Abstract
Background
Restrictive red blood cell (RBC) transfusion strategies are increasingly used during venovenous extracorporeal membrane oxygenation (VV ECMO), but the prognostic implications of sustained exposure to low hemoglobin concentrations remain uncertain. We investigated the associations of longitudinal hemoglobin exposure with 60-day mortality and successful ECMO weaning in patients receiving VV ECMO.
Methods
We retrospectively analyzed adults receiving VV ECMO for at least 48 hours at a single center between January 2015 and May 2024. Patients who were pregnant or had received ECMO for >3 days at an outside hospital before transfer were excluded. Time-weighted average hemoglobin (TWA-Hb) was calculated from serial venous complete blood count measurements by weighting each hemoglobin value according to the time interval represented by that measurement. Patients were categorized as <70, 70–<80, 80–<90, and ≥90 g/L. The proportion of ECMO support time with hemoglobin <70 g/L was calculated as an exploratory measure of cumulative severe-anemia exposure. The primary outcome was 60-day mortality.
Results
Among 182 patients, the median age was 59 years, 74.2% were male, and 97.3% received RBC transfusions (median 20 units). Seventy-five patients (41.21%) died within 60 days. Compared with the ≥90 g/L group, TWA-Hb <70 g/L was associated with a higher hazard of 60-day mortality [adjusted hazard ratio (HR) 3.47, 95% confidence interval (CI): 1.00–12.03, P=0.049]. In adjusted logistic regression, the odds ratios (ORs) for successful ECMO weaning within 60 days were 0.13 (95% CI: 0.04–0.38, P<0.001) for the <70 g/L group and 0.26 (95% CI: 0.09–0.72, P=0.01) for the 70–<80 g/L group, compared with the ≥90 g/L group.
Conclusions
TWA-Hb <70 g/L during VV ECMO was associated with higher 60-day mortality and a lower likelihood of successful ECMO weaning. Sustained exposure to severe anemia below 70 g/L warrants particular clinical attention and should be minimized when feasible. These findings support further prospective evaluation of hemoglobin-management strategies aimed at reducing prolonged severe-anemia exposure during VV ECMO.
Keywords: Venovenous extracorporeal membrane oxygenation (VV ECMO), hemoglobin, anemia, transfusion, prognosis
Highlight box.
Key findings
• Time-weighted average hemoglobin (TWA-Hb) <70 g/L during venovenous extracorporeal membrane oxygenation (VV ECMO) was associated with higher 60-day mortality and a lower likelihood of successful ECMO weaning.
What is known and what is new?
• Restrictive transfusion strategies are increasingly used in VV ECMO, but the balance between anemia tolerance and transfusion-related risk remains uncertain. Most previous studies have focused on transfusion thresholds or isolated hemoglobin measurements rather than longitudinal hemoglobin exposure.
• This study evaluated longitudinal hemoglobin exposure using TWA-Hb and the proportion of ECMO support time with hemoglobin <70 g/L, thereby incorporating both hemoglobin concentration and duration of exposure. The findings suggest that very low longitudinal hemoglobin exposure may identify patients at increased risk, while the continuous analyses did not support a uniform dose-response relationship across the entire TWA-Hb range.
What is the implication, and what should change now?
• Sustained exposure to severe anemia below 70 g/L warrants particular clinical attention and should be minimized when feasible during VV ECMO. These findings support prospective evaluation of hemoglobin-management strategies aimed at reducing prolonged severe-anemia exposure and improving clinical outcomes.
Introduction
Venovenous extracorporeal membrane oxygenation (VV ECMO) is an advanced life-support intervention for patients with severe acute respiratory distress syndrome (ARDS) when conventional treatments such as lung-protective ventilation, prone positioning, and optimized mechanical ventilation fail to ensure adequate gas exchange (1). While VV ECMO can provide effective extracorporeal gas exchange, it also introduces distinct physiological and technical challenges.
One of the most frequent complications during ECMO is anemia, caused by hemolysis, anticoagulation-related bleeding, cannulation trauma, and frequent blood sampling (2). As hemoglobin is a major determinant of arterial oxygen content, reductions in hemoglobin may substantially decrease oxygen-carrying capacity and contribute to impaired systemic oxygen delivery (DO2), which is jointly determined by hemoglobin, arterial oxygen saturation, and cardiac output. In clinical practice, more than 90% of VV ECMO patients receive at least one red blood cell (RBC) transfusion to maintain systemic DO2 (3). While transfusions may improve DO2, they are also associated with complications such as transfusion-related immunomodulation, fluid overload, and increased risk of infection (4).
From a physiological perspective, oxygen consumption (VO2) remains stable despite a reduction in DO2 until a critical threshold is crossed; beyond this point, VO2 declines and anaerobic metabolism develops (5). This principle has encouraged the adoption of restrictive transfusion strategies, aiming to limit unnecessary blood exposure while maintaining adequate DO2 (6). Growing evidence supports the safety of restrictive transfusion strategies in VV ECMO. For example, Martucci et al. (7) reported that mortality increased only when hemoglobin concentrations fell below 70 g/L, with transfusion-associated survival benefit observed only below this threshold. Similarly, Pratt et al. (8) found that the lowest mortality in VV ECMO patients occurred when transfusion thresholds ranged from 70 to 80 g/L.
