The Hyperoxia Paradox: The Missing Red Blood Cell Axis in Stress Hyperglycaemia

1. Introduction

Stress hyperglycaemia is among the most reproducible metabolic findings in critical illness. Within the first 48 hours of intensive care unit (ICU) admission, more than half of non-diabetic patients develop blood glucose elevations exceeding 180 mg/dL [1]. The phenomenon is robustly associated with mortality, organ failure, infection risk, and post-ICU diabetes incidence [2,3]. Despite four decades of investigation, the mechanistic account has remained narrowly focused: a triad of counter-regulatory hormones (cortisol, catecholamines, glucagon, growth hormone), pro-inflammatory cytokines (TNF-α, IL-1, IL-6), and the resulting peripheral insulin resistance dominates the textbook explanation [1,4]. Even sophisticated recent treatments emphasising muscle insulin receptor substrate 1 (IRS1) downregulation and immune-cell glucose consumption remain within this hormonal-inflammatory paradigm [5,6].

This perspective argues that an entire cellular compartment has been systematically overlooked: the red blood cell (RBC). The case for re-examination has become unavoidable following the 2026 demonstration by Martí-Mateos et al. that, under chronic hypoxia, RBCs serve as a primary glucose sink—accounting for the majority of systemic glucose disposal not explained by visceral organ uptake [7]. Erythrocyte glucose transporter 1 (GLUT1) abundance approximately doubles in newly synthesised RBCs born in hypoxic bone marrow, and per-cell glucose uptake increases roughly threefold. Phlebotomy and transfusion experiments establish RBCs as both necessary and sufficient for the systemic glycaemic effect.

If hypoxia upregulates this RBC glucose sink, what does the opposite condition do? Critically ill patients spend large fractions of their admission exposed to moderate-to-severe hyperoxia—a state of erythrocyte physiology that has, to our knowledge, never been examined in the stress hyperglycaemia literature. We propose that ICU hyperoxia silences the RBC glucose sink through a bidirectional molecular switch first characterised in the context of red cell storage biology, and that this silencing constitutes a hitherto-unrecognised contributor to stress hyperglycaemia.

2. The Band 3 N-Terminus as a Bidirectional Metabolic Switch

Mature human erythrocytes lack nuclei and mitochondria. Their metabolic flexibility therefore depends entirely on post-translational regulation, particularly the reversible sequestration of glycolytic enzymes by the cytoplasmic N-terminus of band 3 (anion exchanger 1, AE1) [8,9]. Issaian and colleagues established that this scaffold functions as a binary metabolic switch governed by haemoglobin oxygenation state [8]. Under deoxygenation or low-oxidant conditions, deoxyhaemoglobin competitively occupies the band 3 N-terminus, displacing glyceraldehyde-3-phosphate dehydrogenase (GAPDH), phosphofructokinase (PFK), and aldolase into the cytosol; this releases the brake on glycolysis, accelerates 2,3-diphosphoglycerate (2,3-DPG) production via the Rapoport-Luebering shunt, and enhances tissue oxygen delivery [8,9]. Conversely, under oxygenated or high oxidant-stress conditions, glycolytic enzymes themselves bind the band 3 N-terminus, partially inhibiting flux through the canonical Embden-Meyerhof pathway and redirecting glucose-6-phosphate into the pentose phosphate pathway (PPP) to generate NADPH for glutathione recycling and antioxidant defence [8].

This is not a theoretical model. D'Alessandro and colleagues demonstrated the switch experimentally using 13C-labelled glucose flux analysis in RBCs stored under controlled, hyperoxic, or hypoxic conditions [10]. Hyperoxic storage induced significantly higher methaemoglobin accumulation during the first three weeks (an index of haemoglobin oxidation), accompanied by markedly elevated PPP flux throughout the storage period—a direct experimental signature of the predicted metabolic shift. GAPDH activity declined under hyperoxia, consistent with its band 3 sequestration. The hypoxia-derived insight that RBCs reprogram glycolysis to feed 2,3-DPG synthesis [7] thus possesses a precise mirror image: in hyperoxia, RBCs reprogram away from glycolysis to feed antioxidant defence.

The implication for ICU physiology is direct and largely unexplored. If sustained hyperoxia locks GAPDH onto band 3 and diverts erythrocyte glucose metabolism toward PPP-NADPH generation, the per-cell glucose disposal capacity is constrained. Moreover, because the chronic hypoxic upregulation of GLUT1 is imprinted at the time of erythropoiesis [7], one would expect the converse to apply: RBCs born under sustained hyperoxia should be generated with a baseline-low GLUT1 phenotype, perpetuating the glucose-sink suppression over the cohort's 120-day lifespan.

