Hypoglycemic energy failure

In adults with T1D or T2D and recurrent or prolonged substantial hypoglycemia, local glucose deprivation causes inadequate usable ATP in distal peptidergic C-fiber axons. Selectively replacing the energetic contribution lost during those episodes would preserve at least 20% of five-year net physical terminal loss, in a baseline group containing at least 10% of reference-care incident or progressive neuropathy cases.

Eligibility requires, during the preceding 90 days, either at least two verified plasma-equivalent glucose episodes below 3.0 mmol/L for ≥30 minutes, or one below 2.2 mmol/L for ≥2 hours. Include insulin- and noninsulin-treated diabetes, any duration, with or without established DPN. A sensor dip alone is insufficient unless it represents true glucose exposure. These are explicit study-scope thresholds, not established nerve-injury boundaries or treatment targets.

Candidate baseline groups are all eligible people, T1D, T2D, established DPN, no DPN, and those meeting the prolonged <2.2 mmol/L criterion. No additional intersections or future responders qualify. This is a secondary-insult/progression scope; it is not the default recently diagnosed T2D population.

Local exposure. R1 requires at least two episodes in follow-up year one in which extracellular glucose immediately serving mapped distal axons/terminals falls to ≤2.5 mmol/L for ≥30 minutes because of the systemic low. This must affect units serving ≥10% of mapped terminals in ≥10% of a candidate group, before an additional baseline-normalized net terminal deficit of ≥5% lasts 30 days. Brief routine turnover is not that sustained deficit. Blood glucose and local extracellular availability are different quantities.

Usable ATP. ATP is the chemical substrate that powers cellular work. Measure both free ATP activity A and the positive magnitude Φ of the chemical free energy released by ATP hydrolysis. Calculate Φ using local ATP, ADP and phosphate activities, pH, free magnesium and temperature. Total ATP extracted from tissue, respiration and mitochondrial membrane potential are not interchangeable with these measurements.

R2 requires A or Φ to fall ≥20% below the same unit's pre-episode baseline and below its matched nondiabetic fifth percentile for ≥30 minutes in the same two episodes. The normal reference matches age, sex, temperature and activity. At least half the integrated shortfall must be attributable to the local glucose drop, as specified by the comparison below. This can occur despite compensatory lactate delivery, glycogen use, reduced firing or other fuels; no particular failed buffer is a separate mandatory mechanism.

U — isolate the glucose contribution. For the mechanistic reference, maintain local extracellular glucose at max(reference-care glucose, 3.9 mmol/L) during systemic episodes below 3.0 mmol/L. Apply this only to the mapped axonal/terminal neighborhoods and their immediately supporting Schwann cells. Preserve systemic glucose, insulin, hormones, perfusion and other incoming substrates. Let local metabolism respond. Comparing U with reference care identifies the part of the measured axonal energetic shortfall attributable to local glucose deprivation.

Q — replace that lost energetic contribution. During each systemic <3.0 mmol/L episode and the following 48 hours, provide a selectively delivered external chemical energy source and ATP-regeneration machinery to the mapped axoplasm. Q restores the joint A/Φ trajectory that would occur under U, including any persistent deficit that U would not correct. It does not normalize unrelated pre-existing energy failure or change systemic hypoglycemia.

Account for the required energy, adenylate, phosphate and reaction products. Preserve extracellular ATP/adenosine at its reference trajectory to exclude a direct purinergic treatment. Intracellular nucleotide sensing, maintenance, transport and repair may respond naturally. Q does not directly prevent inflammation, replace lost axons or suppress an execution enzyme. Removing useful metabolic adaptations can offset benefit.

This source-specific energetic replacement is an ideal causal intervention, not an available drug. It tests the fuel-deprivation route without crediting it with every effect of preventing systemic hypoglycemia, such as avoiding a fall or a cardiovascular event. U defines the energy contribution; Q's physical outcome determines whether that contribution matters.

Physical outcome. Track stable-identity peptidergic terminals in fixed bilateral distal-leg and dorsal-foot territories for five years. Net loss is disappearance minus newly established connected terminals, including later loss of replacements. Peptide downregulation, electrical silence and detached debris do not count as physical loss or preservation. Normalize to entry terminal amount, average territories within people and then people equally. For positive reference-care loss L₀ and Q loss L_Q, define B=(L₀−L_Q)/L₀.

C is the share of reference-care people with objectively incident or progressive neuropathy belonging to the qualifying baseline group. Incident disease requires new bilateral length-dependent neuropathy with objective small-fiber loss or abnormal nerve conduction; progression requires new objective deterioration in established disease. Pain alone is insufficient. A complete qualifying group must satisfy every preceding requirement; different groups cannot supply different pieces.

