Safety and Effectiveness of Direct Oral Anticoagulants Versus Low-Molecular-Weight Heparin for Cancer-Associated Thrombosis: A Systematic Review and Meta-analysis

Abstract

Background

Cancer-associated thrombosis is a condition associated with high mortality rates, yet limited evidence exists regarding the safety and effectiveness of low-molecular-weight heparin (LMWH) and direct oral anticoagulants (DOACs), focusing on a fixed follow-up period based on clinical practice guideline recommendations.

Objective

This study aimed to compare the safety and effectiveness of DOACs versus LMWH in patients with cancer-associated thrombosis over a 6-month follow-up period.

Methods

PubMed, Embase, and Cochrane Library databases were systematically searched up to 30 June, 2025. Recurrent venous thromboembolism, major bleeding, and all-cause mortality were pooled using a random-effects meta-analysis.

Results

Seven randomized controlled trials and 28 cohort studies were included in our systematic review. After applying the criteria for a 6-month follow-up period, five randomized controlled trials and 16 cohort studies with 49,824 patients were analyzed in the meta-analysis. In randomized controlled trials, DOACs showed a lower incidence of venous thromboembolism recurrence (relative risk [RR] 0.66, 95% confidence interval [CI] 0.49–0.87) compared with LMWH, with a non-significant increase in major bleeding (RR 1.28, 95% CI 0.87–1.88) and no significant difference in all-cause mortality (RR 1.00, 95% CI 0.86–1.18). Cohort studies demonstrated a lower incidence of venous thromboembolism recurrence (RR 0.69, 95% CI 0.62–0.76) with DOACs, a non-significant reduction in major bleeding (RR 0.85, 95% CI 0.68–1.07), and a lower risk of all-cause mortality (RR 0.47, 95% CI 0.31–0.72).

Conclusions

In patients with cancer-associated thrombosis, DOACs demonstrated a decrease in recurrent venous thromboembolism without increasing the risk of all-cause mortality. A non-significant increase in the risk of major bleeding was recorded in randomized controlled trials, but not in cohort studies. DOACs may provide greater effectiveness for cancer-associated thrombosis compared with LMWH.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s40264-026-01684-4.

Key Points

In patients with cancer-associated thrombosis, direct oral anticoagulants were associated with a significantly lower risk of recurrent venous thromboembolism than low-molecular-weight heparin, without a significantly increased risk of major bleeding.

Compared with low-molecular-weight heparin, direct oral anticoagulants were associated with a significantly lower risk of all-cause mortalities in cohort studies, but not in randomized controlled trials.

Overall, direct oral anticoagulants may provide greater effectiveness for cancer-associated thrombosis than low-molecular-weight heparin, with comparable safety.

Introduction

Cancer-associated thrombosis (CAT) is a serious complication in patients with cancer. It is characterized by repeat occurrences of venous thromboembolism (VTE), including deep vein thrombosis and pulmonary embolism. Patients with cancer face a nine-fold higher risk of VTE compared with the general population [1]. In a recent study conducted in Denmark, the cumulative incidence of death among patients with CAT at 1 and 5 years after diagnosis was 45% and 69%, respectively [2].

Anticoagulants, such as vitamin K antagonists (e.g., warfarin), heparin, and its derivatives (including low-molecular-weight heparin [LMWH]), and direct oral anticoagulants (DOACs), are widely recognized to play a crucial role in preventing the formation of new blood clots and reducing the risk of existing clots becoming larger or causing complications [3]. Low-molecular-weight heparins have been the preferred choice owing to greater effectiveness and similar safety profile compared to vitamin K antagonists [4–7]. Since 2010, DOACs, which include direct thrombin inhibitors and direct factor Xa inhibitors, have emerged as an alternative for CAT management, offering patients with cancer the convenience of oral administration and reduced need for therapeutic monitoring.

Currently, both LMWH and DOACs are recommended as first-line treatments for patients with CAT for 3–6 months, according to clinical guidelines published by the American Society of Clinical Oncology and the European Society for Medical Oncology [8, 9]. However, the randomized controlled trials (RCTs) referenced in the guidelines may have been underpowered to detect certain clinical outcomes such as bleeding events, and their controlled environments may not accurately represent the complexity of real-world practice [10, 11]. In cohort studies, the safety and effectiveness of DOACs compared to LMWH remain controversial, especially in bleeding events [12–14]. Variations in study design, including follow-up periods, inclusion criteria, treatment duration, and outcome measurements, contribute to the heterogeneity observed across studies [15, 16]. Previous meta-analyses were limited in integrating and analyzing data from specific follow-up periods because of a scarcity of relevant research [17, 18].

Furthermore, differences may exist when comparing the results between RCTs and real-world studies, and these discrepancies have yet to be investigated [18]. Therefore, this systematic review and meta-analysis aims to quantitatively analyze the overall safety and effectiveness of DOACs compared to LMWH among patients with CAT. Analysis was performed separately for RCTs and cohort studies, specifically focusing on a fixed follow-up period of 6 months, based on current clinical practice guideline recommendations.

