Nafamostat mesilate as anticoagulant for blood purification: a systematic review and meta-analysis Authors: Caihong Liu, Yingying Yang, Jinglei Ren, Yongxiu Huang, Wei Wei, Xu Li, Fang Wang, Xue Tang, Ping Fu, Ling Zhang, Yuliang Zhao Published: 2025 ABSTRACT ObjectiveThe nafamostat mesilate (NM) has shown good safety and effectiveness as anticoagulant for blood purification. This systematic review and meta-analysis aims to comprehensively compare NM with other anticoagulation strategies during blood purification.MethodWe searched the electronic databases of PubMed, Web of Science, Cochrane and Embase before 6 May 2025 on the application of NM in comparison with conventional anticoagulants, including unfractionated heparin, low-molecular-weight heparin, and regional citrate, or non-anticoagulant (NA) during blood purification. The protocol was registered in PROSPERO (CRD420251110062). We compared the incidence of bleeding, clotting and mortality with risk ratio (RR) and confidence intervals (95% CIs). Mean differences (MDs) with 95% CIs were employed for the comparisons of filter life and hospitalization length. Subgroup analysis was conducted according to the specific type of conventional anticoagulants and extracorporeal modalities of blood purification. Assessment of study quality, and sensitivity and publication bias analyses were conducted.ResultIn total, 16 studies consisting of 4349 participants were included in this systematic review and meta-analysis. Of them, 3956 patients received renal replacement therapy, and 393 underwent extracorporeal membrane oxygenation. NM showed a significantly lower risk of bleeding (RR: 0.53, 95% CI: 0.29–0.97, P = 0.040) and comparable filter life (MD: -1.23 h, 95% CI: -4.27–1.81, P = 0.428) compared with conventional anticoagulants. In comparison with NA, NM demonstrated a substantially longer filter life (MD: 4.84 h, 95% CI: 0.18–9.51, P = 0.042) and similar risk for bleeding complication (RR = 0.49, 95% CI: 0.18–1.35, P = 0.170). Of note, it was also associated with lower in-hospital mortality compared with NA (RR: 0.87, 95% CI: 0.77–0.97, P = 0.015).ConclusionNM is a safe and effective anticoagulant for blood purification, showing lower bleeding risk than conventional anticoagulants and longer filter life than NA, which is particularly suitable for patients at high risk of bleeding.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13054-025-05813-w. Introduction Blood purification therapies, encompassing hemodialysis, continuous renal replacement therapy (CRRT), and extracorporeal membrane oxygenation (ECMO), etc., constitute indispensable life-support modalities for patients suffering from acute kidney injury (AKI) and/or other critical illnesses. The efficacy and continuity of these therapies are highly contingent upon the patency of extracorporeal filters and circuits, which in turn necessitate optimal anticoagulation strategies. Traditionally, unfractionated heparin (UFH) and low-molecular-weight heparin (LMWH) have served as first-line anticoagulants because of their cost-effectiveness, easy monitoring, and reversibility [1]. Nonetheless, their clinical utility is frequently compromised by the risk of major bleeding events, the potential development of heparin-induced thrombocytopenia (HIT), and contraindications in specific high-risk populations, particularly among critically ill individuals. Over the past decade, regional citrate anticoagulation (RCA) has increasingly been recognized as a safer and more efficacious alternative to UFH and LMWH, and is currently endorsed as the preferred anticoagulation modality for CRRT [2]. Despite its favorable safety profile, it necessitates meticulous monitoring due to the risk of citrate accumulation, metabolic alkalosis, and hypocalcemia [3]. RCA use should be treated with caution for patients with liver disease and shock, where impaired citrate metabolism may exacerbate systemic complications [4]. In clinical scenarios where classic anticoagulation is contraindicated or poses an unacceptable risk, alternative approaches such as non-anticoagulation (NA) and the use of nafamostat mesilate (NM) have emerged as critical alternatives. Inevitably, NA is inherently associated with a heightened risk of circuit clotting, often resulting in significantly reduced circuit lifespan, typically ranging from 12 to 32 h [5, 6]. NM is a synthetic serine protease inhibitor that was first developed in Japan in 1981 and initially used for the prevention and treatment of acute pancreatitis and disseminated intravascular coagulation [7, 8]. It began to be used as an anticoagulant for blood purification in 1986 in Japan, and was introduced to China and Korea afterwards [9]. A nationwide survey in Japan in 2018 reported that NM was the most commonly used anticoagulant in blood purification, accounting for 78.6% [10]. Pharmacodynamically, it inhibits key coagulation factors, including VIIa, Xa, and XIIa, and combines with plasmin to extend the fibrinolysis time, thus playing the anticoagulant role during blood purification (Fig. 1) [11]. In addition, the dual metabolism pathways by the liver and blood render NM a very short half-life of 5 to 8 min [7, 12]. These properties make NM particularly advantageous for patients with a heightened risk of bleeding, such as those with recent surgery, trauma, or coagulopathy. Over the past years, multiple clinical studies and small randomized controlled trials (RCTs) have investigated the comparative efficacy and safety of NM versus conventional anticoagulants across diverse blood purification settings. Some