Effectiveness and safety of rivaroxaban vs warfarin in patients with pulmonary embolism and right ventricular dysfunction: a retrospective cohort study

Abstract

Background

Limited data exist comparing the effectiveness and safety of rivaroxaban and warfarin in patients with right ventricular dysfunction (RVD), a common acute pulmonary embolism (PE) complication.

Objectives

To assess the effectiveness and safety of rivaroxaban compared with warfarin among patients with PE and RVD.

Methods

Adult patients newly prescribed rivaroxaban or warfarin during PE-related hospitalization with evidence of RVD were identified from Mass General Brigham’s Research Patient Data Registry database (January 2013-May 2023). Outcomes included time-to-first recurrent venous thromboembolism (VTE) and time-to-first major bleeding event. The proportion of international normalized ratio (INR) measurements within therapeutic range (INR: 2-3) while on warfarin was described. Kaplan–Meier analysis described event rates at 6-month intervals up to 36 months, which were compared using hazard ratios, 95% CIs, and P values from Cox proportional hazards models.

Results

Overall, 246 rivaroxaban and 315 warfarin users were included (mean age, 63 years; female: 53%). Median time of treatment was 270 and 235 days for rivaroxaban and warfarin users, respectively; 50.9% of INR measurements among warfarin users were within therapeutic range. Rivaroxaban was associated with significantly lower risk of VTE recurrence than warfarin at all-time points, including 41% lower risk at 36 months (20.4% vs 30.3%; hazard ratio [95% CI], 0.59 [0.38, 0.92]). There was no significant difference in risk of major bleeding between cohorts up to 36 months of treatment (8.2% vs 13.6%).

Conclusion

Rivaroxaban was associated with lower risk of recurrent VTE compared with warfarin, without a significant difference in risk of major bleeding.

Essentials

• •Do clinical outcomes differ between patients treated with rivaroxaban vs warfarin for PE/RVD?.
• •This study evaluated treatment effectiveness and safety using Mass General Brigham data.
• •Rivaroxaban was more effective than warfarin without a significant increase in safety.
• •These results may help guide future research on-treatment choices for patients with PE/RVD.

1. Introduction

Right ventricular dysfunction (RVD) may occur as a consequence of volume and/or pressure overload in the setting of pulmonary embolism (PE) due to thromboembolic obstruction, circulating pulmonary vasoconstrictors, and hypoxemic vasoconstriction [1]. Patients with PE and evidence of RVD have a higher risk of adverse outcomes, with a more than 2-fold higher risk of PE-related mortality than those without RVD [2]. Additionally, persistence of RVD is associated with significantly higher risk of venous thromboembolism (VTE) recurrence [3]. As such, early detection and treatment of PE and RVD are crucial for improving outcomes [1].

The American Society of Hematology, American College of Chest Physicians, and European Society of Cardiology guidelines recommend use of direct oral anticoagulants (DOACs), such as rivaroxaban, over vitamin K antagonists for the treatment of PE, in large part as a result of more convenient and easier use that makes the transition to and follow-up in an outpatient setting easier while preserving efficacy and reducing the risk of major bleeding [[4], [5], [6], [7]]. However, guidance on the choice of DOAC among the subgroup of patients with both PE and RVD remains unclear, as there is limited evidence on the comparative effectiveness of rivaroxaban vs warfarin in this patient population. Therefore, the objective of this study was to compare the effectiveness and safety, as evaluated using the risk of recurrent VTE and major bleeding, respectively, of rivaroxaban vs warfarin among patients with PE and evidence of RVD using real-world data.

