Highlights
– In a 13‑ICU Korean cohort of 2,613 adults with sepsis, 45.3% received RBC transfusions during ICU days 1–3; mean nadir hemoglobin on the transfusion day was 8.8 ± 1.7 g/dL.
– Transfusion was more likely in patients with greater physiologic derangement (higher SAPS 3, septic shock, higher lactate, multi‑vasopressor use, coagulopathy, renal replacement therapy).
– In propensity‑matched analyses there was no overall 60‑day mortality difference between transfused and non‑transfused patients, but a qualitative interaction at an ICU day‑1 hemoglobin threshold of 10 g/dL (p for interaction = 0.0124): transfusion associated with higher mortality when pre‑transfusion Hb ≥10 g/dL and a trend toward lower mortality when Hb <10 g/dL.
– Clinical implication: avoid liberal transfusion when Hb ≥10 g/dL; decisions below ~9–10 g/dL should integrate physiologic markers and individualized judgment, pending confirmatory trials.
Background
Red blood cell (RBC) transfusion is a common intervention during the early resuscitation of critically ill patients with sepsis. Historically, transfusion practice varied widely, from liberal strategies guided by symptom relief and Hb targets near 10 g/dL to restrictive policies favoring thresholds around 7 g/dL. Randomized trials in general critical illness (TRICC) and septic shock (TRISS) support restrictive transfusion thresholds in many settings, and international guidelines (Surviving Sepsis Campaign) generally recommend a hemoglobin threshold of 7 g/dL for hemodynamically stable patients without active ischemia or severe hypoxemia. Nonetheless, clinicians frequently transfuse based on severity, perfusion markers, and comorbid conditions, creating a persistent practice–guideline gap.
Study design
This report is a secondary analysis of the Korean Sepsis Alliance prospective multicenter cohort (September 2019–December 2022), including 13 intensive care units across South Korea. Adults aged ≥19 years admitted to the ICU with sepsis were classified as transfused or non‑transfused according to whether they received RBC transfusion during ICU days 1–3 (the early resuscitation phase). No experimental interventions were applied. The primary analysis examined associations between early transfusion and 60‑day mortality, using propensity score matching to address confounding by indication. The investigators also explored interactions according to hemoglobin thresholds and reported clinical and laboratory correlates of transfusion.
Key findings
Population and transfusion frequency
The cohort comprised 2,613 patients (mean age 71.9 ± 13.3 years). Nearly half (45.3%) received at least one RBC transfusion during ICU days 1–3, indicating that transfusion during early sepsis care remains common in contemporary practice.
Hemoglobin values and transfusion practice
The mean lowest hemoglobin on the day of transfusion was 8.8 ± 1.7 g/dL. Despite this average, transfusion decisions were not driven solely by hemoglobin: patients receiving RBCs exhibited greater physiologic derangement compared with non‑transfused patients.
Clinical correlates of transfusion
Variables independently associated with receiving RBC transfusion included higher Simplified Acute Physiology Score 3 (SAPS 3), higher Sepsis‑Induced Coagulopathy scores, presence of septic shock, requirement for two or more vasopressors, continuous renal replacement therapy (CRRT), lower platelet counts, higher lactate and bilirubin levels, and undergoing surgical source control. These associations illustrate that clinicians integrate measures of organ dysfunction, perfusion, and bleeding/coagulopathy risk when deciding to transfuse.
Mortality outcomes and threshold interaction
In the propensity score‑matched cohort (matching transfused to non‑transfused patients on observed baseline covariates), there was no overall difference in 60‑day mortality between groups. However, stratified analyses revealed a significant qualitative interaction when patients were grouped by ICU day‑1 hemoglobin at a 10 g/dL threshold (p for interaction = 0.0124). Specifically:
- When baseline (ICU day‑1) hemoglobin was ≥10 g/dL, RBC transfusion was associated with higher 60‑day mortality.
- When baseline hemoglobin was <10 g/dL, transfusion showed a trend toward reduced mortality, though this was not definitively statistically significant in the matched analysis.
The implication is that transfusion may be neutral overall but potentially harmful when applied at higher Hb levels and possibly beneficial when used in patients with more pronounced anemia.
Safety and additional outcomes
As an observational study, detailed adjudication of transfusion‑related complications (e.g., TRALI, TACO, transfusion‑associated immunomodulation) and unit characteristics (leukoreduction status, storage duration) were not central to the report and thus are not fully characterized. The analysis focused on mortality as the primary patient‑centered outcome.
Expert commentary and interpretation
The findings by Song et al. reinforce several clinically relevant points. First, RBC transfusion remains common early in sepsis management, reflecting clinicians’ responses to hemodynamic instability and organ dysfunction beyond static hemoglobin numbers. Second, the lack of a clear overall mortality advantage with early transfusion aligns with randomized evidence favoring restrictive thresholds in critical illness but complicates the translation to heterogeneous septic populations.