Global blood product shortages since 2019 have further pressured clinicians to implement more conservative transfusion strategies (9). However, the lower safety limit for hemoglobin concentrations in VV ECMO patients remains unclear. Most existing studies have relied on single-point hemoglobin values rather than integrating measurements over time, which may underestimate the clinical impact of prolonged exposure to low hemoglobin levels.
To address this gap, we conducted a retrospective cohort study using time-weighted average hemoglobin (TWA-Hb) to reflect cumulative anemia burden in VV ECMO patients. Unlike an isolated or nadir value, a time-weighted measure incorporates both hemoglobin concentration and the interval between measurements. We additionally quantified the proportion of ECMO support time with hemoglobin <70 g/L as an exploratory measure of cumulative severe-anemia exposure. We aimed to examine the associations of these longitudinal hemoglobin metrics with 60-day mortality and successful ECMO weaning. Given the observational design, these analyses were intended to characterize prognostic associations rather than to define a causal transfusion threshold. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1038/rc).
Methods
Study design and setting
This retrospective cohort study was conducted in the Department of Intensive Care Medicine at The First Affiliated Hospital of Guangzhou Medical University. All eligible patients who received VV ECMO support between January 1, 2015, and May 31, 2024, were included. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The First Affiliated Hospital of Guangzhou Medical University (No. ES-2024-K089-01); informed consent for this retrospective analysis was waived.
Inclusion and exclusion criteria
Eligible patients were adults (≥18 years) who received VV ECMO support for at least 48 hours. For patients who underwent multiple ECMO episodes, only the first episode was included. Exclusion criteria included pregnancy and ECMO support exceeding 3 days at an external hospital prior to transfer (Figure 1).
Figure 1.
Flowchart of the study population. Patients were grouped according to their time-weighted average hemoglobin concentrations. VV ECMO, venovenous extracorporeal membrane oxygenation.
Data collection
We extracted baseline data from electronic medical records, including demographic information, comorbidities, ECMO indications, pre-ECMO laboratory values (hemoglobin, creatinine, blood gas), and ECMO-related interventions. During ECMO, we recorded hemoglobin concentrations, central venous oxygen saturation (ScvO2), transfusion volumes, and major complications including bleeding, venous thrombosis, and acute kidney injury (AKI). Bleeding was defined according to the 2014 Extracorporeal Life Support Organization (ELSO) anticoagulation guidelines (10). Venous thrombosis was diagnosed using Doppler ultrasonography, and AKI was defined according to the 2012 KDIGO criteria.
RBC transfusion protocol
During the study period, RBC transfusion decisions were made by the attending intensivists based on a combination of hemoglobin concentration, hemodynamic status, DO2 indicators, and clinical judgment. In general, a restrictive transfusion strategy was adopted, with a hemoglobin threshold of 70 g/L commonly used as a trigger for transfusion in stable patients. However, transfusion could be considered at higher hemoglobin levels in the presence of active bleeding, ongoing hemolysis, hemodynamic instability, or signs of inadequate tissue oxygenation (e.g., elevated lactate, low ScvO2). The final decision was individualized according to the patient’s clinical condition.
Outcomes
The primary outcome was 60-day mortality after ECMO initiation. For patients discharged or transferred before day 60, vital status at day 60 was ascertained through telephone follow-up. Complete 60-day vital status was available for all included patients, with no loss to follow-up. Secondary outcomes included successful ECMO weaning within 60 days and in-hospital mortality.
Hemoglobin concentrations and ScvO2
Only hemoglobin concentrations obtained from venous complete blood count measurements were included; hemoglobin values derived from blood gas analyses were excluded to ensure consistency in the measurement source. During VV ECMO support, complete blood count testing was routinely performed at least once daily as part of standard clinical care. All available eligible hemoglobin measurements from venous complete blood count testing during ECMO support were included in the analysis. TWA-Hb was calculated using an interval-weighted approach: each hemoglobin measurement was assumed to represent the interval from its sampling time until the subsequent measurement. TWA-Hb was calculated using the following formula (11,12):
| [1] |
where Hbi represents the hemoglobin value measured at each time point, ωi represents the duration of each hemoglobin value, and n represents the number of hemoglobin data points.
To assess the patient’s DO2 status during ECMO, we simultaneously performed time-weighted averaging of ScvO2.
Statistical analysis
All analyses were conducted using R (version 4.4.1) and SPSS (version 27.0). Continuous variables were described as mean ± standard deviation (SD) or median with interquartile range (IQR), and categorical variables as counts and percentages. Group comparisons used one-way analysis of variance (ANOVA) or Kruskal-Wallis tests for continuous variables and chi-square or Fisher’s exact tests for categorical variables. When post hoc pairwise comparisons were performed, Bonferroni correction was applied for multiple comparisons; Bonferroni-adjusted P values are reported in Table S1. All statistical tests were two-sided, and P<0.05 was considered statistically significant.
Missing baseline Sequential Organ Failure Assessment (SOFA) scores were handled using multiple imputation by chained equations with predictive mean matching. Thirty imputed datasets were generated. Regression models were fitted separately within each imputed dataset, and estimates were combined according to Rubin’s rules. Variables other than SOFA were not imputed solely for descriptive reporting.
