SvO₂ response to red blood cell transfusion in cardiovascular surgical icu patients: a retrospective observational study
Authors: Kentaro Okamoto, Kimito Minami, Chase Donaldson, Jean Deschamps, Marcelo Gama de Abreu, Shigeki Fujitani, Takaki Naito, Tatsutoshi Shimatani, Muneyuki Takeuchi
Published: 2026

ABSTRACT
BackgroundRed blood cell (RBC) transfusion decisions after cardiovascular surgery require integration of hemoglobin (Hb), hemodynamics, bleeding status, and oxygen supply–demand balance. Mixed venous oxygen saturation (SvO₂) reflects the global relationship between oxygen delivery and oxygen consumption, but an increase in SvO₂ does not necessarily indicate improved tissue oxygenation or clinical benefit. We evaluated acute SvO₂ changes after RBC transfusion in cardiovascular surgical ICU patients.MethodsWe conducted a single-center retrospective cohort study of adult cardiovascular surgical ICU patients who received one RBC-equivalent transfusion with pre-transfusion Hb ≥ 7.5 g/dL and paired pre- and post-transfusion SvO₂ measurements. The primary outcome was an individual-level SvO₂ response, defined a priori as ΔSvO₂ ≥5% points from baseline to approximately 60 min after transfusion initiation. Multivariable regression was used to evaluate predictors selected a priori based on physiologic relevance to oxygen delivery and consumption. Receiver operating characteristic analysis was used to evaluate the discriminative ability of pre-transfusion SvO₂.Results1,352 unique patients met the final inclusion criteria. Mean Hb increased from 9.82 ± 0.92 to 10.24 ± 0.98 g/dL, corresponding to a paired mean difference of 0.42 g/dL (95% CI, 0.38 to 0.46; P < 0.001). Mean SvO₂ changed minimally at the cohort level, from 73.79 ± 9.91% to 73.86 ± 9.37%, corresponding to a paired mean difference of 0.08% points (95% CI, -0.28 to 0.43; P = 0.671). Lower pre-transfusion SvO₂ was associated with a higher probability of an individual-level SvO₂ response. In multivariable logistic regression, pre-transfusion SvO₂ was associated with SvO₂ response (adjusted odds ratio, 0.89 per 1% increase; 95% confidence interval, 0.86–0.91; P < 0.001). Pre-transfusion SvO₂ showed moderate discrimination for SvO₂ response, with an area under the curve of 0.778 (95% confidence interval, 0.740–0.816). The Youden-derived cutoff was 69%, which should be interpreted as an exploratory predictor of acute SvO₂ rise rather than a transfusion trigger.ConclusionsLower pre-transfusion SvO₂ was associated with acute individual-level SvO₂ increase after transfusion, but mean SvO₂ change was minimal. The association did not establish tissue oxygenation or clinical benefit and does not support using SvO₂, including the 69% cutoff, as a transfusion indication or trigger.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13054-026-06149-9.

Background

Red blood cell (RBC) transfusion is frequently performed in the postoperative management of cardiovascular surgery. Recent international guidelines generally recommend a restrictive transfusion strategy with a threshold of hemoglobin (Hb) 7.5 g/dL for cardiac surgery patients [1]. Yet transfusion decisions beyond hemoglobin concentration, particularly those incorporating physiological indicators, remain poorly established. Mixed venous oxygen saturation (SvO₂) reflects the balance between oxygen delivery (DO₂) and oxygen consumption (VO₂) and may provide physiologic context for transfusion decisions, although it cannot by itself establish improved tissue oxygenation or clinical benefit. Nevertheless, studies evaluating transfusion efficacy using venous oxygen saturation (SvO₂ or ScvO₂) are limited [2, 3].

Previous studies have evaluated venous oxygen saturation as an adjunct to hemoglobin-based transfusion decisions after cardiac surgery. In an observational study of cardiothoracic and vascular ICU patients with anemia, Zeroual et al. reported that patients with lower pre-transfusion ScvO₂ were more likely to show an increase in ScvO₂ after RBC transfusion [4]. Subsequently, in a randomized trial of post-cardiac surgery patients who developed anemia with Hb < 9 g/dL and no active bleeding, Zeroual et al. compared a control strategy in which patients were transfused for Hb < 9 g/dL with an ScvO₂-guided strategy in which transfusion was given only when ScvO₂ was ≤ 65% [5]. This strategy reduced RBC exposure compared with the Hb < 9 g/dL control strategy. Similarly, Fischer et al. evaluated an individualized ScvO₂-based transfusion strategy in post-cardiac surgery patients with Hb < 9 g/dL and reported reduced RBC transfusion without a clear difference in clinical outcomes [6].

These prior studies suggest that transfusion strategies integrating hemoglobin levels with a marker of oxygen supply–demand mismatch, such as ScvO₂, can reduce RBC exposure in cardiac surgery patients compared with relatively liberal Hb-based strategies, typically using a transfusion threshold of approximately 9 g/dL, without clear evidence of increased adverse events [7].

Importantly, these trials evaluated ScvO₂-guided transfusion against relatively liberal hemoglobin-based control strategies using an Hb threshold of approximately 9 g/dL, rather than against contemporary restrictive strategies around 7–7.5 g/dL. Current international guidelines recommend restrictive RBC transfusion for most hemodynamically stable adults and note that clinicians may choose a threshold of 7.5 g/dL for patients undergoing cardiac surgery [1]. Therefore, whether venous oxygen saturation provides physiologic context for transfusions performed above contemporary restrictive thresholds remains uncertain. In this context, we evaluated routinely measured pulmonary artery catheter-derived SvO₂ in cardiovascular surgical ICU patients who received one RBC-equivalent transfusion at Hb ≥ 7.5 g/dL. Our aim was not to validate an SvO₂-based transfusion trigger but to characterize acute SvO₂ changes and determine whether lower pre-transfusion SvO₂ was associated with individual-level SvO₂ rise.

