Prolonged Thromboelastography R Time and Thrombocytopenia as Early Warning Indicators of Bleeding in ECMO Patients Who Developed Sepsis During Support: A Retrospective Cohort Study

Introduction

Extracorporeal membrane oxygenation (ECMO) serves as a salvage therapy for refractory respiratory or circulatory failure. Infection is a common complication during ECMO support,^1,2 which can severely lead to sepsis. Sepsis can not only cause organ dysfunction³ but also lead to complex coagulopathies, even triggering disseminated intravascular coagulation (DIC).⁴ On the other hand, systemic anticoagulation (eg, with heparin) is required during ECMO to prevent circuit thrombosis. Despite technical improvements, bleeding remains one of the most common and devastating complications in ECMO patients;^5,6 an analysis of adverse events in 11,984 venoarterial ECMO (VA-ECMO) patients from the ELSO registry showed that bleeding accounted for up to 62.1%.⁷ Another retrospective analysis of 7579 venovenous ECMO (VV-ECMO) patients in the ELSO database also indicated that 23.4% of patients experienced bleeding.⁸

When sepsis complicates ECMO support, patients face a “double hit”: one stemming from ECMO-related coagulation activation and anticoagulation therapy, and the other from sepsis-induced coagulopathy. This complex pathophysiological state makes the management of bleeding risk extremely challenging. Currently, anticoagulation monitoring during ECMO predominantly relies on activated partial thromboplastin time (aPTT) or activated clotting time (ACT),^9,10 with doses typically titrated to maintain levels at 1.5–2 times the normal value and between 180–210 seconds, respectively.¹⁰ However, in septic patients, aPTT and ACT only reflect parts of the coagulation cascade and fail to provide a comprehensive assessment of the overall balance of the coagulation system.¹¹

Thromboelastography (TEG)¹² is a point-of-care viscoelastic coagulation assay that provides a global assessment of the entire haemostatic process, from clot initiation and formation to fibrinolysis, thereby offering a more comprehensive view of coagulation status than conventional tests. A retrospective study of 32 patients with severe respiratory failure receiving ECMO therapy¹³ observed that despite patients’ aPTT and ACT remaining within the therapeutic anticoagulation range, their TEG tracings already displayed a significant heparin effect. Given this finding and considering that excessive anticoagulation is a persistent risk factor for bleeding during ECMO,^14,15 the question of whether a TEG-guided strategy could serve as a safe and feasible alternative to the “conventional” aPTT- or ACT-based monitoring has been an area of active investigation. However, routine anticoagulation titration driven primarily by TEG is not universally recommended in current guidance documents, largely due to limited prospective evidence demonstrating clinical benefit, substantial heterogeneity in viscoelastic testing protocols, and the lack of standardized therapeutic targets across centers.^16,17 Nevertheless, TEG may still be clinically valuable as an adjunct tool because it captures the dynamic, whole-blood evolution of coagulation and fibrinolysis that conventional assays (aPTT/ACT) may not reflect, especially in ECMO patients complicated by sepsis where the hemostatic balance is highly unstable This gap between guideline-level evidence and bedside need provides a strong rationale to explore whether specific TEG patterns, together with platelet dynamics, can serve as early warning indicators of bleeding risk in this high-risk population.