Methods

Literature Source and Search Strategy

We conducted a systematic literature search from the inception of relevant databases up to 30 June, 2025, to identify studies that assessed the safety and effectiveness of DOACs compared to LMWH in patients with CAT. The literature search was performed using PubMed (MEDLINE database), Excerpta Medical Database (Embase), and Cochrane Library. The search combined keywords related to: (1) DOACs; (2) LMWH; (3) cancer or CAT; and (4) VTE. The complete detailed search strategies are provided in the Electronic Supplementary Material (ESM). We further examined the reference lists of review articles to identify any additional relevant studies.

Study Eligibility

Two reviewers (WK and RTLL) independently reviewed eligible studies using a standardized methodology. In cases of uncertainty or discrepancy between the two reviewers, the third reviewer (STHL) intervened to resolve through consensus. Through discussion and deliberation, a consensus was reached to resolve discrepancies and ensure the accuracy and integrity of the study selection process. After a full-text review, articles that met the following inclusion criteria were included: (1) study design: an RCT or retrospective/prospective cohort study; (2) language: English; (3) population: adult patients with CAT; (4) interventions: DOACs (apixaban, edoxaban, rivaroxaban, or dabigatran); (5) control: LMWH (dalteparin, enoxaparin, tinzaparin, bemiparin, nadroparin, reviparin, parnaparin, or certoparin); and (6) outcomes: recurrent VTE, major bleeding, and all-cause mortality (if included).

Articles were excluded if they met any of the following criteria: (1) duplicated articles or presented previously published results; (2) other article types such as conference abstracts, reviews, case reports, editorials, guidelines, and study protocols; (3) DOACs and LMWH were used for prophylactic purposes; (4) patients received both LMWH and DOACs during the study period, with the possibility of crossover, i.e., patients initially assigned to the LMWH group could switch to DOACs, or vice versa; and (5) articles with incomplete data and deemed insufficient for a systematic review and meta-analysis.

Outcome Definition

The effectiveness outcome was recurrent VTE, and the safety outcomes were major bleeding and all-cause mortality (if included). In terms of the follow-up duration, studies that explicitly specified a 6-month follow-up duration were included in the meta-analysis. In cases where the follow-up duration was not explicitly specified, these studies were still included in our study but their data were not considered in the subsequent meta-analysis.

Quality Assessment

WK and RTLL independently assessed the study quality. In cases of uncertainty or discrepancy, STHL intervened to reach a consensus, ensuring accuracy and integrity in the quality assessment process. The revised Cochrane risk-of-bias tool (RoB 2, 22 August, 2019 version) was used to assess the quality and risk of bias in RCTs [19]. Risk of bias was evaluated based on five criteria: the randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Each criterion was rated as “low risk,” “some concerns,” or “high risk,” leading to an overall risk‑of‑bias judgment for each trial. The Newcastle Ottawa Scale was applied to assess the risk of bias in observational studies [20]. The Newcastle Ottawa Scale evaluates cohort studies based on selection, comparability, and outcome. For selection, it considers whether the exposed cohort was representative of the target population, whether the non-exposed cohort was drawn from the same source population, how exposure was ascertained, and whether the outcome of interest was absent at baseline. The selection criteria can be awarded up to four stars. For comparability, it evaluates whether key confounders were controlled for; this domain can receive up to two stars. For outcome, it assesses the method of outcome ascertainment, whether the follow-up was sufficiently long, and whether the follow-up was adequately complete; this domain can receive up to three stars. Studies could therefore receive a maximum of nine stars, with a higher score indicating greater methodological quality.

Data Extraction

WK and RTLL independently recorded the following details from each study: author, publication year, study design, countries, data sources, study period, follow-up duration, cancer type, sample size, age in the DOAC group, metastatic cancer rate, interventions and comparators, and the number of events for each outcome. In cases of uncertainty or discrepancy, the two reviewers discussed and deliberated to reach a consensus.

Statistical Analysis

Meta-analyses were conducted in two ways to assess safety and effectiveness outcomes. First, RCTs and cohort studies were analyzed separately to examine whether discrepancies exist between controlled and real-world settings. Second, RCTs and cohort studies were pooled in a single random‑effects meta‑analysis to examine whether results were consistent when all available evidence was synthesized together.

Among the studies included in the systematic review, only articles with eligible data for quantitative synthesis and a well-defined 6-month follow-up period were included in the meta-analysis. To ensure consistency across studies, we used intention-to-treat (ITT) results when both ITT and per-protocol analyses were available. An exception was made for two cohort studies in which crossover between DOACs and LMWH was present, and therefore the on-treatment analyses were used to align with our predefined eligibility criteria [21, 22]. Unadjusted event counts were used for both cohort studies and RCTs. The pooled relative risk (RR) and corresponding 95% confidence intervals (CIs) were calculated for each outcome based on the Mantel–Haenszel random-effects model. Statistical significance was determined as P < 0.05. The heterogeneity among the included studies was evaluated using the I² statistic. Statistically significant heterogeneity was considered when P < 0.10 and I² > 50%. Funnel plots and Egger’s test were employed to check for potential publication bias when more than ten studies were available in a single analysis [23]. WK conducted analyses and BPYL independently cross-checked them for quality assurance. All statistics and forest plots were generated using R software (Version 4.2.2).

The present study adheres to the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [24]. The study protocol was registered in PROSPERO (International Prospective Register of Systematic Reviews), available from https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42024594517.

