Publication

• Title: Lower versus Higher Hemoglobin Threshold for Transfusion in Septic Shock
• Acronym: TRISS — Transfusion Requirements in Septic Shock
• Year: 2014
• Journal published in: New England Journal of Medicine
• Citation: 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.

Context & Rationale

• Background

  • Anaemia is frequent during septic shock because of inflammation, haemodilution, impaired erythropoiesis, phlebotomy and bleeding; red-cell transfusion was therefore common even without active haemorrhage.
  • Transfusion increases arterial oxygen-carrying capacity, but septic tissue hypoxia may reflect heterogeneous microvascular flow and impaired cellular oxygen use rather than inadequate global oxygen delivery. Red-cell transfusion also carries risks including circulatory overload, lung injury, haemolysis, immune effects and consumption of a finite resource.
  • The Rivers early goal-directed therapy protocol included transfusion to a haematocrit above 30% when central venous oxygen saturation remained below 70% after initial resuscitation. The 2012 Surviving Sepsis Campaign retained this higher target during early resuscitation, while recommending a 7 g/dL transfusion threshold after hypoperfusion had resolved.¹²
  • TRICC had supported a restrictive transfusion strategy in a broad, resuscitated ICU population, but its septic subgroup was not definitive, the trial did not specifically enrol patients with septic shock, and the red cells were not routinely prestorage leucoreduced.³
• Research Question/Hypothesis

  • In adults with septic shock and haemoglobin ≤9 g/dL, would transfusing at a lower threshold of 7 g/dL, rather than a higher threshold of 9 g/dL, alter 90-day mortality?
  • The trial also examined whether reducing red-cell exposure would increase ischaemic events, severe transfusion reactions or the requirement for life support.
• Why This Matters

  • The two thresholds represented common practice but implied markedly different use of donor blood.
  • A reliable septic-shock-specific answer could remove an unproven component of resuscitation, reduce transfusion exposure and simplify clinical guidance.
  • The question tested whether a persuasive physiological rationale—raising haemoglobin and calculated oxygen delivery—translated into better patient-centred outcomes.