of these studies suggest that NM may reduce bleeding complications while maintaining circuit patency. However, the current evidence remains inconclusive and limited by small sample sizes, heterogeneity in study design, and variability in outcome reporting [13–18]. In light of these uncertainties, we conducted a comprehensive systematic review and meta-analysis to assess the efficacy and safety profile of NM as an anticoagulant during blood purification. The PICOS framework is designed as follows: P (population), patients undergoing blood purification; I (intervention), NM as anticoagulant; C (control), treated with conventional anticoagulants like UFH, LMWH, RCA, or NA; O (outcome), bleeding and clotting events, filter lifespan, and clinical outcomes; S (study), randomized controlled study or cohort study. Primary outcomes of interest include bleeding risks, filter lifespan, and clotting risks. The secondary outcome focused on clinical prognosis encompassing mortality, lengths of hospitalization, and intensive care unit (ICU) stay. By pooling data from available studies comparing NM and conventional treatments, this research aims to inform clinical decision-making, particularly for patients with high bleeding risk, and to contribute to the refinement of anticoagulation strategies in blood purification. Fig. 1Anticoagulation mechanisms, functional processes, and metabolism of NM during blood purification. NM, nafamostat mesilate; TF, tissue factor; PL, phospholipid Anticoagulation mechanisms, functional processes, and metabolism of NM during blood purification. NM, nafamostat mesilate; TF, tissue factor; PL, phospholipid Method Protocol and registration The study protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD420251110062). The review methodology adheres to the guidelines outlined in the Cochrane Handbook and complies with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. Search methodology A comprehensive literature search was conducted using four major electronic databases: PubMed, Web of Science, Cochrane Library, and Embase. The search was restricted to studies published in English up to 6 May 2025, focusing on the use of NM as an anticoagulant in blood purification. We retrieved relevant articles through the combination of Medical Subject Headings (MeSH) terms comprising “nafamostat”, and “anticoagulants” or “blood purification”/“renal replacement therapy” and all the corresponding free words. Complete and detailed search strategies of each database are provided in Supplementary Appendix S1. No limitations were imposed regarding publication date, patient sex, or ethnicity. Inclusion and exclusion criteria Two independent reviewers (Caihong Liu and Yingying Yang) screened the titles and abstracts of all identified records to remove irrelevant documents. Publications that explicitly referred to the employment of NM in patients receiving blood purification were collected for evaluation. Next, full-text reviews were conducted independently, and studies meeting the following criteria were candidates for inclusion: (1) studies involving adult patients (≥ 18 years) undergoing any form of blood purification therapy with NM as an anticoagulant, in comparison with conventional anticoagulant like UFH, LMWH, and RCA, or NA; (2) studies reported the extracorporeal circuits’ lifespan, bleeding and/or clotting complications, or patients’ clinical outcomes like mortality and hospital stay; (3) availability of sufficient quantitative data (e.g., event rates, effect sizes, or raw counts) in the full text to enable data extraction. The exclusion criteria were as follows: (1) publication types such as reviews, meta-analyses, case reports, conference abstracts, letters to the editor, or clinical guidelines; (2) the study was conducted based on animals; (3) studies where NM was not used specifically for anticoagulation purposes (e.g., used only as a protease inhibitor or antiviral); (4) non-English or duplicate publications. Any disagreement in study selection was resolved by discussion with the third researcher (Jinglei Ren). Data extraction and quality assessment Data extraction was independently carried out by two authors. The following data were extracted from the included studies: (1) study characteristics, including first author, publication year, country or region, study design, and enrollment period; (2) participant characteristics, including sample size, sex ratio (male), and median or mean age; (3) intervention details, type of blood purification, like CRRT and ECMO, conventional anticoagulants used, and primary outcome; and (4) clinical outcomes, including the filter lifespan, incidence of bleeding and clotting events, mortality and other relevant outcomes. According to the Newcastle − Ottawa Scale (NOS) designed for cohort studies and case − control studies and the Cochrane Handbook designed for RCTs, two researchers (Caihong Liu and Yongxiu Huang) independently assessed the risk of bias. For 14 observational studies, selection (0–4 scores), comparability (0–2 scores), and outcome or exposure (0–3 scores) are the three evaluation domains containing an eight-item checklist. Each study was scored on a 9-point scale, with scores ranging from 6 to 9 considered indicative of high methodological quality. Randomized studies were appraised using the Cochrane Collaboration Risk of Bias Tool. The following characteristics will be evaluated: (1) sequence generation; (2) allocation concealment; (3) blinding; (4) incomplete outcome data; (5) selective outcome reporting; and (6) other potential threats to