2. Methods

2.1. Data source

This study used data from the Mass General Brigham (MGB) Research Patient Data Registry (RPDR) PE Data Mart, spanning from January 2013 to May 2023. MGB RPDR aggregates hospital inpatient and outpatient clinical information from electronic health record (EHR) data from 8 major hospitals affiliated with the Harvard Medical School in Massachusetts (Massachusetts General Hospital, Brigham and Women’s Hospital, Brigham and Women’s Faulkner Hospital, Massachusetts Eye and Ear Hospital, McLean Hospital, Newton–Wellesley Hospital, Salem Hospital [formerly North Shore Medical Center], and Spaulding Rehabilitation Hospital). The database stores clinical information for >7 million patients and >3 billion records. The data elements include demographics, providers, visits, diagnoses, medications, procedures, and laboratories, along with all prescription orders placed by providers throughout the MGB hospital system, including those from inpatient and outpatient facilities. The MGB RPDR PE Data Mart also includes clinicians’ notes and reports from cardiology, pulmonary, and radiology, which can be analyzed using natural language processing (NLP) to extract clinical information such as evidence of RVD.

2.2. Study design and population

A retrospective observational design was used to evaluate and compare outcomes among patients with diagnosis of PE in any position (see Supplementary Table S1 for International Classification of Diseases [ICD]-9/10 diagnosis codes) and evidence of RVD treated with rivaroxaban or warfarin (Figure 1). Evidence of RVD was identified from clinical notes using NLP in combination with RVD-related search terms in Table 1 without negation words (eg, “none” and “without”) within the same sentence. The date of the first record of rivaroxaban or warfarin use per review of EHR data was defined as the initiation date, and the 6-month period prior to the initiation date was defined as the baseline period. Patients’ demographics and clinical characteristics were evaluated during the baseline period, and treatment patterns were described between the first and the last prescription of the index agent (ie, rivaroxaban or warfarin). The effectiveness and safety outcomes were evaluated using an on-treatment approach, with the observation (follow-up) period spanning from the day after discharge from the index PE hospitalization (index date) until the earliest of treatment discontinuation, switch to or addition of another oral anticoagulant (OAC), end of clinical activity, end of data availability, active cancer, or death. Information on availability of days of supply in the prescription data was limited; therefore, the discontinuation date was defined as the date of the last prescription of the index agent plus the median time between consecutive prescriptions of the index agent.

Figure 1.

Study design. Data source: Mass General Brigham Research Patient Data Registry, January 2013 to May 2023. NLP, natural language processing; PE, pulmonary embolism; RVD, right ventricular dysfunction; VTE, venous thromboembolism. ^aOnly 15 mg and 20 mg prescriptions were considered for identifying rivaroxaban initiation. ^bDiscontinuation was defined as the last prescription of the index agent plus the median gap between prescriptions (ie, 64 and 33 days for rivaroxaban and warfarin, respectively). ^cObservation was censored at the time of active cancer, which was identified by the presence of both cancer diagnosis and treatment (at the later of the 2 dates). ^dDiagnosis of PE was identified in any position using the following diagnosis codes: International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM): 415.1x; International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM): I26.0x, I26.90, I26.92, and I26.99. Among the study population, >96% of patients in each cohort had a primary diagnosis of PE or a secondary diagnosis of PE combined with an admission diagnosis of PE or a PE-related procedure during the index PE hospitalization [8]. ^eEvidence of RVD was identified using NLP, and the list of terms presented in Table 1. ^fVTE recurrence was defined as having ≥1 primary diagnosis of VTE in an inpatient setting during the on-treatment period.

Table 1.

List of right ventricular dysfunction search terms.

Search termsa	Variations
Acute cor pulmonaleb	Acute core pulmonale, acute cor pulmonal, acute core pulmonal.
Increased RV to LV ratio	Increased RV to LV, increased RV/LV, increased RV:LV, increased RV-LV, increase in RV to LV, increase in RV/LV, increase in RV:LV, increase in RV-LV, RV to LV ratio increase, RV/LV ratio increase, RV:LV ratio increase, RV-LV ratio increase, RV to LV increase, RV/LV increase, RV:LV increase, RV-LV increase.
McConnell’s signb	McConnell sign, McConnells sign, McConnel’s sign, McConnel sign, McConnels sign, McConell’s sign, McConells sign, McConell sign.
Right heart strain	Right-sided heart strain, right-sided heart strain.
Right ventricular dilatation	RV dilatation, right ventricle dilatation.
Right ventricular dilation	RV dilation, right ventricle dilation, RV dilated, right ventricle dilated.
Right ventricular dysfunction	RV dysfunction, right ventricle dysfunction.
Right ventricular enlargement	RV enlargement, right ventricle enlargement.
Right ventricular failure	RV failure, right ventricle failure.
Right ventricular hypokinesis	RV hypokinesis, right ventricle hypokinesis.
Right ventricular pressure	RV pressure, right ventricle pressure, right ventricular pressure overload, right ventricle pressure overload, RVPO.
Right ventricular strain	RV strain, right ventricle strain.
Right-sided heart dilatation	Right-sided heart dilatation, right heart dilatation.
Right-sided heart dilation	Right-sided heart dilation, right heart dilation.
Right-sided heart enlargement	Right-sided heart enlargement, right heart enlargement.
Right-sided heart failure	Right-sided heart failure, right heart failure.

LV, left ventricle; RV, right ventricle; RVPO, right ventricular pressure overload.

^aRight ventricular dysfunction terms with a negation word (eg, “none” and “without”) within the same sentence were omitted to limit the occurrence of false positives.

^bIncludes typos as variations. These typos do not signify another condition.

Adult patients were included in the study population if they met the following criteria: initiated rivaroxaban (15 mg or 20 mg) or warfarin (index agent) during hospitalization with ≥1 diagnosis of PE in any position (index PE hospitalization; see Supplementary Table S1 for ICD-9/10 diagnosis codes); had a prescription of the index agent as the first OAC prescribed after discharge from the index PE hospitalization, which was used to confirm use of the index agent; had evidence of RVD during the index PE hospitalization; and had ≥6 months of continuous clinical activity prior to the index PE hospitalization. Patients were excluded if they met any of the following criteria: had a prescription for an OAC (excluding rivaroxaban 2.5 mg twice a day) in the 6 months prior to the index PE hospitalization, a diagnosis of VTE in any position (see Supplementary Table S1 for ICD-9/10 diagnosis codes) in the 3 months prior to the index PE hospitalization, and a prescription for rivaroxaban 2.5 mg twice a day during the index PE hospitalization; had active cancer (defined as having both a cancer diagnosis and treatment), mechanical heart valve procedure, or diagnosis of antiphospholipid syndrome in the 6 months prior to or during the index PE hospitalization; had a knee or hip replacement surgery in the 35 days prior to or on the initiation date or had pregnancy or renal dysfunction (estimated glomerular filtration rate [eGFR] < 30) during the baseline period or on the initiation date [9,10].

Patients meeting the study selection criteria were classified into mutually exclusive regimens of rivaroxaban or warfarin cohorts based on the medication prescribed during the index PE hospitalization and the first OAC prescription observed after discharge, which was used to confirm medication use.

2.3. Study outcomes

Effectiveness was assessed using time-to-first recurrent VTE, which was defined as hospital readmission with a principal discharge diagnosis of VTE after the index PE hospitalization. Safety was assessed using time-to-first major bleeding event identified by the Cunningham algorithm, which identifies hospitalizations with diagnoses and procedures indicating a current episode of bleeding (excluding bleeding due to major trauma) [11]. In addition, the proportion of international normalized ratio (INR) measurements within therapeutic range (INR of 2-3) while on treatment was described among warfarin users.