Objective

To characterize heterogeneity in acute SvO₂ changes after one RBC-equivalent transfusion in cardiovascular surgical ICU patients with pre-transfusion Hb ≥ 7.5 g/dL, and to evaluate whether lower pre-transfusion SvO₂ is associated with an individual-level SvO₂ increase.

Methods

Study design and setting

This retrospective observational study utilized electronic medical record data from patients admitted to the Cardiovascular Intensive Care Unit (CVICU) at the National Cerebral and Cardiovascular Center between July 1, 2019, and December 31, 2024. The study was approved by the Ethics Review Committee (Approval No. R24108). Given the retrospective design, consent was obtained using an opt-out approach, and the requirement for individual written informed consent was waived.

Subjects

Inclusion criteria were: (1) age ≥ 18 years, (2) admission to the CVICU during the study period, (3) receipt of an eligible index RBC transfusion, defined as the first standard Japanese 2-unit RBC product administered during CVICU admission, (4) pre-index Hb ≥ 7.5 g/dL, (5) obtainable paired pre- and post-index SvO₂ measurements, and (6) SvO₂ monitoring via pulmonary artery catheterization.

The study was not designed to evaluate RBC transfusion during active major bleeding or ongoing massive transfusion. Episodes were excluded when electronic transfusion records indicated simultaneous administration of multiple blood products, suspected massive transfusion for postoperative bleeding, inaccurate blood product records, or additional RBC or other blood product administration during the prespecified observation window. The prespecified observation window for concomitant transfusion-related interventions was defined as the period from the start of the index transfusion to the post-transfusion SvO₂ measurement for the primary analysis and to 6 h for the 0–6-hour trajectory analysis. Suspected active major bleeding was operationally defined using available electronic transfusion records as any of the following: simultaneous administration of RBCs with plasma or platelet products, ≥ 4 RBC-equivalent products within 12 h after the index transfusion start, or documentation suggesting massive transfusion for postoperative bleeding. Detailed chest drainage volume and clinician-documented bleeding indications were not uniformly available in the electronic dataset. Patients who declined participation through the opt-out process were also excluded. For patients with multiple RBC transfusion episodes during the CVICU stay, only the first eligible RBC transfusion episode was considered the index transfusion. Repeated transfusions beyond the first eligible episode were not analyzed as independent observations, to avoid within-patient duplication and to standardize the physiologic observation window.

In our institution, red blood cells were routinely administered as a leukoreduced RBC product derived from a 400-mL whole-blood donation. In Japan, this product is labeled as 2 units in the Japanese system; its approximate volume is approximately 280 mL, making it roughly comparable to 1 unit of packed red blood cells in the United States, and it is commonly infused over approximately 1 h in routine adult practice.

Data acquisition and variable definition

The following variables were extracted: transfusion time, Hb from blood gas analysis, SvO₂, SpO₂, heart rate, blood pressure, cardiac index, pulmonary artery pressure, right atrial pressure, and systemic vascular resistance from ACSYS® monitor records. The thermal filament pulmonary artery catheter (Edwards Lifesciences) was used. Pre-transfusion values were defined as the closest recorded value (immediate value) to the transfusion start time. Post-transfusion values were defined as the closest recorded value (immediate value) to approximately 60 min after transfusion start. ΔSvO₂ was calculated as post-transfusion SvO₂ minus pre-transfusion SvO₂.

Outcomes

The primary outcome was SvO₂ responsiveness, defined a priori as ΔSvO₂ ≥5% points after transfusion. This cutoff was selected based on previous work by Zeroual et al. [4], which used a ≥ 5% increase in ScvO₂ after RBC transfusion as a marker of transfusion response, allowing comparison with the existing literature. In addition, we evaluated the predictive performance of pre-transfusion SvO₂ for SvO₂ responsiveness and performed prespecified subgroup and sensitivity analyses.

Statistical analysis

Continuous variables were presented as mean ± standard deviation or median (interquartile range). Categorical variables were presented as number (%). Paired pre- and post-transfusion variables were compared using paired t tests, and paired mean differences with 95% confidence intervals were reported. For variables with markedly skewed distributions, Wilcoxon signed-rank tests were additionally performed as sensitivity analyses. Because of the large sample size, statistical significance was interpreted together with the magnitude and clinical relevance of the absolute change. A two-sided p value < 0.05 was considered statistically significant. To examine factors associated with SvO₂ responsiveness, multivariable logistic regression analysis was performed using variables related to oxygen supply (age, sex, pre-transfusion Hb, pre-transfusion SvO₂, SpO₂, cardiac index, left ventricular ejection fraction) as explanatory variables. The predictive ability of pre-transfusion SvO₂ for SvO₂ responsiveness was evaluated using ROC curves. The cutoff value for pre-transfusion SvO₂ was derived using the Youden index, defined as sensitivity + specificity − 1. This cutoff was considered exploratory and internally derived. It was used only to describe the discrimination of pre-transfusion SvO₂ for an acute individual-level SvO₂ rise and was not intended to define a clinical transfusion trigger. For exploratory analysis, we compared the area under the receiver operating characteristic curves for pre-transfusion SvO₂, Hb, SpO₂, cardiac index, left ventricular ejection fraction, and DO₂i when available. For subgroup analysis, ΔSvO₂ was described by pre-transfusion SvO₂ category in populations with lower limits of pre-transfusion Hb set at 9, 10, 11, and 12 g/dL. For sensitivity analysis, we constructed a risk set anchored at transfusion initiation to compare the transfusion and no-transfusion groups. To adjust for confounding, we evaluated propensity score matching, inverse probability of treatment weighting (IPTW), and overlap weighting and selected overlap weighting because it provided the best covariate balance (Table S1). Trajectories of SvO₂, Hb, and DO₂i over 0–6 h were analyzed using models that included group, time, and a group-by-time interaction term. To evaluate the potential influence of concomitant transfusion-related interventions, we assessed additional RBC transfusion or other blood product administration during two time windows: from the index transfusion start to the post-transfusion SvO₂ measurement, and from the index transfusion start to 6 h. Other blood products included plasma, platelet products, and cryoprecipitate. The former window was used to assess potential contamination of the primary outcome measurement, whereas the latter was used to describe potential concomitant transfusion-related interventions during the 0–6-hour trajectory analysis. These assessments were based on available electronic transfusion records. Furthermore, to reinforce the results of the main analysis, a multivariable linear regression analysis was also performed using ΔSvO₂ as a continuous variable. Statistical analyses were performed using R version 4.5.2 (R Foundation for Statistical Computing, Vienna, Austria). This study is reported in accordance with the STROBE statement for observational studies.