Design & Methods

• Research Question: Among adults in the ICU with septic shock and haemoglobin ≤9 g/dL, does a red-cell transfusion threshold of 7 g/dL, compared with 9 g/dL, improve 90-day survival while avoiding excess ischaemia or organ-support requirements?
• Study Type:
  • Investigator-initiated, pragmatic, international, multicentre, stratified, parallel-group, partially blinded randomised clinical trial.
  • Conducted in 32 general ICUs in Denmark, Sweden, Norway and Finland between 3 December 2011 and 26 December 2013.
  • Central computer-generated 1:1 allocation used permuted blocks of 6, 8 or 10 and stratification by study site and active haematological cancer.
  • The trial used a conventional two-sided difference-testing framework; it was not designed to establish non-inferiority or equivalence.⁴
• Population:
  • Randomised: 1005 patients; 998 were included in the primary modified intention-to-treat analysis—502 in the lower-threshold group and 496 in the higher-threshold group.
  • Inclusion criteria: age ≥18 years; ICU admission; septic shock; and point-of-care haemoglobin ≤9 g/dL.
  • Septic shock definition: at least two systemic inflammatory response syndrome criteria, suspected or confirmed infection, and hypotension despite fluid therapy or vasopressor/inotropic therapy to maintain blood pressure.
  • Important exclusions: acute myocardial ischaemia or acute coronary syndrome; life-threatening bleeding; red-cell transfusion during the current ICU admission; acute burn injury; previous serious transfusion reaction; documented refusal of transfusion; withdrawal of active treatment or brain death; previous enrolment; or inability to obtain consent under national law.
  • Baseline severity: median age 67 years in both groups; median SOFA score 10 in both groups; median SAPS II 51 versus 52; and mechanical ventilation in 68.7% versus 70.6%.
• Intervention:
  • One unit of cross-matched, prestorage leucoreduced red cells suspended in saline–adenine–glucose–mannitol whenever haemoglobin was ≤7 g/dL.
  • Haemoglobin was reassessed within 3 hours after completing each unit or before another transfusion was started.
  • The allocated threshold applied throughout the ICU stay, including ICU readmission, to a maximum of 90 days after randomisation.
• Comparison:
  • The same single-unit, prestorage leucoreduced red-cell product was given whenever haemoglobin was ≤9 g/dL, with identical reassessment after each unit.
  • For life-threatening bleeding, ischaemia or extracorporeal membrane oxygenation, the attending clinician could temporarily suspend the protocol and select the transfusion threshold; the allocated strategy was resumed when appropriate.
  • Transfusion during surgery and after ICU discharge, and all other co-interventions, were at clinicians’ discretion. Following publication of the 6S trial, starch solutions were discouraged.
• Blinding: Patients, bedside clinicians and site investigators were aware of allocation because the haemoglobin thresholds could not realistically be concealed. Mortality ascertainment, the trial statistician and the data and safety monitoring committee were blinded; this protected the objective primary outcome more effectively than the clinician-diagnosed secondary safety outcomes.
• Statistics:
  • Power calculation: A total of 1000 patients was required to detect a 9-percentage-point absolute reduction in 90-day mortality—from 45% to 36%, corresponding to a 20% relative reduction—with 80% power at a two-sided 5% significance level.
  • The primary analysis used logistic regression adjusted for the stratification variables—site and active haematological cancer—with conversion of odds ratios to relative risks.
  • The authors termed the primary population intention-to-treat; because seven randomised patients were removed from all analyses, it is more accurately described as modified intention-to-treat. Fully adjusted and three per-protocol sensitivity analyses were prespecified.
  • One interim analysis was planned after 500 patients had completed 90-day follow-up, using a Haybittle–Peto stopping criterion of P<0.001.
• Follow-Up Period: The primary report followed patients for 90 days after randomisation; the allocated ICU transfusion strategy continued until ICU discharge, death or day 90. Prespecified one-year mortality and health-related quality-of-life outcomes were reported subsequently.

Key Results

This trial was not stopped early. It continued after the prespecified interim analysis and completed the planned sample.

• The lower threshold produced substantial and sustained separation in haemoglobin and red-cell exposure: median 1 versus 4 units, with approximately half as many transfusions overall.
• Routine transfusion at 9 g/dL did not produce a statistically significant improvement in 90-day survival, life-support use, ischaemic events or days alive outside hospital.
• The primary confidence interval did not establish exact equivalence: it remained compatible with a 22% relative mortality reduction or a 9% relative mortality increase with the lower threshold.

