Reviewer3 (AI review)

September 28, 2026

AI Review (reviewer3.com)

Comments (4)

1. Inconsistent author initials in Author Contributions — Editorial

Author contributions. A.K.K. conceived the hypothesis, assembled the quantitative model and wrote the manuscript. K.A. contributed to the clinical framing, the

The Author Contributions section attributes the conception of the hypothesis, assembly of the quantitative model, and writing of the manuscript to 'A.K.K.'. However, the authors listed on the title page are 'Amr Ahmed' (whose initials would be A.A.) and 'Khulood Almatuiri' (K.A.). There is no author with the initials 'A.K.K.' in the manuscript.

2. HbA1c confounded by shortened RBC lifespan — Major

Chronic hyperglycemia attacks both. Non - enzymatic glycation of the hemo globin β - chain N - terminal valine, the modification measured as HbA1c, raises oxygen affinity and has recently been quantified at about - 0.19 mmHg of P50 per 1% HbA1c, the effect proposed as glycohypoxia. Methylglyoxal, the dicarbonyl by

The model uses HbA1c as the primary quantitative metric of chronic hyperglycemia to calculate the glycation-induced P50 shift. However, active tuberculosis is strongly associated with anemia of chronic disease and systemic inflammation, which are well-documented to shorten red blood cell lifespan. Because non-enzymatic glycation of hemoglobin is a slow, time-dependent process, a shortened erythrocyte lifespan systematically reduces the measured HbA1c relative to the actual average blood glucose level. This means that in patients with active tuberculosis, the measured HbA1c will underestimate the true degree of tissue glycation and the corresponding left-shift in P50. The authors should address this confounding factor by discussing how shortened red cell survival affects the model's predictions and whether alternative glycemic markers, such as fructosamine or glycated albumin, would provide a more reliable representation of the glycohypoxia effect in this population.

3. Omission of hypoxemia from P50 equation — Major

This is the observation on which the first axis turns. Using the two published slopes, the expected position of a patient is P50 ≈ 26.8 + 0.30 × (15 − Hb) − 0.19 × (HbA1c − 5) with hemoglobin in g/dL and HbA1c in percent. Table 1 gives worked values, and

The quantitative model proposed to predict the net P50 of a patient relies solely on hemoglobin concentration and HbA1c. However, pulmonary tuberculosis frequently causes regional or systemic arterial hypoxemia due to lung tissue damage. Hypoxemia is a potent, independent physiological stimulus for erythrocyte 2,3-DPG synthesis and the subsequent right-shift of the oxygen dissociation curve. By omitting arterial oxygen tension or saturation from the quantitative equation, the model fails to account for a major driver of the hematological adaptation in the very patients being modeled. The authors should modify the equation to incorporate a term for hypoxemia or explicitly discuss how varying degrees of arterial hypoxemia would alter the predicted net P50 and the threshold for a net left-shift.

4. Leptin receptor deficiency confounds animal model evidence — Major

Hyperglycemic Leprdb/db mice carried higher lung bacterial loads with reduced expression of HIF-1 target genes.

The in vivo evidence supporting the cellular axis relies heavily on studies using hyperglycemic Leprdb/db mice. However, these mice possess an intrinsic deficiency in leptin receptor signaling. Leptin is a critical immunomodulator, and its signaling is independently required for normal Th1 immune responses, macrophage activation, and the control of Mycobacterium tuberculosis infection. Consequently, the increased bacterial loads and impaired immune responses observed in these mice are heavily confounded by the lack of leptin signaling, making it impossible to attribute these effects solely to hyperglycemia or methylglyoxal-induced HIF-1α suppression. The authors should acknowledge this major confounding factor in their discussion of the animal model evidence and temper their conclusions regarding the in vivo validation of the cellular axis accordingly.

xPeerd (AI review)

September 28, 2026

AI Review (xPeerd — DAReview)

Summary

This hypothesis/theory preprint proposes that the excess mortality observed in tuberculosis patients with diabetes arises from a two-axis “glycohypoxia” mechanism: first, hyperglycemia impairs oxygen unloading from hemoglobin by shifting P50 leftward; second, hyperglycemia-derived methylglyoxal suppresses HIF-1–dependent macrophage responses that are important for tuberculosis control. The manuscript’s strengths are its clear central question, explicit effort to formulate a falsifiable mechanistic hypothesis, and its attempt to connect hematologic physiology with host-pathogen immunobiology. It also commendably acknowledges contrary literature and states several refutable predictions.