Results

Identification of subject patients

Among patients aged 18 years or older admitted to the CVICU, 18,117 transfusion episodes with Hb ≥ 7.5 g/dL were identified. Of these, 6,962 episodes had paired pre- and post-transfusion SvO₂ values. After excluding repeated transfusions beyond the first transfusion per patient (n = 5,555) and transfusions other than 2 units of RBCs (n = 55), 1,352 patients were included in the analysis population (Fig. 1).

Fig. 1Study flow diagram. Among 18,117 RBC transfusion episodes with Hb ≥ 7.5 g/dL identified in the CVICU between July 1, 2019, and December 31, 2024, we identified patients who received one standard Japanese 2-unit RBC product, approximately equivalent to one international packed RBC unit, and had paired pre- and post-transfusion SvO₂ measurements. After exclusion of repeated transfusions beyond the first transfusion per patient and episodes not corresponding to one standard Japanese 2-unit RBC product, 1,352 unique patients were included in the final analysis

Study flow diagram. Among 18,117 RBC transfusion episodes with Hb ≥ 7.5 g/dL identified in the CVICU between July 1, 2019, and December 31, 2024, we identified patients who received one standard Japanese 2-unit RBC product, approximately equivalent to one international packed RBC unit, and had paired pre- and post-transfusion SvO₂ measurements. After exclusion of repeated transfusions beyond the first transfusion per patient and episodes not corresponding to one standard Japanese 2-unit RBC product, 1,352 unique patients were included in the final analysis

Patient background and pre- and post-transfusion parameters

Baseline characteristics are summarized in Table 1. The cohort included 1,352 patients with a mean age of 66.6 ± 15.2 years, and 822 patients (60.8%) were male. The cohort included a broad range of cardiovascular surgical procedures, most commonly isolated valve surgery, isolated open aortic/vascular surgery, isolated CABG/coronary surgery, and isolated endovascular procedures. Detailed demographic, echocardiographic, mortality, and procedural characteristics are shown in Table 1.

Table 1Baseline characteristics of the study population (N = 1352)CharacteristicOverall Demographics Age at ICU admission, years66.6 ± 15.2Male sex, n (%)822 (60.8)Body mass index, kg/m²23.0 ± 4.0ICU length of stay, days4 [3–7] Preoperative echocardiography Left ventricular ejection fraction (visual), % (n = 776)49.79 ± 17.38Left ventricular ejection fraction (modified Simpson), % (n = 602)45.89 ± 15.02 Mortality outcomes ICU mortality, n (%)48 (3.6)In-hospital mortality, n (%)213 (15.8) Procedure type (mutually exclusive) Isolated valve surgery, n (%)377 (27.9)Isolated open aortic/vascular surgery, n (%)218 (16.1)Isolated CABG/coronary surgery, n (%)202 (14.9)Isolated endovascular procedure (EVAR/TEVAR), n (%)154 (11.4)Isolated heart transplant/VAD, n (%)76 (5.6)CABG + valve surgery, n (%)48 (3.6)Valve + open aortic surgery, n (%)35 (2.6)CABG + open aortic surgery, n (%)14 (1.0)Other combined, n (%)24 (1.8)Unclassified, n (%)204 (15.1)Values are presented as mean ± SD, median [interquartile range], or n (%), as appropriatePercentages are based on the total cohort (N = 1352). ICU, intensive care unit; CABG, coronary artery bypass grafting; EVAR, endovascular aneurysm repair; TEVAR, thoracic endovascular aortic repair; VAD, ventricular assist device

Table 1Baseline characteristics of the study population (N = 1352)CharacteristicOverall Demographics Age at ICU admission, years66.6 ± 15.2Male sex, n (%)822 (60.8)Body mass index, kg/m²23.0 ± 4.0ICU length of stay, days4 [3–7] Preoperative echocardiography Left ventricular ejection fraction (visual), % (n = 776)49.79 ± 17.38Left ventricular ejection fraction (modified Simpson), % (n = 602)45.89 ± 15.02 Mortality outcomes ICU mortality, n (%)48 (3.6)In-hospital mortality, n (%)213 (15.8) Procedure type (mutually exclusive) Isolated valve surgery, n (%)377 (27.9)Isolated open aortic/vascular surgery, n (%)218 (16.1)Isolated CABG/coronary surgery, n (%)202 (14.9)Isolated endovascular procedure (EVAR/TEVAR), n (%)154 (11.4)Isolated heart transplant/VAD, n (%)76 (5.6)CABG + valve surgery, n (%)48 (3.6)Valve + open aortic surgery, n (%)35 (2.6)CABG + open aortic surgery, n (%)14 (1.0)Other combined, n (%)24 (1.8)Unclassified, n (%)204 (15.1)Values are presented as mean ± SD, median [interquartile range], or n (%), as appropriatePercentages are based on the total cohort (N = 1352). ICU, intensive care unit; CABG, coronary artery bypass grafting; EVAR, endovascular aneurysm repair; TEVAR, thoracic endovascular aortic repair; VAD, ventricular assist device

Baseline characteristics of the study population (N = 1352)

Values are presented as mean ± SD, median [interquartile range], or n (%), as appropriate

Percentages are based on the total cohort (N = 1352). ICU, intensive care unit; CABG, coronary artery bypass grafting; EVAR, endovascular aneurysm repair; TEVAR, thoracic endovascular aortic repair; VAD, ventricular assist device

Table 2 shows circulatory and respiratory parameters before and after transfusion. The median time from transfusion initiation to post-transfusion SvO₂ measurement was 70 min (interquartile range [IQR], 58–86 min; range, 45–120 min; mean, 72.3 min). Hb increased modestly after transfusion, with a paired mean difference of 0.42 g/dL (95% CI, 0.38 to 0.46; P < 0.001). By contrast, the cohort-level mean change in SvO₂ was minimal and not statistically significant, with a paired mean difference of 0.08% points (95% CI, − 0.28 to 0.43; P = 0.671). Several variables showed statistically significant paired differences, including SpO₂, systolic arterial pressure, heart rate, blood temperature, cardiac index, cardiac output, and systemic vascular resistance; however, the absolute changes were small. These findings indicate that one RBC-equivalent transfusion did not produce a uniform cohort-level increase in SvO₂, supporting the need to evaluate heterogeneity of individual-level SvO₂ changes. When ΔSvO₂ was summarized by pre-transfusion SvO₂ category, the category-level distributions did not uniformly exceed the responder threshold of ΔSvO₂ ≥5% points (Fig. 2A). In contrast, the proportion of individual responders decreased progressively as pre-transfusion SvO₂ increased, from 42.6% in patients with SvO₂ <60% to 3.8% in those with SvO₂ ≥75% (Fig. 2B). The predicted probability curve showed a similar inverse association (Fig. 2C). Thus, lower pre-transfusion SvO₂ should be interpreted as increased probability of individual-level response, rather than indicating a uniform category-level response.

Table 2Hemodynamic and respiratory variables before and after one RBC-equivalent transfusionVariablePre-transfusionPost-transfusionPaired meandifference95% CIP valueHemoglobin, g/dL9.82 ± 0.9210.24 ± 0.980.420.38 to 0.46< 0.001Oxygen saturation (SpO₂), %98.57 ± 1.5298.34 ± 2.87-0.22-0.39 to -0.060.009Arterial blood pressure, systolic, mmHg102.56 ± 16.43103.37 ± 16.780.810.09 to 1.530.028Arterial blood pressure, diastolic, mmHg53.59 ± 9.8353.64 ± 9.980.05-0.35 to 0.460.798Mean arterial pressure, mmHg68.97 ± 11.1869.07 ± 9.060.10-0.47 to 0.670.723Heart rate, beats/min94.76 ± 14.0793.89 ± 14.50-0.87-1.34 to -0.41< 0.001Respiratory rate, breaths/min14.88 ± 4.9514.91 ± 5.000.03-0.24 to 0.300.841Blood temperature, °C37.05 ± 0.8837.33 ± 0.770.280.26 to 0.31< 0.001Mixed venous oxygen saturation (SvO₂), %73.79 ± 9.9173.86 ± 9.370.08-0.28 to 0.430.671Pulmonary artery pressure, systolic, mmHg27.15 ± 6.9327.15 ± 6.660.00-0.27 to 0.270.995Pulmonary artery pressure, diastolic, mmHg17.19 ± 4.6917.14 ± 4.55-0.05-0.27 to 0.170.668Pulmonary artery pressure, mean, mmHg21.01 ± 6.0520.96 ± 5.23-0.05-0.31 to 0.220.727Mean right atrial pressure, mmHg11.59 ± 16.7112.39 ± 22.060.80-0.51 to 2.110.229Cardiac index, L/min/m²3.03 ± 2.662.87 ± 1.17-0.16-0.29 to -0.020.020Cardiac output, L/min4.79 ± 1.574.67 ± 1.49-0.12-0.16 to -0.07< 0.001Systemic vascular resistance, dyn·s·cm⁻⁵1253.58 ± 511.191275.16 ± 492.0821.580.47 to 42.690.045Values are presented as mean ± SD unless otherwise indicated. Paired mean differences were calculated as post-transfusion minus pre-transfusion values. P values were obtained using paired t tests. Analyses were performed using patients with paired pre- and post-transfusion values for each variable. SpO₂, oxygen saturation measured by pulse oximetry; SvO₂, mixed venous oxygen saturation