2.4. Statistical analysis

Descriptive statistics included mean, SD, and median values for continuous variables and relative frequencies and proportions for categorical variables. To address potential confounding, cohorts were weighted using propensity score (PS) overlap weighting (PS-OW) [12,13]. In PS-OW, patients who are the most similar between cohorts are given the most weight, leading to a pseudopopulation that emphasizes patients at clinical equipoise. PS-OW achieves an exact balance of every variable included in the PS model without being prone to extreme weights and has been shown to outperform standard adjustment methods (eg, inverse probability of treatment weighting and PS matching) for the estimation of the treatment effect in heterogeneous cohorts [12]. The following patient characteristics were included in the PS-OW calculation: age; sex (male/female); race; year of index agent initiation date; RVD terms identified during the index PE hospitalization; unprovoked index PE (yes/no); baseline healthcare resource utilization; Quan–Charlson Comorbidity Index (0-24) [14]; modified Registro Informatizado de Enfermedad TromboEmbólica bleeding score (0-8) [15]; major bleeding; comorbidities, medications, and procedures with prevalence ≥10% in either cohort; eGFR; hemoglobin level; and red blood cell, white blood cell, and platelet counts. Patients with missing values for race and insurance plan type were considered as “unknown.” Differences in baseline characteristics before and after PS-OW were assessed between cohorts using standardized differences.

To account for truncation at the end of follow-up (ie, treatment discontinuation, switch to another OAC, end of clinical activity, end of data, or death), time-to-first recurrent VTE (effectiveness) and major bleeding (safety) were described in PS-OW–adjusted cohorts using Kaplan–Meier analysis, and event rates at 6, 12, 18, 24, 30, and 36 months of treatment were compared between weighted cohorts using hazard ratios (HR), 95% CIs, and P values derived from Cox proportional hazards regression models.

All analyses were conducted using software SAS Enterprise Guide version 7.15 (SAS Institute). This study was approved by the MGB Institutional Review Board.

3. Results

3.1. Patient characteristics

A total of 1727 and 3458 patients with PE were screened for inclusion in the rivaroxaban and warfarin cohorts, respectively (Supplementary Figure S1). Approximately 38% of those discharged with confirmed use of rivaroxaban (384/1005) or warfarin (726/1866) had evidence of RVD (clinical notes supporting evidence of RVD among 20 randomly selected patients are presented in Supplementary Table S2). After applying the remaining selection criteria, 246 rivaroxaban and 315 warfarin patients were included in the study population. Among the study population, >96% of patients in each cohort had a primary diagnosis of PE or a secondary diagnosis of PE combined with an admission diagnosis of PE or a PE-related procedure during the index PE hospitalization [8].

After PS-OW, the mean age was 63 years, with 53% of patients being female. During baseline, the mean Quan–Charlson Comorbidity Index and modified Registro Informatizado de Enfermedad TromboEmbólica bleeding scores were 1.8 and 2.7, respectively. Most patients had at least 1 risk factor for both VTE and bleeding (77%); the most prevalent comorbidities were obesity (44%) and coronary artery disease (28%). Most patients were prescribed non-OACs (95%)—including unfractionated heparin and low-molecular-weight heparin—and cardiovascular medications (82%), and 54% of patients had mild or moderate loss of kidney function (eGFR < 90; all standardized differences = 0%; Table 2). The full list of patient characteristics before and after PS-OW is presented in Supplementary Tables S3–S6.

Table 2.

Baseline demographics and clinical characteristics.