However, the development of such a TEG-guided strategy necessitates a deep understanding of the predictive value and interrelationships of various TEG parameters in complex clinical scenarios. In patients with ECMO complicated by sepsis, accurate prediction of bleeding risk may require integrated assessment of two critical dimensions: coagulation initiation and platelet function. From a pathophysiological perspective, ECMO- and sepsis-related coagulopathy often involves early impairment of clot initiation due to dysregulated coagulation factor activity and anticoagulant exposure, together with platelet consumption and dysfunction driven by systemic inflammation and circuit-related activation. TEG parameters such as R time provide a direct functional readout of the initiation phase of coagulation, whereas platelet count represents a key cellular determinant of effective hemostasis and is frequently one of the earliest components to deteriorate during ECMO support. Although other biomarkers (eg, fibrinogen) are also clinically relevant, their trajectories in sepsis can be highly heterogeneous because fibrinogen may initially increase as an acute-phase reactant before later consumption, making single measurements less informative for early bleeding warning. Therefore, this study aimed to investigate how TEG parameters and platelet count evolve dynamically and work in concert to predict bleeding. We hypothesize that TEG parameters reflecting coagulation initiation and a decreased platelet count may function as complementary early-warning signals. Notably, unlike most prior ECMO studies that evaluated viscoelastic parameters in heterogeneous populations or at isolated time points, our study specifically focuses on the high-risk subgroup of patients who develop sepsis during ECMO support and examines the longitudinal evolution from sepsis onset to the pre-bleeding phase. This dynamic, sepsis-centered perspective provides novel insight into bleeding-specific early warning patterns beyond conventional anticoagulation monitoring. By retrospectively testing this hypothesis, our study seeks to provide a critical theoretical foundation for future development of individualized TEG-based anticoagulation management protocols.

Methods

Ethical

This study was approved by the Ethics Committee of the First Affiliated Hospital, Zhejiang University School of Medicine (Approval No. 2025BIIT0961). As this was a retrospective observational study, the requirement for informed consent was waived by the ethics committee. All patient data were anonymized prior to analysis to ensure confidentiality. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Study Design

Study Participants

This single-center, retrospective study consecutively included patients who received ECMO support in the ICU of the First Affiliated Hospital, Zhejiang University School of Medicine between January 2021 and December 2024, and who developed sepsis during ECMO support. Sepsis was diagnosed according to the Sepsis-3.0 criteria,³ defined as the presence of suspected or confirmed infection, plus an increase in the Sequential Organ Failure Assessment (SOFA) score of ≥2 points from baseline (ΔSOFA≥2).

Inclusion Criteria

1. Aged between 18 and 70 years.
2. ECMO support duration ≥ 24 hours after initiation.
3. For patients who received multiple ECMO runs, only the data from the first ECMO run were analyzed.

Exclusion Criteria

1. Absence of any thromboelastography (TEG) testing during ECMO support.
2. Use of nafamostat mesylate, bivalirudin, or argatroban for anticoagulation for > 72 hours during ECMO support.
3. Presence of severe bleeding prior to ECMO initiation.
4. Pregnancy or lactation.
5. No diagnosis of sepsis during the ECMO support period.
6. Presence of severe bleeding prior to the onset of sepsis.
7. Known congenital or acquired hematologic disorders (eg, hemophilia, immune thrombocytopenia), or diagnosis of heparin-induced thrombocytopenia (HIT) during ECMO support.

Major Bleeding Was Defined per ELSO criteria

1. Overt bleeding with a ≥2 g/dL drop in hemoglobin over 24 hours or the need for RBC transfusion (≥10 mL/kg) ¹⁰;
2. Any intracranial or retroperitoneal hemorrhage, or pulmonary hemorrhage with radiologic evidence of infiltration, or any bleeding necessitating surgical repair.

Grouping

Enrolled patients were divided into two groups based on the occurrence of major bleeding during ECMO support: the bleeding group (patients with major bleeding) and the Non-bleeding group (patients without major bleeding).

Thromboelastography (TEG) Assay

TEG analysis was performed using a TEG 5000 or TEG 6s analyzer (Haemonetics Corporation, USA) according to the manufacturer’s instructions. Fresh whole blood samples were activated with kaolin and analyzed within 4 minutes after blood collection. The following six TEG parameters were recorded:

R-value (Reaction time): representing the latency of initial fibrin formation (clotting factor activity). Reference range: 5–10 min.

K-value (Kinetics time): representing the speed of clot formation. Reference range: 1–3 min.

α-angle (Alpha angle): representing the rapidity of fibrin cross-linking and clot strengthening. Reference range: 53–72°.

MA (Maximum Amplitude): representing the ultimate strength of the clot. Reference range: 50–70 mm.

LY30 (Lysis at 30 minutes): representing the percentage of clot amplitude reduction 30 minutes after MA, indicating fibrinolysis. Reference range: 0–7.5%.