Sensitivity Analysis

We conducted a sensitivity analysis to assess the robustness of our results by repeating the meta‑analysis for cohort studies after excluding six studies that used per-protocol or on-treatment analyses, restricting the analysis to ITT analyses only, to evaluate the potential impact of mixing ITT with per-protocol effect estimates.

Results

Study Selection and Characteristics

As shown in Fig. 1, 4691 articles were identified as potentially relevant. Among these articles, one study could not be retrieved using keywords and was identified through a systematic review article. After eliminating duplicates and conducting screening based on the title and abstract, 49 articles were identified for a comprehensive full-text review. Among the excluded articles, the populations reported in three articles did not meet the inclusion criteria, three articles were excluded because of inappropriate interventions (irrelevant intervention, or lack of a control group), one article did not report the specific outcomes of interest, one article was a cross-sectional study, and six articles involved a crossover between the two treatment groups, where patients switched between LMWH or DOACs during the study period.

Fig. 1.

Flow diagram. RCT randomized controlled trial

A total of 35 articles were included for a systematic review, including seven RCTs, two prospective cohort studies, and 26 retrospective cohort studies, comprising a total of 74,892 patients with cancer. The characteristics of the included studies are shown in Table 1, and the quality assessment for RCTs and cohort studies is shown in Table S1 and Table S2 of the ESM, respectively. Although all included RCTs were open‑label, the risk of bias associated with the lack of blinding was judged to be low because there were no important deviations from the intended interventions related to awareness of allocation, and the outcomes (recurrent VTE, major bleeding, and death) were objective and unlikely to be influenced by knowledge of treatment assignment. Most studies (n = 28) recruited patients with all cancer sites, while other studies focused on specific cancer types, including lung cancer (n = 1), gastrointestinal cancer (n = 3), gynecologic cancer (n = 2), and urologic cancer (n = 1). Regarding DOACs, five studies (including two RCTs) compared the safety and effectiveness of apixaban and LMWH in preventing the recurrence of VTE in patients with CAT [10, 25–28], while one RCT evaluated the use of edoxaban [11], and 17 studies (including three RCTs) assessed the use of rivaroxaban [13, 14, 16, 21, 22, 29–40]. No study solely investigated dabigatran. Additionally, 12 articles included multiple oral anticoagulants in their studies [12, 41–51]. The duration of follow-up in these studies ranged from 3 to 24 months, with some studies lacking clear specifications of the follow-up duration.

Table 1.

Study characteristics

Author (publication year)	Study design	Countries	Data sources	Study period	Follow-up duration (months)	Cancer type	Sample size	Age in the DOAC group (years)	Metastatic cancer rate (%)	Intervention (DOACs)	Comparator (LMWH)	Outcome contribution to meta-analysis
Ross et al. (2017) [41]	Retrospective cohort study	USA	1 center	20140101–20151231	6/12	All cancer sites	153	Median (IQR): 64 (53–69)	NA	Rivaroxaban (mainly)	Enoxaparin (mainly)	rVTE (ITT)
Ageno et al. (2017) [35]	Prospective cohort study	Austria, Belgium, Canada, Czech Republic, Denmark, France, Germany, Greece, Hungary, Israel, Italy, Moldova, Netherlands, Norway, Portugal, Slovenia, Spain, Sweden, Switzerland, Ukraine, UK	Multicenter	20120626–20140313	12	All cancer sites except brain cancer and acute leukemia	316	Mean (SD): 69.3 (11.9)	NA	Rivaroxaban	LMWH	Not included because of follow-up period
Young et al. (2018) [29]	RCT	UK	50 centers	20130906–20161222	6	All cancer sites except basal cell or squamous cell skin cancer	406	Median (IQR): 67 (22–87)	58.1	Rivaroxaban	Dalteparin	rVTE, MB, ACM (ITT)
Raskob et al. (2018) [11]	RCT	USA, Belgium, France, Hungary, Italy, Netherlands, Australia, Austria, Canada, Czech Republic, Germany, New Zealand	112 centers	20150701–20161231	6/12	All cancer sites except basal cell or squamous cell skin cancer	1046	Mean (SD): 64.3 (11.0)	53.0	Edoxaban	Dalteparin	rVTE, MB, ACM (ITT)
Chaudhury et al. (2018) [30]	Retrospective cohort study	India	1 center	20100501–20150630	6 for recurrent VTE	All cancer sites	286	Mean (SD): 62.20 (12.34)	51.0	Rivaroxaban	Dalteparin	rVTE (ITT)
Simmons et al. (2018) [36]	Retrospective cohort study	USA	1 center	20130301–20160731	3/12	All cancer sites	266	Mean (SD): 62.6 (12.1)	55.6	Rivaroxaban	Enoxaparin	Not included because of follow-up period
Streiff et al. (2018) [13]	Retrospective cohort study	USA	Humana database	20130101–20150531	6/12	All cancer sites	1367	Mean (SD): 72.7 (8.8)	NA	Rivaroxaban	LMWH	rVTE, MB (ITT)
Alzghari et al. (2018) [42]	Retrospective cohort study	USA	1 center	20130601–20150930	3/6	All cancer sites	71	Mean (SD): 62.0 (13.8)	45.1	Rivaroxaban or apixaban	Enoxaparin	rVTE, MB, ACM (ITT)
Lee et al. (2019) [38] Lee et al. (2019) [38]	Retrospective cohort study	South Korea	1 center	20120101–20171231	NA	Primary urologic cancer (including bladder cancer, prostate cancer, kidney cancer, testicular cancer, and other urinary tract cancers)	124	Mean (SD): 68.05 (11.17)	69.4	Rivaroxaban	Dalteparin	Not included because of follow-up period
Lee et al. (2019) [37]	Retrospective cohort study	South Korea	1 center	20120101–20161231	NA	Primary lung cancer	204	Mean (SD): 66.79 (11.32)	83.8	Rivaroxaban	Dalteparin	Not included because of follow-up period
Lee et al. (2019) [14]	Retrospective cohort study	South Korea	1 center	20120101–20161231	6	Gastrointestinal tract cancer (including stomach cancer, colorectal cancer, and pancreaticobiliary cancers)	281	Mean (SD): 66.5 (11.6)	79.7	Rivaroxaban	Dalteparin, enoxaparin, or nadroparin	rVTE, MB, ACM (per-protocol)
Signorelli et al. (2019) [31]	Retrospective cohort study	USA	1 center	20130701–20150630	6	Gynecologic cancers	44	Mean (SD): 60.4 (13.1)	NA	Rivaroxaban	Enoxaparin	rVTE, MB (ITT)
Phelps et Al. (2019) [43]	Retrospective cohort study	USA	1 center	20101201–20160131	6	All cancer sites except carcinoma in-situ (including non-metastatic basal cell and squamous cell skin carcinomas)	480	Mean (SD): 58 (13)	53.0	Rivaroxaban, apixaban, or dabigatran	Enoxaparin or dalteparin	rVTE, MB, ACM (per-protocol)
Wysokinski et al. (2019) [44]	Prospective cohort study	USA	1 center	20130301–20180130	3/6	All cancer sites	750	Mean (SD): 64 (12) in apixaban group, and 62 (13) in rivaroxaban group	55.2	Apixaban or rivaroxaban	Enoxaparin	rVTE, MB, ACM (ITT)
Faqah et al. (2020) [32]	Retrospective cohort study	Pakistan	1 center	20120101–20171231	6	All cancer sites	150	Mean (SD): 51.84 (13.49)	48.6	Rivaroxaban	Enoxaparin	rVTE, MB (ITT)
Lee et al. (2020) [39]	Retrospective cohort study	South Korea	1 center	20120101–20171231	NA	Gynecologic cancer (including ovarian, cervical, uterine, vaginal, vulvar, endometrial, and fallopian tube cancers)	162	Mean (SD): 56.90 (12.32)	79.6	Rivaroxaban	Dalteparin	Not included because of follow-up period
Kim et al. (2020) [33]	Retrospective cohort study	South Korea	1 center	20040101–20141231	6	Advanced upper gastrointestinal tract and hepatopancreatobiliary cancer	174	Median (IQR): 67 (59–72)	NA	Rivaroxaban	Dalteparin or enoxaparin	rVTE, MB (ITT)
McBane et al. (2020) [25]	RCT	USA	28 centers	20151120–20171002	6	All cancer sites except non-melanoma skin cancer	287	Mean (SD): 64.4 (11.3)	64.3	Apixaban	Dalteparin	rVTE, MB, ACM (ITT)