Characteristics	Unweighted cohorts							Weighted cohortsa
	Rivaroxaban n = 246			Warfarin n = 315			Std. diff.b (%)	Rivaroxaban n = 246			Warfarin n = 315			Std. diff.b (%)
On-treatment period, d, mean ± SD (median)	682	± 796	(287)	676	± 829	(293)	0.7	699	± 857	(270)	599	± 764	(235)	12.3
Demographics
Age,c y, mean ± SD (median)	61.1	± 16.2	(64)	63.5	± 16.4	(66)	14.5	63.1	± 15.1	(65)	63.1	± 17.2	(64)	0.0
Female, n (%)	117	(47.6)		165	(52.4)		9.7	130	(52.6)		166	(52.6)		0.0
Race, n (%)
White	180	(73.2)		250	(79.4)		14.6	188	(76.5)		241	(76.5)		0.0
Black	38	(15.4)		38	(12.1)		9.8	38	(15.3)		48	(15.3)		0.0
Other	20	(8.1)		18	(5.7)		9.5	14	(5.7)		18	(5.7)		0.0
Unknown	8	(3.3)		9	(2.9)		2.3	6	(2.5)		8	(2.5)		0.0
RVD terms during the index PE hospitalization, n (%)
Right ventricular strain	182	(74.0)		229	(72.7)		2.9	177	(71.8)		226	(71.8)		0.0
Right ventricular dilation	105	(42.7)		145	(46.0)		6.7	98	(40.0)		126	(40.0)		0.0
Right ventricular dysfunction	73	(29.7)		138	(43.8)		29.6	83	(33.7)		106	(33.7)		0.0
Acute cor pulmonale	46	(18.7)		37	(11.7)		19.4	34	(13.8)		44	(13.8)		0.0
McConnell’s sign	21	(8.5)		31	(9.8)		4.5	22	(9.1)		29	(9.1)		0.0
Other termsd	20	(8.1)		29	(9.2)		3.8	24	(9.8)		20	(6.5)		12.4
Quan-CCI,emean ± SD (median)	1.42	± 1.76	(1)	2.33	± 1.98	(2)	48.5	1.77	± 1.88	(1)	1.77	± 1.79	(2)	0.0
Modified RIETE bleeding score,fmean ± SD (median)	2.41	± 1.58	(2)	3.13	± 1.72	(3)	43.7	2.71	± 1.69	(3)	2.71	± 1.64	(3)	0.0
Major bleeding,gn (%)	11	(4.5)		34	(10.8)		24.0	15	(6.3)		20	(6.3)		0.0
eGFR,h,imean ± SD (median)	90	± 18	(90)	80	± 24	(82)	44.9	86	± 18	(87)	86	± 23	(88)	0.0
Healthcare resource utilization, mean ± SD (median)
Hospitalizations	1.27	± 0.76	(1)	1.38	± 0.85	(1)	14.0	1.30	± 0.85	(1)	1.30	± 0.76	(1)	0.0
Emergency room visits	0.26	± 0.82	(0)	0.20	± 0.59	(0)	7.4	0.21	± 0.66	(0)	0.21	± 0.58	(0)	0.0
Outpatient visits	6.54	± 9.39	(3)	8.73	± 10.56	(5)	21.9	7.97	± 11.12	(5)	7.97	± 10.14	(5)	0.0
Lab values of interest,hn (%) patients with normal values
Hemoglobin level (females: 12-16 g/dL; males: 14-18 g/dL)	99	(40.2)		84	(26.7)		29.1	82	(33.3)		105	(33.3)		0.0
Red blood cell count (4-6 × 106 cells/μL)	159	(64.6)		150	(47.6)		34.8	144	(58.7)		185	(58.7)		0.0
White blood cell count (4.5-11 × 103 cells/μL)	209	(85.0)		231	(73.3)		28.9	199	(80.7)		254	(80.7)		0.0
Platelet count (150-450 × 103 cells/μL)	198	(80.5)		247	(78.4)		5.1	199	(81.0)		255	(81.0)		0.0
Medication use, n (%)
Non-OACs	229	(93.1)		301	(95.6)		10.7	234	(95.2)		300	(95.2)		0.0
Cardiovascular medications	187	(76.0)		275	(87.3)		29.5	202	(82.1)		259	(82.1)		0.0
NSAIDs	127	(51.6)		127	(40.3)		22.8	114	(46.4)		146	(46.4)		0.0
Diabetes medications	42	(17.1)		122	(38.7)		49.8	56	(22.8)		72	(22.8)		0.0
Glucocorticoids	60	(24.4)		109	(34.6)		22.5	74	(30.1)		95	(30.1)		0.0
SSRIs	26	(10.6)		54	(17.1)		19.1	35	(14.4)		45	(14.4)		0.0
Cardiovascular procedure, n (%)	40	(16.3)		66	(21.0)		12.1	34	(13.8)		43	(13.8)		0.0
Comorbidities of interest, n (%)
Obesity (BMI ≥30 or diagnosis)	92	(37.4)		145	(46.0)		17.6	108	(43.9)		138	(43.9)		0.0
Coronary artery disease	82	(33.3)		102	(32.4)		2.0	68	(27.6)		87	(27.6)		0.0
Sleep apnea	40	(16.3)		55	(17.5)		3.2	49	(19.9)		63	(19.9)		0.0
Asthma	34	(13.8)		52	(16.5)		7.5	35	(14.1)		44	(14.1)		0.0
Pulmonary hypertension	30	(12.2)		55	(17.5)		14.9	32	(12.9)		41	(12.9)		0.0
Hypotension	29	(11.8)		81	(25.7)		36.3	37	(15.2)		48	(15.2)		0.0
Atrial fibrillation	17	(6.9)		68	(21.6)		42.9	27	(10.9)		34	(10.9)		0.0
Stroke/embolism	10	(4.1)		38	(12.1)		29.7	15	(6.2)		19	(6.2)		0.0
Shock	9	(3.7)		52	(16.5)		43.7	18	(7.3)		23	(7.3)		0.0
Risk factors, n (%)
VTE and bleeding	167	(67.9)		258	(81.9)		32.8	188	(76.5)		241	(76.5)		0.0
VTE only	200	(81.3)		288	(91.4)		29.8	216	(87.8)		277	(87.8)		0.0
Bleeding only	208	(84.6)		277	(87.9)		9.8	211	(85.8)		270	(85.8)		0.0

