemostasis (ISTH) criteria, which includes fatal bleeding, symptomatic bleeding in a critical area or organ (intracranial, intraspinal, intraocular, retroperitoneal, intra-articular, or pericardial), bleeding causing a fall in hemoglobin level of 20 g/L (1.24 mmol/L) or more, or leading to transfusion of two or more units of whole blood or red cells [21]. Minor bleeding was defined as any overt bleeding that did not meet the criteria for major bleeding but required medical intervention, hospitalization, or caused a temporary interruption of the study drug.

Sample size calculation

The sample size was determined using Cochran’s formula. With a 90% confidence level and a margin of error ranging from 7 to 12%, the estimated sample size required was between 48 and 139 participants. A total of 70 participants were enrolled in this study, a number chosen to balance the feasibility of recruitment with the need for adequate statistical power to achieve the study objectives.

Statistical analysis

Data were analyzed using IBM SPSS v24, with statistical significance set at a p-value of less than or equal to 0.05. Normality was assessed via the Shapiro-Wilk test. For comparisons, Student’s t-test was used for parametric data, and the Mann-Whitney U test was used for non-parametric data. Multivariate analysis involved logistic regression adjusted for age, eGFR, and CHADS-VASc score. ROC analysis was performed to determine anti-Xa cutoffs for bleeding and thrombosis risk, with AUC, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) reported.

Ethical considerations

The study received approval from the Biomedical Research Ethics Committee (BMREC; ID: MD-011024-049054-02) at Damascus University. Informed consent was obtained in writing from all participants after explaining their rights to voluntary participation and withdrawal. Confidentiality was maintained by not collecting personal identifiers, and all data was accessible only to the investigators..

Results

Baseline characteristics of participants

This study enrolled 70 participants with a mean age of 66 ± 10.9 years, consisting of 37 females (53%) and 33 males (47%). Participants were recruited from four centers: National Hospital (n = 24), Al-Mowasat University Hospital (n = 21), Damascus Hospital (n = 19), and Red Crescent Hospital (n = 6). The mean estimated glomerular filtration rate (eGFR) was 78.7 ± 18.3 mL/min. A significantly lower eGFR was observed in the bleeding group (59.3 ± 11.6 mL/min) compared to the non-bleeding group (82.2 ± 18.1 mL/min), with a p-value of 0.0052, suggesting an association between renal impairment and bleeding risk. All patients received rivaroxaban 20 mg/day, with treatment durations ranging from 1 month to 5 years.

The average CHADS-VASc score for patients without bleeding was 2.82 ± 1.64. For patients with bleeding generally, the average CHADS-VASc score was 3.16 ± 1.91 (P = 0.4022). Specifically, for minor bleeding, the average score was 3.00 ± 1.63 (P = 0.6622), and for major bleeding, it was 3.13 ± 2.03 (P = 0.6133). The average HAS-BLED index for patients without bleeding was 1.48 ± 0.94. For patients with bleeding generally, the average HAS-BLED index was 2.44 ± 1.47 (P = 0.0014). Specifically, for minor bleeding, the average index was 2.30 ± 1.54 (P = 0.0135), and for major bleeding, it was 2.50 ± 1.30 (P = 0.0103). The average duration of rivaroxaban use for patients without bleeding was 15.6 ± 11.58 months. For patients with minor bleeding, the average duration was 15.22 ± 11.47 months (NS). For patients with major bleeding, the average duration was 5.52 ± 5.37 months (P = 0.0153).

Below is a summary table for CHADS-VASc Score, HAS-BLED Index, and Duration of Rivaroxaban Use. Subgroup Comparisons.