For 60-day mortality, Cox proportional hazards regression was used. Model 1 was unadjusted; Model 2 was adjusted for baseline SOFA score; and Model 3 was additionally adjusted for ECMO indication. Baseline SOFA score and ECMO indication were selected on clinical grounds to account for baseline illness severity and major differences in the underlying indication for VV ECMO rather than according to univariable statistical significance. Given the 75 observed 60-day deaths, the multivariable model was deliberately kept parsimonious. The proportional hazards assumption for the mortality model was assessed using scaled Schoenfeld residuals separately within each of the 30 imputed datasets.
Kaplan-Meier curves and the log-rank test were used to compare 60-day survival across TWA-Hb groups. For successful ECMO weaning, cumulative incidence functions were estimated with death before weaning treated as a competing event, and group differences were assessed using Gray’s test. Multivariable logistic regression was used to evaluate successful ECMO weaning within 60 days using the same three model specifications as the mortality analysis. To explore the continuous association between TWA-Hb and 60-day mortality, an unadjusted restricted cubic spline (RCS) Cox model with four knots placed at the 5th, 35th, 65th, and 95th percentiles of the TWA-Hb distribution was fitted. P values for the overall and nonlinear associations were evaluated separately.
To further quantify the burden of severe anemia, we calculated for each patient the proportion of time with hemoglobin <70 g/L (exposure proportion), defined as the cumulative duration of hemoglobin <70 g/L divided by total ECMO support duration, expressed as a percentage. Cumulative duration was estimated by linear interpolation between consecutive hemoglobin measurements, followed by numerical integration. The association between exposure proportion and 60-day mortality was examined as a continuous exploratory exposure using adjusted RCS regression with baseline SOFA score and ECMO indication.
Results
Patient characteristics
A total of 182 patients who received VV ECMO support were included. Based on TWA-Hb concentrations during ECMO, patients were stratified into four groups: <70 g/L (n=37), 70–<80 g/L (n=44), 80–<90 g/L (n=50), ≥90 g/L (n=51) (Figure 1).
The majority of patients were male (n=135, 74.18%), with a median age of 59 years (IQR: 46–68) and a median body mass index (BMI) of 22.55 kg/m2 (IQR: 20.03–24.77). Comorbidities included immunosuppression (47.80%), hypertension (30.22%), diabetes (22.53%), coronary heart disease (14.29%), chronic obstructive pulmonary disease (12.64%), chronic kidney disease (7.14%), and hyperlipidemia (6.59%).
ARDS due to pulmonary infection was the leading ECMO indication (n=124, 68.13%), followed by bridge to lung transplantation (n=40, 21.98%). Significant differences were observed across groups for age, Acute Physiology and Chronic Health Evaluation II (APACHE II) score, SOFA score, ECMO indication, chronic kidney disease prevalence, and pre-ECMO hemoglobin concentration (all P<0.05). The median pre-ECMO hemoglobin concentration was 98 g/L (IQR: 81–115), and the median platelet count was 163.5×109/L (IQR: 97.25–240.50).
During ECMO support, 177 patients (97.25%) received RBC transfusions; the median number of units transfused during ECMO was 20 units (IQR: 10–38.75), averaging 1.11 units/day. Forty-six patients (25.27%) underwent continuous renal replacement therapy (CRRT). Major complications included bleeding (n=85, 46.70%), venous thrombosis (n=78, 42.86%), and AKI in nearly 60% of patients. Rates of oxygenator exchange, bleeding events, AKI incidence, and CRRT use differed significantly across groups (Table 1).
Table 1. Characteristics and outcomes according to time-weighted average hemoglobin group.
| Variables | Total (n=182) | <70 g/L (n=37) | 70–<80 g/L (n=44) | 80–<90 g/L (n=50) | ≥90 g/L (n=51) | P value | Missing data (%) |
|---|---|---|---|---|---|---|---|
| Sex, male | 135 (74.18) | 28 (75.68) | 33 (75.00) | 35 (70.00) | 39 (76.47) | 0.88 | 0.000 |