Internal Validity

• Randomisation and Allocation: The central, concealed, computer-generated sequence, variable block sizes and stratification by site and active haematological cancer make selection bias unlikely. Prognostic balance was good.
• Dropout or Exclusions: Five patients were excluded immediately after randomisation—one had been randomised in error and four withdrew consent—and two further patients withdrew consent for use of their data. The resulting 998-patient analysis was therefore modified rather than strict intention-to-treat. Study treatment was discontinued in 38 lower-threshold and 24 higher-threshold patients, but 90-day vital status was obtained for all 998 analysed patients; secondary-outcome data were missing for only 21 patients.
• Performance/Detection Bias: Open-label care created potential for co-intervention and diagnostic bias. This matters little for registry-derived all-cause mortality, but more for clinician-recognised ischaemia and transfusion reactions. The statistician and mortality assessors remained blinded.
• Protocol Adherence: Among 933 patients assessed for violations, transfusion above the assigned threshold occurred in 45/463 (10%) lower-threshold patients versus 16/470 (3%) higher-threshold patients; omission of transfusion on a day with haemoglobin below the assigned threshold occurred in 42/463 (9%) versus 104/470 (22%). These deviations tended to make the groups more alike rather than create a false separation.⁵
• Baseline Characteristics: Median age was 67 years in both groups, SOFA score 10 in both, SAPS II 51 versus 52, chronic cardiovascular disease 14.9% versus 13.3%, mechanical ventilation 68.7% versus 70.6%, and renal replacement therapy 13.5% versus 10.7%. The observed 90-day mortality of 43.0% versus 45.0% was close to the event rate assumed for sample-size planning.
• Heterogeneity: The trial deliberately enrolled a pragmatic mixture of medical and surgical septic shock across 32 ICUs, with pulmonary and abdominal infection predominating. This clinical heterogeneity strengthens the estimate of the average treatment effect; stratification, baseline balance and absence of convincing subgroup interaction reduce concern that it concealed a large common effect.
• Timing: Median time from ICU admission to randomisation was 23 hours (IQR 7–50) versus 20 hours (IQR 7–43). TRISS therefore tested transfusion during the ICU course rather than a narrowly defined first six-hour resuscitation window. Baseline median lactate was 2.4 versus 2.7 mmol/L and central venous oxygen saturation, where measured, was 69% versus 68%, so the cohort was not enriched for profound ongoing oxygen-delivery failure.
• Dose: Single-unit transfusion with haemoglobin reassessment within 3 hours was a clinically sensible and blood-conserving way to implement the thresholds. It tested two haemoglobin-trigger strategies—not transfusion guided by symptoms, microcirculatory variables or physiological evidence of oxygen debt, and not thresholds below 7 g/dL.
• Separation of the Variable of Interest: Median lowest haemoglobin on day 2 was 7.7 g/dL (IQR 7.1–8.3) versus 9.0 g/dL (IQR 8.8–9.5), and on day 7 was 7.6 g/dL (IQR 7.1–8.1) versus 9.3 g/dL (IQR 8.9–9.8). Median cumulative red-cell exposure was 1 versus 4 units, and 36.1% versus 1.2% received no ICU transfusion. Treatment separation was therefore clinically substantial despite violations.⁵
• Key Delivery Aspects: On day 1, 81/488 (17%) lower-threshold patients and 455/489 (93%) higher-threshold patients were transfused. The product was modern, prestorage leucoreduced red cells. Transfusion during surgery was not protocolised; 43 patients—28 versus 15—were transfused above their allocated threshold during surgery, which could modestly attenuate the treatment contrast.
• Crossover: Temporary protocol suspension occurred in 29/488 (5.9%) lower-threshold patients and 11/489 (2.2%) higher-threshold patients, usually for life-threatening bleeding or ischaemia. This safety mechanism was appropriate but again biased towards convergence of the strategies rather than exaggerating a lower-threshold benefit.
• Adjunctive Therapy Use: Major co-interventions were broadly balanced: albumin was given to 306/487 (63%) versus 303/489 (62%), synthetic colloids to 16/487 (3%) versus 15/489 (3%), plasma to 113/488 (23%) versus 127/489 (26%), and platelets to 79/488 (16%) versus 96/489 (20%). Median cumulative non-blood fluid volume was 14,128 versus 14,778 mL. Not every possible co-intervention was captured.
• Outcome Assessment: All-cause mortality at 90 days was objective, clinically important and nearly immune to ascertainment bias. Life support was clearly defined. Ischaemic events required clinical recognition and treatment, myocardial surveillance was not protocolised, and silent cerebral ischaemia or cognitive injury was not measured.
• Statistical Rigour: The event rate and final sample met the assumptions for detecting the prespecified large effect. The final analysis plan preceded database analysis, and unadjusted, fully adjusted and three per-protocol analyses were concordant.⁵ The trial was not powered to rule out smaller mortality differences, rare adverse events or clinically important subgroup effects; secondary and subgroup analyses were not multiplicity-adjusted.