Parameter	Non-Bleeding (n=45) (Mean ± SD)	Bleeding (General) (Mean ± SD)	Minor Bleeding (Mean ± SD)	Major Bleeding (Mean ± SD)	Significance (p-value)
CHADS-VASc score	2.82 ± 1.64	3.16 ± 1.91	3.00 ± 1.63	3.13 ± 2.03	0.4022
HAS-BLED index	1.48 ± 0.94	2.44 ± 1.47	2.30 ± 1.54	2.50 ± 1.30	0.0014
Duration of Rivaroxaban (months)	15.6 ± 11.58	-	15.22 ± 11.47	5.52 ± 5.37	NS (for minor), 0.0153 (for major)

Anti-Xa levels were measured at both peak (1–3 h post-dose) and trough (pre-dose). Mean peak levels were significantly higher in patients who experienced bleeding (399 ± 78 ng/mL) compared to those who did not (206 ± 67 ng/mL), with a p-value of less than 0.0001. Similarly, trough levels were also elevated in bleeders (41 ± 14 ng/mL) versus non-bleeders (20 ± 10 ng/mL), also with a p-value of less than 0.0001. Multivariate analysis confirmed that anti-Xa levels independently predicted bleeding after adjusting for eGFR, with an odds ratio of 1.02 (95% CI: 1.01–1.03). Of the 70 patients, 25 (35.7%) experienced bleeding events. Specifically, 8 (11.4%) experienced major bleeding, and 17 (24.2%) experienced minor bleeding.

Bleeding outcomes

Of the 70 patients, 25 (35.7%) experienced bleeding events, with 8 (11.4%) experiencing major bleeding and 17 (24.2%) experiencing minor bleeding.Subgroup Comparisons.

Parameter	Non-Bleeding (n = 45)	Bleeding (n = 25)	p-value
Age (years)	65.9 ± 10.4	69.3 ± 8.3	0.1721
Weight (kg)	84.4 ± 15.2	79.8 ± 15.7	0.2302
Hemoglobin (g/dL)	12.4 ± 1.3	11.3 ± 2.0	0.0083
Platelets (×10³/mm³)	260 ± 84	291 ± 99	0.1699
Creatinine (mg/dL)	1.09 ± 0.33	1.28 ± 0.39	0.0015
eGFR (mL/min)	82.2 ± 18.1	59.3 ± 11.6	0.0052
PT (sec)	16.6 ± 1.5	21.6 ± 7.2	0.0002
PTT (sec)	34.0 ± 3.5	40.6 ± 10.6	0.0010

Predictive performance of anti-xa assays

The predictive performance of peak anti-Xa levels was excellent, with an Area Under the Curve (AUC) of 0.985, indicating a strong ability to differentiate between patients with and without bleeding events (p < 0.001). An optimal cutoff for peak anti-Xa was determined to be 298 ng/mL (Fig. 1). At this cutoff, the assay demonstrated a sensitivity of 89.5% (95% CI: 66.9–98.7%), meaning it correctly identified most patients who experienced bleeding. Its specificity was 93.3% (95% CI: 81.7–98.6%), indicating a high rate of correctly identifying patients who did not bleed. The positive predictive value (PPV) was 85.0% (95% CI: 65.3–94.5%), suggesting that a high proportion of patients with levels above the cutoff experienced bleeding. Finally, the negative predictive value (NPV) was 95.5% (95% CI: 85.0–98.7%), meaning that most patients with anti-Xa levels below the cutoff did not experience bleeding.

Fig. 1

Peak time anti-Xa assay for predicting bleeding ROC curve

Trough anti-xa levels

The predictive performance of trough anti-Xa levels was found to be good, with an Area Under the Curve (AUC) of 0.887, indicating its strong capability in distinguishing between patients with and without bleeding events (p < 0.001). The optimal cutoff for trough anti-Xa was determined to be 27.5 ng/mL (Fig. 2). At this cutoff, the assay demonstrated a sensitivity of 86.4% (95% CI: 65.1–97.1%), meaning it correctly identified a high percentage of patients who experienced bleeding. Its specificity was 76.3% (95% CI: 59.8–88.6%), indicating a good rate of correctly identifying patients who did not bleed. The positive predictive value (PPV) was 67.9% (95% CI: 53.8–79.3%), suggesting that a considerable proportion of patients with levels above the cutoff experienced bleeding. Lastly, the negative predictive value (NPV) was 90.6% (95% CI: 76.9–96.6%), indicating that most patients with anti-Xa levels below the cutoff did not experience bleeding.CI: 76.9–96.6%)

Fig. 2

Trough time anti-Xa assay for predicting bleeding ROC curve

Interpretation: Peak anti-Xa demonstrated superior predictive accuracy for bleeding compared to trough levels.