EPL (Estimated Percent Lysis): an estimated percentage of fibrinolysis. Reference range: 0–15%.

Anticoagulation Management During ECMO

In our institution, unfractionated heparin is the standard anticoagulant used during ECMO support unless contraindicated. Heparin was administered as a continuous intravenous infusion and titrated to maintain the activated partial thromboplastin time (aPTT) at approximately 1.5–2.0 times the normal value (target range: 60–80 s).

Anticoagulation intensity was primarily guided by conventional coagulation assays, while thromboelastography (TEG) was performed as an adjunctive tool rather than serving as the sole determinant for heparin dose adjustments. In the event of clinically significant bleeding or suspected over-anticoagulation, temporary dose reduction or discontinuation of heparin was considered at the treating physician’s discretion.

Data Collection

The following data were extracted from the electronic medical records:

Baseline Characteristics

Including demographic information, medical history, and admission details: age, sex, body mass index (BMI), comorbidities (eg, hypertension, diabetes), time of hospital admission, time of ICU admission, time of ECMO initiation, ECMO mode (eg, VA-ECMO, VV-ECMO), time of sepsis diagnosis, use of intra-aortic balloon pump (IABP), and use of antiplatelet medications.

Laboratory Parameters

Time point A (Pre-sepsis): During ECMO support but before the diagnosis of sepsis. Parameters included: thromboelastography (TEG) parameters (R-time, K-time, α-angle, MA, CI, EPL, LY30), hemoglobin (Hb) level, platelet count (PLT), and coagulation profile (prothrombin time [PT], activated partial thromboplastin time [aPTT], thrombin time [TT], international normalized ratio [INR], fibrinogen [Fib], and D-dimer).

Time point B (Post-sepsis): Within 24 hours after the diagnosis of sepsis. The same parameters as Time point A were collected.

Time point C (Pre-bleeding, Bleeding group only): For patients in the bleeding group, the latest available parameters before the bleeding event were collected. In our ICU, coagulation profiles and thromboelastography are routinely monitored on a daily basis in ECMO patients complicated by sepsis, with blood samples consistently collected each morning between 05:00 and 05:30. Accordingly, Time point C was defined as the most recent routinely obtained measurement within 24 hours prior to the onset of major bleeding.

Outcome Measures

The primary outcome was the occurrence of major bleeding during ECMO support. For patients who experienced bleeding, the time and anatomical site of the bleeding event were recorded.

Statistical Analysis

All statistical analyses were performed using SPSS software (Version 22.0; IBM Corp., Armonk, NY, USA). All hypothesis tests were two-sided, and a significance level (α) of 0.05 was adopted. Continuous variables were tested for normality using the Shapiro–Wilk test and are presented as median (interquartile range, IQR) since all were non-normally distributed. Comparisons between groups (Bleeding vs Non-bleeding) were made using the Mann–Whitney U-test. Comparisons of repeated measures within the same group (eg, pre- vs post-sepsis) were analyzed using the Friedman test; if the overall comparison was significant, post hoc pairwise comparisons were conducted. For paired comparisons between two time points (eg, sepsis onset vs pre-bleeding), the Wilcoxon signed-rank test was applied. Categorical variables are presented as number (percentage, %). Group differences in categorical variables were assessed using the Chi-square test or Fisher’s exact test (when more than 20% of expected cell counts were <5). A P-value of < 0.05 was considered statistically significant.

Result

A total of 418 patients received ECMO support during the study period (2021–2024). After initial screening, 127 patients were excluded for not meeting the inclusion criteria, leaving 291 patients who initially met the eligibility requirements. After applying the predefined exclusion criteria, 64 patients were finally enrolled in the study (Figure 1). Baseline characteristics at sepsis onset are presented first, followed by longitudinal changes in coagulation parameters preceding major bleeding. The reasons for exclusion were as follows: 61 patients due to the absence of TEG testing during ECMO support, 25 due to anticoagulation with nafamostat or bivalirudin for >72 hours, 12 due to pre-existing major bleeding before ECMO initiation, 123 due to no diagnosis of sepsis during ECMO, and 6 due to major bleeding occurring prior to the onset of sepsis.