3. The Septic Erythrocyte: A Converging Line of Evidence

Although the systemic stress of sepsis is multifactorial and not reducible to hyperoxia alone, the resulting RBC phenotype provides corroborating evidence that critical illness silences the hypoxia-adapted erythrocyte programme. Bateman and Ellis's comprehensive review of the septic erythrocyte enumerates the following alterations: decreased intracellular 2,3-DPG, decreased deformability, redistribution of membrane phospholipids (including externalisation of phosphatidylserine), altered antioxidant status, dysregulated intracellular Ca²⁺ homeostasis, modified band 3 phosphorylation patterns, and impaired oxygen-dependent ATP efflux—a mechanism normally engaged in microvascular autoregulation [11]. Crucially, the authors note that 2,3-DPG depletion and metabolic alterations are partly reversible with appropriate treatment, whereas membrane and deformability changes appear permanent and drive accelerated clearance.

Kopterides et al. provided what may be regarded as an inadvertent rescue experiment: in 37 patients with severe sepsis or septic shock, RBC transfusion improved the microdialysis-assessed interstitial lactate-to-pyruvate ratio in subcutaneous adipose tissue, suggesting that infusing functionally competent erythrocytes ameliorated tissue metabolic stress [12]. While the original interpretation emphasised oxygen delivery, the present framework suggests an alternative or complementary explanation: transfused RBCs from healthy donors restored the dynamic glucose-handling capacity that the recipient's own circulating RBCs had lost.

Plasma metabolomic studies in sepsis consistently report reduced glycolytic intermediates and pentose phosphate pathway metabolites [13]. While these signatures are typically interpreted as systemic metabolic collapse, the erythrocyte contribution—given that RBCs constitute approximately 85% of all cells in the human body [7]—is rarely quantified. Selective metabolomic interrogation of the erythrocyte compartment in ICU cohorts remains a striking research gap.

4. The HBO Paradox and Its Resolution

A potential counter-argument arises from hyperbaric oxygen therapy (HBOT). Multiple studies report that HBOT improves glycaemic control in patients with type 2 diabetes, reducing fasting glycaemia by approximately 36% and ameliorating insulin resistance [14,15]. If sustained hyperoxia suppresses RBC glucose disposal, why does HBOT improve systemic glucose handling?

Three considerations dissolve this apparent paradox without invalidating the central hypothesis:

4.1. Dose, Duration, and Pattern

Standard HBOT regimens consist of brief intermittent exposures (typically 60–90 minutes at 2.0–2.4 ATA) [14], in marked contrast to the continuous moderate hyperoxia of ICU oxygen therapy delivered over days to weeks. The temporal patterns are biologically non-equivalent. Intermittent supraphysiological oxygen may engage hormetic adaptations that sustained moderate hyperoxia does not.

4.2. Extra-Erythrocyte Mechanisms

HBOT improves tissue-specific (skeletal muscle, hepatic, and adipose) insulin sensitivity and mitochondrial capacity in people with type 2 diabetes [16], and may transiently suppress carotid body chemoreceptor activity—a potent glucose and insulin sensor [14]. These peripheral and neural effects can dominate the net glycaemic outcome, particularly during brief exposures, even if the RBC glucose sink is partially constrained.

4.3. The ICU Patient Is Not the HBOT Patient

HBOT is applied to ambulatory patients with intact stress-hormone axes and baseline RBC physiology. The ICU patient combines hyperoxia with high-cortisol, high-catecholamine, systemic inflammation, frequent hyperglycaemic nutrition, and accelerated erythrocyte turnover with neocytolysis-type clearance [17]. Under these conditions, the RBC glucose sink contribution becomes both more vulnerable and clinically more salient.

The paradox, properly framed, is therefore not a contradiction but a dose-pattern question. ICU oxygen therapy and HBOT occupy non-overlapping regions of the oxygen exposure spectrum. The hypothesis advanced here applies specifically to the former.

5. The Red Cell Hypoxic Metabolic Index (RHMI)

We propose a composite biomarker—the Red Cell Hypoxic Metabolic Index (RHMI)—as a clinically tractable index of erythrocyte metabolic phenotype in critical illness. The proposed components, each feasible with existing critical-care laboratory infrastructure or modest additions thereto, are summarised in Table 1.

The RHMI is intended to function as a continuous score rather than a binary classifier, capturing the position of an individual patient's erythrocyte population along the hypoxia-adapted ↔ hyperoxia-suppressed continuum. The detailed algorithmic specification, including weighting, normalisation, and reference values, is developed in a companion technical document.

6. Testable Predictions

The framework generates four predictions amenable to observational and interventional studies:

Prediction 1: ICU patients exposed to sustained moderate-to-severe hyperoxia (e.g., FiO₂ ≥ 0.40 or PaO₂ > 100 mmHg for > 48 hours) will demonstrate lower RHMI than time-matched ICU controls maintained on conservative oxygen targets (SpO₂ 88–94%).

Prediction 2: Within an ICU cohort, RHMI will correlate inversely with peak blood glucose, glycaemic variability (coefficient of variation), and total insulin requirement, after adjustment for established covariates including illness severity, corticosteroid use, dextrose-containing nutrition, and pre-morbid HbA1c.