Subgroup analysis by age

We performed a subgroup analysis to compare anti-Xa levels between patients aged less than 65 years and those aged 65 years or older.

For patients aged less than 65 years, the mean peak anti-Xa level for all patients was 238.68 ± 108.17 ng/mL, and the mean trough anti-Xa level was 24.52 ± 18.91 ng/mL. In this younger group, patients experiencing minor bleeding had mean peak anti-Xa levels of 377.75 ± 14.45 ng/mL and mean trough anti-Xa levels of 41.25 ± 25.26 ng/mL. Patients experiencing major bleeding in this age group had mean peak anti-Xa levels of 408.5 ± 30.40 ng/mL and mean trough anti-Xa levels of 61 ± 8.48 ng/mL.

For patients aged 65 years or older, the mean peak anti-Xa level for all patients was 300.13 ± 118.23 ng/mL, and the mean trough anti-Xa level was 33.89 ± 18.46 ng/mL. In this older group, patients experiencing minor bleeding had mean peak anti-Xa levels of 387.54 ± 68.44 ng/mL and mean trough anti-Xa levels of 40.77 ± 15.95 ng/mL. Patients experiencing major bleeding in this age group had mean peak anti-Xa levels of 437.33 ± 124.46 ng/mL and mean trough anti-Xa levels of 50.5 ± 9.65 ng/mL. No statistically significant difference was observed in the overall distribution of anti-Xa levels between the age groups (less than 65 vs. 65 or older) for either minor or major bleeding events.

Age Group (< 65 years)	Anti-Xa Peak (ng/mL) (Mean ± SD)	Anti-Xa Trough (ng/mL) (Mean ± SD)
All Patients	238.68 ± 108.17	24.52 ± 18.91
Minor Bleeding	377.75 ± 14.45	41.25 ± 25.26
Major Bleeding	408.50 ± 30.40	61.00 ± 8.48

Age Group (≥ 65 years)	Anti-Xa Peak (ng/mL) (Mean ± SD)	Anti-Xa Trough (ng/mL) (Mean ± SD)
All Patients	300.13 ± 118.23	33.89 ± 18.46
Minor Bleeding	387.54 ± 68.44	40.77 ± 15.95
Major Bleeding	437.33 ± 124.46	50.50 ± 9.65

Predictive performance of anti-xa assays for major bleeding

To specifically address the predictive performance for major bleeding events, we performed a separate ROC analysis excluding minor bleeding cases from the threshold calculations.

Peak anti-xa levels for major bleeding

The ROC analysis for peak anti-Xa levels in predicting major bleeding yielded an Area Under the Curve (AUC) of 0.875 (95% CI: 0.768–0.982, p = 0.001). An optimal predictive threshold of 309 ng/mL was identified. At this cutoff, the assay demonstrated a sensitivity of 87.50% and a specificity of 71% for predicting major bleeding. The ROC curve for peak anti-Xa assay in predicting major bleeding is presented in Fig. 3.

Fig. 3

Peak time anti-Xa assay for predicting major bleeding ROC curve

Trough anti-xa levels for major bleeding

For trough anti-Xa levels in predicting major bleeding, the ROC analysis showed an AUC of 0.895 (95% CI: 0.816–0.974, p < 0.0001). The optimal predictive threshold was determined to be 47 ng/mL. At this cutoff, the assay demonstrated a sensitivity of 87.50% and a specificity of 87.10% for predicting major bleeding. The ROC curve for trough anti-Xa assay in predicting major bleeding is presented in Fig. 4.