Table 2Hemodynamic and respiratory variables before and after one RBC-equivalent transfusionVariablePre-transfusionPost-transfusionPaired meandifference95% CIP valueHemoglobin, g/dL9.82 ± 0.9210.24 ± 0.980.420.38 to 0.46< 0.001Oxygen saturation (SpO₂), %98.57 ± 1.5298.34 ± 2.87-0.22-0.39 to -0.060.009Arterial blood pressure, systolic, mmHg102.56 ± 16.43103.37 ± 16.780.810.09 to 1.530.028Arterial blood pressure, diastolic, mmHg53.59 ± 9.8353.64 ± 9.980.05-0.35 to 0.460.798Mean arterial pressure, mmHg68.97 ± 11.1869.07 ± 9.060.10-0.47 to 0.670.723Heart rate, beats/min94.76 ± 14.0793.89 ± 14.50-0.87-1.34 to -0.41< 0.001Respiratory rate, breaths/min14.88 ± 4.9514.91 ± 5.000.03-0.24 to 0.300.841Blood temperature, °C37.05 ± 0.8837.33 ± 0.770.280.26 to 0.31< 0.001Mixed venous oxygen saturation (SvO₂), %73.79 ± 9.9173.86 ± 9.370.08-0.28 to 0.430.671Pulmonary artery pressure, systolic, mmHg27.15 ± 6.9327.15 ± 6.660.00-0.27 to 0.270.995Pulmonary artery pressure, diastolic, mmHg17.19 ± 4.6917.14 ± 4.55-0.05-0.27 to 0.170.668Pulmonary artery pressure, mean, mmHg21.01 ± 6.0520.96 ± 5.23-0.05-0.31 to 0.220.727Mean right atrial pressure, mmHg11.59 ± 16.7112.39 ± 22.060.80-0.51 to 2.110.229Cardiac index, L/min/m²3.03 ± 2.662.87 ± 1.17-0.16-0.29 to -0.020.020Cardiac output, L/min4.79 ± 1.574.67 ± 1.49-0.12-0.16 to -0.07< 0.001Systemic vascular resistance, dyn·s·cm⁻⁵1253.58 ± 511.191275.16 ± 492.0821.580.47 to 42.690.045Values are presented as mean ± SD unless otherwise indicated. Paired mean differences were calculated as post-transfusion minus pre-transfusion values. P values were obtained using paired t tests. Analyses were performed using patients with paired pre- and post-transfusion values for each variable. SpO₂, oxygen saturation measured by pulse oximetry; SvO₂, mixed venous oxygen saturation

Hemodynamic and respiratory variables before and after one RBC-equivalent transfusion

Values are presented as mean ± SD unless otherwise indicated. Paired mean differences were calculated as post-transfusion minus pre-transfusion values. P values were obtained using paired t tests. Analyses were performed using patients with paired pre- and post-transfusion values for each variable. SpO₂, oxygen saturation measured by pulse oximetry; SvO₂, mixed venous oxygen saturation

Fig. 2Association between pre-transfusion SvO₂ and acute individual-level SvO₂ response after one RBC-equivalent transfusion. Panel A shows the distribution of ΔSvO₂ across pre-transfusion SvO₂ categories. The dashed horizontal line represents the individual-level responder definition of ΔSvO₂ ≥ 5 percentage points and does not imply that the mean or median ΔSvO₂ within each category exceeded 5 percentage points. Panel B shows the proportion of individual responders within each pre-transfusion SvO₂ category. Panel C shows the predicted probability of individual-level SvO₂ response according to pre-transfusion SvO₂ using a logistic spline model. The vertical dashed line indicates the exploratory Youden-derived cutoff of 69%, which should not be interpreted as a clinical transfusion trigger

Association between pre-transfusion SvO₂ and acute individual-level SvO₂ response after one RBC-equivalent transfusion. Panel A shows the distribution of ΔSvO₂ across pre-transfusion SvO₂ categories. The dashed horizontal line represents the individual-level responder definition of ΔSvO₂ ≥ 5 percentage points and does not imply that the mean or median ΔSvO₂ within each category exceeded 5 percentage points. Panel B shows the proportion of individual responders within each pre-transfusion SvO₂ category. Panel C shows the predicted probability of individual-level SvO₂ response according to pre-transfusion SvO₂ using a logistic spline model. The vertical dashed line indicates the exploratory Youden-derived cutoff of 69%, which should not be interpreted as a clinical transfusion trigger

Prediction of SvO₂ response

The predictive ability of pre-transfusion SvO₂ for individual-level SvO₂ response was moderate, with an AUC of 0.778 (95% CI, 0.740–0.816) (Fig. 3). The exploratory Youden-derived cutoff was 69%, with a sensitivity of 0.68 and specificity of 0.75. For reference, applying the 65% cutoff used in the prior study by Zeroual et al. using ScvO₂ guidance to this dataset yielded a sensitivity of 0.52 and a specificity of 0.86 [4]. In an exploratory comparison of candidate predictors, pre-transfusion SvO₂ showed higher discriminative ability than pre-transfusion DO₂i (AUC, 0.684), cardiac index (AUC, 0.674), Hb (AUC, 0.548), LVEF (AUC, 0.519), and SpO₂ (AUC, 0.478) (Figure S4).

Fig. 3ROC curve predicting individual-level SvO₂ response after one RBC-equivalent transfusion based on pre-transfusion SvO₂. The area under the curve (AUC) was 0.778 (95% CI, 0.74–0.816). The plotted cutoff of 69% corresponded to a sensitivity of 0.68 and a specificity of 0.75. The cutoff of 69% was derived using the Youden index and should be interpreted as an exploratory internally derived cutoff, not as a clinical transfusion trigger

ROC curve predicting individual-level SvO₂ response after one RBC-equivalent transfusion based on pre-transfusion SvO₂. The area under the curve (AUC) was 0.778 (95% CI, 0.74–0.816). The plotted cutoff of 69% corresponded to a sensitivity of 0.68 and a specificity of 0.75. The cutoff of 69% was derived using the Youden index and should be interpreted as an exploratory internally derived cutoff, not as a clinical transfusion trigger

Multivariable analysis

In a multivariable logistic regression analysis with SvO₂ responsiveness as the outcome, only pre-transfusion SvO₂ was independently associated with SvO₂ responsiveness (Table 3; Fig. 4).