validity. Disagreements were resolved through discussion with a third reviewer (Wei Wei). Statistical analysis In this meta-analysis, estimated effects were reported as risk ratios (RRs) with 95% confidence intervals (95% CIs) for dichotomous outcomes and mean differences (MDs) with 95% CIs for continuous outcomes. Between-study heterogeneity was assessed for each pooled summary estimate using Cochran’s Q statistic and the I2 statistic, with I2 > 50%, P < 0.05 indicating potential heterogeneity. A random-effects model was applied to pool the results across the studies of statistical heterogeneity, while the fixed-effect model was employed for a lower level of heterogeneity. To assess publication bias, funnel plot asymmetry was visually inspected and statistically evaluated using Begg’s and Egger’s tests. Sensitivity analyses were performed by sequentially omitting individual studies to examine the stability and robustness of the pooled results. All statistical analyses were performed using Stata Statistical Software (Version 16.0) and Python (Version 3.12.3). A two-tailed P < 0.05 was considered to be statistically significant. Confidence in evidence The quality of evidence was graded using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach, considering the methodological quality of studies, total sample size, and consistency of results. The quality of evidence was rated as either high, moderate, low, or very low. Result Study selection Totally, we extracted 1125 studies, with 325 from PubMed, 451 from Web of Science, 11 from Cochrane, and 338 from Embase. After removing 537 duplicates, 588 studies remained for title and abstract screening. Of these, 263 studies were excluded for being clearly irrelevant. Then, 103 case reports, 66 reviews, 35 animal-based studies, 16 focusing on pediatrics and pregnancy, 5 non-English publications, and 35 identified as abstract, protocol, letter, guideline, or patents were deleted. Then, 65 full texts were scrutinized for inclusion, and 49 were excluded for irrelevance and insufficient data. Finally, 16 literatures were carried out in this systematic review and meta-analysis (Fig. 2) [14–29]. Fig. 2Flowchart of the selection process Flowchart of the selection process Study characteristics The detailed characteristics of the 16 included studies are summarized in Table 1. A total of 4349 participants undergoing different modalities of blood purification were investigated in our systematic review and meta-analysis. Notably, these investigations were mainly conducted in Korea, China, and Japan, with the earliest study dating back to 1985. Most of them were single-center, retrospective observational studies, with only two RCTs and one prospective observational study. The sample size varied between 43 and 1216, and the median age of recorded patients ranged from a minimum of 49.7 years to a maximum of 77.5 years. CRRT was used as the most common modality for blood purification in 10 studies, and AKI was the primary indication for blood purification. In comparison with NM, UFH was reported by 10 studies, both citrate and NA by 4 studies, LMWH by 2 studies, and systematic UFH (UFH-S) by only one study (received continuous infusion of heparin intravenously due to acute coronary syndrome or cerebral infarction). Filter life was the core consideration in the majority of these studies, followed by bleeding and mortality. Table 1The characteristics of included studiesStudyStudy designRegionStudy periodSample sizeAge, yearsMale, %Blood purification indicationBlood purification modalityConventional anticoagulantPrimary outcome Han, 2011single-center, retrospective cohort studyKoreaJanuary 2006 to December 200973 patients55.572.1acute cardiac or respiratory failureECMOUFHduration of ECMO use Lee, 2014single-center RCTKoreaSeptember 2007 to August 201073 patients55.360.3NRCRRTNAmortality Lim, 2016single-center, retrospective cohort studyKoreaJanuary 2005 to November 2014320 patients58.463.1NRECMOUFHthromboembolic or bleeding complications Makino, 2016single-center, retrospective observational studyJapanJanuary 2011to December 2013101 patients77.570.3NRCRRTUFHincidence of bleeding complications Liu, 2024single-center, retrospective observational studyChinaJanuary 2020 to August 202175 patients67.153.8ESRDhemodialysisUFHcoagulation event Choi, 2015single-center RCTKoreaJuly 2010 to June 201360 patients61.665.5AKICRRTNAfilter life Hwang, 2013single-center, retrospective cohort studyKoreaJanuary 2004 to December 2008222 patients, 1236 treatments67.355.4AKI or other diseasesCRRTUFH, UFH-S, NAfilter life Kamijo, 2020multi-center, retrospective cohort studyJapanJanuary 2011 and December 20131216 patients70.061.7severe sepsis or septic shockblood purificationconventional treatmenthospital mortality Li, 2025single-center, retrospective cohort studyChinaJanuary 2022 to December 2023196 patients, 651 treatments64.865.8NRdialysiscitratemajor clotting Liu, 2024single-center, retrospective cohort studyChinaJanuary 2021 to December 202246 patients, 172 treatments68.567.4NRCRRTcitrateduration of CRRT Kim, 2024single-center, retrospective cohort studyKoreaMarch 2013 to December 2022490 patients63.265.8ESRDdialysisNAmortality Kameda, 2023single-center, retrospective observational studyJapanSeptember 2013 to December 2020286 patients68.074.8NRCRRTUFHfilter life Zhan, 2025single-center, retrospective observational studyChinaMarch 2022 to January 202480 patients58.967.5AKICRRTUFHduration of the first filter use Yue, 2023single-center, retrospective observational studyChinaMay 2015 to August 202262 patients, 501 treatments49.791.9severe burnCRRTUFH, citratefilter life Ohtake, 1991single-center, retrospective cohort studyJapanOctober 1985 to July 199043 patientsNRNRanuriaCRRTUFH, LMWHbleeding incidence Uchino, 2007multi-center, prospective observational studyTwenty-three countriesSeptember 2000 to December 20011006 patients66.065.8ARFCRRTUFH, citrate, LMWHincidence of bleeding complicationsECMO, extracorporeal membrane oxygenation; UFH, unfractionated heparin; RCT, randomized controlled trial; NR, not reported; CRRT, continuous renal replacement therapy; NA, non-anticoagulant; ESRD, end-stage renal disease; AKI, acute kidney injury; UFH-S, systematic UFH; LMWH, low-molecular-weight heparin; ARF, acute renal failure Table 1The characteristics of included studiesStudyStudy designRegionStudy periodSample sizeAge, yearsMale, %Blood purification indicationBlood purification modalityConventional anticoagulantPrimary outcome Han, 2011single-center, retrospective cohort studyKoreaJanuary 2006 to December 200973 patients55.572.1acute cardiac or respiratory failureECMOUFHduration of ECMO use Lee, 2014single-center RCTKoreaSeptember 2007 to August 201073 patients55.360.3NRCRRTNAmortality Lim, 2016single-center, retrospective cohort studyKoreaJanuary 2005 to November 2014320 patients58.463.1NRECMOUFHthromboembolic or bleeding complications Makino, 2016single-center, retrospective observational studyJapanJanuary 2011to December 2013101 patients77.570.3NRCRRTUFHincidence of bleeding complications Liu, 2024single-center, retrospective observational studyChinaJanuary 2020 to August 202175 patients67.153.8ESRDhemodialysisUFHcoagulation event Choi, 2015single-center RCTKoreaJuly 2010 to June 201360 patients61.665.5AKICRRTNAfilter life Hwang, 2013single-center, retrospective cohort studyKoreaJanuary 2004 to December 2008222 patients, 1236 treatments67.355.4AKI or other diseasesCRRTUFH, UFH-S, NAfilter life Kamijo, 2020multi-center, retrospective cohort studyJapanJanuary 2011 and December 20131216 patients70.061.7severe sepsis or septic shockblood purificationconventional treatmenthospital mortality Li, 2025single-center, retrospective cohort studyChinaJanuary 2022 to December 2023196 patients, 651 treatments64.865.8NRdialysiscitratemajor clotting Liu, 2024single-center, retrospective cohort studyChinaJanuary 2021 to December 202246 patients, 172 treatments68.567.4NRCRRTcitrateduration of CRRT Kim, 2024single-center, retrospective cohort studyKoreaMarch 2013 to December 2022490 patients63.265.8ESRDdialysisNAmortality Kameda, 2023single-center, retrospective observational studyJapanSeptember 2013 to December 2020286 patients68.074.8NRCRRTUFHfilter life Zhan, 2025single-center, retrospective observational studyChinaMarch 2022 to January 202480 patients58.967.5AKICRRTUFHduration of the first filter use Yue, 2023single-center, retrospective observational studyChinaMay 2015 to August 202262 patients, 501 treatments49.791.9severe burnCRRTUFH, citratefilter life Ohtake, 1991single-center, retrospective cohort studyJapanOctober 1985 to July 199043 patientsNRNRanuriaCRRTUFH, LMWHbleeding incidence Uchino, 2007multi-center, prospective observational studyTwenty-three countriesSeptember 2000 to December 20011006 patients66.065.8ARFCRRTUFH, citrate, LMWHincidence of bleeding complicationsECMO, extracorporeal membrane oxygenation; UFH, unfractionated heparin; RCT, randomized controlled trial; NR, not reported; CRRT, continuous renal replacement therapy; NA, non-anticoagulant; ESRD, end-stage renal disease; AKI, acute kidney injury; UFH-S, systematic UFH; LMWH, low-molecular-weight heparin; ARF, acute renal failure The characteristics of included studies ECMO, extracorporeal membrane oxygenation; UFH, unfractionated heparin; RCT, randomized controlled trial; NR, not reported; CRRT, continuous renal replacement therapy; NA, non-anticoagulant; ESRD, end-stage renal disease; AKI, acute kidney injury; UFH-S, systematic UFH; LMWH, low-molecular-weight heparin; ARF, acute renal failure Bleeding complication A total of 10 studies reported bleeding events related to anticoagulation during blood purification. Overall, NM was associated with a significantly lower risk of bleeding compared with conventional anticoagulants. Pooled analysis using a random-effects model due to moderate heterogeneity (I² = 61%) showed an RR of 0.53 (95% CI: 0.29–0.97, P = 0.040, n = 13) (Fig. 3A). To further explore this effect, a subgroup analysis was conducted based on the type of anticoagulant comparator, identifying five subgroups: UFH, UFH-S, conventional treatment (CT, unspecified anticoagulation strategies), LMWH, and citrate. Among these, the comparison between NM and LMWH demonstrated a statistically significant difference in favor of NM, with a pooled RR of 0.22 (95% CI: 0.06–0.79, P = 0.020, n = 2), and no heterogeneity across studies (I² = 0%, P = 0.614) (Fig. 3B). Other subgroups also seemed to favor NM, although without reaching statistical significance, likely due to smaller sample sizes and wider confidence intervals. Though the studies covered a broad time range, the cumulative meta-analysis demonstrated a stable pooled effect over time. While earlier studies showed wider confidence intervals due to smaller sample sizes, the overall conclusion remained consistent as new evidence was incorporated (Supplementary Fig. 1). Fig. 3Forest plots of the bleeding risk of NM in comparison with conventional treatments: (A) overall analysis compared with conventional anticoagulants, (B) subgroup analysis compared with different types of anticoagulants, (C) compared with NA. NM, nafamostat