Prediction Score for Major Bleeding in Patients With Venous Thromboembolism Receiving Direct Oral Anticoagulants　― Insights From the COMMAND VTE Registry-2 ―

Abstract

Background: Predicting the bleeding risk during anticoagulation therapy is a key clinical challenge in patients with venous thromboembolism (VTE). However, there is no established prediction score for major bleeding (MB) in patients with VTE treated with direct oral anticoagulants (DOACs).

Methods and Results: Using the COMMAND VTE Registry-2, which enrolled 5,197 patients with acute symptomatic VTE between 2015 and 2020 among 31 centers in Japan, we investigated the risk factors for MB beyond 7 days and within 180 days in patients who received DOACs. A prediction score was developed in the derivation cohort (n=1,618), and prediction performance was evaluated in the validation cohort (n=809). Multivariate logistic regression analysis in the derivation cohort identified factors associated with MB. Based on β coefficients for each factor, the prediction score assigned 2 points to active cancer, history of MB, and thrombocytopenia, and 1 point to creatinine >1.2 mg/dL and anemia, summing them. The C statistic of the prediction score was 0.74 (95% confidence interval [CI] 0.68–0.80) in the derivation cohort and 0.74 (95% CI 0.67–0.81) in the validation cohort (P=0.98). When a cut-off value of 3 was used for the risk score, the sensitivity and specificity were 56.1% and 79.2%, respectively.

Conclusions: The prediction score developed for MB during DOAC therapy (COMMAND-BLEED score) could be clinically useful for decision-making regarding anticoagulation strategies with DOACs.

Conventionally, venous thromboembolism (VTE) has been treated with initial parenteral anticoagulants (unfractionated or low-molecular-weight heparin) followed by oral anticoagulants (vitamin K antagonists [VKAs]). After the introduction of direct oral anticoagulants (DOACs) for the treatment of VTE, they have become widely used because of their clinical benefits, such as no need for frequent dose adjustment or blood tests and fewer drug and food interactions,¹ resulting in a paradigm shift in the management of VTE.²

Bleeding events were the main drawback of antithrombotic therapies, with major bleeding (MB) being a predictor of short- and long-term mortality.³ A meta-analysis showed that patients receiving anticoagulants for VTE treatment have a 2% risk of MB in the first 3 months.⁴ Compared with VKAs, DOACs demonstrated a better safety profile in randomized clinical trials, with an absolute MB risk of 1.1% during the first 3–12 months of treatment.^5,6 DOACs could offer many benefits over VKAs; however, they still carry a considerable risk of MB.

Estimating the bleeding risk before starting anticoagulation therapy in individual patients with VTE helps clinicians determine the appropriate intensity and duration of anticoagulation therapy.⁷ Thus far, several clinical prediction scores for bleeding have been proposed to identify high-risk patients, mostly based on cohorts receiving VKAs.^8–10 In contrast, the VTE-BLEED score, derived from the RE-COVER and RE-COVER II studies using dabigatran, seems to be the only well-validated score during DOAC therapy.^11,12 However, no prediction score has been established for bleeding in patients with VTE treated with DOACs, including Factor Xa inhibitors. In addition, bleeding risk scores for patients with atrial fibrillation (AF) generally have limited accuracy in the setting of VTE.^8,13–15 Consequently, further development of a prediction score for bleeding during DOAC therapy including Factor Xa inhibitors is needed for patients with VTE in the current era. Therefore, we sought to develop and validate a new prediction score using a large-scale Japanese observational database of VTE in the DOAC era.

Methods

Study Population

The Contemporary Management and Outcomes in Patients with Venous Thromboembolism (COMMAND VTE) Registry-2 is a physician-initiated multicenter retrospective cohort study that enrolled consecutive patients with acute symptomatic VTE objectively confirmed by imaging or autopsy at 31 centers in Japan between January 2015 and August 2020 after the introduction of DOACs for VTE in Japan. The design of the registry has been reported in detail elsewhere.¹⁶ Briefly, the registry enrolled consecutive patients who met the definitions of acute symptomatic VTE diagnosed within 31 days from symptom onset during the study period.¹⁷

The research protocol was approved by the relevant review boards or ethics committees at all 31participating centers listed in Supplementary Appendix 1. The study was conducted in accordance with the principles of the Declaration of Helsinki. The requirement for written informed consent was waived because we only used clinical information obtained from routine practice in accordance with the Ministry of Health, Labour, and Welfare guidelines for epidemiological studies in Japan.

Of the 51,313 patients screened for suspected VTE, 5,197 with acute symptomatic VTE were enrolled in the registry. The population of the present study consisted of 2,427 patients who received DOACs (Factor Xa inhibitors) for more than 180 days, after excluding 1,075 patients who did not receive DOACs (Factor Xa inhibitors; edoxaban, rivaroxaban, and apixaban), 203 patients with missing values for body weight, creatine, or hemoglobin, 1,418 patients who received DOACs for less than 180 days without MB, and 74 patients who developed MB within 7 days after VTE diagnosis or without DOAC therapy (Figure 1). Because MB events in the acute phase could be significantly influenced by acute treatment strategies other than DOAC therapy, we excluded patients who developed MB within 7 days after VTE diagnosis. To develop and validate the prediction score, the study population was randomly divided 2:1 into 2 cohorts: a derivation cohort (n=1,618) and a validation cohort (n=809). A prediction score was developed in the derivation cohort, and its predictive performance was investigated in the validation cohort.

Figure 1.

Study flow chart. Venous thromboembolism (VTE) includes pulmonary embolism and/or deep vein thrombosis. DOAC, direct oral anticoagulant; BW, body weight; Cr, creatinine; Hb, hemoglobin.

Data Collection and Definitions of Patient Characteristics

Data on patient characteristics were collected from hospital charts or databases using prespecified definitions via an electronic case report form in a web-based system. Expert physicians for VTE at each institution, trained in data collection before enrollment, were responsible for data entry. The system automatically checked for missing, contradictory, or out-of-range values. Additional manual data audits were performed at the general office of the registry, and any data discrepancies were resolved by reconfirming with the original hospital charts or databases at each institution.