| Age (years) | 59.00 [46.00, 68.00] | 67.00 [53.00, 73.00] | 60.50 [48.50, 68.25] | 59.00 [43.25, 67.00] | 53.00 [42.50, 64.50] | 0.01 | 0.000 |
| BMI (kg/m2) | 22.55 [20.03, 24.77] | 21.72 [20.06, 24.22] | 22.88 [19.53, 24.75] | 21.85 [19.99, 24.19] | 23.67 [20.63, 25.33] | 0.24 | 0.000 |
| Height (cm) | 166.05±7.88 | 163.86±9.47 | 165.43±6.84 | 165.58±7.09 | 168.63±7.75 | 0.03 | 0.000 |
| Comorbidities | |||||||
| Hypertension | 55 (30.22) | 16 (43.24) | 13 (29.55) | 10 (20.00) | 16 (31.37) | 0.14 | 0.000 |
| Diabetes | 41 (22.53) | 8 (21.62) | 16 (36.36) | 8 (16.00) | 9 (17.65) | 0.08 | 0.000 |
| Hyperlipidemia | 12 (6.59) | 3 (8.11) | 4 (9.09) | 4 (8.00) | 1 (1.96) | 0.38 | 0.000 |
| Coronary heart disease | 26 (14.29) | 8 (21.62) | 7 (15.91) | 5 (10.00) | 6 (11.76) | 0.43 | 0.000 |
| Chronic kidney disease | 13 (7.14) | 7 (18.92) | 4 (9.09) | 2 (4.00) | 0 (0.00) | 0.003 | 0.000 |
| COPD | 23 (12.64) | 3 (8.11) | 8 (18.18) | 7 (14.00) | 5 (9.80) | 0.50 | 0.000 |
| Immunosuppression | 87 (47.80) | 17 (45.95) | 21 (47.73) | 27 (54.00) | 22 (43.14) | 0.74 | 0.000 |
| Indication | 0.002 | ||||||
| ARDS caused by pulmonary infection | 124 (68.13) | 33 (89.19) | 32 (72.73) | 36 (72.00) | 23 (45.10) | 0.000 | |
| ARDS caused by other reasons | 18 (9.89) | 1 (2.70) | 4 (9.09) | 5 (10.00) | 8 (15.69) | 0.000 | |
| Bridge to lung transplantation | 40 (21.98) | 3 (8.11) | 8 (18.18) | 9 (18.00) | 20 (39.21) | 0.000 | |
| APACHE II score on admission to the ICU | 23.69±7.45 | 26.76±6.95 | 22.85±5.99 | 22.77±9.42 | 22.15±6.23 | 0.046 | 26.370 |
| SOFA score on admission to the ICU | 10.05±3.49 | 11.47±3.84 | 8.93±2.72 | 10.12±3.72 | 9.83±3.24 | 0.02 | 28.570 |
| Pre-ECMO | |||||||
| P/F ratio (mmHg) | 100.38 [74.06, 141.99] | 112.40 [77.75, 158.10] | 112.14 [75.60, 142.83] | 96.56 [77.85, 126.48] | 93.85 [73.19, 139.42] | 0.53 | 8.240 |
| PaO2 (mmHg) | 81.00 [66.70, 102.00] | 93.20 [63.20, 116.00] | 81.05 [66.75, 95.48] | 81.95 [68.78, 97.90] | 77.05 [66.20, 106.50] | 0.62 | 7.140 |
| PaCO2 (mmHg) | 53.30 [45.00, 65.00] | 52.70 [44.35, 64.80] | 50.55 [45.47, 67.20] | 54.30 [45.85, 61.88] | 54.75 [44.10, 66.90] | 0.97 | 7.140 |
| pH | 7.33 [7.26, 7.38] | 7.31 [7.26, 7.37] | 7.34 [7.29, 7.42] | 7.31 [7.26, 7.36] | 7.33 [7.24, 7.39] | 0.32 | 7.140 |
| Lactate (mmol/L) | 1.72 [1.29, 2.58] | 1.78 [1.45, 2.25] | 1.65 [1.21, 2.39] | 2.10 [1.37, 2.68] | 1.42 [1.23, 2.28] | 0.17 | 12.640 |
| Hemoglobin (g/L) | 98.00 [81.00, 115.00] | 81.00 [66.50, 105.50] | 95.50 [81.75, 116.00] | 104.00 [86.00, 114.00] | 107.50 [88.50, 132.75] | <0.001 | 6.590 |
| Platelets (×109/L) | 163.50 [97.25, 240.50] | 110.00 [73.50, 218.00] | 182.00 [119.00, 280.75] | 157.00 [88.00, 216.00] | 171.00 [103.50, 232.75] | 0.09 | 6.590 |
| Blood flow rate after ECMO initiation (L/min) | 4.00 [3.66, 4.29] | 3.94 [3.72, 4.29] | 4.00 [3.64, 4.29] | 4.00 [3.70, 4.28] | 4.00 [3.61, 4.24] | 0.99 | 17.03 |
| Time-weighted average ScvO2 (%) | 69.59 [63.32, 75.89] | 65.64 [61.80, 77.66] | 66.19 [61.01, 70.75] | 70.41 [65.67, 75.83] | 72.42 [64.84, 78.44] | 0.054 | 9.89 |
| Oxygenator exchange | 74 (40.66) | 22 (59.46) | 24 (54.55) | 16 (32.00) | 12 (23.53) | <0.001 | 0.000 |
| Red blood cell transfusion | 177 (97.25) | 37 (100.00) | 43 (97.73) | 49 (98.00) | 48 (94.12) | 0.30 | 0.000 |
| Total RBC units during ECMO (units) | 20.00 [10.00, 38.75] | 32.00 [18.00, 53.00] | 20.00 [10.00, 44.50] | 20.00 [12.00, 34.75] | 10.50 [6.50, 22.00] | <0.001 | 0.000 |
| RBC/ECMO day (units/day) | 1.11 [0.67, 1.90] | 1.38 [0.82, 1.90] | 0.98 [0.61, 1.93] | 1.12 [0.76, 1.69] | 1.09 [0.59, 2.58] | 0.76 | 0.000 |
| ECMO duration (days) | 16.00 [8.25, 30.00] | 27.00 [16.00, 45.00] | 21.00 [12.00, 30.25] | 16.50 [10.00, 33.00] | 10.00 [6.00, 16.00] | <0.001 | 0.000 |
| CRRT | 46 (25.27) | 28 (75.68) | 16 (36.36) | 1 (2.00) | 1 (1.96) | <0.001 | 0.000 |
| Complications | |||||||
| Bleeding | 85 (46.70) | 26 (70.27) | 17 (38.64) | 27 (54.00) | 15 (29.41) | <0.001 | 0.000 |
| Venous thrombosis | 78 (42.86) | 21 (56.76) | 18 (40.91) | 22 (44.00) | 17 (33.33) | 0.18 | 0.000 |
| AKI | 0.009 | ||||||
| None | 73 (40.11) | 7 (18.92) | 22 (50.00) | 22 (44.00) | 22 (43.14) | 0.000 | |