Source assessment

Best judgment: about 10%; uncertainty range (sensitivity): 0.46–56%. Prolonged hypoglycemia can damage peripheral axons. The uncertain bridge is from actual human low-glucose episodes to a local energy deficit that causes material, irreversible loss of peptidergic terminals. These are subjective probabilities of the complete proposition, not the risk attached to one low glucose reading. Reviewed 14 September 2026. Probability method · Collection index.

Biological reasoning

An electrically quiet axon still needs energy for ion gradients, membrane renewal and maintenance. Yet a low blood glucose value does not establish an energy deficit in that axon. Local delivery, endogenous stores, alternative fuels and reduced work can maintain the necessary chemical state.

Production and consumption remain balanced at steady state. A damaged energy system may meet reduced work demand while supporting less maintenance; it need not consume more ATP than can indefinitely be produced. Conversely, low total ATP can reflect fewer surviving cells rather than depletion within living axons.

Function and survival are also separate. Conduction can stop well before irreversible destruction and can return when glucose or lactate is supplied. Small unmyelinated axons can lose conduction sooner than large myelinated axons in one preparation, while prolonged in-vivo hypoglycemia preferentially damages large motor fibers. That difference warns against translating a conduction-latency ordering directly into a structural-vulnerability ordering.

Hypoglycemia can additionally trigger inflammation, altered perfusion or injury during glucose recovery. Those may contribute to neuropathy without the defined local energetic replacement preserving terminals. This proposition asks about the energy component specifically.

Primary evidence

Peripheral glycogen buffers conduction, with a fiber-class distinction. In isolated adult mouse sciatic nerve, blocking glycogen breakdown with 1 mM DAB shortened the latency to a 5% decline in A-fiber compound action potential during zero-glucose exposure from 2.10±0.34 to 0.64±0.12 hours (mean±SD), reported n=4 preparations per condition. The C-fiber comparison was 0.58 versus 0.50 hours, P=0.13. These are onset latencies, not times until all axons fail or die.

Glycogen remained higher with DAB after deprivation, consistent with blocked use. Exogenous lactate supported both fiber classes. Preparations, rather than independently identified donor animals, were reported; possible bilateral pairing is not recoverable. The study establishes acute functional support, not irreversible human terminal loss. Absence of a significant C-fiber DAB effect does not prove that all glial fuels fail to support C fibers. Brown et al., 2012.

Prolonged hypoglycemia produces actual axonal degeneration. Jensen randomized 36 young nondiabetic rats, six of each sex per arm, to vehicle or two insulin-infusion doses for 29 days. Glucose was generally 2–4 mmol/L in treated groups. There were 11 premature deaths, including hypoglycemic and catheter-related losses.

Including all animals with examined nerves, axonal degeneration occurred in 18/23 insulin-treated rats versus 0/11 controls; Wilson intervals for those proportions are approximately 58–90% and 0–26%. Among terminal survivors alone, the corresponding numbers were 14/16 and 0/9. The structural signal therefore does not depend solely on dropping premature deaths.

However, this comparison changes insulin, weight and feeding as well as glucose, lacks an insulin-matched euglycemic arm, and evaluates predominantly myelinated sciatic pathology. Mixed exposure durations and small dose/sex strata limit precision. It establishes injury under prolonged insulin-associated hypoglycemia, not a human C-terminal threshold or an effect of Q. Jensen et al., 2016.

A spontaneously diabetic WBN/Kob-rat study also found severe axonal degeneration in the most hypoglycemic insulin-implant group after 40 days, tending to affect large myelinated fibers. Group mean nerve-conduction velocity did not significantly distinguish the three exposure groups. Its morphology supports physical injury despite an inconclusive functional group contrast; it is not a selective small-fiber or ATP-rescue experiment. Ozaki et al., 2010.

Human prospective evidence is larger than the corneal study, but the endpoint is broad. Hypo-RESOLVE pooled 84 insulin trials containing 39,373 people. Fully adjusted neuropathy analyses involved 5,292 T1D participants with 35 events and 17,125 T2D participants with 112 events. Each standard-deviation increase in log(1+cumulative reported hypoglycemic episodes) predicted rate ratios 2.81 (1.48–5.34) and 2.38 (1.63–3.49), respectively.

These were time-updated observational exposure analyses, not randomized hypoglycemia contrasts. Median trial duration was approximately 40 weeks in T1D and 26 weeks in T2D. Neuropathy was an adverse-event category including new or worsening peripheral sensory, motor, sensorimotor, poly- and mononeuropathy. There was no serial terminal count.

Adjustment included baseline neuropathy, glycemia, insulin and medication variables, but ascertainment, unavailable confounders and rapid improvement remain alternatives. