HbA1c Reconsidered: Glycated Haemoglobin in the Era of Dynamic Erythrocyte Glucose Metabolism

1. Introduction: The Hidden Assumption Beneath HbA1c

Glycated haemoglobin (HbA1c) is among the most consequential biomarkers in modern medicine. First validated for diabetes monitoring in the late 1970s [1] and codified as a diagnostic criterion by the American Diabetes Association in 2010 [2], HbA1c is now measured several hundred million times annually worldwide. Treatment intensification, drug approvals, cardiovascular outcome trials, public-health diabetes surveillance, and individual prognosis all rest on the interpretation of this single number.

The molecular basis is straightforward: glucose binds non-enzymatically to the N-terminal valine of the β-chain of haemoglobin A0, forming a Schiff base that undergoes Amadori rearrangement to a stable ketoamine [3]. The reaction is irreversible during the erythrocyte’s lifespan. The amount of glycated haemoglobin accumulated by an individual red cell is therefore the time-integral of its intracellular glucose exposure over its circulating life.

Translating this molecular picture into a clinically usable biomarker requires three simplifying assumptions:

• Erythrocyte intracellular glucose concentration approximates plasma glucose concentration
• Erythrocyte lifespan is reasonably uniform (~120 days)
• Glycation rate constants are similar across individuals

Decades of clinical experience have shown that all three assumptions can fail. Inter-individual variation in HbA1c at any given mean blood glucose is well documented [4]; the same true mean glucose may yield HbA1c values differing by more than 1 percentage point, corresponding to estimated mean glucose differences exceeding 60 mg/dL [5]. Mechanistic models incorporating subject-specific red cell age and glycation kinetics have improved the situation modestly [5,6], but unexplained variance remains substantial.

Figure 1. HbA1c reflects the time-integral of intracellular erythrocyte glucose exposure, scaled by red-cell lifespan and glycation kinetics—not the time-integral of plasma glucose. Under stable erythrocyte glucose transport the two track closely (panel A, left); once GLUT1 becomes a programmable variable set by oxygen-exposure history (panel A, right), HbA1c and true glycaemia diverge in directionally predictable ways (panel B). Panel C outlines the clinical consequences. Directions shown reflect the mechanism argued in the text and the cited discordance literature; they are advanced as testable predictions, not established quantitative corrections.

Figure 1. HbA1c reflects the time-integral of intracellular erythrocyte glucose exposure, scaled by red-cell lifespan and glycation kinetics—not the time-integral of plasma glucose. Under stable erythrocyte glucose transport the two track closely (panel A, left); once GLUT1 becomes a programmable variable set by oxygen-exposure history (panel A, right), HbA1c and true glycaemia diverge in directionally predictable ways (panel B). Panel C outlines the clinical consequences. Directions shown reflect the mechanism argued in the text and the cited discordance literature; they are advanced as testable predictions, not established quantitative corrections.

This perspective argues that a fourth, hitherto-unrecognised source of variability has now become visible: erythrocyte glucose uptake itself is not constant. The 2026 demonstration by Martí-Mateos and colleagues [7] that chronic hypoxia upregulates erythrocyte GLUT1 protein abundance approximately twofold and per-cell glucose uptake approximately threefold, together with the broader recognition that band 3 acts as a bidirectional regulator of glycolytic flux [8,9], requires us to revisit the most fundamental assumption underlying HbA1c. If the GLUT1 transporter is a dynamic variable rather than a fixed constant, then HbA1c does not measure what we have always assumed it measures.

2. The Dynamic Erythrocyte: A 2026 View

2.1. Erythrocyte GLUT1 as a Programmable Variable

The Martí-Mateos et al. study used murine chronic hypoxia (8% O₂ × 4 weeks) combined with positron emission tomography, isotopic glucose flux analysis, phlebotomy, and red cell transfusion experiments to establish red cells as a primary glucose sink [7]. Key quantitative findings:

• Approximately 70% of the hypoxia-induced increase in systemic glucose uptake could not be accounted for by visceral organ uptake
• Newly synthesised red cells produced under hypoxic erythropoiesis showed approximately 2-fold higher GLUT1 protein per cell
• Per-cell glucose uptake rate increased approximately 3-fold and persisted ex vivo under normoxic conditions—a stable cellular phenotype, not an acute signalling response
• Manipulating red cell number through phlebotomy or transfusion directly altered plasma glucose, establishing red cells as both necessary and sufficient for the glycaemic effect

In parallel, Issaian et al. characterised the band 3 N-terminus as a bidirectional metabolic switch [8]. Under deoxygenation, deoxyhaemoglobin competitively displaces glycolytic enzymes (GAPDH, PFK, aldolase) from band 3, releasing them into the cytosol and accelerating glycolysis with downstream 2,3-DPG accumulation. Under oxygenation or oxidant stress, glycolytic enzymes themselves bind band 3, partially inhibiting glycolysis and redirecting flux through the pentose phosphate pathway to generate NADPH for antioxidant defence. Storage biology experiments by Reisz et al. confirmed this switching behaviour experimentally: hyperoxic storage produced higher methaemoglobin accumulation and elevated PPP flux throughout the storage period [9].