| Stage 1 | 12 (6.59) | 1 (2.70) | 5 (11.36) | 3 (6.00) | 3 (5.88) | 0.000 | |
| Stage 2 | 4 (2.20) | 0 (0.00) | 2 (4.55) | 0 (0.00) | 2 (3.92) | 0.000 | |
| Stage 3 | 93 (51.10) | 29 (78.38) | 15 (34.09) | 25 (50.00) | 24 (47.06) | 0.000 | |
| Proportion of time with hemoglobin <70 g/L (%) | 9.09 [0.00, 46.50] | 83.58 [66.21, 88.76] | 28.31 [15.50, 41.19] | 3.33 [0.00, 7.56] | 0.00 [0.00, 0.21] | <0.001 | 0.000 |
| Outcomes | |||||||
| 60-day mortality | 75 (41.21) | 20 (54.05) | 17 (38.64) | 22 (44.00) | 16 (31.37) | 0.18 | 0.000 |
| Successfully weaned off ECMO within 60 days | 112 (61.54) | 13 (35.14) | 24 (54.55) | 32 (64.00) | 43 (84.31) | <0.001 | 0.000 |
| In-hospital mortality | 97 (53.30) | 24 (64.86) | 26 (59.09) | 27 (54.00) | 20 (39.22) | 0.08 | 0.000 |
Categorical variables are presented as n (%). Continuous variables are presented as mean ± standard deviation or median [interquartile range]. AKI, acute kidney injury; APACHE II, Acute Physiology and Chronic Health Evaluation II; ARDS, acute respiratory distress syndrome; BMI, body mass index; COPD, chronic obstructive pulmonary disease; CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; FiO2, fraction of inspired oxygen; ICU, intensive care unit; P/F, PaO2/FiO2; PaCO2, arterial partial pressure of carbon dioxide; PaO2, arterial partial pressure of oxygen; RBC, red blood cell; ScvO2, central venous oxygen saturation; SOFA, sequential organ failure assessment.
Hemoglobin concentrations and clinical outcomes
A total of 112 patients (61.54%) were successfully weaned from ECMO within 60 days. Seventy-five patients (41.21%) died within 60 days of ECMO initiation, and 97 patients (53.30%) died during hospitalization. Significant differences were found among groups in terms of 60-day ECMO weaning success and RBC transfusion volumes (Table 1). Specifically, patients in the <70 g/L and 70–<80 g/L groups received higher transfusion volumes and had lower weaning success rates compared to those in the ≥90 g/L group (Table S1). However, there were no significant differences in unadjusted 60-day mortality across groups (Table 1).
In the unadjusted RCS analysis, the fitted association between continuous TWA-Hb and 60-day mortality did not show a clear monotonic or nonlinear dose-response pattern (Figure 2). Neither the overall association nor the nonlinear component was statistically significant (P for overall association =0.267, P for nonlinearity =0.541), providing no evidence of a distinct risk threshold in the continuous analysis. Kaplan-Meier analysis similarly showed no significant difference in unadjusted 60-day survival among the four TWA-Hb groups (log-rank P=0.13, Figure 3A).
Figure 2.
Unadjusted restricted cubic spline analysis of the association between continuous TWA-Hb concentration and 60-day mortality. The solid line represents the estimated hazard ratio and the shaded area represents the 95% CI. CI, confidence interval; TWA-Hb, time-weighted average hemoglobin.
Figure 3.
Sixty-day survival and successful ECMO weaning according to time-weighted average hemoglobin concentration. (A) Kaplan-Meier curves for 60-day survival; differences among groups were assessed using the log-rank test. (B) Cumulative incidence functions for successful ECMO weaning within 60 days, with death before ECMO weaning treated as a competing event; differences among groups were assessed using Gray’s test. ECMO, extracorporeal membrane oxygenation.
Multivariable Cox regression analysis (Table 2) showed that, after adjusting for SOFA score and ECMO indication, TWA-Hb <70 g/L was associated with a higher hazard of 60-day mortality compared to the ≥90 g/L group [adjusted hazard ratio (HR) 3.47, 95% confidence interval (CI): 1.00–12.03, P=0.049]. The associations for the 70–<80 g/L group (HR 2.75, 95% CI: 0.79–9.51, P=0.11) and the 80–<90 g/L group (HR 2.53, 95% CI: 0.72–8.89, P=0.15) did not reach statistical significance. No evidence of violation of the proportional hazards assumption was observed for the 60-day mortality Cox model; the Schoenfeld residual tests for TWA-Hb group, SOFA score, ECMO indication, and the global model were nonsignificant in all 30 multiply imputed datasets.