The main weaknesses are that the manuscript’s quantitative core depends on combining coefficients drawn from heterogeneous external contexts rather than directly observed TB-diabetes data; key mechanistic links remain inferential rather than demonstrated in the target population; and some clinical inferences extend beyond what the cited evidence can currently support. The paper is coherent as a hypothesis article, but it should sharpen the distinction between evidence, extrapolation, and speculation, temper some claims of clinical implication, and improve reproducibility by making the modeling assumptions and figure-generation methods fully transparent. The anti-fraud profile is overall reassuring: the manuscript fits a recognizable biomedical paradigm, cites testable prior literature, and explicitly states limitations and falsification criteria. However, there is also a notable self-referential dependence on very recent author-linked “glycohypoxia” publications, so the novelty and independence of the evidentiary base should be presented with greater caution.

Potential Major Revisions

1. Clarify the article type and recalibrate the strength of causal language

The manuscript is labeled “Hypothesis and Theory” , yet several passages move from proposing a mechanism to implying explanatory adequacy. For example, the abstract states, “We propose a different reading, built on a single physiological currency: the oxygen the tissue actually receives” and later, “Our claim is about their conjunction” . The conclusion similarly states, “We propose something narrower and easier to test: that uncontrolled hyperglycemia removes, at the same moment, the hematological adaptation that delivers oxygen to a hypoxic lesion and the transcriptional program that lets a macrophage act on hypoxia.”

These are reasonable hypothesis statements, but elsewhere the manuscript edges toward explanatory overcommitment. In the introduction, the authors write that patients “often look worse than their numbers” and that “the clinical trajectory is that of a hypoxic patient,” which is then aligned with “what the recently proposed glycohypoxia framework describes.” This clinical observation is suggestive, but it is anecdotal and not systematically analyzed here. The paper should more consistently mark where it is hypothesizing rather than evidencing. I recommend revising language throughout to distinguish among: established findings, plausible mechanistic inference, and untested downstream consequence.

2. Strengthen justification for combining heterogeneous coefficients into a unified quantitative model

The core quantitative relation is: “P50 ≈ 26.8 + 0.30 × (15 − Hb) − 0.19 × (HbA1c − 5)” . This equation is elegant, but the two coefficients come from substantially different evidentiary settings: an older anemia adaptation literature and a recent glycohypoxia analysis in diabetes. The authors do acknowledge this partially in Section 9, noting that “The blood axis rests on a coefficient derived from six older studies, and it is contradicted by whole-blood measurements in compensated diabetes [8,9,10].”

That limitation is central, not peripheral. The paper should explicitly discuss whether these coefficients are commensurable across assay conditions, populations, sample handling, pH control, and disease states. The statement, “Our additive treatment of the two P50 coefficients assumes independence” should be expanded into a fuller methodological subsection because that assumption is doing much of the work. In particular, since the authors themselves state that glycation and 2,3-BPG likely interact “at the same site,” the additive formulation may not merely be approximate but structurally misspecified. The manuscript should either present a sensitivity analysis around this assumption or clearly present the current model as a heuristic toy model rather than a physiologically parameterized estimate.

3. Address the manuscript’s dependence on author-originated recent literature

A substantial portion of Axis 1 depends on recent work cited as references [4] and [5], including the claim that “Recent work synthesizing six human datasets (N = 450) gave a pooled slope of -0.19 mmHg of P50 per 1% HbA1c” and that “The authors named the resulting state glycohypoxia.” The reference list identifies [4] and [5] as publications by Akl and/or Ahmed, overlapping with the current manuscript authorship.

This is not invalid, but it raises a review concern about evidentiary independence. The manuscript should disclose more explicitly that an important coefficient and conceptual framework come from the authors’ own recent work, and it should better balance this with external validation literature. Without that, readers may overestimate the consensus behind the glycohypoxia construct.

4. Reconcile the contradiction between purified hemoglobin evidence and whole-blood evidence more rigorously

Section 3.4 is commendably candid: “Samaja and colleagues measured P50... and found it higher, not lower, in diabetes,” and subsequent studies found either compensation or no difference. The response offered is plausible: compensated diabetes in otherwise healthy subjects may differ from TB-diabetes, where reserve is already “spent.”