Conclusion on Internal Validity: Internal validity is strong for the primary comparison of 90-day mortality and for the effect on transfusion exposure. It is moderate for ischaemic safety, rare transfusion reactions and subgroup claims because those outcomes were less completely blinded, less systematically ascertained and substantially underpowered.

External Validity

• Population Representativeness: The cohort was clinically recognisable: median age 67 years, all had septic shock, median SOFA score 10, mechanical ventilation in 68.7% versus 70.6%, mixed medical and surgical admissions, and 90-day mortality of 43.0% versus 45.0%. Both university and non-university hospitals participated, and 1005 of 1224 screened patients were randomised.
• Important Exclusions: The findings do not directly apply to acute coronary syndrome, active life-threatening haemorrhage, burns, previous serious transfusion reactions, patients already transfused during the current ICU admission, or those for whom active treatment was being withdrawn.
• Phase of Illness: Because randomisation generally occurred after initial ICU resuscitation rather than exclusively during the first hours of shock, the trial is most applicable to ongoing ICU transfusion decisions. It does not definitively answer whether a selected patient with severe anaemia and persistent early oxygen-delivery failure benefits from an earlier higher target.
• Healthcare Setting and Product: All sites were in high-income Scandinavian systems using point-of-care haemoglobin testing, modern blood banking and prestorage leucoreduced red cells. A 7 g/dL strategy is attractive where blood is scarce, but exact safety may be less certain where monitoring, rescue transfusion or blood-product quality differ materially.
• Contemporary Sepsis Definition: Enrolment used SIRS-era septic-shock criteria. A subsequent exploratory analysis applying Sepsis-3 shock criteria found no interaction, but this cannot fully substitute for a prospectively enrolled contemporary cohort.

Conclusion on External Validity: Generalisability is high for non-bleeding adults with septic shock in modern ICUs who do not have acute myocardial ischaemia. It is limited for acute coronary syndromes, major bleeding, very early refractory hypoperfusion, highly selected oncological populations and settings without comparable blood products or monitoring.

Strengths & Limitations

• Strengths:
  • Large, investigator-initiated, multicentre randomised trial conducted across 32 ICUs and four countries.
  • Central concealed allocation with clinically appropriate stratification and good baseline balance.
  • Objective, patient-important primary outcome with complete 90-day vital-status ascertainment in the analysed cohort.
  • Pragmatic delivery using single units of modern prestorage leucoreduced red cells.
  • Excellent separation in haemoglobin and transfusion exposure, despite protocol deviations that generally biased towards the null.
  • Prespecified analysis plan, planned interim monitoring, blinded statistician and consistent adjusted and per-protocol analyses.
  • Direct quantification of blood conservation: median 1 versus 4 units and 1545 versus 3088 transfusions.
• Limitations:
  • Open-label bedside care, with residual risk of performance bias and detection bias for secondary outcomes.
  • Seven randomised patients were excluded from all analyses, making the primary population modified rather than strict intention-to-treat.
  • Designed to detect a large 9-percentage-point mortality difference, not to establish non-inferiority or exact equivalence.
  • Underpowered for ischaemic events, rare transfusion reactions and subgroup effects; the numerical myocardial-ischaemia imbalance was post hoc and imprecise.
  • No routine myocardial-ischaemia surveillance, brain imaging or dedicated cognitive assessment.
  • Excluded acute coronary syndrome and life-threatening bleeding—the two settings in which a higher haemoglobin may be most biologically plausible.
  • Median randomisation 20–23 hours after ICU admission and relatively modest baseline lactate limited direct inference about profound early hypoperfusion.
  • Protocol suspensions, treatment discontinuations and non-protocolised perioperative transfusions reduced the purity of the allocated strategies.
  • Scandinavian, high-resource practice and universal leucoreduction may limit transfer to materially different blood systems.