Table 2. Multivariable Cox regression analysis of 60-day mortality according to time-weighted average hemoglobin group.
| Groups | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| HR (95% CI) | P | HR (95% CI) | P | HR (95% CI) | P | |
| ≥90 g/L | Reference | Reference | Reference | |||
| <70 g/L | 1.74 (0.91–3.34) | 0.10 | 3.53 (1.03–12.11) | 0.045 | 3.47 (1.00–12.03) | 0.049 |
| 70–<80 g/L | 1.20 (0.61–2.39) | 0.59 | 2.77 (0.80–9.57) | 0.11 | 2.75 (0.79–9.51) | 0.11 |
| 80–<90 g/L | 1.31 (0.69–2.50) | 0.41 | 2.69 (0.77–9.40) | 0.12 | 2.53 (0.72–8.89) | 0.15 |
Model 1: crude; Model 2: adjusted for SOFA score; Model 3: adjusted for SOFA score and ECMO indication. Missing SOFA scores were handled using multiple imputation, and regression estimates were pooled according to Rubin’s rules. CI, confidence interval; ECMO, extracorporeal membrane oxygenation; HR, hazard ratio; SOFA, sequential organ failure assessment.
The cumulative incidence of successful ECMO weaning within 60 days differed significantly across TWA-Hb groups when death before weaning was treated as a competing event (Gray’s test P<0.001, Figure 3B). In multivariable logistic regression adjusted for baseline SOFA score and ECMO indication, patients with TWA-Hb <70 g/L had a substantially lower likelihood of successful ECMO weaning than those with TWA-Hb ≥90 g/L [adjusted odds ratio (OR) 0.13, 95% CI: 0.04–0.38, P<0.001]. A lower likelihood of successful weaning was also observed in the 70–<80 g/L group (OR 0.26, 95% CI: 0.09–0.72, P=0.01) (Table 3).
Table 3. Multivariable logistic regression analysis of successful ECMO weaning within 60 days according to time-weighted average hemoglobin group.
| Groups | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| OR (95% CI) | P | OR (95% CI) | P | OR (95% CI) | P | |
| ≥90 g/L | Reference | Reference | Reference | |||
| <70 g/L | 0.10 (0.04–0.28) | <0.001 | 0.10 (0.04–0.28) | <0.001 | 0.13 (0.04–0.38) | <0.001 |
| 70–<80 g/L | 0.22 (0.08–0.59) | 0.003 | 0.22 (0.08–0.59) | 0.003 | 0.26 (0.09–0.72) | 0.01 |
| 80–<90 g/L | 0.33 (0.13–0.86) | 0.02 | 0.33 (0.13–0.87) | 0.03 | 0.40 (0.15–1.08) | 0.07 |
Model 1: crude; Model 2: adjusted for SOFA score; Model 3: adjusted for SOFA score and ECMO indication. Missing SOFA scores were handled using multiple imputation, and regression estimates were pooled according to Rubin’s rules. CI, confidence interval; ECMO, extracorporeal membrane oxygenation; OR, odds ratio; SOFA, sequential organ failure assessment.
Subgroup analysis
Subgroup analyses showed no significant interaction between hemoglobin concentration and 60-day mortality by sex, age, or BMI (Figure S1).
Association between exposure proportion (proportion of time with hemoglobin <70 g/L) and 60-day mortality
To further investigate the impact of cumulative exposure to severe anemia, we calculated the proportion of time during ECMO support in which hemoglobin levels remained below 70 g/L (exposure proportion). The median exposure proportion for the entire cohort was 9% (IQR: 0–46%).
After adjustment for baseline SOFA score and ECMO indication, RCS analysis showed no statistically significant overall association between exposure proportion and 60-day mortality (P for overall association =0.110) and no evidence of a nonlinear association (P for nonlinearity =0.221) (Figure 4). Accordingly, exposure proportion is interpreted as a continuous exploratory marker, and the present analysis does not define a discrete clinical cut-off.
Figure 4.
Restricted cubic spline analysis of the association between the proportion of ECMO support time with hemoglobin <70 g/L and 60-day mortality. The model was adjusted for baseline SOFA score and ECMO indication. Solid lines represent point estimates and shaded areas represent 95% CIs. CI, confidence interval; ECMO, extracorporeal membrane oxygenation; HR, hazard ratio; SOFA, sequential organ failure assessment.
Discussion
In this retrospective cohort of patients receiving VV ECMO, very low longitudinal hemoglobin exposure was associated with less favorable clinical outcomes. Patients with TWA-Hb <70 g/L had higher 60-day mortality, and lower TWA-Hb was also associated with a lower likelihood of successful ECMO weaning. Importantly, however, the continuous spline analysis did not show a clear graded association between TWA-Hb and mortality across the entire exposure range or identify a distinct inflection point. This pattern suggests that the prognostic signal may be concentrated among patients with sustained, very low hemoglobin exposure rather than reflecting a uniform dose-response relationship. Accordingly, our findings identify severe longitudinal anemia as a potential high-risk phenotype during VV ECMO, but do not establish 70 g/L as a statistically defined transfusion threshold.
Hemoglobin is a critical determinant of DO2, particularly in VV ECMO patients whose circulatory function still relies on native cardiac output. Although ECMO provides extracorporeal respiratory support, adequate tissue oxygenation depends largely on hemoglobin concentration (13). When hemoglobin falls below a critical threshold, oxygen-carrying capacity becomes inadequate, potentially resulting in tissue hypoxia and multi-organ dysfunction. Notably, patients with lower hemoglobin required significantly more transfusions, exposing them to cumulative transfusion-related risks (4). This illustrates the double-edged nature of anemia management in ECMO: too little hemoglobin impairs DO2, while frequent transfusions carry their own risks.