However, the rebuttal remains conceptual rather than evidential. Since this contrary literature directly challenges the first axis, the manuscript should provide a more rigorous reconciliation framework. For example, a structured contrast table could compare study populations, anemia status, inflammatory burden, assay method, pH normalization, and 2,3-BPG handling. As written, the paper admits the contradiction but does not yet resolve whether the proposed mechanism is likely in vivo in TB-diabetes or remains one of several competing models.

5. Temper therapeutic implications and separate them more clearly from hypothesis generation

Section 8 states, “Four implications follow, in descending order of confidence,” then discusses glycemic control, metformin, HIF-1 support, and carbonyl scavenging. This is acceptable if framed strictly as translational speculation. However, the wording occasionally risks sounding quasi-recommendatory, especially where the manuscript says, “Metformin is the agent best positioned for testing” and “Pharmacological support of HIF-1 signalling is the direct test of axis 2.”

The evidence base the manuscript itself summarizes remains incomplete: “The evidence that this translates into survival is currently incomplete,” and the meta-analysis cited did not show significant mortality benefit for glycemic control. The paper should therefore avoid any implication that these are clinically actionable recommendations. A clearer “research implications only—not clinical guidance” disclaimer would improve epistemic discipline.

6. Improve reproducibility of the modeling and figures

The declarations state: “No new data were generated. The code used to compute the curves in Figures 1 and 2 is available from the corresponding author on request.” For a quantitative hypothesis paper, “on request” is weaker than current reproducibility standards. Since the figures and Table 1 are arithmetic consequences of the model and not empirical measurements, the underlying code and formulas should ideally be provided directly in a supplement or repository. This is especially important because the paper emphasizes falsifiability and quantitative transparency. The current arrangement hinders verification of curve construction, parameter choices, and any implicit assumptions behind the plotted relationships.

7. Expand discussion of confounding mechanisms that could rival or overwhelm the proposed axes

The manuscript notes that it is “silent on strain virulence, host genetics of the interferon-γ axis, HIV coinfection, malnutrition and the pharmacokinetics of rifampicin with oral hypoglycemic agents.” It also introduces thrombosis as a modifier rather than a third pillar. This is appropriate, but the manuscript could better explain why the proposed two-axis mechanism deserves priority among these alternatives.

In its current form, the model risks appearing underdetermined: multiple mechanisms may plausibly contribute to the same mortality excess. Since the paper’s value depends on identifying a mechanism “specific enough to be wrong,” the authors should specify what empirical pattern would distinguish their model from rival explanations beyond the predictions already listed.

8. Refine the novelty claim

The abstract states, “Neither axis is new on its own. Our claim is about their conjunction.” This is a fair novelty claim. However, the manuscript should be more explicit that the novelty lies in integrative synthesis, not in demonstrating a new mechanism in patients. This matters because some readers may otherwise infer that the paper has provided a direct mechanistic explanation of TB-diabetes mortality. It has not; it has proposed one. Stating this more directly would improve coherence and reduce risk of overinterpretation.

Potential Minor Revisions

1. Typographic, wording, and grammar issues

There are multiple mechanical issues that should be corrected.

• In Section 9, the manuscript states, “Each of these influences' mortality on its own.” The apostrophe is incorrect; this should read “influences mortality” or “influences mortality on its own.”
• In the conclusion, “BPG2.3” appears in “P50, BPG2.3, methylglyoxal...” This should likely be “2,3-BPG” for consistency with the rest of the manuscript.
• In the references, entry 10 ends with “PMID: 7834160..” containing a duplicated period.
• The manuscript alternates between “2,3-DPG” and “2,3-BPG.” While the authors explain these are alternative names, the dual naming is repeated frequently and may reduce readability. Standardize usage after first definition.
• “preprint not peer reviewed” and SSRN watermark text appear repeatedly within extracted text blocks and may indicate formatting clutter in the PDF that should be minimized in the final version if possible.