Fig. 4

Trough time anti-Xa assay for predicting major bleeding ROC curve

It is important to note that for these specific ROC analyses, patients experiencing only minor bleeding were excluded from the calculation of threshold values; only major bleeding events were considered. This approach was taken to distinguish the predictive value for severe bleeding outcomes. While minor bleeding events were recorded, they were not included in this particular threshold determination, as minor bleeding may not always necessitate intervention unless it meets criteria for clinically relevant non-major bleeding (CRNMB). Our study focused on identifying thresholds for major bleeding due to its significant clinical implications.

Below is a summary table for the ROC analysis for major bleeding.

Parameter	Cutoff (ng/mL)	Sensitivity (%)	Specificity (%)	AUC	p-value	95% Confidence Interval (AUC)
Peak Anti-Xa (Major Bleeding)	309	87.5	71	0.875	0.001	0.768–0.982
Trough Anti-Xa (Major Bleeding)	47	87.5	87.1	0.895	< 0.0001	0.816–0.974

ROC curves are graphical representations of the diagnostic ability of a binary classifier system as its discrimination threshold is varied. They are essential for visualizing the trade-off between sensitivity and specificity and for identifying optimal cutoff points. Including these figures directly supports and illustrates the statistical analysis you’ve presented for the predictive performance of anti-Xa levels for major bleeding. They are standard practice in papers presenting ROC analysis.

Discussion

Our study evaluated rivaroxaban (20 mg/day) in patients with non-valvular atrial fibrillation (NVAF), revealing an overall bleeding rate of 35.7%, with a notably high proportion of major bleedings at 11.4%. This rate is higher than reported in many large randomized controlled trials and real-world registries for rivaroxaban. Several interconnected factors likely contribute to this observation and justify the observed high bleeding rate.

Primarily, our cohort may have included a higher proportion of inherently high-risk patients, a common characteristic of real-world studies outside controlled trial settings. This includes individuals who were hospitalized or potentially had active bleeding at baseline, conditions that are often exclusion criteria in pivotal clinical trials, thereby leading to lower observed bleeding rates in those more selected populations.

Furthermore, we observed significant indicators of increased bleeding risk within our patient group. Patients who experienced bleeding, including major events, demonstrated significantly higher creatinine levels (1.28 ± 0.39 mg/dL vs. 1.09 ± 0.33 mg/dL, p = 0.0015) and lower eGFR (59.3 ± 11.6 mL/min vs. 82.2 ± 18.1 mL/min, p = 0.0052) compared to the non-bleeding group. Kidney insufficiency is a well-established strong risk factor for bleeding in patients on anticoagulants. Although our study protocol included a rivaroxaban dose reduction from 20 mg/day to 15 mg/day for patients with an eGFR less than 50 mL/min, it is plausible that some patients with eGFR just above this threshold, or those with fluctuating kidney function not captured by single measurements, may still have experienced higher-than-intended rivaroxaban exposure. The significantly higher peak (399 ± 78 ng/mL vs. 206 ± 67 ng/mL, p < 0.0001) and trough (41 ± 14 ng/mL vs. 20 ± 10 ng/mL, p < 0.0001) anti-Xa levels in the bleeding group (including those with major bleeding) further support the notion of increased drug exposure contributing directly to this elevated bleeding risk. Importantly, our multivariate analysis confirmed that anti-Xa levels independently predicted bleeding even after adjusting for age and eGFR, with an odds ratio of 1.02 (95% CI: 1.01–1.03). This suggests that anti-Xa levels can provide additional predictive value for bleeding risk, reflecting individual pharmacokinetic profiles beyond what is captured by eGFR alone.

Another contributing factor could be the patient cohort demonstrating significant gender-based differences, with males experiencing bleeding events more frequently than females (40% versus 20.5%), though the specific underlying reasons for this difference were not further explored in our study. Lastly, while not explicitly measured, potential patient adherence issues or undetected concomitant use of other bleeding-risk-increasing medications (e.g., NSAIDs, SSRIs, which were exclusion criteria but could theoretically have been used unknowingly to investigators), could also contribute to the observed bleeding rates. However, our strict exclusion criteria aimed to mitigate these confounding factors. These factors, particularly the high prevalence of renal impairment and likely increased systemic drug exposure as indicated by anti-Xa levels, collectively offer a clear justification for the unexpectedly high proportion of major bleeding events observed in our study compared to other cohorts.