2.2. Single-Cell Heterogeneity

Recent single-cell imaging using the fluorescent glucose analogue 2-NBDG has revealed substantial heterogeneity in erythrocyte glucose uptake both within and between individuals [10]. The authors of that work explicitly note that this transporter-level variation will affect HbA1c formation, and propose that more personalised diagnostic strategies may be warranted. Their work was published independently of the chronic hypoxia framework, but the two observations converge on the same conclusion: erythrocyte glucose handling is not a fixed background parameter.

3. Reinterpreting HbA1c

Combining these observations leads to a refined conceptualisation of HbA1c. We propose the following:

HbA1c does not measure the time-integral of plasma glucose. It measures the time-integral of intracellular erythrocyte glucose exposure, scaled by erythrocyte lifespan and glycation kinetics. Under conditions of stable erythrocyte glucose transport, these two quantities track closely. Under conditions of altered erythrocyte oxygen environment or accelerated red cell turnover, they diverge in directionally predictable ways.

Formally, the steady-state HbA1c concentration follows:

HbA1c ∝ k_glycation × [glucose]_intracellular × τ_RBC

where k_glycation is the second-order rate constant for haemoglobin glycation, [glucose]_intracellular is the time-averaged intracellular glucose concentration, and τ_RBC is the mean red cell lifespan. The traditional simplification asserts that [glucose]_intracellular ≈ [glucose]_extracellular, which holds when GLUT1-mediated transport is rapid relative to glycation and when GLUT1 expression is uniform. In the dynamic erythrocyte framework, this approximation fails:

[glucose]_intracellular = f(GLUT1_expression, glycolytic_flux, band3-GAPDH binding state)

The first variable (GLUT1 expression) governs influx; the second (glycolytic flux) governs efflux through metabolism; the third (band 3 state) modulates both. Each of these varies with the erythrocyte’s oxygen exposure history.

4. Empirical Evidence: HbA1c Discordance in Specific Populations

Several patient populations have long shown HbA1c values that diverge from expected glycaemic patterns. We argue that the dynamic erythrocyte framework provides a unified explanation.

4.1. High-Altitude Populations: A Natural Experiment

Multiple studies in Tibetan and Andean highlanders have documented HbA1c values that do not track expected relationships with plasma glucose [11,12,13]. A 2025 cross-sectional study in 410 patients with type 2 diabetes living at high altitude in China found that whilst HbA1c and fasting plasma glucose were both elevated, the HbA1c-to-FPG ratio decreased markedly and the correlation weakened [14]. Most strikingly:

• Subjects with haemoglobin >160 g/L showed relatively lower HbA1c, consistent with polycythaemia-induced HbA1c dilution
• Glycated albumin (GA) was significantly lower in the high-altitude group (2.47 ± 0.63% vs 3.78 ± 1.52%, p < 0.001)—a trend opposite to HbA1c

The opposite-direction movement of GA and HbA1c is a critical observation. If HbA1c elevation reflected true hyperglycaemia, GA should track it. The reversal of direction indicates that HbA1c is responding to factors other than mean plasma glucose—precisely the erythrocyte-level mechanisms the dynamic framework predicts.

A Tibetan population study has likewise demonstrated that the optimal HbA1c cutoff for detecting abnormal glucose metabolism differs between high- and low-altitude residents [15], and the authors recommend population-specific thresholds. The framework advanced here provides the mechanistic basis for this recommendation.

4.2. Obstructive Sleep Apnoea: A Complex Case

HbA1c is positively associated with OSA severity in a dose-response manner [16,17]. A 2,139-patient study showed odds ratios for HbA1c > 6.0% rising from 1.0 (AHI < 5) to 2.96 (AHI ≥ 50) after adjustment for traditional confounders [16]. The conventional interpretation attributes this to intermittent hypoxia-driven insulin resistance.