Agnelli et al. (2020) [10]	RCT	USA, Belgium, France, Germany, Israel, Italy, Netherlands, Poland, Portugal, Spain, UK	119 centers	20170401–20190630	6	All cancer sites except basal cell or squamous cell carcinoma of the skin, brain cancer, and acute leukemia	1155	Mean (SD): 67.2 (11.3)	NA	Apixaban	Dalteparin	rVTE, MB, ACM (ITT)
Cohen et al. (2021) [26]	Retrospective cohort study	USA	4 US commercial claims databases	20140901–20170630	6	All cancer sites except nonmelanoma skin cancer	9501	Mean (SD): 64.6 (12.6)	51.6	Apixaban	LMWH	rVTE, MB (per-protocol)
Douros et al. (2021) [45]	Retrospective cohort study	Canada	3 linked electronic healthcare databases	20120101–20151231	6	All cancer sites	3211	Mean (SD): 72.0 (11.6)	NA	Rivaroxaban, apixaban, or dabigatran	LMWH	rVTE, MB, ACM (ITT)
Chen et al. (2021) [12]	Retrospective cohort study	Taiwan	Chang Gung Research Database	20120101–20190131	12	All cancer sites	1109	Mean (SD): 67.7 (13.3)	48.3	Rivaroxaban, apixaban, edoxaban, or dabigatran	Enoxaparin	Not included because of follow-up period
Fovel et al. (2021) [46]	Retrospective cohort study	USA	1 center	20161001–20181031	NA	All cancer sites	156	Median (IQR): 68.0 (59.8–75.0)	64.1	Rivaroxaban, apixaban, or dabigatran	Enoxaparin	Not included because of follow-up period
Planquett et al. (2022) [16]	RCT	France	18 centers	20160901–20180131	3	All cancer sites	158	Median (IQR): 68.6 (62.9–77.8)	72.8	Rivaroxaban	Dalteparin	Not included because of follow-up period
Coleman et al .(2023) [34]	Retrospective cohort study	USA	EHR database	20120101–20201231	3/6/12	All cancer sites except esophageal, gastric, unresected colorectal, bladder, and noncerebral central nervous system cancers and leukemia	3708	NA	49.7	Rivaroxaban	Enoxaparin (mainly)	rVTE, MB, ACM (ITT)
Cohen et al. (2023) [27]	Retrospective cohort study	USA	4 US commercial claims databases and Medicare database	20140301–20180331	6	Leukemia, bladder, brain tumor, gynecologic, lung, lymphoma, pancreas, renal cell carcinoma, stomach, testicular, gastrointestinal cancer, female breast cancer, prostate cancer, multiple myeloma	18999	Mean (SD): 72.7 (11.4)	63.5	Apixaban	LMWH	rVTE, MB (per-protocol)
Guan et al. (2023) [47]	Retrospective cohort study	USA	EHR from 3 centers	20180101–20191231	12	Primary gastrointestinal cancer (including stomach, esophagus, pancreas, liver, colon, rectal)	141	Mean (SD): 62.8 (12.6)	64.5	Rivaroxaban or apixaban	Enoxaparin (mainly)	Not included because of follow-up period
Riaz et al. (2023) [48]	Retrospective cohort study	USA	OptumLabs Data Warehouse	20120101–20190930	12	Primary cancer diagnosis (except skin cancer)	3640	Mean (SD): 67.4 (12.1)	60.1	DOAC	LMWH	Not included because of follow-up period
Schrag et al. (2023) [49]	RCT	USA	EHR from 67 centers	20161201–20201131	6	Any invasive solid tumor, lymphoma, multiple myeloma, or chronic lymphocytic leukemia	671	Median (IQR): 64 (56–70)	69.1	Apixaban, dabigatran, edoxaban, or rivaroxaban	Enoxaparin, dalteparin, or fondaparinux	Not included as absolute event numbers were not reported
Cohen et al. (2024) [28]	Retrospective cohort study	USA	Komodo Health Healthcare Map	20160701–20220430	6 for DOAC; 4 for LMWH	Brain tumor, breast cancer, gastrointestinal, gynecologic, lung, lymphoma, prostate cancer	16441	Mean (SD): 64.6 (12.6)	39.2	Apixaban	LMWH	Not included because of follow-up period
Kengkla et al. (2024) [50]	Retrospective cohort study	Thailand	EHR database from 3 centers	20170101–20211231	12	Gynecologic, lung, hepatobiliary, head and neck, gastrointestinal, hematologic	1449	Mean (SD): 63.2 (13.3)	60.5	Dabigatran, rivaroxaban, apixaban, or edoxaban	LMWH	Not included because of follow-up period
Linder et al. (2024) [21]	Retrospective cohort study	Sweden	Nationwide health register data in Sweden	20130101–20201231	3/6/12/24	All types of cancers except lip/upper gastrointestinal cancer, malignant immunoproliferative diseases, leukemia, and non-melanoma skin cancer	5464	< 65: 28% ≥ 65: 72%	NA	Rivaroxaban	LMWH	rVTE, MB, ACM (on-treatment)
Chung et al. (2025) [51]	Retrospective cohort study	USA	NYU Langone Health system	20140101–20200229	3	Gynecologic, lung, breast, colon and rectal, lymphoma, multiple myeloma, pancreatic, prostate, head and neck, and sarcoma	102	Mean (SD): 68 (14) in apixaban group, and 69 (12) in rivaroxaban group	NA	Apixaban, rivaroxaban	Enoxaparin	Not included because of follow-up period
Cohen et al. (2025) [22]	Retrospective cohort study	UK	UK Clinical Practice Research Datalink GOLD and Aurum databases	20130101–20201031	3/6/12	All types of cancer except non-brain central nervous system, unresected colorectal/lower gastrointestinal tract, hematologic, esophagus, stomach, and bladder	2173	Mean (SD): 72.4 (12.1)	NA	Rivaroxaban	LMWH	rVTE, MB, ACM (on-treatment)
Sinn et al. (2025) [40]	RCT	Germany	Multicenter	20160301–20190630	3 for rVTE; 6 for MB and ACM	All cancer sites except skin cancer	247	Mean (SD): 63 (11)	69.5	Rivaroxaban	Enoxaparin, tinzaparin, dalteparin, certoparin, or nadroparin	MB, ACM (ITT)