Through statistical analysis using Student’s t-test, we observed highly significant differences in anti-Xa assay results between bleeding and non-bleeding groups (p < 0.0001) at both peak and trough times. Receiver operating characteristic (ROC) curve analysis identified clinically relevant cutoff values for bleeding prediction: 298 ng/mL at peak time (sensitivity 89.47%, specificity 93.33%) and 27.5 ng/mL at trough time (sensitivity 86.36%, specificity 76.32%), suggesting these thresholds may serve as important indicators of bleeding risk during rivaroxaban therapy.

The trough cutoff value of 27.5 ng/mL is relatively low and close to the lower limit of quantification of the assay. It is important to clarify that this threshold was identified through ROC analysis specifically for predicting an increased risk of bleeding events, rather than as a therapeutic target for stroke prevention. While very low trough levels might indeed raise concerns about insufficient anticoagulation for stroke prevention, our study’s primary objective was focused on identifying markers for bleeding complications. The observation that even a relatively low trough level can be associated with bleeding in our cohort may reflect individual patient susceptibility to bleeding at lower drug concentrations, or variations in drug metabolism and clearance within the population. This underscores the complexity of rivaroxaban pharmacodynamics, where even residual drug levels can confer a bleeding risk in sensitive individuals. Clinicians must consider both bleeding and stroke risks when interpreting anti-Xa levels and making dose adjustment decisions.

These findings align with but also differ from Japanese studies (16,17), reflecting important population-specific variations in pharmacokinetics and clinical outcomes.

Japanese guidelines, for example, recommend lower rivaroxaban doses (15 mg/day for NVAF, 10 mg/day for patients with CrCl 15–49 mL/min) due to documented pharmacokinetic differences in Asian populations, including 31–43% lower drug clearance compared to Western populations. The J-ROCKET AF trial validated the 15 mg/day dose as non-inferior to warfarin for stroke prevention while demonstrating reduced gastrointestinal bleeding, though the XAPASS subanalysis revealed that underdosing (10 mg/day in patients with CrCl ≥ 50 mL/min) increased thromboembolic events despite similar bleeding rates. Our anti-Xa thresholds showed numerical differences from Japanese studies such as Sawa et al. (2019), which reported a higher cutoff of 400 ng/mL with 100% sensitivity and an excellent AUC of 0.908, and Sakaguchi et al. (2017), which found a cutoff of 493 ng/mL (2.19 IU/mL) with more modest predictive performance (AUC: 0.73). These variations underscore the need for population-specific thresholds, as supported by additional evidence showing that Japan-specific dosing regimens yield more predictable anti-Xa levels than standard doses in Asian populations.

Real-world evidence from Japan further informs these findings, with the EXPAND study demonstrating that 10 mg/day reduced bleeding but was less effective in preventing coronary events in elderly patients (≥ 65 years), highlighting important dose-dependent efficacy trade-offs. Similarly, the J-EINSTEIN program for venous thromboembolism showed that 15 mg/day rivaroxaban achieved efficacy and safety comparable to warfarin. (16)

While our study offers valuable insights as the first of its kind in the Arabic world, addressing a critical gap in regional pharmacovigilance data, several limitations must be acknowledged, with a particular emphasis on their potential impact on the study’s conclusions. The modest sample size (n = 70) and the single-region design (four tertiary care hospitals in Damascus, Syria) significantly constrain the generalizability of our findings. The modest sample size, constrained by strict exclusion criteria such as concomitant antiplatelet use, may particularly limit the broader applicability of our identified thresholds. Antiplatelet use is common in NVAF management, and by excluding these patients, our cohort may not fully represent the real-world population of NVAF patients on rivaroxaban who might also be on antiplatelet therapy. This exclusion means our findings might not be directly transferable to patients requiring dual antithrombotic therapy, potentially underestimating or overestimating bleeding risks in such complex cases. Additionally, the single-region design, while providing valuable local data, means that our results are specific to the Syrian population and healthcare setting. Population-specific pharmacokinetic variations, as evidenced by the differences observed when comparing our thresholds to Japanese studies, underscore the need for local validation of anti-Xa thresholds. Therefore, the optimal thresholds identified in our study (298 ng/mL for peak and 27.5 ng/mL for trough) may not be universally applicable and highlight the potential impact of ethnic and demographic differences in drug metabolism and response on observed bleeding risks. Furthermore, differences in calibration methods across studies underscore the need for standardized protocols.