The dynamic erythrocyte framework adds nuance. OSA partially activates erythrocyte hypoxic reprogramming—not enough to fully replicate the chronic high-altitude phenotype (only ~2% of OSA patients develop overt polycythaemia [18]), but enough to alter erythrocyte glucose handling. Partial GLUT1 upregulation may increase intracellular glucose exposure relative to extracellular, contributing to HbA1c elevation independent of true plasma hyperglycaemia. The proportion attributable to insulin resistance versus erythrocyte-level glycation enhancement has not been disentangled.

This has direct clinical implications. The American Diabetes Association threshold of HbA1c ≥ 6.5% for diabetes diagnosis may overestimate diabetes prevalence in OSA populations if a substantial fraction of HbA1c elevation reflects erythrocyte-level rather than systemic glucose dysregulation. CGM-based assessment, which bypasses erythrocyte processing entirely, would help quantify the discrepancy.

4.3. Cyanotic Congenital Heart Disease: A Developmental Layer

Cyanotic CHD children represent perhaps the cleanest model of chronic erythrocyte hypoxic adaptation. Versmold et al. demonstrated decades ago that 2,3-DPG and P50 rise significantly in CCHD patients after the foetal-to-adult haemoglobin switch (>3 months age) [19], and Gidding and Stockman characterised the high-haematocrit/low-EPO phenotype [20]. These observations describe an erythrocyte population with substantially altered metabolic state.

If the framework is correct, CCHD children should exhibit HbA1c values that progressively decouple from true plasma glucose as the haemoglobin switch completes. Hypoglycaemia during fasting (originally reported by Haymond et al. in 1979 [21]) and concurrently relatively low HbA1c would both reflect the same underlying mechanism: erythrocyte-level glucose consumption exceeding plasma supply. We are not aware of any study that has directly measured HbA1c discordance with mean glucose in CCHD children stratified by HbA0/HbF ratio—a striking gap.

4.4. Critical Illness: A Bidirectional Perturbation

Critical illness adds multiple perturbing factors: sustained hyperoxia (suppressing the glucose sink [22]), neocytolysis (accelerating red cell turnover [23]), and frequent transfusion (introducing donor erythrocytes with different histories). HbA1c interpretation in critically ill or recently-discharged-from-ICU patients is widely recognised as unreliable. The framework here provides a mechanistic basis: the ICU environment alters every term in the HbA1c equation simultaneously, in directions that are not easily predicted without measurement of erythrocyte parameters themselves.

4.5. Chronic Kidney Disease: A Documented Underestimate

CKD patients on dialysis exhibit HbA1c values that systematically underestimate glycaemia, and glycated albumin has been proposed as the preferred alternative [24]. The conventional explanation invokes shortened red cell lifespan from uraemia and EPO deficiency. The dynamic framework adds that the angiotensin-SphK1-S1P axis—described in CKD by Xie et al. [25]—directly reprograms erythrocyte glucose metabolism toward glycolysis, paradoxically lowering steady-state intracellular glucose available for glycation despite preserved plasma levels. This is a population in which traditional and dynamic explanations converge to predict the same direction of HbA1c bias, both contributing.

5. Testable Predictions

The framework generates several falsifiable predictions amenable to clinical investigation:

Prediction 1: In any population with altered chronic oxygen environment (high altitude, OSA, CCHD, COPD, ICU survivor), the discrepancy between HbA1c-estimated and CGM-measured average glucose will correlate with erythrocyte GLUT1 expression as measured by flow cytometry.

Prediction 2: Glycated albumin will track CGM-measured glucose more faithfully than HbA1c in patients with high-altitude residence, severe OSA, polycythaemic CHD, and CKD.

Prediction 3: Therapeutic interventions that alter erythrocyte oxygen environment—initiation of CPAP in OSA, surgical correction of CCHD, descent from altitude, transition from ICU to ward—will produce changes in HbA1c that lag behind changes in true glucose by approximately one red cell turnover (~3–4 months), as the existing red cell cohort is progressively replaced.

Prediction 4: A composite erythrocyte metabolic biomarker—we propose the Red Cell Hypoxic Metabolic Index (RHMI), incorporating GLUT1 expression, ex vivo glucose uptake, 2,3-DPG, and methaemoglobin [26]—can serve as a personalised correction factor for HbA1c interpretation. Patients with high RHMI (hypoxia-adapted erythrocytes) will require downward correction of HbA1c-derived estimated average glucose; patients with low RHMI (hyperoxia-suppressed erythrocytes) will require upward correction.