ACM all-cause mortality, DOACs direct oral anticoagulants, EHR electronic health record, IQR interquartile range, ITT intention-to-treat, , LMWH low-molecular-weight heparin, MB major bleeding, NA not available, RCT randomized controlled trial, rVTE recurrent venous thromboembolism, SD standard deviation

After restricting the follow-up period to 6 months, 21 articles with eligible data were selected for the meta-analysis, including five RCTs, and 16 cohort studies. Most are ITT studies, except for Lee et al., Phelps et al., Cohen et al., and Cohen et al., which are per-protocol studies that only included outcomes during the anticoagulation treatment period [14, 26, 27, 43]. Furthermore, data from the on-treatment analyses were used for Linder et al. and Cohen et al., where patients who discontinued their initial treatment or switched to a different anticoagulant class were censored [21, 22]. Only recurrent VTE data from Ross et al. and Chaudhury et al. were included, as the 6-month data for major bleeding were unavailable, whereas only major bleeding data from Sinn et al. were included, as the 6-month data for recurrent VTE were unavailable [30, 40, 41]. Among these 49,824 patients, 16,599 (33.3%) received DOAC and 33,225 (66.7%) received LMWH.

Effectiveness of DOACs Compared with LMWH for CAT

We conducted a meta-analysis to examine the RR of DOACs compared to LMWH for preventing the recurrence of VTE in patients with CAT. As shown in Fig. 2, recurrent VTE occurred in 75 of 1446 (5.2%) patients receiving DOACs versus 119 of 1448 (8.2%) patients receiving LMWH in RCTs, and 536 of 15,030 (3.6%) patients receiving DOACs versus 1686 of 31,653 (5.3%) patients receiving LMWH in cohort studies. The pooled analysis revealed that DOACs were significantly associated with a reduced risk of VTE recurrence in both RCTs (RR = 0.66; 95% CI 0.49–0.87; P = 0.24; I² = 29%) and cohort studies (RR = 0.69; 95% CI 0.62–0.76; P = 0.57; I² = 0%).

Fig. 2.

Forest plot of risk ratios of direct oral anticoagulants (DOACs) versus low-molecular-weight heparin (LMWH) for recurrent venous thromboembolism in randomized controlled trials (RCTs) and cohort studies. CI confidence interval, MH Mantel–Haenszel

Furthermore, the funnel plot was visually inspected, and no clear sign of asymmetry was observed for recurrent VTE when comparing the DOACs and LMWH groups in cohort studies (Fig. S1 of the ESM). Further analysis using Egger’s tests indicated no significant publication bias (P = 0.17). The combined analysis of RCTs and cohort studies yielded results consistent with the separate analyses, demonstrating a significantly reduced risk of recurrent VTE in the DOAC group compared with LMWH (RR: 0.69, 95% CI 0.62–0.75, P = 0.54, I² = 0%).

Safety of DOACs Compared with LMWH for CAT

A meta-analysis was conducted to examine the risk of DOACs compared to LMWH for major bleeding in patients with CAT. As shown in Fig. 3, major bleeding occurred in 67 of 1569 (4.3%) and 53 of 1572 (3.4%) patients treated with DOACs versus LMWH in RCTs, and 502 of 14,940 (3.4%) and 1211 of 31,433 (3.9%) patients treated with DOACs versus LMWH in cohort studies. Our analyses did not detect statistically significant differences in major bleeding between DOAC and LMWH. The pooled analysis showed a non-significant higher risk of major bleeding in RCTs (RR = 1.28; 95% CI 0.87–1.88; P = 0.40; I² = 1.3%) and a non-significant lower risk in cohort studies (RR = 0.85; 95% CI 0.68–1.07; P = 0.02; I² = 48%). The funnel plot of cohort studies was visually inspected, and no clear sign of asymmetry was observed for major bleeding when comparing the DOACs and LMWH groups (Fig. S2 of the ESM). Further analysis using Egger’s tests indicated no significant publication bias in these cohort studies (P = 0.53). The combined analysis of RCTs and cohort studies were generally consistent with the separate analyses, with no statistically significant differences in major bleeding (RR: 0.92, 95% CI 0.75–1.15, P = 0.0091, I² = 49%).

Fig. 3.

Forest plot of risk ratios of direct oral anticoagulants (DOACs) versus low-molecular-weight heparin (LMWH) for major bleeding in randomized controlled trials (RCTs) and cohort studies. CI confidence interval, MH Mantel–Haenszel

Furthermore, a separate meta-analysis was conducted to investigate the risk of all-cause mortality in patients with CAT treated with DOACs or LMWH. As shown in Fig. 4, all-cause mortality occurred in 364 of 1569 (23.2%) and 365 of 1572 (23.2%) patients treated with DOACs versus LMWH in RCTs, and 318 of 2898 (11.0%) and 2730 of 13,240 (20.6%) patients treated with DOACs versus LMWH in cohort studies. The pooled analysis indicated no significant difference between the DOAC and LMWH groups in RCTs (RR: 1.00, 95% CI 0.86–1.18, P = 0.25, I² = 26%). However, in cohort studies, DOAC use was associated with a significantly lower risk of death compared with LMWH in patients with CAT (RR: 0.47, 95% CI 0.31–0.72, P < 0.0001, I² = 91%). The combined analysis of RCTs and cohort studies also demonstrated a reduction in all-cause mortality in patients receiving DOACs compared to those receiving LMWH (RR: 0.65, 95% CI 0.47–0.92, P < 0.0001, I² = 90%).

Fig. 4.

Forest plot of risk ratios of direct oral anticoagulants (DOACs) versus low-molecular-weight heparin (LMWH) for death in randomized controlled trials (RCTs) and cohort studies. CI confidence interval, MH Mantel–Haenszel

Sensitivity Analysis

When restricting to ITT effect estimates from cohort studies, the results were largely consistent with our main analysis (recurrent VTE: RR: 0.74, 95% CI 0.64–0.87, P = 0.42, I² = 2.4%; major bleeding: RR: 1.10, 95% CI 0.79–1.53, P = 0.09, I² = 43%; death: RR: 0.47, 95% CI 0.23–0.99, P < 0.0001, I² = 95%) [Fig. S3 of the ESM].