Figure 1 Flowchart.

Table 1 summarizes the baseline characteristics of the 64 enrolled patients at the time of sepsis onset during ECMO support. The median age was 54 (44–60) years, and 81.3% (52/64) were male. VA-ECMO was the predominant mode (62.5%, 40/64). Twenty-three patients (35.9%) were on antiplatelet therapy. Major bleeding occurred in 26 patients (40.6%), with gastrointestinal bleeding (9.4%, 6/64) and intracranial bleeding (7.8%, 5/64) being the most common types. All patients exhibited coagulopathy at sepsis onset, as detailed by TEG and coagulation parameters in Table 1.

Table 1 Baseline and Clinical Characteristics of the Study Cohort (N=64)

As shown in Table 2, no statistically significant differences were found between the Bleeding group (n=26) and the Non-bleeding group (n=38) in baseline parameters at sepsis onset, including demographic data, ECMO mode, use of IABP support, antiplatelet therapy, TEG parameters, coagulation profiles, and platelet counts.

Table 2 Comparison of Baseline Parameters at Sepsis Onset Between Bleeding and Non-Bleeding Groups

Dynamic assessment of the Bleeding group revealed significant changes in TEG parameters and platelet counts from sepsis onset to the pre-bleeding phase (Figure 2 and Table 3). Specifically, the R-time was significantly prolonged in the pre-bleeding phase compared with sepsis onset [10.1 (5.9–29.4) min vs 7.5 (4.0–12.5) min; P=0.03]; in parallel, platelet count was significantly decreased before bleeding events [93.5 (41.5–144.3) ×10⁹/L vs 110.0 (56.5–224.0) ×10⁹/L; P=0.02], highlighting a progressive temporal worsening in coagulation initiation and platelet availability prior to bleeding events. In the non-bleeding group, using a matched observation window (Day 4 after sepsis onset, based on the median 4.3-day interval from sepsis onset to the pre-bleeding sample in the bleeding group), neither R-time (P = 0.547) nor platelet count (P = 0.085) showed a statistically significant longitudinal change. These parallel analyses are provided in Supplementary Table S1.

Table 3 Dynamic Changes in Parameters from Sepsis Onset to Pre-Bleeding Event in the Bleeding Group (n=26)

Figure 2 Dynamic changes in thromboelastography (TEG) R-time and platelet count in patients with bleeding. (A) R-time at sepsis onset and within 24 h prior to the bleeding event. (B) Platelet count at sepsis onset and within 24 h prior to the bleeding event. Data are presented as median (IQR). P<0.05 indicates a significant difference between time points.

Discussion

This study aimed to investigate the factors associated with bleeding in patients who developed sepsis during extracorporeal membrane oxygenation (ECMO) support. Our results showed a bleeding incidence of 40.6%. Notably, patients in the bleeding group exhibited a significant progressive prolongation of the thromboelastography (TEG) R-time and a significant decline in platelet count from sepsis onset to the pre-bleeding phase, compared to their levels at sepsis onset. Although the absolute magnitude of these changes was modest, in the context of ECMO complicated by sepsis, where hemostasis is highly fragile and rapidly evolving, even subtle longitudinal shifts in coagulation initiation and platelet availability may reflect clinically relevant deterioration toward bleeding. Importantly, dynamic trends may provide greater warning value than isolated measurements relative to reference ranges. This combined pattern is mechanistically plausible because R-time primarily reflects the initiation phase of coagulation, driven by clotting factor activity and anticoagulant exposure, whereas platelet count represents a distinct cellular determinant of clot propagation and mechanical stability. In ECMO patients complicated by sepsis, bleeding risk is rarely attributable to a single abnormality but rather to the convergence of delayed clot initiation and impaired platelet-mediated clot strengthening. Therefore, concurrent longitudinal deterioration in these two dimensions may function as complementary early warning indicators of impending major bleeding.