Patients with active cancer were defined as those receiving treatment for cancer, such as chemotherapy or radiotherapy, those scheduled to undergo cancer surgery, those with metastasis to other organs, and/or those with terminal cancer (expected life expectancy ≤6 months) at the time of diagnosis.¹⁸ A history of MB was diagnosed if the patient had a history of International Society of Thrombosis and Hemostasis (ISTH) MB, which consisted of fatal bleeding, symptomatic bleeding in a critical area or organ, and bleeding causing a reduction in the hemoglobin level by at least 2 g/dL or leading to the transfusion of at least 2 units of whole blood or red cells.¹⁹ Anemia was defined as a hemoglobin level <13 g/dL in men and <12 g/dL in women according to the World Health Organization classification of anemia.²⁰ Thrombocytopenia was defined as a platelet count <10×10⁴/μL.²¹ Detailed definitions of other patient characteristics are provided in Supplementary Appendix 2.

Clinical Follow-up and Outcomes

Follow-up information was primarily collected via hospital chart reviews and additional data were obtained by contacting patients, relatives, and/or referring physicians via telephone and/or mail. This follow-up included questions regarding vital status, clinical events, invasive procedures, and anticoagulation therapy status.

Current guidelines recommend anticoagulation for 3–6 months after the onset of VTE, with extended anticoagulation of unknown optimal duration recommended for patients with unprovoked VTE, VTE induced by persistent risk factors, or active cancer.^22–24 Even in patients with provoked VTE, the median duration of anticoagulant therapy with DOAC is 180 days in real-world practice.²⁵ The COMMAND VTE Registry-2 is a retrospective cohort study in which the majority of patients have unprovoked VTE and active cancer.²⁶ Given that the VTE-BLEED score was designed to predict bleeding events within 6 months,¹¹ the clinical endpoint in the present study was MB during anticoagulation therapy within 180 days after VTE onset.

The independent clinical event committee (Supplementary Appendix 3), unaware of patient characteristics, reviewed all detailed clinical courses and adjudicated clinical events. In the case of inconsistencies, final adjudication of clinical events was based on a full consensus of the independent clinical event committee.

Statistical Analysis

Categorical variables are presented as numbers and percentages and continuous variables are presented as the median with interquartile range. Categorical variables were compared using the Chi-squared test or Fisher’s exact test, as appropriate, and continuous variables were compared using Wilcoxon’s rank sum test. Kaplan-Meier analysis was used to estimate the cumulative incidence of events. To develop the prediction score, we constructed a multivariable logistic regression model to estimate the odds ratio (OR) and 95% confidence interval (CI) of potential factors associated with MB. Continuous variables were dichotomized using clinically relevant cut-off values.

First, based on clinical relevance and previous reports,^{10,14,15,27–30} we selected a comprehensive list of 26 potential variables, including age >75 years, sex, body weight <50 kg, a history of MB, a history of VTE, a history of stroke, thrombus location at diagnosis, comorbidities (hypertension, diabetes, dyslipidemia, liver cirrhosis, heart disease, chronic lung disease, autoimmune diseases, and active cancer), creatinine >1.2 mg/dL, anemia, thrombocytopenia, and concomitant medication use (antiplatelet agents, non-steroidal anti-inflammatory drugs [NSAIDs], corticosteroids, and histamine H₂ receptor blockers or proton pump inhibitors). Second, candidate variables that were associated with MB in the univariate logistic regression models with significance level of P<0.10 were included in the multivariable logistic regression model. We then conducted a backward stepwise model selection procedure for potential candidates, using a significance level of 0.05, to eliminate variables with higher P values. We finally constructed a multivariable logistic regression model using those variables with P<0.05.³¹ A prediction score was constructed based on β coefficients. The β coefficient for each variable was divided by the smallest β coefficient and rounded to the nearest integer, which was used as the point score for the variable. The risk score for each patient was determined by summing the scores for each variable.

We assessed the discriminatory performances of the models using receiver operator characteristic (ROC) curve analysis in the derivation and validation cohorts. The C statistic was calculated and compared between the 2 cohorts. The Youden Index was used to determine the optimal cut-off point for MB. We also compared the C statistic between our bleeding risk assessment score (COMMAND-BLEED score) and previously reported bleeding risk assessment scores for VTE patients (i.e., VTE-BLEED score¹¹ and RIETE score¹⁰) using Delong’s test. Calibration of the risk score was assessed by a calibration plot showing the relationship between the observed and predicted incidence of MB.

All statistical analyses were conducted using IBM SPSS Statistics version 28.0 (SPSS, Chicago, IL, USA) or EZR (Jichi Medical University Saitama Medical Center, Saitama, Japan), a modified version of R commander for biostatistics (R Foundation for Statistical Computing, Vienna, Austria). All P values reported are 2-tailed, and P<0.05 was considered statistically significant.

Results

Patient Characteristics

The clinical characteristics of the derivation cohort compared between patients with and without MB are presented in Table 1. Patients with MB were more likely to have a history of MB, active cancer, lower hemoglobin levels, and higher alkaline phosphatase and D-dimer levels, and were more likely to be using NSAIDs. The clinical characteristics of the validation cohort are presented in Supplementary Table 1, with comparisons between the derivation and validation cohorts presented in Supplementary Table 2. There were no significant differences in clinical characteristics between the derivation and validation cohorts. The detailed types and doses of DOACs are presented in Supplementary Table 3.

Table 1.