The observed interaction at an Hb of 10 g/dL is noteworthy. It suggests potential harm when transfusing patients who are not substantially anemic and who may not benefit from increased oxygen‑carrying capacity. Known harms from transfusion—volume overload (TACO), immune modulation, transfusion reactions, and potential pro‑inflammatory effects—provide biological plausibility for worse outcomes when transfusion is applied liberally. Conversely, in more anemic patients (lower Hb), transfusion may restore oxygen delivery and thus plausibly reduce organ ischemia risk, which could explain the trend toward benefit below the 10 g/dL threshold.
Yet this study cannot establish causality. Residual confounding by indication is the dominant threat: patients selected for transfusion are sicker in ways that may not be fully captured by observed covariates despite propensity matching. The strong associations between transfusion and severity (shock, high lactate, multi‑organ support) indicate that transfusion was a marker of critical illness. Even well‑constructed propensity models are limited to measured variables.
Clinical implications
Given the totality of randomized and observational data, the following pragmatic recommendations are reasonable for most clinicians managing sepsis in the ICU:
- Adhere to restrictive transfusion strategies in the absence of active ischemia, severe hypoxemia, or massive hemorrhage. For hemodynamically stable septic patients, a hemoglobin threshold of 7 g/dL is supported by randomized trials (e.g., TRISS) and guideline statements.
- Avoid routine transfusion when hemoglobin is ≥10 g/dL. The current study suggests potential harm when transfusing above this threshold.
- Individualize transfusion decisions in patients with hemoglobin between ~7–10 g/dL by integrating physiologic indicators (persistent elevated lactate, poor perfusion, ongoing bleeding, myocardial ischemia, high vasopressor requirements) rather than relying on a single numeric Hb target.
- When transfusion is given, follow best practices to minimize harm: use leukoreduced blood when available, administer the minimum effective units, monitor for volume overload, and reassess need frequently.
Limitations and research gaps
Key limitations of the study include its observational design (vulnerable to confounding by indication), lack of granular data on transfusion units (volume, storage age, leukoreduction), absence of detailed transfusion indications, and focus on early transfusions only (ICU days 1–3). The cohort was from South Korea and had an older average age (mean 71.9 years), which may affect generalizability to other settings and younger populations.
Important research questions remain: Should certain subgroups of septic patients (for example, profound anemia, refractory shock with high lactate, or coronary artery disease) be managed with higher transfusion thresholds? Can physiologic or biomarker‑guided transfusion algorithms (e.g., using mixed venous oxygen, lactate clearance, microcirculatory measures) improve outcomes? Pragmatic randomized trials targeting the intermediate Hb zone (7–10 g/dL) or comparing physiology‑guided vs. fixed‑threshold strategies would be especially informative.
Conclusion
Song et al. provide contemporary, real‑world data showing that early RBC transfusions are frequent in sepsis and closely associated with illness severity. While propensity‑matched analyses show no overall mortality difference, a meaningful interaction at a hemoglobin threshold of 10 g/dL suggests harm from transfusion when Hb is ≥10 g/dL and a potential benefit when Hb is lower. These findings support continued adherence to restrictive transfusion practices and individualized, physiology‑based decision making for patients with intermediate anemia during sepsis. Confirmatory randomized trials addressing thresholds and physiologic guidance are needed.
Funding and clinicaltrials.gov
Funding and trial registration: As reported by the authors — see the original publication for detailed funding disclosures. This analysis was a secondary analysis of the Korean Sepsis Alliance (KSA) cohort. No clinicaltrials.gov registration is indicated for this observational cohort analysis.
References
1. Song MJ, Seo WI, Jang Y, Park S, Ko R, Suh GY, Oh DK, Lee SY, Park MH, Lim CM, Jung SY, Lim SY; Korean Sepsis Alliance (KSA) Investigators. RBC Transfusion Practices in Critically Ill Patients With Sepsis. Crit Care Med. 2025 Dec 1;53(12):e2596-e2606. doi: 10.1097/CCM.0000000000006892 IF: 6.0 Q1 . Epub 2025 Oct 15. PMID: 41090988 IF: 6.0 Q1 .
2. Holst LB, Haase N, Wetterslev J, et al. Lower versus higher hemoglobin threshold for transfusion in septic shock. N Engl J Med. 2014;371(15):1381–1391. doi:10.1056/NEJMoa1406617 IF: 78.5 Q1 .
3. Hébert PC, Wells G, Blajchman MA, et al. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. N Engl J Med. 1999;340(6):409–417. doi:10.1056/NEJM199902113400601 IF: 78.5 Q1 .
4. Evans L, Rhodes A, Alhazzani W, et al. Surviving Sepsis Campaign: international guidelines for management of sepsis and septic shock 2021. Crit Care Med. 2021;49(11):e1063–e1143. doi:10.1097/CCM.0000000000005337 .
5. Carson JL, Guyatt G, Heddle NM, et al. Clinical practice guidelines from the AABB: red blood cell transfusion thresholds and storage. JAMA. 2016;316(19):2025–2035. doi:10.1001/jama.2016.9185 .