Importantly, these hemostatic changes should be interpreted within the infection-specific framework of sepsis-induced coagulopathy, where pathogen-driven inflammation and dysregulated host responses can rapidly destabilize coagulation balance during ECMO support. The R-time prolongation observed in our study is comparable to findings in previous literature. A study in pediatric ECMO patients¹⁸ reported a median R-time of 9.8 minutes but did not evaluate its association with bleeding outcomes. The range of R-times (7.9–10.8 min) reported by Panigada et al¹³ is similar to our findings; however, that study, which had higher platelet counts (119–121 ×10⁹/L), reported a lower incidence of major bleeding (22%). Furthermore, an RCT¹⁷ that targeted an R-time of 16–24 minutes for anticoagulation found a very high bleeding complication rate (59.5%). Collectively, these findings, along with our results, suggest that dynamic monitoring of R-time and platelet count may aid in the early identification of bleeding risk in septic patients on ECMO. Clinically, such progressive deterioration may warrant closer surveillance and individualized reassessment of anticoagulation intensity and transfusion support, although prospective validation is required before routine implementation.

Although the activated partial thromboplastin time (aPTT) is the conventional monitoring parameter for anticoagulation management during ECMO, it only reflects part of the coagulation cascade and fails to provide a comprehensive assessment of hemostatic status.^17,19–21 We found no significant differences in baseline demographics, ECMO configuration, support devices, medication use, TEG parameters, or conventional coagulation assays between the bleeding and non-bleeding groups. Notably, even within the bleeding group, traditional coagulation parameters before bleeding did not differ significantly from their baseline values at sepsis onset. This indicates that the occurrence of bleeding may be more closely related to the dynamic progression of coagulopathy rather than the initial state. The significant prolongation of R-time prior to bleeding suggests that TEG may be more sensitive than conventional tests in detecting this evolving hemostatic abnormality.

Several considerations are important when interpreting these findings. To clarify whether the observed temporal deterioration in R-time and platelet count represents a bleeding-specific signal rather than a common trajectory during ECMO support, parallel longitudinal analyses were also conducted in the non-bleeding cohort and are provided in the Supplementary Materials (Supplementary Table S1). In brief, no significant longitudinal change was observed in non-bleeding patients for R-time (P = 0.547) or platelet count (P = 0.085) over the matched Day-4 window. Although the absolute magnitude of R-time prolongation and platelet decline was modest, the hemostatic balance in ECMO patients complicated by sepsis is particularly vulnerable, and even borderline dynamic shifts may precede clinically significant bleeding. From a practical perspective, these parameters should not be viewed as definitive treatment thresholds, but rather as early warning indicators that may prompt intensified surveillance, repeat viscoelastic testing, and individualized reassessment of anticoagulation intensity and supportive transfusion strategies. Prospective studies are needed to determine whether incorporating such dynamic monitoring into anticoagulation management protocols can reduce bleeding complications.

Our study has several limitations. First, this was a single-center, retrospective study with a limited sample size, which may introduce selection bias and leaves residual confounding (eg, individual variations in anticoagulant dosage) uncontrolled. Given the retrospective design and the lack of multivariable adjustment, our findings should be interpreted as associative and hypothesis-generating rather than conclusive evidence of independent prediction. Second, the timing of TEG measurements was not standardized, and the observed dynamic changes might be influenced by monitoring frequency. Future large-scale, prospective studies are warranted to validate our findings and to further investigate whether a TEG-guided individualized anticoagulation strategy can effectively reduce bleeding complications. In addition, bleeding events may also occur in ECMO patients without sepsis, and future studies should compare septic versus non-septic bleeding phenotypes to clarify sepsis-specific mechanisms.

Conclusion

In patients developing sepsis during ECMO support, bleeding events are preceded by a dynamic prolongation of TEG R-time and a progressive decrease in platelet count. Dynamic combined monitoring of TEG parameters (especially R-time) and platelet count may provide early warning indicators associated with major bleeding risk. However, these findings should be interpreted as early warning indicators rather than definitive independent predictors, intended to support risk stratification and enhanced surveillance, and should be considered hypothesis-generating, requiring confirmation in larger prospective studies before routine clinical implementation.