Table 3Multivariable logistic regression analysis for SvO₂ response (ΔSvO₂ ≥ 5% points)VariableAdjusted OR95% CIP valueAge (per year)1.000.99–1.020.86Male sex0.860.56–1.330.49Hemoglobin (g/dL)0.970.74–1.260.83Baseline SvO₂ (per 1% increase)0.890.86–0.91< 0.001SpO₂ (%)0.950.84–1.090.45Cardiac index (L/min/m²)0.950.90–1.080.17Ejection fraction (%)1.000.98–1.010.83The outcome variable was post-transfusion SvO₂ response (ΔSvO₂ ≥ 5% points). All covariates were measured pre-transfusion (baseline). Left ventricular ejection fraction was obtained from the most recent transthoracic or transesophageal echocardiogram (n = 1204/1352)

Table 3Multivariable logistic regression analysis for SvO₂ response (ΔSvO₂ ≥ 5% points)VariableAdjusted OR95% CIP valueAge (per year)1.000.99–1.020.86Male sex0.860.56–1.330.49Hemoglobin (g/dL)0.970.74–1.260.83Baseline SvO₂ (per 1% increase)0.890.86–0.91< 0.001SpO₂ (%)0.950.84–1.090.45Cardiac index (L/min/m²)0.950.90–1.080.17Ejection fraction (%)1.000.98–1.010.83The outcome variable was post-transfusion SvO₂ response (ΔSvO₂ ≥ 5% points). All covariates were measured pre-transfusion (baseline). Left ventricular ejection fraction was obtained from the most recent transthoracic or transesophageal echocardiogram (n = 1204/1352)

Multivariable logistic regression analysis for SvO₂ response (ΔSvO₂ ≥ 5% points)

The outcome variable was post-transfusion SvO₂ response (ΔSvO₂ ≥ 5% points). All covariates were measured pre-transfusion (baseline). Left ventricular ejection fraction was obtained from the most recent transthoracic or transesophageal echocardiogram (n = 1204/1352)

Fig. 4Forest plot of multivariable logistic regression with SvO₂ response as the target variable. Points indicate adjusted odds ratios, and whiskers indicate 95% confidence intervals. Continuous variables are shown as odds ratios per unit increase

Forest plot of multivariable logistic regression with SvO₂ response as the target variable. Points indicate adjusted odds ratios, and whiskers indicate 95% confidence intervals. Continuous variables are shown as odds ratios per unit increase

Subgroup analysis and sensitivity analysis

Based on available electronic transfusion records, no included patients received additional RBC transfusion or other blood product administration between the index transfusion start and the post-transfusion SvO₂ measurement or within 6 h after the index transfusion start (0/1,352 patients for both windows). Therefore, the final cohort met the prespecified clean-window criterion for concomitant transfusion-related interventions. No patients met the operational definition of suspected active major bleeding based on available electronic transfusion records. Subgroup analyses were performed by modifying the lower limit of pre-transfusion Hb (Hb ≥ 9, Hb ≥ 10, Hb ≥ 11, and Hb ≥ 12 g/dL), and ΔSvO₂ was evaluated by pre-transfusion SvO₂ category within each stratum (Figure S1). In exploratory subgroup analyses, the inverse association between lower pre-transfusion SvO₂ and greater individual-level SvO₂ increase was visually preserved in the Hb ≥ 9 g/dL and Hb ≥ 10 g/dL strata. However, these subgroup findings should be interpreted cautiously because transfusion indications were unavailable and sample sizes decreased at higher Hb thresholds. In the Hb ≥ 11 g/dL and Hb ≥ 12 g/dL strata, the number of patients was limited, particularly in the lower pre-transfusion SvO₂ categories, precluding robust interpretation. In the exploratory overlap-weighted transfusion-versus-no-transfusion trajectory analysis, covariate balance substantially improved after weighting, with all weighted absolute standardized mean differences below 0.1 (Table S1A and Figure S2A). Over 0–6 h, trajectories of SvO₂, Hb, and DO₂i differed between groups (all group-by-time interaction P < 0.001) (Figure S2B). In an additional exploratory check within the overlap-weighted cohort, the baseline SvO₂-dependent 1-hour ΔSvO₂ pattern was more pronounced in transfused patients than in non-transfused controls (Supplementary Table S3). However, because these comparisons may be affected by postoperative physiologic trajectory, clinician selection, and unmeasured time-dependent interventions, they should be interpreted as exploratory rather than as causal estimates of transfusion effect. Furthermore, in multivariable linear regression analyzing ΔSvO₂ as a continuous variable, CI and SpO₂ were significantly associated with ΔSvO₂, in addition to baseline SvO₂, consistent with the primary analysis (Table S2, Figure S3). Additional supplementary analyses showed low variance inflation factors in the primary multivariable model and preserved the association between lower pre-transfusion SvO₂ and individual-level SvO₂ response across alternative model specifications (Figures S5 and S6).