Interpretation & Why It Matters

• Bedside Practice

  For a non-bleeding adult with septic shock, no acute myocardial ischaemia and no other compelling indication, transfusion can usually be deferred until haemoglobin is approximately 7 g/dL.
• What TRISS Established

  A 7 g/dL threshold markedly reduces donor-blood exposure without evidence of worse 90-day mortality, organ-support use, overt ischaemia or hospital-free survival compared with a 9 g/dL threshold.
• What It Did Not Establish

  TRISS did not prove that the strategies are exactly equivalent, that 7 g/dL is optimal for every individual, or that the lower threshold is safe in acute coronary syndrome, major haemorrhage, severe hypoxaemia or profound persistent oxygen debt.
• Physiological Lesson

  Increasing haemoglobin and calculated oxygen delivery does not necessarily increase tissue oxygen use or improve outcomes in sepsis; randomisation captured the total balance of possible benefit, transfusion harm and resource use.
• Resource Stewardship

  Median red-cell exposure was 1 versus 4 units, and total red-cell use was approximately halved, creating a substantial system-level benefit without a detectable clinical penalty in the studied population.

Controversies & Other Evidence

• “Definitive evidence” versus the formal statistical claim: The accompanying editorial declared 7 g/dL the “new normal” and characterised TRISS as definitive.⁶ The practice conclusion is persuasive because the higher threshold used much more blood without a favourable signal, but the trial was powered for a large between-group difference. Failure to detect superiority at 9 g/dL is not formal proof of equivalence or non-inferiority.
• Was the population likely to benefit from increased oxygen-carrying capacity? Correspondence highlighted baseline haemoglobin around 8.4 g/dL, central venous oxygen saturation around 68–69% and lactate around 2.4–2.7 mmol/L, indicating that many patients were not globally oxygen-delivery limited.⁷ Combined with randomisation 20–23 hours after ICU admission, this limits inference about the small subgroup with severe anaemia and persistent early hypoperfusion. The later pooled ARISE, ProCESS and ProMISe evidence found no overall survival advantage from ScvO₂-guided early goal-directed therapy, reducing—but not eliminating—the plausibility of a selective early benefit.⁸
• Silent cerebral ischaemia and cognition: TRISS measured overt, clinically recognised ischaemic events rather than systematic brain imaging or cognitive outcomes.⁷ At one year, mortality was 53.5% versus 54.6% (RR 0.97; 95% CI 0.85 to 1.09; P=0.62), and SF-36 physical and mental component scores were similar among 629 Danish patients, but SF-36 is not a dedicated neurocognitive assessment and does not exclude silent infarction.⁹
• Myocardial ischaemia remains the crucial boundary condition: The post hoc myocardial-ischaemia count was 13/488 versus 6/489 (RR 2.17; 95% CI 0.83 to 5.67; P=0.10), without routine surveillance and with wide uncertainty. Acute coronary syndrome had been excluded. MINT later found recurrent myocardial infarction or death at 30 days in 16.9% with a restrictive strategy versus 14.5% with a liberal strategy (RR 1.15; 95% CI 0.99 to 1.34; P=0.07), leaving clinically important benefit from liberal transfusion plausible.¹⁰ The 2025 AABB acute-myocardial-infarction guideline consequently conditionally recommends a liberal strategy when haemoglobin is below 10 g/dL.¹¹
• Age of stored blood: The correspondence argued that storage lesion might have obscured benefit from transfusion.⁷ ABLE and TRANSFUSE subsequently found no outcome advantage from preferentially fresher rather than standard-issue red cells in critically ill adults, making storage duration an unlikely explanation for the TRISS findings.¹²¹³
• Threshold versus individualised transfusion: A subsequent published debate framed the residual disagreement explicitly. Sakr and Vincent argued that comorbidity, microcirculation and physiological response justify individualisation; Holst, Carson and Perner defended a restrictive haemoglobin threshold as the safest evidence-based default; Docherty and Walsh concluded that the evidence was insufficient to resolve the issue fully.¹⁴¹⁵¹⁶ The defensible synthesis is a restrictive default with explicit clinical exceptions, not a haemoglobin number divorced from context.
• Subgroups did not reveal a reproducible beneficiary of 9 g/dL: Beyond the prespecified age, cardiovascular-disease and SAPS II analyses, exploratory TRISS analyses found no interaction for chronic lung disease (P=0.31), haematological malignancy (P=0.47), metastatic cancer (P=0.51), surgery (P=0.99) or Sepsis-3 septic shock (P=0.20).¹⁷ These analyses were post hoc and underpowered, so absence of interaction is not proof of a uniform treatment effect.
• Oncological septic shock produced discordant evidence: The single-centre TRICOP trial randomised 300 cancer patients with septic shock within six hours of ICU admission. Mortality with the 9 g/dL versus 7 g/dL strategy was 45% versus 56% at 28 days (HR 0.74; 95% CI 0.53 to 1.04; P=0.08) and 59% versus 70% at 90 days (HR 0.72; 95% CI 0.53 to 0.97; P=0.03).¹⁸ Its smaller, single-centre, oncology-only population and discordance between the primary 28-day and later 90-day analyses prevent it from overturning TRISS, but it weakens claims of universal applicability.
• Meta-analytic synthesis: A sepsis-specific meta-analysis of three trials and 1516 patients found 28- or 30-day mortality of 36.4% with liberal versus 36.2% with restrictive transfusion (pooled OR 0.99; 95% CI 0.67 to 1.46), with substantial heterogeneity (I²=61%).¹⁹ The 2025 Cochrane update across clinical settings likewise found that restrictive strategies substantially reduce red-cell exposure without evidence of worse mortality in most populations, while preserving uncertainty in selected groups such as acute myocardial infarction.²⁰
• Current guideline position: The 2025 CHEST guideline strongly recommends a restrictive strategy for most critically ill adults and conditionally suggests against adding a permissive threshold to usual care in septic shock with end-organ hypoperfusion.²¹ The 2026 Surviving Sepsis Campaign strongly recommends a restrictive over a liberal transfusion strategy with moderate-certainty evidence.²² These recommendations preserve clinical judgement for active bleeding, myocardial ischaemia, severe hypoxaemia and other circumstances not adequately tested by TRISS.