Discussion

In this systematic review and meta-analysis comparing DOACs and LMWH in patients with CAT, we conducted an analysis on the risk of recurrent VTE, major bleeding, and all-cause mortality over a 6-month follow-up period. Our data indicate that DOACs, compared with LMWH, were associated with a significantly reduced risk of recurrent VTE. This finding was observed in both RCTs and cohort studies. Regarding major bleeding, RCTs showed a non-significant higher incidence of bleeding with DOACs, whereas cohort studies suggested a non-significant reduction. For all-cause mortality, we observed a significant reduction in the DOAC group compared with the LMWH group in cohort studies, while no significant differences were observed in RCTs. The combined results for both RCTs and cohort studies indicate DOACs were associated with a significantly decreased risk of recurrent VTE and all-cause mortality in patients with CAT, while suggesting no significant difference in the risk of major bleeding. However, it should be noted that significant heterogeneity may exist between RCTs and cohort studies. Overall, the combined analyses and sensitivity analyses were concordant with our main analyses, which supported the robustness of our findings.

The use of anticoagulant therapy poses a complex clinical challenge in patients with cancer, as it requires a delicate balance between effectively controlling VTE and minimizing the risk of bleeding and mortality. Based on evidence from RCT studies, current guidelines recommend both LMWH and DOACs as first-line options for patients with CAT for at least 3 months, except for those with renal disease, who should avoid LMWH, and patients with a higher risk of major bleeding, who require cautious consideration for DOACs [8, 9, 52, 53]. It is important to note that each study in this field had a unique study design. Treatment duration and follow-up periods varied among these studies (range 3–12 months), with some studies extending beyond 12 months (Table 1), so that these results are neither comparable nor combinable. Considering the higher occurrence of recurrent VTE and bleeding events within the first 6 months [11], only studies with a 6-month follow-up period were included in the meta-analysis based on international clinical practice guidelines and previous studies [8–11]. However, further categorization of observational studies based on patients’ treatment duration was impossible because of limited studies. Our meta-analysis found that DOACs were associated with a reduced recurrence of VTE and a comparable risk of major bleeding compared with LMWH among patients with CAT, using a fixed 6-month follow-up period. These findings are consistent with previous studies [17, 18, 54, 55]. However, it is worth noting that a remarkable discrepancy exists in the risk of all-cause mortality between RCTs and cohort studies. Previous research has highlighted that cancer progression and recurrent VTE contribute to all-cause mortality among patients with CAT [56]. The observed difference in mortality outcomes may be attributed to several factors.

RCTs often have strict inclusion and exclusion criteria that differ from real-world observational studies, such as limitations regarding cancer sites and comorbidities. This can lead to differences in the populations studied and the generalizability of the results. Additionally, RCTs adhere to a pre-specified protocol and undergo rigorous monitoring, which minimizes confounding, ensures a standardized outcome assessment, and enhances internal validity. In contrast, observational cohort studies often include larger and more diverse patient populations, better reflecting real-world clinical practice and providing greater statistical power to detect less frequent outcomes; however, they are inherently more prone to methodological bias. Residual confounding and selection bias cannot be entirely eliminated, even after statistical adjustment. In clinical settings, physicians may preferentially prescribe DOACs to patients with better performance status, fewer comorbidities, or less advanced malignancies, potentially conferring an apparent survival advantage unrelated to treatment efficacy. Furthermore, immortal time bias may arise when the interval between cancer diagnosis and treatment initiation is misclassified, artificially inflating survival in one group. Patients receiving DOACs in real-world settings may differ from those treated with LMWH in terms of baseline cancer stage, prognosis, or concomitant anticancer therapies, which could partially explain the observed differences in mortality outcomes. Moreover, cohort studies exhibited higher heterogeneity, suggesting that the risk of all-cause mortality may vary among patients from different ethnicities or across healthcare systems. Although the cohort studies adjusted for baseline characteristics, we pooled unadjusted event counts across studies, as the degree of adjustment and adjustment methods varied across studies, precluding meaningful pooling of adjusted estimates. Thus, variability in treatment adherence, dosing, and monitoring in clinical practice could further contribute to these discrepancies. Further research is essential to clarify these methodological and clinical factors, ultimately improving our understanding of CAT treatment strategies and enhancing patient care in real-world settings.

Unlike the safety and effectiveness of each LMWH for patients with CAT, which has been extensively explored over the years with no differences found among them, the risk of VTE and major bleeding associated with each DOAC for patients with CAT requires further investigation [57]. Although our overall findings support the use of DOACs over LMWH for CAT, because of limited data, a subgroup analysis comparing each DOAC with LMWH was not feasible in this study. Previous studies suggest there may be clinically relevant differences among DOAC agents, with some data indicating lower bleeding risks for apixaban and dabigatran compared with rivaroxaban [58–62]. However, current evidence, particularly for dabigatran and edoxaban, remains limited, thus, more comparative research is needed to guide the choice of specific DOACs for CAT in clinical practice.

Owing to the convenience of DOACs, more patients are switching from LMWH to DOACs for CAT [63]. In our meta-analysis of the safety and effectiveness of LMWH versus DOACs in patients with CAT, we specifically excluded studies that allowed patients to switch between LMWH and DOACs during the study period [64–66]. Although typically included in previous meta-analyses, these studies are not suitable for directly comparing LMWH and DOACs [17, 18, 55]. In addition, the available evidence on the safety and effectiveness of switching from LMWH to DOACs remains limited. Nevertheless, reasons for switching in real-world practice include better adherence and a lower incidence of local adverse events of oral DOACs compared with injectable LMWH [67–69].