Clinical Characteristics for All Patients in the Derivation Cohort and Those With and Without Major Bleeding Separately

	All patients (n=1,618)	No major bleeding (n=1,544)	Major bleeding (n=74)	P value
Baseline characteristics
Age (years)	71 [60–79]	71 [60–79]	70.5 [55–76]	0.12
Age >75 years	575 (35.5)	554 (35.9)	21 (28.7)	0.19
Sex (women)	944 (58.3)	900 (58.3)	44 (59.5)	0.84
Body weight (kg)	59.0 [50.0–68.0]	59.0 [50.0–68.0]	58.8 [47.2–65.7]	0.36
Body weight <50 kg	587 (24.2)	556 (24.0)	31 (27.0)	0.48
Comorbidities
History of VTE	120 (7.4)	117 (7.6)	3 (4.1)	0.26
History of major bleeding	94 (5.8)	81 (5.2)	13 (17.6)	<0.001
History of stroke	124 (7.7)	115 (7.4)	9 (12.2)	0.14
Hypertension	732 (45.2)	695 (45.0)	37 (50.0)	0.40
Diabetes	240 (14.8)	230 (14.9)	10 (13.5)	0.74
Dyslipidemia	460 (28.4)	439 (28.4)	21 (28.4)	0.99
Chronic heart disease	153 (9.5)	143 (9.3)	10 (13.5)	0.22
History of MI	43 (2.7)	39 (2.5)	4 (5.4)	0.13
Heart failure	61 (3.8)	60 (3.9)	1 (1.4)	0.52
Atrial fibrillation	71 (4.4)	66 (4.3)	5 (6.8)	0.37
Chronic lung disease	180 (11.1)	174 (11.3)	6 (8.1)	0.40
Liver cirrhosis	10 (0.6)	8 (0.5)	2 (2.7)	0.07
Autoimmune diseases	202 (12.5)	195 (12.6)	7 (9.5)	0.42
Hereditary thrombophilia	59 (3.6)	57 (3.7)	2 (2.7)	1.00
Active cancer	430 (26.6)	385 (24.9)	45 (60.8)	<0.001
Laboratory tests at diagnosis
WBC count (/μL)	7,300 [5,700–9,500]	7,300 [5,700–9,460]	7,405 [5,195–9,615]	0.74
Hemoglobin (g/dL)	12.4 [10.8–13.9]	12.5 [10.9–13.9]	10.9 [8.6–12.8]	<0.001
Anemia	756 (46.7)	703 (45.5)	53 (71.6)	<0.001
Platelet count (×104/μL)	20.2 [15.9–26.2]	20.2 [16.0–26.1]	20.0 [13.9–27.8]	0.84
Thrombocytopenia	69 (4.3)	61 (4.0)	8 (10.8)	0.01
Creatinine (mg/dL)	0.78 [0.62–0.95]	0.78 [0.62–0.95]	0.77 [0.61–1.04]	0.59
Creatinine >1.2 mg/dL	128 (7.9)	117 (7.6)	11 (14.9)	0.02
Creatinine clearance (mL/min)	67.5 [49.2–93.8]	67.5 [49.3–93.9]	64.4 [45.3–88.9]	0.45
AST (IU/L)	23 [17–32]	22 [17–32]	24 [19–33]	0.07
ALT (IU/L)	18 [12–28]	18 [12–28]	20 [13–28]	0.24
D-dimer (μg/mL)	9.1 [4.5–17.3]	8.9 [4.4–16.9]	13.8 [5.9–24.0]	<0.001
Site of VTE
PE with or without DVT	977 (60.4)	937 (60.7)	40 (54.1)	0.25
DVT only	641 (39.6)	607 (39.3)	34 (45.9)
Treatment in the acute phase other than OACs
Initial parenteral anticoagulants	810 (50.1)	769 (49.8)	41 (55.4)	0.35
Thrombolysis	126 (7.8)	123 (8.0)	3 (4.1)	0.22
Inferior vena cava filter	167 (10.3)	155 (10.0)	12 (16.2)	0.09
Ventilator support	24 (1.5)	20 (1.3)	4 (5.4)	0.02
Percutaneous cardiopulmonary support	10 (0.6)	8 (0.5)	2 (2.7)	0.07
Types of DOAC
Edoxaban	742 (45.9)	693 (44.9)	49 (66.2)	0.001
Rivaroxaban	500 (30.9)	489 (31.7)	11 (14.9)
Apixaban	376 (23.2)	362 (23.4)	14 (18.9)
Concomitant medications
NSAIDs	119 (7.4)	109 (7.1)	10 (13.5)	0.04
Corticosteroids	210 (13.0)	196 (12.7)	14 (18.9)	0.12
PPI or H2 receptor blockers	853 (52.7)	810 (52.5)	43 (58.1)	0.34
Statins	327 (20.2)	314 (20.3)	13 (17.6)	0.56
Antiplatelet agents	132 (8.2)	125 (8.1)	7 (9.5)	0.68

Unless indicated otherwise, data are given as the median [interquartile range] or n (%). Chronic heart disease was defined as persistent heart disorders, including heart failure, history of myocardial infarction (MI), and atrial fibrillation. Chronic lung disease was defined as persistent lung disorders, such as asthma, chronic obstructive pulmonary disease, and restrictive lung diseases. Autoimmune disorder was defined as immune-mediated diseases, such as inflammatory bowel disease, rheumatoid arthritis, and antiphospholipid syndrome. History of major bleeding was diagnosed if the patient had a history of International Society of Thrombosis and Hemostasis major bleeding. Anemia was defined as hemoglobin <13 g/dL in men and <12 g/dL in women. Thrombocytopenia was defined as a platelet count <10×10⁴/μL. Hereditary thrombophilia included protein C deficiency, protein S deficiency, and antithrombin III deficiency. ALT, alanine aminotransferase; AST, aspartate aminotransferase; DOAC, direct oral anticoagulant; DVT, deep vein thrombosis; NSAIDs, non-steroidal anti-inflammatory drugs; OACs, oral anticoagulants; PE, pulmonary embolism; PPI, proton pump inhibitor; VTE, venous thromboembolism; WBC, white blood cell.

MB During DOAC Therapy in the Derivation Cohort

During DOAC therapy, 74 (4.6%) patients developed MB events. The cumulative incidence of MB was 3.1% at 90 days and 4.6% at 180 days (Supplementary Figure 1). The incidence of MB was especially remarkable within 30 days compared with beyond 30 days. The most common site of MB was the lower gastrointestinal tract, followed by the upper gastrointestinal tract, intracranial, and genitourinary tract (Supplementary Table 4).

Logistic Regression Analysis and Score Derivation

Univariate logistic regression analysis of the risk factors for MB during DOAC therapy revealed that liver cirrhosis, active cancer, history of MB, anemia, thrombocytopenia, creatinine >1.2 mg/dL, and the concomitant use of NSAIDs were candidate variables for the multivariable logistic regression model (Table 2). Multivariate logistic regression analysis identified active cancer, a history of MB, thrombocytopenia, creatinine >1.2 mg/dL, and anemia as independent risk factors (Table 3). Based on the β coefficients for each variable, 2 points each were assigned to active cancer, a history of MB, and thrombocytopenia, and 1 point each was assigned to creatinine ≥1.2 mg/dL and anemia; the sum of these scores was used as the prediction score.

Table 2.

Univariate Logistic Regression Analysis of Risk Factors for Major Bleeding in the Derivation Cohort (n=1,618)

	OR (95% CI)	P value
Age >75 years	0.71 (0.42–1.19)	0.19
Female sex	1.05 (0.65–1.69)	0.84
Body weight <50 kg	1.22 (0.73–2.05)	0.45
Hypertension	1.22 (0.77–1.95)	0.40
Diabetes	0.89 (0.45–1.76)	0.74
Dyslipidemia	1.00 (0.60–1.67)	0.99
Liver cirrhosis	5.33 (1.11–25.57)	0.04
Congenital thrombogenicity	0.73 (0.17–3.03)	0.66
Heart diseases	1.53 (0.77–3.05)	0.23
History of MI	2.21 (0.77–6.34)	0.14
Heart failure	0.34 (0.05–2.48)	0.29
Atrial fibrillation	1.62 (0.63–4.16)	0.31
Chronic lung diseases	0.70 (0.30–1.63)	0.40
Autoimmune diseases	0.72 (0.33–1.60)	0.42
Active cancer	4.67 (2.89–7.56)	<0.001
History of VTE	0.52 (0.16–1.66)	0.27
History of major bleeding	3.85 (2.03–7.30)	<0.001
History of stroke	1.72 (0.84–3.54)	0.14
Creatinine >1.2 mg/dL	2.13 (1.09–4.15)	0.03
Anemia	3.02 (1.80–5.05)	<0.001
Thrombocytopenia	2.95 (1.36–6.41)	0.006
DVT only	1.31 (0.82–2.10)	0.26
Concomitant use of antiplatelet agents	1.19 (0.53–2.64)	0.68
Concomitant use of NSAIDs	2.06 (1.03–4.20)	0.04
Concomitant use of corticosteroids	1.61 (0.88–2.93)	0.12
Concomitant use of H2 receptor blocker or PPI	1.26 (0.78–2.02)	0.34

Anemia was defined as hemoglobin <13 g/dL in men and <12 g/dL in women. Thrombocytopenia was defined as a platelet count <10×10⁴/μL. CI, confidence interval; OR, odds ratio. Other abbreviations as in Table 1.

Table 3.

Multivariate Logistic Regression Analysis of Risk Factors for Major Bleeding in the Derivation Cohort

Variables	Multivariate analysis				COMMAND-BLEED score (points)
	β	SE	OR (95% CI)	P value
Active cancer	1.39	0.26	4.02 (2.40–6.74)	<0.001	2
History of major bleeding	1.41	0.35	4.08 (2.07–8.04)	<0.001	2
Thrombocytopenia	0.94	0.41	2.57 (1.15–5.77)	0.02	2
Creatinine >1.2 mg/dL	0.78	0.36	2.18 (1.08–4.39)	0.03	1
Anemia	0.58	0.28	1.78 (1.03–3.10)	0.04	1

A history of major bleeding was diagnosed if the patient had a history of International Society of Thrombosis and Hemostasis major bleeding. Anemia was defined as hemoglobin <13 g/dL in men and <12 g/dL in women. Thrombocytopenia was defined as platelet count <10×10⁴/μL. CI, confidence interval; OR, odds ratio; SE, standard error.

The incidence rates of MB for each score in the derivation cohort are shown in Figure 2A. The C statistic based on the ROC curve in the derivation group was 0.74 (95% CI 0.68–0.80; Figure 2B). The Youden Index provided the highest sum of sensitivity plus specificity at total score of 3 points, with the incidence rate of MB among those with total score of ≥3 being >10%.

Figure 2.

Predictive performance of the developed prediction score for major bleeding (MB) in the derivation cohort. (A) Incidence of MB for each COMMAND-BLEED score. (B) Receiver operating characteristic curve. CI, confidence interval.

Internal Validation of the Newly Developed Prediction Score

The C statistic for predicting MB using the newly developed prediction score in the validation cohort was 0.74 (95% CI 0.67–0.81; Figure 3), which was not significantly different from that in the derivation cohort (P=0.98; Supplementary Figure 2). Patients with a score of ≥3 points experienced more MB events than those with a score of <3 points (Supplementary Figure 3A), with an OR of 4.86 (95% CI 2.56–9.22; P <0.001). Based on a cut-off value of 3 points, the sensitivity of our COMMAND-BLEED score was 56.1%, specificity was 79.2%, positive predictive value (PPV) was 12.6%, and negative predictive value (NPV) was 97.1%. Figure 3 also shows the sensitivity, specificity, PPV, NPV, positive likelihood ratio, and negative likelihood ratio at alternative cut-off values (2 and 4 points). In contrast, the calibration performance of the COMMAND-BLEED score was not satisfactory (Supplementary Figure 3B).

Figure 3.

Predictive performance of the developed prediction score for major bleeding in the validation group using receiver operator characteristic (ROC) curve analysis. The table below the ROC curve shows the predictive performance of COMMAND-BLEED scores of ≥2, ≥3, and ≥4 points for major bleeding. LHR, likelihood ratio; PV, predictive value.

Discussion

The present study from a large real-world VTE registry in the DOAC era successfully developed a new prediction score for MB in patients with VTE during DOAC (Factor Xa inhibitors) therapy. The newly developed prediction score (COMMAND-BLEED score) consisted of 2 points each for active cancer, a history of MB, and thrombocytopenia, and 1 point each for creatinine >1.2 mg/dL and anemia. The score seemed to be internally well validated, and the C statistic was 0.74 with a specificity of 79.2% and an NPV of 97.1% at a high-risk cut-off of ≥3 points.

Several clinical prediction scores for bleeding in patients with VTE have been reported. The RIETE score was generated for MB within 3 months after VTE treatment, primarily with VKA.¹⁰ The VTE-BLEED score was generated for all bleeding events between 30 days and 6 months after VTE treatment based on data from randomized trials with dabigatran.¹¹ Currently, there is no established prediction score for bleeding in patients with VTE treated with DOACs, including Factor Xa inhibitors, which have been common oral anticoagulants in daily clinical practice. Lecumberri et al.³² compared the predictive performance of the VTE-BLEED and RIETE scores for bleeding risk at different time points (first 30 days after VTE diagnosis, and Days 31–90, 91–180, and 181–360 after VTE diagnosis) and revealed that the RIETE score showed marginal superiority over the VTE-BLEED score in assessing risk in the first month of treatment; however, the difference in the area under the curve of 0.02 was not clinically relevant. That study also suggested that the usefulness of the bleeding risk score may vary depending on patient population characteristics and the time frame evaluated.³² External validation of VTE-BLEED and RIETE scores suggested modest predictive performance, with C statistics ranging from 0.66 to 0.71^12,33,34 for the VTE-BLEED score and from 0.51 to 0.60^8,35,36 for the RIETE score, particularly in the DOAC-treated population, in which the C statists were 0.67 and 0.60, respectively.³³ Validation of the VTE-BLEED score in patients receiving edoxaban in the Hokusai-VTE trial showed that the C statistic for the predictive performance of the score for MB was 0.63 in the edoxaban group and 0.69 in the warfarin group.¹² In the observational XALIA study comparing rivaroxaban with conventional therapy, external validation showed that a 2-point threshold on the VTE-BLEED score predicted MB in the rivaroxaban group with a C statistic of 0.69 and an NPV of 99.7%.³⁷ Thus, the VTE-BLEED score could be useful for identifying patients being treated with a Factor Xa inhibitor who are at low risk of bleeding, although the predictive performance of the VTE-BLEED score seems to be modest, partly because the score was derived from the data of patients treated with dabigatran, which is not a Factor Xa inhibitor.³⁸

Considering the potentially different impacts of bleeding risk factors depending on the anticoagulant being used,³⁹ a score to predict bleeding based on data with Factor Xa inhibitors specifically has been needed. In the comparison between the COMMAND-BLEED, VTE-BLEED, and RIETE scores using ROC, the C statistic was the highest for the COMMAND-BLEED score in our validation cohort (Supplementary Table 5; Supplementary Figure 4). Statistically significant differences were observed between the COMMAND-BLEED and VTE-BLEED scores, but not between the COMMAND-BLEED and RIETE scores. The COMMAND-BLEED score could potentially be useful as a risk assessment tool for MB events in patients with VTE treated with DOACs (Supplementary Table 5; Supplementary Figure 4).

Previous studies have reported several common factors associated with bleeding, such as advanced age, anemia, a history of bleeding, renal dysfunction, and active cancer. Notably, advanced age was not included in the COMMAND-BLEED score, whereas thrombocytopenia was.

In patients with AF, age correlates more with hemorrhagic events than ischemic events, with apixaban and edoxaban showing greater benefits over warfarin in older rather than younger patients.⁴⁰ A meta-analysis conducted in AF patients aged >75 years showed that apixaban reduced systemic embolism, MB, and intracerebral hemorrhage compared with warfarin.⁴¹ In patients with VTE, data from the RIETE registry suggest that the use of DOACs may be safer than standard therapy in VTE fragile patients (defined as age ≥75 years and/or estimated glomerular filtration rate ≤50 mL/min/1.73 m² and/or body weight ≤50 kg).⁴² In the treatment of VTE, DOACs have similar efficacy and improved safety compared with VKAs in both young and older patients.^43,44 The COMMAND-BLEED score, based on DOAC-treated patients with VTE, suggests that older age is not a risk factor for MB, unlike previous scores based on VKA-treated patients. In addition, the effect of age on bleeding has differed between studies of DOACs-treated patients, with a few studies suggesting a preference for apixaban and edoxaban for bleeding in older adult populations. In fact, an increased risk of gastrointestinal bleeding with dabigatran and rivaroxaban, but not with apixaban or edoxaban, has been reported in patients with AF aged >75 years.⁴⁵ The difference in the safety profile of each DOAC may have affected the relationship between age and bleeding.

Data from the RIETE registry showed varying rates of MB based on platelet count during a 3-month follow-up period, and multivariable analysis identified thrombocytopenia as a risk factor for MB.⁷ A previous study in the DOAC era also reported that the cumulative 5-year incidence of MB was significantly higher in patients with moderate/severe thrombocytopenia (platelet count <100×10⁹/L) than in the other groups (platelet count >150×10⁹/L and 100–150×10⁹/L).⁴⁶ In fact, thrombocytopenia is included as a risk factor for bleeding in some risk assessment tools.^27,47 Thrombocytopenia could be a notable risk factor for bleeding in patients with VTE during DOAC therapy.

Study Limitations

The present study has several limitations. First, being an observational study, it was subject to various biases regarding therapeutic decisions; initial parenteral anticoagulant use and DOAC selection and dosages were at the discretion of the attending physicians. Second, the risk factors for MB events among DOACs were not compared because the number of MB events for each DOAC was small when divided into the 3 DOACs groups. Third, we evaluated the risk factors at baseline; however, bleeding risk factors, such as renal/liver function and cancer status, may change during the course of therapy. Fourth, racial and ethnic differences in drug efficacy and adverse reactions could influence clinical outcomes.⁴⁸ Finally, there were relatively few patients with a COMMAND-BLEED score >4 points, and the risk score did not have a satisfactory calibration performance, although the discrimination performance based on the ROC curve in the validation cohort was not different from that of the derivation cohort. Therefore, external validation will be required before the wider clinical use of this prediction score.

Conclusions

The newly developed prediction score for MB during DOAC therapy (COMMAND-BLEED score) could be clinically useful for decision-making regarding anticoagulation strategies with DOACs, although further external validation is needed.

Acknowledgments

The authors thank the coinvestigators who participated in the COMMAND VTE Registry-2 for their support and collaboration.

Sources of Funding

The COMMAND VTE Registry-2 is supported, in part, by JSPS KAKENHI (Grant no. JP 21K16022).

Disclosures

S.I. has received lecture fees from Bayer Healthcare, Bristol-Myers Squibb, and Daiichi-Sankyo. Y.Y. has received lecture fees from Bayer Healthcare, Bristol-Myers Squibb, Pfizer, and Daiichi-Sankyo, and grant support from Bayer Healthcare and Daiichi-Sankyo. T.M. has received lecture fees from Bristol-Myers Squibb, Daiichi Sankyo, Japan Lifeline, Kowa, Kyocera, Novartis, and Toray; manuscript fees from Bristol-Myers Squibb and Kowa; and has served on the advisory board for Sanofi. K. Maemura has received lecture fees from Bayer Healthcare, Bristol-Myers Squibb, and Daiichi-Sankyo and is a member of Circulation Journal’s Editorial Team. Y.N. has received lecture fees from Bayer Healthcare, Bristol-Myers Squibb, Pfizer, and Daiichi-Sankyo. N.I. has received lecture fees from Bayer Healthcare, Bristol-Myers Squibb, and Daiichi-Sankyo. Y.O. has received lecture fees from Bayer Healthcare, Bristol-Myers Squibb, Pfizer, and Daiichi-Sankyo, and research funds from Bayer Healthcare and Daiichi-Sankyo. N.K. has received lecture fees from Bayer Healthcare and grant support from Pfizer. K. Kaneda has received lecture fees from Bristol-Myers Squibb, Pfizer, and Daiichi-Sankyo. K. Ono is a member of Circulation Journal’s Editorial Team. All other authors report no relationships relevant to the contents of this paper.

IRB Information

The research protocol was approved by the relevant review boards or ethics committees at all 31 participating centers listed in Supplementary Appendix 1.

Supplementary Files

Please find supplementary file(s);

https://doi.org/10.1253/circj.CJ-25-0186

References

• 1.   van Es N, Coppens M, Schulman S, Middeldorp S, Büller HR. Direct oral anticoagulants compared with vitamin K antagonists for acute venous thromboembolism: Evidence from phase 3 trials. Blood 2014; 124: 1968–1975, doi:10.1182/blood-2014-04-571232.
• 2.   Lui B, Wee B, Lai J, Khattak Z, Kwok A, Donarelli C, et al. A ten-year review of the impact of the transition from warfarin to direct oral anticoagulant: Has venous thromboembolism treatment become safer? Thromb Res 2022; 219: 112–120, doi:10.1016/j.thromres.2022.09.006.
• 3.   Lecumberri R, Alfonso A, Jiménez D, Fernández Capitán C, Prandoni P, Wells PS, et al. Dynamics of case-fatalilty rates of recurrent thromboembolism and major bleeding in patients treated for venous thromboembolism. Thromb Haemost 2013; 110: 834–843, doi:10.1160/th13-02-0132.
• 4.   Linkins LA, Choi PT, Douketis JD. Clinical impact of bleeding in patients taking oral anticoagulant therapy for venous thromboembolism: A meta-analysis. Ann Intern Med 2003; 139: 893–900, doi:10.7326/0003-4819-139-11-200312020-00007.
• 5.   van der Hulle T, Kooiman J, den Exter PL, Dekkers OM, Klok FA, Huisman MV. Effectiveness and safety of novel oral anticoagulants as compared with vitamin K antagonists in the treatment of acute symptomatic venous thromboembolism: A systematic review and meta-analysis. J Thromb Haemost 2014; 12: 320–328, doi:10.1111/jth.12485.
• 6.   Cohen AT, Hamilton M, Mitchell SA, Phatak H, Liu X, Bird A, et al. Comparison of the novel oral anticoagulants apixaban, dabigatran, edoxaban, and rivaroxaban in the initial and long-term treatment and prevention of venous thromboembolism: Systematic review and network meta-analysis. PLoS One 2015; 10: e0144856, doi:10.1371/journal.pone.0144856.
• 7.   Di Micco P, Monreal M. Platelet count and bleeding in patients receiving anticoagulant therapy for venous thromboembolism: Lesson from the RIETE Registry. J Blood Med 2019; 10: 453–456, doi:10.2147/jbm.s234053.
• 8.   Riva N, Bellesini M, Di Minno MN, Mumoli N, Pomero F, Franchini M, et al. Poor predictive value of contemporary bleeding risk scores during long-term treatment of venous thromboembolism: A multicentre retrospective cohort study. Thromb Haemost 2014; 112: 511–521, doi:10.1160/th14-01-0081.
• 9.   Piovella C, Dalla Valle F, Trujillo-Santos J, Pesavento R, López L, Font L, et al. Comparison of four scores to predict major bleeding in patients receiving anticoagulation for venous thromboembolism: Findings from the RIETE registry. Intern Emerg Med 2014; 9: 847–852, doi:10.1007/s11739-014-1073-8.
• 10.   Ruíz-Giménez N, Suárez C, González R, Nieto JA, Todolí JA, Samperiz AL, et al. Predictive variables for major bleeding events in patients presenting with documented acute venous thromboembolism: Findings from the RIETE Registry. Thromb Haemost 2008; 100: 26–31, doi:10.1160/th08-03-0193.
• 11.   Klok FA, Hosel V, Clemens A, Yollo WD, Tilke C, Schulman S, et al. Prediction of bleeding events in patients with venous thromboembolism on stable anticoagulation treatment. Eur Respir J 2016; 48: 1369–1376, doi:10.1183/13993003.00280-2016.
• 12.   Klok FA, Barco S, Konstantinides SV. External validation of the VTE-BLEED score for predicting major bleeding in stable anticoagulated patients with venous thromboembolism. Thromb Haemost 2017; 117: 1164–1170, doi:10.1160/th16-10-0810.
• 13.   Klok FA, Niemann C, Dellas C, Hasenfuß G, Konstantinides S, Lankeit M. Performance of five different bleeding-prediction scores in patients with acute pulmonary embolism. J Thromb Thrombolysis 2016; 41: 312–320, doi:10.1007/s11239-015-1239-x.
• 14.   Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ, Lip GY. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: The Euro Heart Survey. Chest 2010; 138: 1093–1100, doi:10.1378/chest.10-0134.
• 15.   Gage BF, Yan Y, Milligan PE, Waterman AD, Culverhouse R, Rich MW, et al. Clinical classification schemes for predicting hemorrhage: Results from the National Registry of Atrial Fibrillation (nRAF). Am Heart J 2006; 151: 713–719, doi:10.1016/j.ahj.2005.04.017.
• 16.   Kaneda K, Yamashita Y, Morimoto T, Chatani R, Nishimoto Y, Ikeda N, et al. Anticoagulation strategies and long-term recurrence in patients with venous thromboembolism in the era of direct oral anticoagulants. Eur J Intern Med 2023; 118: 59–72, doi:10.1016/j.ejim.2023.08.007.
• 17.   Goldhaber SZ, Visani L, De Rosa M. Acute pulmonary embolism: Clinical outcomes in the International Cooperative Pulmonary Embolism Registry (ICOPER). Lancet 1999; 353: 1386–1389, doi:10.1016/s0140-6736(98)07534-5.
• 18.   Chatani R, Yamashita Y, Morimoto T, Mushiake K, Kadota K, Kaneda K, et al. Cancer-associated venous thromboembolism in the direct oral anticoagulants era: Insight from the COMMAND VTE Registry-2. Thromb Res 2024; 234: 86–93, doi:10.1016/j.thromres.2023.12.016.
• 19.   Schulman S, Kearon C; Subcommittee on Control of Anticoagulation of the Scientific and Standardization Committee of the International Society on Thrombosis and Haemostasis. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients. J Thromb Haemost 2005; 3: 692–694, doi:10.1111/j.1538-7836.2005.01204.x.
• 20.   Nutritional anaemias. Report of a WHO scientific group. World Health Organ Tech Rep Ser 1968; 405: 5–37.
• 21.   Yamashita Y, Morimoto T, Amano H, Takase T, Hiramori S, Kim K, et al. Influence of baseline platelet count on outcomes in patients with venous thromboembolism (from the COMMAND VTE Registry). Am J Cardiol 2018; 122: 2131–2141, doi:10.1016/j.amjcard.2018.08.053.
• 22.   Ortel TL, Neumann I, Ageno W, Beyth R, Clark NP, Cuker A, et al. American Society of Hematology 2020 guidelines for management of venous thromboembolism: Treatment of deep vein thrombosis and pulmonary embolism. Blood Adv 2020; 4: 4693–4738, doi:10.1182/bloodadvances.2020001830.
• 23.   National Institute for Health and Care Excellence (NICE). Venous thromboembolic diseases: Diagnosis, management and thrombophilia testing. NICE guideline. 2020. https://www.nice.org.uk/guidance/ng158/resources/venous-thromboembolic-diseases-diagnosis-management-and-thrombophilia-testing-pdf-66141847001797 (accessed February 28, 2025).
• 24.   Stevens SM, Woller SC, Kreuziger LB, Bounameaux H, Doerschug K, Geersing GJ, et al. Antithrombotic therapy for VTE disease: Second update of the CHEST guideline and expert panel report. Chest 2021; 160: e545–e608, doi:10.1016/j.chest.2021.07.055.
• 25.   Joyce E, Haymart B, Kong X, Ali MA, Carrigan M, Kaatz S, et al. Length of anticoagulation in provoked venous thromboembolism: A multicenter study of how real-world practice mirrors guideline recommendations. J Am Heart Assoc 2022; 11: e025471, doi:10.1161/jaha.122.025471.
• 26.   Kaneda K, Yamashita Y, Morimoto T, Chatani R, Nishimoto Y, Ikeda N, et al. Temporal changes in long-term outcomes of venous thromboembolism from the warfarin era to the direct oral anticoagulant era. J Am Heart Assoc 2024; 13: e034412, doi:10.1161/jaha.124.034412.
• 27.   Kearon C, Akl EA, Comerota AJ, Prandoni P, Bounameaux H, Goldhaber SZ, et al. Antithrombotic therapy for VTE disease: Antithrombotic therapy and prevention of thrombosis, 9th ed: American College of Chest Physicians evidence-based clinical practice guidelines. Chest 2012; 141(Suppl): e419S–e496S, doi:10.1378/chest.11-2301.
• 28.   Nieto JA, Solano R, Ruiz-Ribó MD, Ruiz-Gimenez N, Prandoni P, Kearon C, et al. Fatal bleeding in patients receiving anticoagulant therapy for venous thromboembolism: Findings from the RIETE registry. J Thromb Haemost 2010; 8: 1216–1222, doi:10.1111/j.1538-7836.2010.03852.x.
• 29.   Kuijer PM, Hutten BA, Prins MH, Büller HR. Prediction of the risk of bleeding during anticoagulant treatment for venous thromboembolism. Arch Intern Med 1999; 159: 457–460, doi:10.1001/archinte.159.5.457.
• 30.   O’Brien EC, Simon DN, Thomas LE, Hylek EM, Gersh BJ, Ansell JE, et al. The ORBIT bleeding score: A simple bedside score to assess bleeding risk in atrial fibrillation. Eur Heart J 2015; 36: 3258–3264, doi:10.1093/eurheartj/ehv476.
• 31.   Morimoto T, Gandhi TK, Fiskio JM, Seger AC, So JW, Cook EF, et al. Development and validation of a clinical prediction rule for angiotensin-converting enzyme inhibitor-induced cough. J Gen Intern Med 2004; 19: 684–691, doi:10.1111/j.1525-1497.2004.30016.x.
• 32.   Lecumberri R, Jiménez L, Ruiz-Artacho P, Nieto JA, Ruiz-Giménez N, Visonà A, et al. Prediction of major bleeding in anticoagulated patients for venous thromboembolism: Comparison of the RIETE and the VTE-BLEED scores. TH Open 2021; 5: e319–e328, doi:10.1055/s-0041-1729171.
• 33.   Vedovati MC, Mancuso A, Pierpaoli L, Paliani U, Conti S, Ascani A, et al. Prediction of major bleeding in patients receiving DOACs for venous thromboembolism: A prospective cohort study. Int J Cardiol 2020; 301: 167–172, doi:10.1016/j.ijcard.2019.11.105.
• 34.   Klok FA, Barco S, Konstantinides SV. Evaluation of VTE-BLEED for predicting intracranial or fatal bleeding in stable anticoagulated patients with venous thromboembolism. Eur Respir J 2018; 51: 1800077, doi:10.1183/13993003.00077-2018.
• 35.   Seiler E, Limacher A, Mean M, Beer HJ, Osterwalder J, Frauchiger B, et al. Derivation and validation of a novel bleeding risk score for elderly patients with venous thromboembolism on extended anticoagulation. Thromb Haemost 2017, doi:10.1160/th17-03-0162.
• 36.   Poli D, Antonucci E, Testa S, Cosmi B, Palareti G, Ageno W. The predictive ability of bleeding risk stratification models in very old patients on vitamin K antagonist treatment for venous thromboembolism: Results of the prospective collaborative EPICA study. J Thromb Haemost 2013; 11: 1053–1058, doi:10.1111/jth.12239.
• 37.   Klok FA, Barco S, Turpie AGG, Haas S, Kreutz R, Mantovani LG, et al. Predictive value of venous thromboembolism (VTE)-BLEED to predict major bleeding and other adverse events in a practice-based cohort of patients with VTE: Results of the XALIA study. Br J Haematol 2018; 183: 457–465, doi:10.1111/bjh.15533.
• 38.   Nopp S, Ay C. Bleeding risk assessment in patients with venous thromboembolism. Hamostaseologie 2021; 41: 267–274, doi:10.1055/a-1339-9987.
• 39.   Loewen P, Dahri K. Risk of bleeding with oral anticoagulants: An updated systematic review and performance analysis of clinical prediction rules. Ann Hematol 2011; 90: 1191–1200, doi:10.1007/s00277-011-1267-3.
• 40.   Kato ET, Goto S, Giugliano RP. Overview of oral antithrombotic treatment in elderly patients with atrial fibrillation. Ageing Res Rev 2019; 49: 115–124, doi:10.1016/j.arr.2018.10.006.
• 41.   Malik AH, Yandrapalli S, Aronow WS, Panza JA, Cooper HA. Meta-analysis of direct-acting oral anticoagulants compared with warfarin in patients >75 years of age. Am J Cardiol 2019; 123: 2051–2057, doi:10.1016/j.amjcard.2019.02.060.
• 42.   López-Núñez JJ, Pérez-Andrés R, Di Micco P, Schellong S, Gómez-Cuervo C, Sahuquillo JC, et al. Direct oral anticoagulants or standard anticoagulant therapy in fragile patients with venous thromboembolism. TH Open 2019; 3: e67–e76, doi:101055/s-0039-1683970.
• 43.   Gross PL, Chan NC. Thromboembolism in older adults. Front Med (Lausanne) 2020; 7: 470016, doi:10.3389/fmed.2020.470016.
• 44.   Ballestri S, Romagnoli E, Arioli D, Coluccio V, Marrazzo A, Athanasiou A, et al. Risk and management of bleeding complications with direct oral anticoagulants in patients with atrial fibrillation and venous thromboembolism: A narrative review. Adv Ther 2023; 40: 41–66, doi:101007/s12325-022-02333-9.
• 45.   Bhandari M, Pradhan A, Vishwakarma P, Di Renzo L, Iellamo F, Ali W, et al. Direct oral anticoagulant use in older adults with atrial fibrillation: Challenges and solutions. Eur Cardiol 2025; 20: e03, doi:10.15420/ecr.2024.17.
• 46.   Ito S, Inoko M, Yamashita Y, Morimoto T, Kimura T. Thrombocytopenia and bleeding events in patients with venous thromboembolism. Eur J Intern Med 2024; 122: 132–134, doi:10.1016/j.ejim.2024.01.007.
• 47.   Decousus H, Tapson VF, Bergmann JF, Chong BH, Froehlich JB, Kakkar AK, et al. Factors at admission associated with bleeding risk in medical patients: Findings from the IMPROVE Investigators. Chest 2011; 139: 69–79, doi:10.1378/chest.09-3081.
• 48.   Tamargo J, Kaski JC, Kimura T, Barton JC, Yamamoto K, Komiyama M, et al. Racial and ethnic differences in pharmacotherapy to prevent coronary artery disease and thrombotic events. Eur Heart J Cardiovasc Pharmacother 2022; 8: 738–751, doi:10.1093/ehjcvp/pvac040.