Discussion

In this single-center retrospective cohort of cardiovascular surgical ICU patients, mean cohort-level SvO₂ changed minimally after one RBC-equivalent transfusion, despite a modest increase in Hb. Therefore, our findings should not be interpreted as demonstrating a uniform physiologic effect of RBC transfusion across the cohort. Rather, the principal finding was heterogeneity of acute SvO₂ change: lower pre-transfusion SvO₂ was associated with a higher probability of an individual-level increase in SvO₂. The wide distribution of ΔSvO₂ across baseline SvO₂ categories further supports substantial interindividual heterogeneity in the acute venous oximetry response. This variability may reflect differences in baseline oxygen extraction reserve, cardiac output response, measurement timing, and unmeasured concurrent non-transfusion interventions, rather than a uniform effect of RBC transfusion. Importantly, an increase in SvO₂ represents a global venous oximetry signal and does not necessarily indicate improved microcirculatory oxygen delivery, cellular oxygen utilization, or patient-centered clinical benefit. Conversely, in a patient with VO₂/DO₂ dependency, an increase in DO₂ after transfusion could be accompanied by increased VO₂ and improved tissue metabolism, such as lactate reduction, with little or no increase in SvO₂. Similarly, attenuation of compensatory tachycardia or improvement in hemodynamic reserve may be clinically relevant even when the SvO₂ change is small. Therefore, absence of a large SvO₂ rise does not necessarily exclude physiologic benefit. The 69% cutoff was exploratory and internally derived for predicting ΔSvO₂ ≥5% points in this cohort and should not be interpreted as an SvO₂-based transfusion trigger. These findings should, therefore, be viewed as descriptive physiologic observations, not as evidence supporting an SvO₂-guided transfusion indication. Although the association appeared to persist in patients with Hb ≥ 9 g/dL and Hb ≥ 10 g/dL, these subgroup analyses were exploratory and should not be interpreted as supporting transfusion at higher Hb levels.

Figure 2 clarifies an important distinction between category-level changes and individual-level response. The ΔSvO₂ ≥5%-point threshold was used as an individual-level responder definition and should not be interpreted as a category-level mean target. Although lower pre-transfusion SvO₂ categories had higher responder proportions, not all patients in these categories responded, and the mean or median ΔSvO₂ within categories did not uniformly exceed 5% points. Therefore, the main observation is not that patients below a specific SvO₂ threshold uniformly responded to transfusion, but that lower pre-transfusion SvO₂ was associated with a higher probability of individual-level SvO₂ increase.

The overlap-weighted comparison with non-transfused patients should also be interpreted cautiously. The downward SvO₂ trajectory observed in non-transfused patients may reflect postoperative physiologic evolution, clinician selection, or residual time-dependent confounding rather than the absence of transfusion itself. Therefore, this analysis was retained as an exploratory sensitivity analysis and was not used to infer a causal effect of transfusion.

Both SvO₂ and ScvO₂ reflect the balance between systemic oxygen delivery (DO₂) and oxygen consumption (VO₂). SvO₂ is measured in the pulmonary artery and represents mixed venous blood from the whole circulation, whereas ScvO₂ is typically obtained from a central venous catheter in the superior vena cava. Although ScvO₂ is often used as a less invasive surrogate, the two measures are not interchangeable [8, 9], and SvO₂ may better reflect whole-body oxygen supply-demand balance [10]. Since DO₂ is primarily determined by cardiac output, arterial oxygen saturation, and Hb, it is theoretically expected that, if VO₂ remains constant, an increase in Hb due to transfusion will increase DO₂, resulting in an observable rise in SvO₂. In this study, lower pre-transfusion SvO₂ was associated with a higher probability of an acute individual-level SvO₂ increase in SvO₂ post-transfusion. This finding is consistent with the physiologic expectation that patients with lower baseline SvO₂ may be more likely to show an acute venous oximetry response after an increase in oxygen-carrying capacity, but it should not be interpreted as support for transfusion based on SvO₂ alone. Conversely, the group with higher pre-transfusion SvO₂ showed minimal change in SvO₂ post-transfusion. Elevated SvO₂ values may indicate either (i) that oxygen supply already sufficiently meets tissue oxygen demand, making additional increases in DO₂ unlikely to enhance oxygen extraction, or (ii) that peripheral oxygen extraction and utilization capacity are impaired, leading to residual oxygen in the venous system [11]. As a pathophysiological explanation for (ii), vasoplegic syndrome may develop in the acute postoperative period after cardiovascular surgery requiring initial transfusion, in the setting of cardiopulmonary bypass exposure and ischemia-reperfusion–related inflammation. This may lead to circulatory abnormalities mediated by excessive nitric oxide (NO) production, as well as impaired oxygen utilization due to mitochondrial dysfunction [12–14]. Stored RBCs may also interact with NO biology, including NO scavenging by cell-free hemoglobin or RBC-derived products, which could influence vascular tone and peripheral oxygen unloading [15]. The volume load associated with an RBC-equivalent transfusion may also influence preload, stroke volume, cardiac output, and arterial oxygenation, thereby affecting SvO₂ independently of changes in Hb or tissue oxygen utilization. These mechanisms were not measured in the present study but may partly explain why changes in Hb, DO₂, and SvO₂ are not directly interchangeable with improved tissue oxygenation. This study was conducted in a high-volume cardiovascular surgical ICU in Japan, where the ICU mortality rate was 3.6% and patient severity was high. Some patients may have had postoperative vasoplegic or inflammatory physiology related to cardiopulmonary bypass exposure and ischemia-reperfusion; however, vasoplegia was not systematically captured in the registry, and this explanation remains speculative. Furthermore, stored RBCs may exhibit a leftward shift in the oxygen dissociation curve (decreased P50) and reduced peripheral oxygen release due to decreased 2,3-DPG levels. Scott et al. reported that stored RBCs exhibit a marked decrease in 2,3-DPG and a reduction in P50, and that it takes several days for the patient’s blood 2,3-DPG to recover after transfusion with stored RBCs [16]. Postoperative anemia may also reflect blood loss anemia and iron-restricted erythropoiesis, which can influence postoperative recovery and anemia management beyond the hemoglobin concentration itself [17, 18]. Because iron status, perioperative iron therapy, and erythropoietic response were not available in our registry, the present analysis could not evaluate these mechanisms. These findings represent one potential mechanism by which elevated Hb does not necessarily directly improve tissue oxygenation. Additionally, circulatory changes associated with transfusion (attenuated compensatory tachycardia and cardiac output) may prevent Hb elevation from leading to increased DO₂. Overall, Table 2 showed small decreases in HR, CO, and CI after transfusion, whereas the exploratory overlap-weighted trajectory analysis showed increases in Hb and DO₂i and maintenance or slight increase in SvO₂ in the transfusion group. This exploratory trajectory analysis should not be interpreted as a causal estimate of transfusion effect. Although restrictive transfusion is generally recommended [1], uncertainty remains in some cardiovascular populations [19], underscoring the need to interpret physiologic markers in the overall clinical context rather than as isolated triggers. This study provides a large real-world description of pulmonary artery catheter-derived SvO₂ changes after one RBC-equivalent transfusion across a broad pre-transfusion Hb range. A key strength is the use of directly measured hemodynamic data obtained during routine cardiovascular surgical ICU practice. The main practical implication is cautionary rather than directive: although lower baseline SvO₂ identified patients more likely to show an acute venous oximetry rise, one RBC-equivalent transfusion produced minimal average SvO₂ change, and the observed response was not linked to tissue oxygenation or clinical outcomes. These data do not support using SvO₂, including the internally derived 69% cutoff, as an indication for RBC transfusion. When already available from pulmonary artery catheter monitoring, SvO₂ may help describe oxygen supply–demand physiology, but transfusion decisions should remain grounded in Hb, bleeding status, hemodynamics, symptoms and signs, and the overall clinical context.