Our findings should be considered in the context of previous ECMO transfusion studies, which have addressed related but not identical clinical questions. PROTECMO was a prospective multicenter observational study describing transfusion practice and outcomes across VV ECMO centers (7), whereas Pratt et al. evaluated outcomes in relation to transfusion thresholds in patients with acute respiratory failure supported with ECMO (8). Hunsicker et al. compared lower and higher hemoglobin strategies using a time-weighted approach in patients with ARDS with and without ECMO (12). More recently, a single-center mixed-ECMO cohort found longer ECMO duration among patients with lower achieved mean hemoglobin levels, although mortality did not differ significantly across hemoglobin groups (14). By contrast, the present study focused primarily on achieved longitudinal hemoglobin exposure over the entire ECMO course rather than a prespecified transfusion trigger. Differences in patient selection, transfusion practice, severity of illness, exposure definitions, and study endpoints make direct comparison difficult and may explain why findings across studies are not fully concordant. Nevertheless, taken together, these studies suggest that very restrictive hemoglobin management during VV ECMO deserves careful evaluation rather than assuming that lower hemoglobin exposure is uniformly benign.
A principal rationale for using TWA-Hb was to move beyond isolated hemoglobin values. A single measurement or nadir captures the depth of anemia at one point in time but provides limited information about how long that exposure persisted. In contrast, TWA-Hb incorporates both the measured concentration and the interval represented by each measurement, thereby summarizing the hemoglobin environment experienced over the ECMO course and reducing the disproportionate influence of unequal sampling intervals. This metric is not intended to serve as a real-time bedside transfusion trigger, because the complete TWA-Hb is only known retrospectively after the observation period. Its value is instead epidemiological: it provides a longitudinal measure of achieved hemoglobin exposure that complements repeated measurements, nadir values, and threshold-based summaries.
Beyond survival, lower TWA-Hb was associated with less favorable ECMO weaning outcomes, with the strongest associations observed in the <70 and 70–<80 g/L groups. To our knowledge, no prior studies have directly examined hemoglobin concentration as a predictor of ECMO liberation. This relationship likely reflects the increased physiological demands placed on the native cardiopulmonary system during weaning, where suboptimal hemoglobin impairs DO2 and compromises organ function. Hunsicker et al. reported no increase in 28-day ECMO weaning failure with restrictive transfusion strategies but did find lower ventilator weaning success in the restrictive group (12), suggesting a potential role for hemoglobin in weaning outcomes. Lai et al. similarly found that hemoglobin levels of 80–100 g/L were associated with higher mechanical ventilation weaning rates (15). Although these studies addressed ventilator weaning rather than ECMO removal, they reinforce the concept that adequate hemoglobin may facilitate successful transition from artificial to native respiratory support.
These findings have practical implications for contemporary VV ECMO care. Restrictive transfusion strategies can reduce donor blood exposure and are particularly relevant when blood products are limited (16), but restriction should not be pursued without considering the duration and severity of anemia. Sustained exposure to hemoglobin concentrations below 70 g/L warrants particular clinical attention and should be minimized when feasible. At the same time, low hemoglobin may reflect active bleeding, circuit-related blood loss, prolonged ECMO support, or greater underlying illness severity, and increasing transfusion itself carries important risks. Transfusion decisions should therefore remain individualized according to bleeding status, hemodynamics, markers of systemic DO2, comorbidities, and the overall clinical trajectory. Blood-conservation strategies, including approaches that reduce avoidable donor exposure (17), may complement individualized transfusion practice. Prospective studies are needed to determine whether strategies aimed at limiting prolonged severe-anemia exposure can improve outcomes without increasing transfusion-related harm.
This study has several limitations. First, its retrospective, single-center design may introduce selection and information bias, and the findings may not be generalizable to other populations or ECMO centers. Second, sample sizes within some subgroups, particularly patients with hemoglobin less than 70 g/L, were relatively small, which may have limited the statistical power to detect differences. Third, although we adjusted for multiple confounders, residual confounding cannot be excluded. Important variables such as fluid balance and precise transfusion indications were not fully captured. In addition, transfusion decisions were made at the discretion of treating clinicians, introducing practice variability, and patients in the <70 g/L group may have had greater baseline disease severity, which multivariable adjustment may not have fully addressed. Fourth, because we used a time-weighted method, the frequency of hemoglobin measurements could have influenced the calculated averages, potentially introducing measurement bias. Finally, although time-weighted averaging was applied to ScvO2, the absence of cardiac output measurements prevented calculation of DO2, limiting further mechanistic exploration of the relationship between anemia and outcomes.
Conclusions
In patients receiving VV ECMO, TWA-Hb <70 g/L was associated with higher 60-day mortality and a lower likelihood of successful ECMO weaning. Sustained exposure to severe anemia below 70 g/L warrants particular clinical attention and should be minimized when feasible. These findings support further prospective evaluation of hemoglobin-management strategies aimed at reducing prolonged severe-anemia exposure during VV ECMO.
Supplementary
The article’s supplementary files as
Acknowledgments
An abstract of this study was previously presented at the 2025 APELSO Conference.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The First Affiliated Hospital of Guangzhou Medical University (No. ES-2024-K089-01); informed consent for this retrospective analysis was waived.