Summary

• TRISS randomised 1005 adults with septic shock and haemoglobin ≤9 g/dL in 32 Scandinavian ICUs to single-unit transfusion at thresholds of 7 or 9 g/dL.
• The lower threshold reduced median red-cell exposure from 4 units to 1 unit and reduced total transfusions from 3088 to 1545.
• Death by day 90 occurred in 43.0% versus 45.0% (RR 0.94; 95% CI 0.78 to 1.09; P=0.44), with no statistically significant differences in life support, overt ischaemia, severe transfusion reactions or hospital-free time.
• Primary-outcome validity was strong, but the trial did not establish formal equivalence and was underpowered for rare harms, myocardial ischaemia and subgroup-specific effects.
• The evidence supports 7 g/dL as the default threshold in non-bleeding septic shock without acute myocardial ischaemia; it should not be extrapolated uncritically to acute coronary syndrome, major haemorrhage or profound persistent oxygen debt.

Overall Takeaway

TRISS is a practice-defining critical care trial because it isolated the transfusion threshold in septic shock, achieved major separation in blood exposure and found no patient-important advantage to routinely transfusing at 9 g/dL. It established 7 g/dL as a highly defensible default—not an absolute physiological law—and sharply narrowed the circumstances in which a higher target requires separate evidence or an individual clinical justification.