mesilate; NA, non-anticoagulant; 95% CI, 95% confidence intervals; CT, conventional treatment Forest plots of the bleeding risk of NM in comparison with conventional treatments: (A) overall analysis compared with conventional anticoagulants, (B) subgroup analysis compared with different types of anticoagulants, (C) compared with NA. NM, nafamostat mesilate; NA, non-anticoagulant; 95% CI, 95% confidence intervals; CT, conventional treatment In addition, two studies directly compared NM with NA. The pooled results indicated no statistically significant difference in bleeding risk between the two groups with fixed-effect model (RR = 0.49, 95% CI: 0.18–1.35, P = 0.170, I² = 43.6%, n = 2) (Fig. 3C). Filter life and clotting risk Filter life, defined as the duration from circuit initiation to treatment suspension due to clotting, was reported in 11 studies. Pooled analysis using a random-effects model showed NM didn’t shorten the filter lifespan in comparison with conventional anticoagulants (pooled MD: −1.23 h, 95% CI: −4.27–1.81, P = 0.428, I2 = 87.9%, n = 10) (Fig. 4A). In light of the significant heterogeneity, subgroup analyses by anticoagulant types and extracorporeal modalities (CRRT and ECMO) were conducted. Compared with filters with any type of anticoagulant, filters in the NM group reached similar survival time (Fig. 4B). No difference regarding filter/circuit life between the two groups was observed across CRRT and ECMO subgroups (Supplementary Fig. 2). Importantly, in comparisons with the NA strategy, NM demonstrated a longer filter lifespan (pooled MD: 4.84 h, 95% CI: 0.18–9.51, P = 0.042; I² = 49%, n=4) (Fig. 4C). Fig. 4Forest plots of the filter life of NM in comparison with conventional treatments: (A) overall analysis compared with conventional anticoagulants, (B) subgroup analysis compared with different types of anticoagulants, (C) compared with NA. NM, nafamostat mesilate; SD, standard deviation; NA, non-anticoagulant; 95% CI, 95% confidence intervals Forest plots of the filter life of NM in comparison with conventional treatments: (A) overall analysis compared with conventional anticoagulants, (B) subgroup analysis compared with different types of anticoagulants, (C) compared with NA. NM, nafamostat mesilate; SD, standard deviation; NA, non-anticoagulant; 95% CI, 95% confidence intervals Given the limited number of studies reporting clotting outcomes, only a qualitative synthesis was conducted (Supplementary Table 1). One study enrolling 320 patients undergoing ECMO found that NM was associated with a significantly lower risk of circuit clotting compared to UFH [19]. Similarly, in a retrospective study by Li et al., NM also showed a significantly reduced clotting risk when compared to RCA (RR = 0.16, 95% CI: 0.05–0.55) [22]. Mortality In-hospital mortality served as a key outcome to assess the clinical impact of anticoagulation strategies during blood purification. Notably, when compared with NA, NM demonstrated a statistically significant reduction in in-hospital death under fixed-effect model, as shown in Fig. 5C (pooled RR: 0.87, 95% CI: 0.77–0.97, P = 0.015, I2 = 28.4%, n = 4). In contrast, mortality rates were comparable between NM and conventional anticoagulants (pooled RR: 0.95, 95% CI: 0.71–1.26, P = 0.704, I2 = 55.6%, n = 5), suggesting no survival advantage in that context, even by subgroup analysis (Fig. 5A and B). Fig. 5Forest plots of the in-hospital mortality of NM in comparison with treatments: (A) overall analysis compared with conventional anticoagulants, (B) subgroup analysis compared with different types of anticoagulants, (C) compared with NA. NM, nafamostat mesilate; NA, non-anticoagulant; 95% CI, 95% confidence intervals Forest plots of the in-hospital mortality of NM in comparison with treatments: (A) overall analysis compared with conventional anticoagulants, (B) subgroup analysis compared with different types of anticoagulants, (C) compared with NA. NM, nafamostat mesilate; NA, non-anticoagulant; 95% CI, 95% confidence intervals Three studies reporting ICU mortality were published within the last five years, yet their findings demonstrated considerable heterogeneity. In two large studies NM was associated with a significantly lower ICU mortality rate compared to conventional treatment and NA [21, 25]. In contrast, when NM was compared directly with UFH, no statistically significant difference was observed [17]. As for 28-day mortality, results remained inconclusive. Two studies conducted in Korea and China, respectively, reported comparable 28-day mortality between NM and UFH, NA groups [18, 26]. Taken together, these findings suggest that NM may confer a survival advantage in ICU settings, particularly compared to NA (Supplementary Table 1). Hospitalization duration Evidence regarding the impact of NM on hospitalization duration remains limited and inconsistent. Among the available studies, findings by Zhan et al. and Kameda et al. indicated a potential trend favoring NM over UFH, with reductions observed in overall hospital and ICU stay, respectively [17, 26]. In contrast, two studies found no significant difference in hospitalization duration and ICU stay between NM and UFH, as well as NA, further highlighting the inconsistency across findings [17, 25]. (Supplementary Table 1). Blood products transfusion Due to the limited number of studies assessing transfusion requirements, a qualitative synthesis was performed (Supplementary Table 2). Three Korean studies (Han et al., Lee et al., and Kim et al.), reported that patients treated with NM required fewer transfused blood