Limitations

This study has several limitations. First, it was a single-center retrospective observational study, and unmeasured confounding and selection bias cannot be excluded. Second, the analysis was limited to patients with pulmonary artery catheter monitoring and paired SvO₂ measurements, which limits the generalizability of the findings to patients without invasive hemodynamic monitoring. Third, transfusion indications were not systematically recorded in the electronic dataset, and no standardized transfusion protocol was applied during the study period. Therefore, we cannot determine whether each transfusion was clinically appropriate, particularly in patients transfused at Hb levels above 10 g/dL. This limitation precludes any inference that patients with higher Hb levels should receive RBC transfusion based on SvO₂ alone. Our findings should be interpreted only as describing acute physiologic SvO₂ changes among patients who were transfused in routine practice. Fourth, although no additional RBC transfusion or other blood product administration was identified between the start of the index transfusion and post-transfusion SvO₂ measurement or within 6 h based on available electronic transfusion records, detailed bleeding status, chest drainage volume, fluid administration, vasopressor or inotrope dose changes, ventilator adjustments, and clinician-documented transfusion indications were not uniformly available. These non-transfusion interventions may have influenced SvO₂, cardiac output, and DO₂i during the observation window, and residual time-dependent confounding could not be excluded, particularly in the exploratory transfusion-versus-no-transfusion trajectory analysis. In addition, several clinically relevant variables requested for physiologic interpretation, including SOFA score, serial lactate or lactate clearance, sepsis or septic shock status, postoperative echocardiographic cardiac function, RBC storage age, coronary artery disease, detailed comorbidities, ischemic events, ventilator liberation, vasopressor weaning, and organ dysfunction outcomes, were not consistently available in the registry used for this study. Therefore, these variables could not be incorporated into the multivariable models or used to determine whether SvO₂ changes were associated with tissue hypoxia, organ perfusion, or clinical recovery. Fifth, post-transfusion SvO₂ measurements were not obtained at a strictly standardized time point after transfusion initiation; the observed measurement times ranged from 45 to 120 min, with a median of 70 min (IQR, 58–86 min). This variability may have affected the estimation of acute SvO₂ changes. Sixth, because ΔSvO₂ includes baseline SvO₂ in its calculation, analyses using ΔSvO₂ may be affected by regression to the mean or mathematical coupling. To address this concern, we performed a sensitivity analysis using continuous ΔSvO₂ as the outcome; however, this issue cannot be fully eliminated in an observational analysis of change from baseline. Finally, this study was primarily intended to characterize acute physiologic SvO₂ changes after transfusion rather than to demonstrate clinical benefit. SvO₂ is a global venous oximetry measure and does not directly assess microcirculatory perfusion, cellular oxygen utilization, lactate clearance, organ dysfunction, or patient-centered outcomes. Accordingly, an increase in SvO₂ should be interpreted as an acute physiologic signal rather than evidence of improved tissue oxygenation or clinical benefit.

Conclusions

In cardiovascular surgical ICU patients with pre-transfusion Hb ≥ 7.5 g/dL, lower pre-transfusion SvO₂ was associated with a greater probability of an acute individual-level SvO₂ increase after transfusion of one standard Japanese 2-unit RBC product, approximately equivalent to one international RBC unit. However, the average cohort-level change in SvO₂ was minimal despite a modest increase in Hb, and the observed association did not establish improved tissue oxygenation or patient-centered benefit. The internally derived 69% threshold should not be used as a stand-alone transfusion trigger. Overall, these findings support SvO₂ as a descriptive marker of post-transfusion physiologic heterogeneity, but do not support using SvO₂ itself as an indication for RBC transfusion.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1.

Supplementary Material 1.