Bibliography

• 1Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med. 2001;345(19):1368-1377.
• 2Dellinger RP, Levy MM, Rhodes A, et al. Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock, 2012. Intensive Care Med. 2013;39(2):165-228.
• 3Hébert 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.
• 4Holst LB, Haase N, Wetterslev J, et al. Transfusion requirements in septic shock (TRISS) trial—comparing the effects and safety of liberal versus restrictive red blood cell transfusion in septic shock patients in the ICU: protocol for a randomised controlled trial. Trials. 2013;14:150.
• 5Holst LB, Haase N, Wetterslev J, et al. Lower versus higher hemoglobin threshold for transfusion in septic shock: supplementary appendix. N Engl J Med. 2014;371(15):1381-1391.
• 6Hébert PC, Carson JL. Transfusion threshold of 7 g per deciliter—the new normal. N Engl J Med. 2014;371(15):1459-1461.
• 7Beerepoot M, Vogt L, Turgeon RD, et al. Hemoglobin threshold for transfusion in septic shock. N Engl J Med. 2015;372(1):90-92.
• 8Angus DC, Barnato AE, Bell D, et al. A systematic review and meta-analysis of early goal-directed therapy for septic shock: the ARISE, ProCESS and ProMISe Investigators. Intensive Care Med. 2015;41(9):1549-1560.
• 9Rygård SL, Holst LB, Wetterslev J, et al. Long-term outcomes in patients with septic shock transfused at a lower versus a higher haemoglobin threshold: the TRISS randomised, multicentre clinical trial. Intensive Care Med. 2016;42(11):1685-1694.
• 10Carson JL, Brooks MM, Hébert PC, et al. Restrictive or liberal transfusion strategy in myocardial infarction and anemia. N Engl J Med. 2023;389(26):2446-2456.
• 11Pagano MB, Stanworth SJ, Dennis J, et al. Red cell transfusion in acute myocardial infarction: AABB international clinical practice guidelines. Ann Intern Med. 2025;178(10):1469-1477.
• 12Lacroix J, Hébert PC, Fergusson DA, et al. Age of transfused blood in critically ill adults. N Engl J Med. 2015;372(15):1410-1418.
• 13Cooper DJ, McQuilten ZK, Nichol A, et al. Age of red cells for transfusion and outcomes in critically ill adults. N Engl J Med. 2017;377(19):1858-1867.
• 14Sakr Y, Vincent JL. Should red cell transfusion be individualized? Yes. Intensive Care Med. 2015;41(11):1973-1976.
• 15Holst LB, Carson JL, Perner A. Should red blood cell transfusion be individualized? No. Intensive Care Med. 2015;41(11):1977-1979.
• 16Docherty AB, Walsh TS. Should blood transfusion be individualised? We are not sure. Intensive Care Med. 2015;41(11):1980-1982.
• 17Rygård SL, Holst LB, Wetterslev J, et al. Higher vs. lower haemoglobin threshold for transfusion in septic shock: subgroup analyses of the TRISS trial. Acta Anaesthesiol Scand. 2017;61(2):166-175.
• 18Bergamin FS, Almeida JP, Landoni G, et al. Liberal versus restrictive transfusion strategy in critically ill oncologic patients: the TRICOP randomized controlled trial. Crit Care Med. 2017;45(5):766-773.
• 19Hirano Y, Miyoshi Y, Kondo Y, Okamoto K, Tanaka H. Liberal versus restrictive red blood cell transfusion strategy in sepsis or septic shock: a systematic review and meta-analysis of randomized trials. Crit Care. 2019;23:262.
• 20Carson JL, Stanworth SJ, Dennis JA, et al. Transfusion thresholds and other strategies for guiding red blood cell transfusion. Cochrane Database Syst Rev. 2025;10(10):CD002042.
• 21Coz Yataco AO, Soghier I, Hébert PC, et al. Red blood cell transfusion in critically ill adults: an American College of Chest Physicians clinical practice guideline. Chest. 2025;167(2):477-489.
• 22Prescott HC, Antonelli M, Alhazzani W, et al. Surviving Sepsis Campaign: international guidelines for management of sepsis and septic shock 2026. Intensive Care Med. 2026;52(5):863-936.