products, including packed red blood cells, platelets, and fresh frozen plasma (FFP), compared to those who received UFH or NA [15, 18, 25]. However, a Japanese study (Kameda et al.) reported a higher volume of transfused blood products within 48 h in the NM group compared to the UFH group, indicating potential inconsistencies across studies and settings [17]. Adverse events In this meta-analysis, seven studies paid attention to adverse events, which were only recorded by two retrospective studies [14, 18, 22–26]. Interestingly, Kim et al. reported high occurrence of agranulocytosis (35% vs. 34.1%, P = 0.85), neutropenia (84.5% vs. 86.3%, P = 0.60), and severe neutropenia (59.5% vs. 63.9%, P = 0.36) during CRRT in both the NM and the NA group [25]. However, given the limited sample size and retrospective nature of this study, the high incidences of these adverse events warrant verification and may not be attributed to NM, as they were comparable between the 2 groups. In Zhan’s study, 2 (5%) patients complained of digestive discomfort like diarrhea and nausea, and hyperkalemia was documented in 1 patient (2.5%) [26]. Quality evaluation The quality assessment for two RCTs was shown in Supplementary Figs. 3 and 4. Neither RCT described allocation concealment. Blinding was not achieved for the overt differences between interventions, thus introducing potential bias and being evaluated as “High risk”. For the remaining 14 observational studies, except for two retrospective studies grading 5, the others were classified as high quality. Among the three domains, comparability was the weakest-performing domain, often due to inadequate control for confounding variables. Detailed information is illustrated in Supplementary Table 3. Sensitivity analysis and publication bias Since the availability of a sufficient number of studies evaluating filter lifespan, bleeding complications, and in-hospital mortality, we conducted sensitivity analyses and assessed publication bias to evaluate the robustness and reliability of their pooled estimates. We conducted leave-one-out sensitivity analyses by omitting 1 study at a time and recalculating the estimates on the remaining studies. The overall pooled effect estimates for bleeding complications, filter life, and in-hospital mortality remained stable across iterations, indicating that no single study exerted a disproportionate influence on the overall outcomes (Fig. 6A-C). As for the publication bias analyses, visual inspection of the funnel plots showed symmetrical distribution of studies around the vertical line representing the pooled effect size, suggesting a low likelihood of publication bias (Fig. 7A-C). To further validate this, we conducted Egger’s and Begg’s statistical tests. No evidence of substantial publication bias existed in this meta-analysis on filter life (Egger test, P = 0.901; Begg test, P = 0.743), bleeding complications (Egger test, P = 0.070; Begg test, P = 0.093) or in-hospital mortality (Egger test, P = 0.367; Begg test, P = 0.348) (Supplementary Table 4). Fig. 6Sensitivity analysis on results stability regarding bleeding risk, filter life, and in-hospital mortality (A) bleeding risk, (B) filter life, (C) in-hospital mortality Sensitivity analysis on results stability regarding bleeding risk, filter life, and in-hospital mortality (A) bleeding risk, (B) filter life, (C) in-hospital mortality Fig. 7Funnel plots of bleeding risk, filter life, and in-hospital mortality (A) bleeding risk, (B) filter life, (C) in-hospital mortality Funnel plots of bleeding risk, filter life, and in-hospital mortality (A) bleeding risk, (B) filter life, (C) in-hospital mortality Quality of evidence assessment (GRADE) Owing to observation studies accounting for the majority and poor quality of RCTs, the initial rating for most outcomes started at the “Low” level. The overall quality of evidence ranged from “Low” to “Very low”, as outlined in Supplementary Table 4. Five outcomes were downgraded one level for substantial heterogeneity (I² ≥ 50%). Apart from that, the insignificant result also contributes to the low quality (Supplementary Table 5). These findings suggest that the certainty of evidence is limited, and future high-quality randomized trials may significantly alter the current conclusions. Discussion This meta-analysis synthesizes evidence from 16 studies involving 4,349 patients and strengthens the evidence that NM is both safe and effective as an anticoagulant during blood purification. With respect to safety, NM was associated with a significantly lower bleeding risk than conventional anticoagulants, most notably compared with LMWH, while achieving comparable circuit survival. Importantly, when compared to NA, NM achieved a similar incidence of bleeding events but demonstrated a longer filter lifespan, thereby overcoming one of the most recognized limitations of the NA strategy. From an efficacy standpoint, NM was associated with reduced in-hospital mortality compared with NA, while its survival outcomes were broadly comparable to conventional anticoagulants. Collectively, these results highlight NM as an attractive alternative anticoagulant, particularly in patients at heightened risk of bleeding, where therapeutic options are otherwise limited. The choice of anticoagulation strategy is pivotal in blood purification [30]. Blood contact with artificial extracorporeal surfaces triggers activation of clotting factors and platelets, leading to circuit thrombotic occlusion. Premature filter failure not only compromises solute clearance and fluid removal but also increases blood