6. Clinical Implications

6.1. Population-Specific HbA1c Thresholds

The Tibet Plateau studies have already moved in this direction with empirical recalibration of HbA1c cutoffs [15]. The framework advanced here suggests systematic extension:

• High-altitude residents (>2,500 m for >3 months): downward revision of HbA1c cutoff (i.e., diagnose diabetes at lower HbA1c)
• Severe OSA (AHI > 30) prior to CPAP: caution in using HbA1c alone for diagnosis
• Unrepaired CCHD: HbA1c likely uninformative for glycaemic status; use CGM
• Post-ICU survivors during first 3–4 months: HbA1c reflects predominantly pre-ICU glycaemia

6.2. Choice of Monitoring Biomarker

In specific populations, glycated albumin or 1,5-anhydroglucitol may better reflect glycaemic state. CGM, which directly measures interstitial glucose, bypasses erythrocyte processing entirely and is the gold standard where available. The next-generation diabetes monitoring framework should explicitly incorporate erythrocyte phenotype as a determinant of biomarker choice.

6.3. Diabetes Complications: A Re-Examination

HbA1c predicts microvascular complications because diabetic retinopathy, nephropathy, and neuropathy occur in cells (retinal pericytes, renal mesangial cells, peripheral neurons) that cannot adjust GLUT transporter expression in response to hyperglycaemia [27]. Traditionally, erythrocytes were considered to fall in the same category—hence HbA1c’s predictive value. In the dynamic framework, this equivalence breaks down: erythrocytes can adjust GLUT1, but the affected complication-prone cells cannot. The prediction is that in populations where the erythrocyte-vulnerable-cell coupling is disrupted (chronic hypoxia, ICU), HbA1c may lose some of its predictive power for microvascular complications. The literature does in fact show such discordances [28]; the framework here provides mechanistic explanation.

6.4. Therapeutic Erythrocyte Modulation

The NIH-sponsored trial NCT04137692 testing red blood cell exchange transfusion for GLUT1 deficiency syndrome explicitly relies on the concept that erythrocytes are a manipulable glucose handling compartment. If therapeutic erythrocyte modulation proves effective in any disease context, HbA1c will become an unreliable monitoring tool for treatment response—because the intervention by design changes the relationship between plasma and intracellular erythrocyte glucose.

7. Limitations and Caveats

Several constraints qualify the framework.

First, the most direct evidence for dynamic GLUT1 expression remains the murine chronic hypoxia model. Whilst single-cell human erythrocyte studies [10] and high-altitude human cohort data [14] are consistent with the framework, direct measurement of GLUT1 across the predicted populations has not been undertaken at scale. The R5 ICU observational study, currently under preparation, will test the framework in critically ill patients [26].

Second, the framework does not eliminate the value of HbA1c. In healthy adults with stable oxygen environment, the constancy assumption holds approximately, and HbA1c remains an excellent biomarker. The argument is for refinement, not replacement.

Third, the quantitative magnitude of GLUT1-driven HbA1c distortion in human populations has not been precisely measured. Whilst the high-altitude data show clinically meaningful effects [14], we cannot yet assign a numerical correction factor with confidence.

Fourth, the foetal-to-adult haemoglobin switch introduces additional complexity in paediatric populations. HbF does not bind 2,3-DPG to the same extent as HbA [19], and the glycation rate of HbF1c versus HbA1c may differ. The framework’s predictions for paediatric CCHD therefore require age-specific calibration.

8. Conclusions

HbA1c is not a measurement of plasma glucose. It is a measurement of erythrocyte intracellular glucose, with all the cellular biology that implies. For four decades, this distinction did not matter clinically because we assumed—reasonably, given the tools available—that the erythrocyte was a passive equilibrium with plasma. The 2026 demonstration that erythrocytes are active, programmable glucose sinks invalidates this assumption.

The implications are not narrowly methodological. HbA1c shapes treatment decisions for hundreds of millions of patients with diabetes, prediabetes, and metabolic risk. Reinterpreting it through the dynamic erythrocyte lens points toward:

• Population-specific diagnostic thresholds
• Biomarker selection guided by erythrocyte phenotype
• Reconciliation of long-standing discordances between HbA1c and clinical outcomes
• Anticipation of new monitoring challenges as erythrocyte-targeting therapies emerge

The traditional reading of HbA1c served medicine well in an era when the erythrocyte was understood as a passive vessel. As that understanding gives way to one of dynamic metabolic activity, the biomarker that defined modern diabetology requires a parallel update in interpretation. HbA1c remains essential. It needs only to be read correctly.