BMI, body mass index; eGFR, estimated glomerular filtration rate; Lab, laboratory; NSAID, nonsteroidal anti-inflammatory drug; OAC, oral anticoagulant; PE, pulmonary embolism; Quan-CCI, Quan–Charlson Comorbidity Index; RIETE, Registro Informatizado de Enfermedad TromboEmbólica; RVD, right ventricular dysfunction; SSRI, selective serotonin reuptake inhibitors; Std. diff., standardized difference; VTE, venous thromboembolism.

^aCohorts were weighted using propensity score overlap weighting, which leads to a Std. diff. = 0% for variables included in the propensity score model.

^bFor continuous variables, the Std. diff. is calculated by dividing the absolute difference in means of the control and the case by the pooled SD of both groups. The pooled SD is the square root of the average of the squared SDs. For dichotomous variables, the Std. diff. is calculated using the following equation, where P is the respective proportion of participants in each group: |(Pcase − Pcontrol)| / √([Pcase{1 − Pcase} + Pcontrol{1 − Pcontrol}]/2).

^cEvaluated on the initiation date.

^dOther terms include “increased RV to LV ratio” and “right ventricular hypokinesis.”

^eQuan et al. [14].

^fRuiz-Gimenez et al. [15].

^gMajor bleeding was identified using the Cunningham algorithm [11].

^hThe last observed baseline measurement was used. If multiple measurements were observed on the same day, the mean value was used.

ⁱEvaluated using the CKD-EPI 2021 equation [16].

3.2. Treatment patterns

The median on-treatment period was 270 days for rivaroxaban users and 235 days for warfarin users, with a median interval from first to last prescription of 333 and 234 days, respectively, and a median interval between consecutive prescriptions of 64 and 33 days, respectively. Rivaroxaban users had a median of 6 prescriptions during follow-up compared with 8 prescriptions for warfarin users. The median interval between initiation of the index agent and discharge from the index PE hospitalization was 1 day for the rivaroxaban cohort and 4 days for the warfarin cohort (Supplementary Table S7). While on treatment, 86.3% (272/315) of warfarin users had 1 or more INR measurements, with an average of 50.9% of measurements per patient being within therapeutic range (INR of 2-3; Supplementary Table S8).

3.3. Effectiveness

Patients treated with rivaroxaban had reduced risk of recurrent VTE up to 36 months of treatment compared with those receiving warfarin (Figure 2). Specifically, risk of VTE recurrence among rivaroxaban users was 45% lower at 12 months (10.8% vs 18.0%; HR [95% CI], 0.55 [0.32, 0.92]; P = .02), 39% lower at 24 months (16.4% vs 22.0%; HR [95% CI], 0.61 [0.38, 0.99]; P = .04), and 41% lower at 36 months of treatment (20.4% vs 30.3%; HR [95% CI], 0.59 [0.38, 0.92]; P = .02).

Figure 2.

Kaplan–Meier venous thromboembolism (VTE) recurrence rates. CI, confidence interval. ^aThe number of patients still observed at the specific point in time, with no evidence of VTE. ^bVTE recurrence was defined as having ≥1 primary diagnosis of VTE in an inpatient setting during the on-treatment period.

3.4. Safety

No significant difference in the risk of major bleeding was observed between patients treated with rivaroxaban and warfarin up to 12 months of treatment (8.0% vs 6.7%; HR [95% CI], 1.49 [0.72, 3.09]; P = .29; Figure 3). However, the proportional hazards assumption did not hold beyond 12 months of treatment, as indicated by visual assessment of the Kaplan–Meier event curves, and major bleeding was reported descriptively beyond this time point. In the rivaroxaban and warfarin cohorts, the Kaplan–Meier rates of major bleeding were 8.0% and 10.4% at 24 months of treatment, respectively, and 8.2% and 13.6% at 36 months of treatment, respectively.

Figure 3.

Kaplan–Meier major bleeding rates. CI, confidence interval. ^aThe number of patients still observed at the specific point in time, with no evidence of venous thromboembolism. ^bMajor bleeding was identified using the Cunningham algorithm [11].

4. Discussion

In this retrospective cohort study, patients with PE and evidence of RVD treated with rivaroxaban experienced reduced risk of VTE recurrence compared with those treated with warfarin, including a 45% lower risk at 12 months and a 41% lower risk at 36 months of treatment. Outside of a clinical trial setting, the observation that warfarin-treated patients achieved an INR within the therapeutic range just >50% of the time further underscores the well-documented challenges associated with its management. In addition, no statistically significant difference was observed in the risk of major bleeding between anticoagulation cohorts up to 36 months of treatment.

Previous observational and postmarketing surveillance studies have observed a lower risk of VTE recurrence and major bleeding among patients with deep vein thrombosis or PE treated with rivaroxaban vs those using vitamin K antagonists [[17], [18], [19], [20]]. However, despite RVD occurring in approximately one-third of patients with an acute PE episode and posing an increased risk of PE-related mortality [1,2,21], no study, to our knowledge, has evaluated the real-world effectiveness and safety of rivaroxaban vs warfarin among patients with both PE and evidence of RVD. Difficulty imaging the geometrically complex right ventricle, uncertainty around the function of the right ventricle with respect to cardiac hemodynamics, and the lack of an ICD diagnosis code to identify RVD in structured data may partly explain this research gap [22,23]. A recent meta-analysis evaluating the prevalence of long-term RVD after acute PE noted that additional research on the management and treatment of RVD is needed [21].

Our findings align with the American Society of Hematology, American College of Chest Physicians, and European Society of Cardiology guidelines, supporting the use of DOACs over vitamin K antagonists to treat patients with PE [4,6,7]. Specifically, the observed effectiveness results, combined with the lack of significant difference in major bleeding, suggest that rivaroxaban may be preferred over warfarin among patients with both PE and evidence of RVD. Given the current lack of specific guidelines for anticoagulation in this understudied population [4,6,7], these results could inform future research, including randomized controlled trials, to provide more targeted treatment recommendations.

Consistent with previous findings, only half of warfarin users’ INR measurements in this study were within therapeutic range during treatment, likely increasing their risk of VTE recurrence [24,25]. Additionally, the median time between treatment initiation and discharge from the index PE hospitalization was 3 days longer with warfarin than with rivaroxaban, potentially indicating challenges in reaching INR therapeutic range before discharge. Optimal efficacy and safety through good INR control, commonly defined as time in therapeutic range (TTR) >60%, can be challenging to achieve outside of clinical trial settings [26,27]. Factors such as dietary interactions, smoking, comorbid conditions, and polypharmacy can affect TTR, requiring regular INR monitoring [[26], [27], [28], [29]]. Despite the importance of INR monitoring, only approximately 10% of patients with PE in the MGB health system are enrolled in their anticoagulation management service, though some may have enrolled in other management programs. This highlights the broader challenge of warfarin management in real-world practice [27], even in well-resourced academic centers with dedicated management services. Socioeconomically disadvantaged communities with fewer resources may face even greater difficulties in achieving effective warfarin management [30].

Although both rivaroxaban and warfarin are administered once daily [29,31], rivaroxaban presents fewer complexities, such as the lack of need for routine testing, dose adjustments, and drug–drug or drug–food interactions [29,32]. These factors, along with greater treatment adherence observed in rivaroxaban users, may contribute to the decreased risk of VTE recurrence compared with warfarin users [[33], [34], [35]]. Nonadherence to warfarin has been linked to an increased risk of VTE recurrence, and a study showed that patients who switch from warfarin to DOACs (including rivaroxaban) report greater satisfaction due to the reduced need for monitoring and medical contacts [36]. The greater logistical requirements associated with warfarin, due to frequent monitoring, negatively impact adherence and TTR and may ultimately influence risk of VTE recurrence.

To our knowledge, this is the first study applying NLP to clinical notes in EHR data to identify patients with PE and evidence of RVD—a method that may be of value to future research. The data used in this analysis include granular information on prescriptions, laboratory tests, diagnoses, and procedures occurring during hospital stays, which allowed for the characterization of index PE hospitalizations. Furthermore, effectiveness and safety were assessed up to 36 months of treatment, allowing for the observation of outcomes that may occur from shorter and longer courses of anticoagulation therapy for VTE.

Some limitations should be taken into consideration when interpreting findings of this study. First, the analysis of EHR data depends on the correct entry of diagnosis, procedure, and drug codes, and coding inaccuracies may lead to misidentification; however, there is no indication that rivaroxaban and warfarin users were impacted differently. Second, patients included in the study were treated in a well-resourced health system in Massachusetts with a dedicated anticoagulation management service. These factors might affect the generalizability of our findings to other regions and environments with lower access to anticoagulation management services, potentially resulting in poorer outcomes among warfarin users due to dose management issues. Third, despite the use of PS-OW to balance patients’ characteristics across cohorts, there is a potential for residual confounding or asymmetric informative censoring due to unobserved heterogeneity. Fourth, given limited information on days of supply in the prescription data, the last prescription of the index OAC, augmented by the median time between consecutive prescriptions, was used as a proxy for discontinuation. Prescriptions may also imperfectly correlate with actual medication use; despite all patients having a prescription for the index agent during the index PE hospitalization and as the first OAC after discharge, some patients may not have taken their medication as prescribed.

5. Conclusions

In conclusion, this real-world study found that rivaroxaban was associated with a significantly decreased risk of recurrent VTE without an increased risk of major bleeding vs warfarin among patients with PE and RVD. These findings contribute to the limited body of evidence on patients with PE and RVD and may help to inform future interventional studies investigating treatments for this understudied population.