Footnotes
Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1038/rc
Funding: This work was supported by the Major Project of Guangzhou National Laboratory Grant (GZNL2023A01003), the National Natural Science Foundation of China (82241003, 82270085), the Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0506503), and the Medical Research Foundation of Guangdong Province (A2023440).
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1038/coif). H.Z. serves as an unpaid editorial board member (Emergency and Critical Care) of Journal of Thoracic Disease. Yonghao Xu reports receiving grants from the Major Project of Guangzhou National Laboratory Grant (GZNL2023A01003), National Natural Science Foundation of China (82241003, 82270085), the Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0506503), and the Medical Research Foundation of Guangdong Province (A2023440). The other authors have no conflicts of interest to declare.
Data Sharing Statement
Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1038/dss
References
- 1. Tonna JE, Abrams D, Brodie D, et al. Management of Adult Patients Supported with Venovenous Extracorporeal Membrane Oxygenation (VV ECMO): Guideline from the Extracorporeal Life Support Organization (ELSO). ASAIO J 2021;67:601-10.. 10.1097/MAT.0000000000001432 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Thomas J, Kostousov V, Teruya J. Bleeding and Thrombotic Complications in the Use of Extracorporeal Membrane Oxygenation. Semin Thromb Hemost 2018;44:20-9.. 10.1055/s-0037-1606179 [DOI] [PubMed] [Google Scholar]
- 3. Hughes T, Zhang D, Nair P, et al. A Systematic Literature Review of Packed Red Cell Transfusion Usage in Adult Extracorporeal Membrane Oxygenation. Membranes (Basel) 2021;11:251.. 10.3390/membranes11040251 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Carson JL, Triulzi DJ, Ness PM. Indications for and Adverse Effects of Red-Cell Transfusion. N Engl J Med 2017;377:1261-72.. 10.1056/NEJMra1612789 [DOI] [PubMed] [Google Scholar]
- 5. Mallat J, Rahman N, Hamed F, et al. Pathophysiology, mechanisms, and managements of tissue hypoxia. Anaesth Crit Care Pain Med 2022;41:101087.. 10.1016/j.accpm.2022.101087 [DOI] [PubMed] [Google Scholar]
- 6. Tonna JE. Patients receiving ECMO are special, but still only need a haemoglobin concentration of 7g/dL. Lancet Respir Med 2023;11:215-7.. 10.1016/S2213-2600(22)00367-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Martucci G, Schmidt M, Agerstrand C, et al. Transfusion practice in patients receiving VV ECMO (PROTECMO): a prospective, multicentre, observational study. Lancet Respir Med 2023;11:245-55.. 10.1016/S2213-2600(22)00353-8 [DOI] [PubMed] [Google Scholar]
- 8. Pratt EH, Pulsipher AM, Moulton NG, et al. Association of RBC Transfusion Thresholds and Outcomes in Medical Patients With Acute Respiratory Failure Supported With Extracorporeal Membrane Oxygenation: A Single-Center Retrospective Cohort Study. Chest 2024;166:1406-16.. 10.1016/j.chest.2024.05.043 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Ramanathan K, Antognini D, Combes A, et al. Planning and provision of ECMO services for severe ARDS during the COVID-19 pandemic and other outbreaks of emerging infectious diseases. Lancet Respir Med 2020;8:518-26.. 10.1016/S2213-2600(20)30121-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lequier L, Annich G, Al-Ibrahim O, et al. ELSO anticoagulation guideline. Extracorporeal Life Support Organization; 2014.
- 11. Xu R, Cao T, Liao Y, et al. Time-Weighted Average Proteinuria and Renal Function Decline in IgA Nephropathy: A Retrospective Cohort Study. Int J Nephrol Renovasc Dis 2025;18:103-10.. 10.2147/IJNRD.S517145 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Hunsicker O, Materne L, Bünger V, et al. Lower versus higher hemoglobin threshold for transfusion in ARDS patients with and without ECMO. Crit Care 2020;24:697.. 10.1186/s13054-020-03405-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Carvalho LT, Mendes PV, Besen BAMP, et al. The hemoglobin level impact on arterial oxygen saturation during venous-venous-extracorporeal membrane oxygenation support of acute respiratory distress syndrome patients: a mathematical marginal approach. Rev Bras Ter Intensiva 2022;34:393-5.. 10.5935/0103-507X.20220465-pt [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Balasubramanian S, Alwakeel M, Sadana D, et al. Red blood cell transfusion practices in extracorporeal membrane oxygenation: A single-center study. Transfus Med 2025;35:337-45.. 10.1111/tme.13154 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Lai YC, Ruan SY, Huang CT, et al. Hemoglobin levels and weaning outcome of mechanical ventilation in difficult-to-wean patients: a retrospective cohort study. PLoS One 2013;8:e73743.. 10.1371/journal.pone.0073743 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Roberts N, James S, Delaney M, et al. The global need and availability of blood products: a modelling study. Lancet Haematol 2019;6:e606-15.. 10.1016/S2352-3026(19)30200-5 [DOI] [PubMed] [Google Scholar]
- 17. Camarda V, Sanderson B, Barrett NA, et al. The effects of blood cell salvage on transfusion requirements after decannulation from veno-venous extracorporeal membrane oxygenation: an emulated trial analysis. Crit Care 2024;28:403.. 10.1186/s13054-024-05177-7 [DOI] [PMC free article] [PubMed] [Google Scholar]