loss, nursing workload, and healthcare costs [31]. UFH and LMWH, while widely available, confer a substantial bleeding risk that may be prohibitive in high-risk populations [32]. Moreover, most critically ill patients have already been burdened with a high hemorrhagic risk. For them, an anticoagulant-related bleeding event is definitely an overwhelming attack. Therefore, a balanced consideration of maintaining an unobstructive circuit and minimizing bleeding complications during blood purification is challenging and significant. The 2012 Kidney Disease Improving Global Outcome (KDIGO) Clinical Practice Guidelines recommended RCA as the first-line choice for AKI patients receiving CRRT due to its outperformance in lower bleeding risk and longer filter lifespan [2, 30]. Nevertheless, RCA has limitations, including the need for specialized protocols, risks of hypocalcemia and metabolic derangements, and restricted applicability in patients with tissue hypoperfusion and severe hepatic dysfunction [3]. Unlike RCA, only a small proportion of NM enters the body from blood purification. While a high-dose iv infusion of NM can indeed produce a systemic anticoagulant effect, NM is more characterized as a regional anticoagulant during blood purification. This is achieved by infusing a moderate-to-low dose into the arterial line, leveraging NM’s rapid degradation, short half-life, and minimal systemic effect to confine the anticoagulation primarily to the extracorporeal circuit. After being administered in arterial access, NM begins to degrade immediately in the blood and is further removed through diffusion, convection, and adsorption when passing the filter (approximately 40%). Consequently, only a small proportion (less than 4%) enters the systemic circulation and is rapidly degraded in the liver (about 80%) and blood (mainly by erythrocytes, about 20%), resulting in a short plasma half-life of approximately 5 to 8 min (Fig. 1) [12, 33–36]. According to the recent Chinese expert consensus, NM is recommended for patients with hepatic failure, particularly when RCA is contraindicated. Although the half-life of NM may be modestly prolonged in this setting, effective anticoagulation can be maintained with conservative dose adjustments and close clinical monitoring [37]. In the included studies, patients with liver dysfunction were also represented, and no remarkable NM-related adverse events were reported. Compared with CRRT, removal of NM by the extracorporeal circuit may be lower during ECMO, raising theoretical concerns about prolonged systemic exposure. However, available pharmacokinetic data indicate that NM remains a short-acting anticoagulant in ECMO, with rapid degradation primarily mediated by plasma esterases and hepatic metabolism [11, 38]. Yet, dedicated comparative pharmacokinetic studies between CRRT and ECMO remain limited. Due to the unique pharmacodynamic characteristics, NM demonstrates a favorable bleeding safety profile comparable to NA while preserving filter function like conventional anticoagulants. This offers a compelling rationale for its use, effectively bridging the safety-efficacy gap between the two strategies. NM is removed within minutes after NM discontinuation, minimizing residual systemic influence and facilitating safe, rapid adjustments during CRRT [14]. Our pooled analysis supports this mechanistic rationale: NM reduced bleeding risk by about half compared with conventional anticoagulants (RR 0.53), in line with the results of lower transfusion requirements in NM-treated patients. Although variation in transfusion thresholds across centers may affect absolute rates, the consistently favorable direction of effect strengthens the overall evidence. Interestingly, although LMWH is considered safer than UFH due to its more predictable pharmacokinetics and reduced nonspecific protein binding [39, 40], our results showed that NM still conferred a lower bleeding risk even against LMWH. This observation is consistent with a prior meta-analysis in hemodialysis patients reporting significantly higher bleeding rates with LMWH than with NM (RR: 4.58, 95% CI: 1.34–15.91, P = 0.02) [41]. Such findings challenge the perception that LMWH is the “safer” heparin-based option, particularly as LMWH is also associated with HIT, which further predisposes patients to hemorrhagic complications [42]. However, a notable proportion of the studies in our meta-analysis were conducted before RCA was widely adopted, so the number of direct NM–RCA comparisons is small and the available data are heterogeneous. A large multinational study reported similar bleeding rates between NM and RCA, whereas two recent Chinese cohorts yielded inconsistent findings for filter lifespan, and comparative mortality data are lacking [23, 27, 29]. Taken together, conclusions regarding the relative performance of NM versus RCA should be interpreted with caution, warranting adequately powered studies comparing NM and RCA in the future. With regard to ECMO and CRRT subgroups, though they represent distinct anticoagulation strategies and clinical implications, the efficacy of NM in maintaining circuit patency is comparable to that of conventional anticoagulants. However, as only two ECMO studies were available, the power of this subgroup analysis is limited. Anticoagulation-free strategies (NA) remain a pragmatic choice for patients with very high bleeding risk, as they effectively reduce hemorrhagic complications. However, this approach is consistently hampered by shortened filter lifespan, often limited to 12–32 h [5, 43–45]. Our meta-analysis