It is important to note that, according to drug licensing and phase 3 trial results, rivaroxaban dose reductions are primarily based on established clinical characteristics and specific laboratory values (such as eGFR) rather than on routine pharmacokinetic monitoring or the measurement of drug levels. Therefore, while our study highlights the potential utility of anti-Xa monitoring in identifying patients at increased bleeding risk, particularly in those with renal impairment and higher drug exposure, these findings currently represent an area of ongoing research and are not yet integrated into standard clinical practice for dose adjustment. The identified anti-Xa thresholds in our study serve as predictive indicators of bleeding risk and may guide investigational clinical decision-making regarding potential dose adjustments, especially in high-risk patients or those with renal impairment, but they do not supersede current guideline-directed therapy.

These collective findings suggest that while anti-Xa monitoring shows significant promise for bleeding prediction, optimal thresholds require local validation. Asian patients may particularly benefit from adjusted dosing regimens to balance efficacy and safety. Future research should focus on establishing standardized protocols that account for ethnic differences in drug metabolism and response, potentially through multinational studies comparing different dosing strategies across diverse populations, to optimize rivaroxaban therapy while minimizing bleeding risks across various patient demographics and clinical settings.

Conclusion

This study highlights the clinical utility of anti-factor Xa (anti-Xa) monitoring in patients with non-valvular atrial fibrillation (NVAF) receiving rivaroxaban, identifying significant associations between elevated anti-Xa levels and bleeding risk. Our findings demonstrate that peak anti-Xa levels > 298 ng/mL and trough levels > 27.5 ng/mL serve as robust predictors of bleeding, with high sensitivity and specificity. These thresholds provide actionable guidance for clinicians in assessing anticoagulation intensity, particularly in high-risk patients or those with renal impairment.

The observed differences in optimal anti-Xa thresholds compared to prior studies—particularly those from Japanese cohorts—underscore the influence of population-specific pharmacokinetics and reinforce the critical need for locally validated reference ranges. While rivaroxaban is traditionally administered in fixed doses, our results suggest that targeted monitoring may enhance safety in select patients, such as those with renal dysfunction or extreme body weight.

Despite limitations, including a modest sample size and single-region design, which inherently limit the generalizability of our findings, this study contributes valuable real-world data from an Arabic population, addressing a critical gap in DOAC pharmacovigilance. Crucially, as patients on antiplatelet therapy were excluded from this study, its conclusions should not be generalized to individuals who are concurrently taking antiplatelets.

To address these limitations and build upon our findings, future multicenter studies with larger, ethnically diverse cohorts are warranted. Specifically, such studies should recruit a significantly larger and more diverse patient population to improve statistical power and generalizability across different demographics and ethnicities. They should also include patients on concomitant antiplatelet therapy to better reflect real-world clinical practice and assess the interaction with rivaroxaban in a higher-risk group. Furthermore, standardizing anti-Xa assay calibration methods across participating centers is crucial to enhance comparability of results. Future research should also conduct head-to-head comparisons of different rivaroxaban dosing strategies (e.g., standard versus adjusted based on anti-Xa levels) to explore their impact on both bleeding and thrombotic outcomes. Finally, investigating the precise mechanisms underlying population-specific pharmacokinetic variations is essential to better inform individualized dosing regimens.

Until such evidence supports changes to prescribing information, rivaroxaban therapy should continue to adhere to established dose reduction criteria. Nevertheless, judicious anti-Xa monitoring, coupled with individualized risk assessment, may serve as a valuable tool to optimize rivaroxaban therapy by balancing efficacy and bleeding risk, particularly in high-risk patients or those with renal impairment.