Moreover, DOACs have potential for long-term cost savings [70]. Thus, further studies are needed to evaluate the safety and effectiveness of switching from LMWH to DOACs compared to continuous use of LMWH. These studies are essential to providing valuable insights and supporting clinicians in making informed decisions that benefit patients with CAT.

There are several strengths in this study. First, to our knowledge, this is the first meta-analysis to comprehensively evaluate the safety and effectiveness of DOACs versus LMWH among patients with CAT in a fixed follow-up period, while synthesizing evidence from both RCTs and real-world cohort studies. Second, in contrast to previous reviews, our study encompasses new articles that surged between 2021 and mid-2025, as well as three RCTs and 13 cohort studies [12, 16, 26, 34, 45, 46]. This expanded scope and broader range of cancer populations enhances the generalizability and statistical power of our analyses, enabling more robust conclusions to be drawn. Third, by restricting to studies in which no crossover between treatment arms occurred, we reduced exposure misclassification and enabled a head-to-head comparison of the effectiveness and safety estimates. Finally, by separately investigating the results of both RCTs and cohort studies, this study sheds light on the differences in major bleeding and all-cause mortality between controlled clinical trials and real-world practice. This approach prompts us to explore the underlying reasons, such as healthcare settings, which can potentially improve patients’ quality of life and survival in real-world practice.

This study has several limitations. First, as an inherent limitation of meta-analyses utilizing aggregated data, there were variations in the study design, definition of outcomes, and baseline characteristics across the studies included in our analysis, such as different cancer types, staging, cancer treatments, and anticoagulation treatment doses. Our future real-world studies should carefully evaluate the comparability of dose and treatment regimens across different anticoagulants to minimize potential biases in effectiveness and safety assessments. Second, it is important to acknowledge that all the studies included in this analysis were open-label trials, which could introduce information bias and influence the interpretation of results. Third, the cohort studies included had variable degrees of confounder adjustment, and potential immortal time bias and confounding by indication remains an inherent limitation of observational data. Fourth, the English-language restriction may have introduced language bias; however, only 2.9% of the initially screened records were excluded on the basis of language, suggesting the potential impact on our findings is limited. Finally, it is pertinent to note that the assessment of publication bias for mortality was underpowered because of the scarcity of literature available in this specific area of research. This limitation may have implications for generalizability and could potentially introduce bias. Further studies are needed to evaluate the safety and effectiveness of individual DOACs in different populations to guide clinical decision making regarding the selection of anticoagulation therapy and duration of use for each patient subgroup, such as elderly patients, patients with a higher risk of major bleeding, and different ethnic groups.

Conclusions

Compared with LMWH, DOACs significantly reduced the risk of recurrent VTE in patients with CAT and showed a comparable risk of major bleeding, with RCTs demonstrating a non-significant trend towards a higher bleeding risk and cohort studies showing a non-significant trend towards a decreased risk. Cohort studies also suggested a lower risk of all‑cause mortality with DOACs; however, this finding remains exploratory and should be interpreted with caution because of heterogeneity and potential residual confounding. Nonetheless, our meta-analysis highlights the important role of DOACs as first-line anticoagulant options for CAT management and reinforces the evidence supporting their use to effectively prevent recurrent VTE.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

This study was supported by the Hong Kong Research Grants Council, General Research Fund, No. 17108621. The funding source had no role in the design and conduct of the study or collection, management, analysis, and interpretation of the data.

Declarations

Conflicts of interest

Esther W. Chan has received grants from the Research Grants Council of the Hong Kong SAR, Research Fund Secretariat of the Health Bureau of the Hong Kong SAR, National Natural Science Fund of China, Bayer, Bristol-Myers Squibb, Pfizer, Janssen, Novartis, Amgen, AstraZeneca, Takeda, the RGA Reinsurance Company, Narcotics Division of the Security Bureau of the Hong Kong SAR, and the National Health and Medical Research Council Australia; consulting fees from AstraZeneca, Pfizer, and Novartis; and an honorarium from the Hospital Authority of the Hong Kong SAR. Wei Kang, Bowie P.Y. Lam, Rinko Tsz Lam Lau, Silvia T.H. Li, Kangrui Wu, Yue Wei, Yu Yang, Kyung Jin Lee, Caige Huang, Vincent K.C. Yan, Hei Hang Edmund Yiu, Shing Fung Lee, Aya El Helali, Victor H.F. Lee, Stephen L. Chan, Rina Y.M. Hui, and Ka On Lam have no conflicts of interest that are directly relevant to the content of this article.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and material

Because this meta-analysis was based on data extracted from previously published research, all the data and study materials are available in the public domain. For further discussions, we invite interested parties to contact the corresponding author.

Code availability

Not applicable.

Authors’ contributions

Conceptualization and design: WK and EWC; acquisition, analysis, or interpretation of data: WK, BPYL, TLL, STHL, KW, and EWC; drafting of the manuscript: WK, BPYL, and TLL; critical revision of the manuscript for important intellectual content: all authors; statistical analysis: WK, BPYL, and STHL; administrative, technical, or material support: EWC; supervision: EWC. All authors read, reviewed, and approved the final manuscript.