demonstrated that NM achieved a comparable bleeding risk to NA, thereby preserving the safety advantage of NA therapy. These findings are in agreement with a prior meta-analysis that reported no excess bleeding with NM compared to NA (RR: 1.03, 95%CI: 0.41–2.56, P = 0.95) with a considerable studies included [41]. In this study, we revealed that NM extended the pooled filter lifespan of 4.84 h than NA and was comparable with conventional anticoagulants. This was more pronounced in another meta-analysis, albeit based on fewer studies (MD: 10.59 h, 95%CI: 5.72–15.45, P < 0.001) [41]. Beyond circuit outcomes, NM also conferred a significant reduction in in-hospital mortality compared with NA. This survival advantage reinforces the clinical significance of improved circuit patency, as prolonged treatment continuity translates into better solute clearance, fluid balance, and overall therapy delivery. A prior meta-analysis also revealed a borderline significant lower risk of in-hospital mortality when compared with NA under random-effect model (RR: 0.79, 95%CI: 0.63–1.00, P = 0.05) [41]. Although our analysis did not identify statistically significant differences in ICU mortality or hospitalization duration, the consistent favorable trends suggest that NM may positively influence broader patient outcomes. Future large-scale, well-designed RCTs are needed to substantiate these findings and clarify their implications for patient-centered endpoints. There are also some matters that need attention when using NM. First, given that NM has a relatively short clinical history, with only several decades in East Asia (since 1986 in Japan and 2020 in China), even older, retrospective, and single-center studies represent valuable sources of safety and efficacy data. Furthermore, considering the significant geographical variation in blood purification practices, we included all available studies to provide a representative and thorough evaluation of NM’s clinical effectiveness. Although the cumulative meta-analysis showed a stable pooled effect over time, the inclusion of older studies may still introduce temporal heterogeneity and limit direct applicability to current CRRT practice. Future large, preferably multicenter studies are needed to confirm these findings. Of note, the efficacy of NM in western populations remains largely unexplored, restricting the generalizability of our conclusions. Second, NM has generally been continuously administered at approximately 20 to 40 mg/h, with dose titration guided by coagulation monitoring. Early protocols relied mainly on activated clotting time (ACT) or activated partial thromboplastin time. But the limitations of ACT have also been clarified, such as underestimation due to adsorption of NM by celite/kaolin reagents and marked site dependence of measurements [46, 47]. A universally accepted standard for monitoring NM has not yet been established, and some heterogeneity in dose and monitoring practice over time and across regions is therefore inevitable. Third, although NM is generally considered as a safe anticoagulation option for blood purification, caution is advised in patients with active bleeding or severe hepatic dysfunction. The upgrade of NM dose should be cautious in hepatic failure patients because the half-life may be slightly extended in them. For patients with life-threatening bleeding in vital organs, NA may be a safer preference [37]. Of note, the antioxidant effects of NM may inhibit linezolid metabolism if administered concurrently during RRT, as revealed in an animal study [48]. Adverse events associated with NM have been reported at a very low incidence, including agranulocytosis, hyperkalemia, and anaphylactoid reactions [49–51], and no specific antidote for NM is currently available. In our systematic review, only two studies documented agranulocytosis, digestive discomfort, and hyperkalemia. Notably, one report described a case of cardiopulmonary arrest following NM administration that resolved immediately after the drug was discontinued. Although this event was considered possibly due to NM-induced anaphylaxis, both the drug lymphocyte stimulation test and basophil activation test were negative, making the diagnosis less convincing [52]. In addition, one study from China in 2024 reported the health economic issue in comparison with heparin and sodium citrate. Taking anticoagulant price and filter life into consideration, the total costs involving circuit replacement and monitoring were ¥3,076.92 (about $439.56) for heparin, ¥3460 (about $494.29) for citrate, and ¥5,493.6 (about $784.8) for NM, respectively [27]. However, because the cost of NM varies substantially across countries and regions due to differences in pricing, reimbursement policies, and healthcare resource use, multicenter data were required to enable a more comprehensive and generalizable economic analysis. Conclusion This systematic review and meta-analysis demonstrate that NM is a safe and effective anticoagulant option for blood purification. Compared with conventional anticoagulants, blood purification with NM significantly reduced bleeding complications, while maintaining comparable filter survival. When contrasted with non-anticoagulation strategy, NM achieved a similar bleeding profile but prolonged filter lifespan and was associated with lower in-hospital mortality. Large, multicenter RCTs are warranted to validate these results and to determine the role of NM in diverse blood purification settings worldwide. Supplementary Information Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2