Prediction 3: Following ICU discharge and normoxic recovery, RHMI will progressively normalise over a timescale matching erythrocyte turnover (approximately 4 months), with the rate of recovery predictive of post-ICU dysglycaemia and diabetes risk.

Prediction 4: RBC transfusion from healthy donors, by introducing erythrocytes with intact baseline GLUT1 expression and band 3 functionality, will transiently improve glycaemic control in critically ill patients, with effect magnitude proportional to recipient RHMI depression. This prediction parallels the rationale of NCT04137692, which is currently testing therapeutic RBC exchange transfusion in GLUT1 deficiency syndrome [18].

7. Clinical and Translational Implications

7.1. A New Rationale for Conservative Oxygen Therapy

Multiple randomised trials and meta-analyses have established that liberal oxygen targeting in critical illness confers no survival advantage over conservative targets and may cause harm [19,20]. The prevailing mechanistic explanations centre on direct reactive oxygen species (ROS)-mediated tissue injury, hyperoxia-induced vasoconstriction, and impaired innate immune function. The framework advanced here provides an additional, complementary mechanism: preservation of the RBC glucose sink as a determinant of glycaemic stability. If validated, this lends new physiological justification to permissive hypoxaemia strategies.

7.2. Therapeutic Implications: From Selection to Design

If a low RHMI predicts severe stress hyperglycaemia, the index becomes a candidate stratification tool for two emerging therapeutic strategies. First, restrictive versus liberal transfusion policies in critical care could be informed by erythrocyte metabolic phenotype rather than haemoglobin concentration alone. Second, the broader concept of "therapeutic erythrocyte modulation"—exemplified by NCT04137692's RBC exchange transfusion for GLUT1 deficiency syndrome—gains a new patient population. Ex vivo erythrocyte preconditioning under controlled hypoxia, followed by autologous or allogeneic infusion, becomes a conceivable adjunct to conventional glycaemic management.

7.3. Implications for Survivor Cohorts

A retrospective cohort study of South Australian ICU survivors demonstrated that stress hyperglycaemia during the initial admission independently predicts incident type 2 diabetes years after discharge [3]. The conventional interpretation invokes "unmasking" of pre-existing insulin resistance. The present framework suggests an alternative or additive mechanism: ICU-induced erythrocyte reprogramming may persist beyond the 4-month turnover window if newly synthesised RBCs continue to be born from a bone marrow exposed to subclinical residual hyperoxia or persistent inflammation. Longitudinal RHMI tracking in survivor cohorts is a feasible follow-up.

8. Limitations and Caveats

Several constraints qualify the hypothesis.

First, the central molecular evidence—the band 3 metabolic switch and its hyperoxia-induced reconfiguration—derives largely from ex vivo red cell biology and storage research. Direct demonstration of the predicted switch in vivo, in human ICU cohorts, has not been undertaken. The proposed observational study using the RHMI framework is the obvious first step.

Second, the relative contribution of hyperoxia-induced RBC suppression to total stress hyperglycaemia is uncertain. Hormonal and inflammatory pathways are unambiguously dominant in absolute magnitude. The hypothesis is most parsimoniously framed not as a replacement for current models but as an additive contributor that may explain the residual hyperglycaemic burden incompletely controlled by insulin therapy.

Third, the haemoglobin-foetal-to-adult switch and the diversity of haemoglobin variants in critically ill populations introduce additional sources of variation. The biomarker panel should be interpreted with awareness of these confounders.

Fourth, the relationship between methaemoglobin elevation and the proposed pathway is non-specific. Methaemoglobin increases in nitric oxide exposure, drug effects, and septic conditions independent of hyperoxia. Multivariate analysis incorporating multiple RHMI components addresses this limitation only partially.

9. Conclusion

Stress hyperglycaemia has been understood for four decades as a hormonal-inflammatory phenomenon centred on the liver, muscle, and immune system. The 2026 demonstration that erythrocytes are quantitatively dominant glucose sinks under chronic hypoxia compels re-examination of the converse condition routinely imposed in intensive care: sustained moderate-to-severe hyperoxia. The molecular switch governing this bidirectionality is identified, the experimental signatures are documented in red cell storage biology, and the clinical predictions are testable with a biomarker panel deployable in existing critical care infrastructure.

We propose that the silencing of the RBC glucose sink in ICU patients constitutes the missing axis of stress hyperglycaemia, and that the Red Cell Hypoxic Metabolic Index provides a tractable means to investigate and ultimately address this hypothesis. The framework also provides a new physiological rationale for conservative oxygen therapy strategies whose clinical superiority has emerged empirically without complete mechanistic explanation. At the broadest level, recognising RBCs as active metabolic participants rather than passive oxygen carriers represents a paradigm shift whose downstream implications—for critical care, transfusion medicine, and metabolic disease therapy—remain to be fully articulated.