Beyond hemoglobin thresholds: A physiology-guided framework for red blood cell transfusion in non-bleeding critically ill patients

Red blood cell transfusion in non-bleeding critically ill patients is traditionally guided by fixed hemoglobin thresholds. Although restrictive strategies are widely recommended based on population-level randomized trials, hemoglobin concentration alone inadequately reflects the complex physiology of oxygen delivery, tissue perfusion, and cellular oxygen utilization in critical illness. Increasing evidence suggests frequent dissociation between hemoglobin increments and meaningful improvements in tissue oxygenation, particularly in the presence of microcirculatory dysfunction, impaired oxygen extraction, or mitochondrial failure. Moreover, recent trials in patients with limited cardiovascular reserve or acute myocardial ischemia challenge the universal safety of restrictive transfusion thresholds, emphasizing interindividual variability in oxygen supply–demand balance. In this mini-review and expert perspective, we synthesize physiological principles, landmark transfusion trials, and emerging monitoring modalities to propose a practical physiology-guided framework for red blood cell transfusion in non-bleeding critically ill patients. We review global, regional, and cellular markers of oxygen balance, including central venous oxygen saturation, oxygen extraction ratio, arterial–venous oxygen content difference, near-infrared spectroscopy–derived tissue oxygenation, microcirculatory flow indices, and mitochondrial oxygen tension, highlighting complementary roles and inherent limitations. We propose a pragmatic bedside approach in which transfusion is considered only after optimization of non-transfusion determinants of oxygen delivery and guided by integrated physiologic evidence of oxygen debt. This paradigm reframes transfusion as targeted therapy rather than numerical correction of anemia. Future research should combine outcome-driven randomized trials with large-scale data-driven and machine learning approaches to validate physiologic transfusion triggers and identify transfusion-responsive phenotypes in critical illness.