2. Formatting and presentation issues

• Table 1 is useful, but it should visually separate measured inputs from modeled outputs more explicitly, because the caption says, “Values are illustrative, not measurements.”
• Table 2 is central to the article’s falsifiability claim and should be moved or highlighted more prominently, possibly with a short explanatory lead-in emphasizing which predictions are considered primary versus exploratory.
• Section numbering appears orderly in the extracted text, but the page footer/header material repeatedly interrupts reading flow. Ensure clean final pagination and figure placement.
• The figures should specify all parameter assumptions directly in the legends, especially the baseline P50, Hill coefficient, normalization assumptions, and whether pH/temperature are fixed. The current legend for Figure 1 gives only partial methodological detail.

3. Precision and consistency of scientific framing

• The manuscript repeatedly refers to “excess mortality” and a “twofold” signal; these should be framed consistently with odds ratio versus hazard ratio terminology, since the cited studies include different effect measures.
• The phrase “tuberculosis is the rare infection whose control depends on a hypoxic tissue response” is rhetorically strong and may overstate singularity. Consider softening to avoid a claim that may be contested by specialists in other intracellular or granulomatous infections.
• The line “Fitting a formula to a published odds ratio is not a prediction, it is decoration” is vivid and memorable, but stylistically sharper than the otherwise formal tone. Depending on target journal, this may need moderation.

4. AI content analysis

The acknowledgments disclose: “The authors used an AI assistant to support drafting, language editing, figure generation and formatting of this manuscript. All scientific content, data interpretation and conclusions are the authors’ own...” This is appropriate disclosure.

Estimated percentage of AI-generated content: moderate, approximately 25–45%.

Reasons for this estimate:

• The prose is unusually uniform, polished, and stylistically controlled across sections.
• The manuscript uses highly balanced rhetorical structures and repeated contrastive framing (“Neither axis is new on its own...”; “A hypothesis that cannot fail is not useful”; “What the number lacks is a mechanism specific enough to be wrong”), which are often associated with AI-assisted scientific drafting.
• The acknowledgments explicitly confirm AI-assisted drafting and figure generation.

Highlighted likely AI-assisted sections:

• Abstract and Introduction, where the prose is especially rhetorically polished and compressed.
• Sections 6–8, which contain crisp, highly templated transitions between prediction, caveat, and implication.
• Conclusion, which has a concise, thesis-restatement style characteristic of assisted synthesis.

Assessed epistemic impact:

• Low to moderate, based on the current text alone.
• The main concern is not AI usage per se, but whether polished language may obscure the speculative status of some claims and make inferential leaps appear firmer than the underlying evidence warrants.
• Because the authors disclosed the AI support, the primary editorial issue is verification of scientific accuracy, reproducibility, and citation fidelity rather than undisclosed authorship substitution.

5. Coherence / fabrication / paper-mill assessment

Warning assessment: no strong paper-mill signature detected.

Reasons:

• The manuscript is thematically coherent and structured around a specific biomedical hypothesis.
• It engages contrary literature instead of ignoring it.
• It specifies falsifiable predictions and explicit refutation criteria.
• It admits major limitations rather than overselling certainty.

Residual caution:

• The framework depends materially on recent author-linked literature and on modeled rather than observed TB-diabetes measurements.
• Because this is a hypothesis manuscript, over-interpretation rather than fabrication is the primary risk.

Recommendations

1. Reframe the manuscript more explicitly as an integrative, testable hypothesis rather than a mechanism nearing confirmation.
2. Add a dedicated methodological note explaining how the P50 equation was assembled, why additive independence was assumed, and how sensitive the conclusions are to alternative coefficient choices.
3. Provide the code for Figures 1 and 2 and the Table 1 calculations in a supplement or public repository rather than only “on request.”
4. Expand the rebuttal to contrary whole-blood literature with a structured comparison table and a clearer statement of what empirical results would favor the authors’ model over competing explanations.
5. Reduce translational overreach in Section 8 by labeling therapeutic discussion strictly as hypothesis-testing implications, not practice suggestions.
6. Disclose more prominently that the glycohypoxia construct and one of the key quantitative coefficients derive from recent author-associated publications, and discuss the current degree of independent replication.
7. Correct the noted typographic and nomenclature inconsistencies before submission.

Overall, I judge the manuscript to be conceptually interesting, reasonably coherent, and suitable as a hypothesis article only if its inferential boundaries are tightened and its quantitative assumptions are made more transparent.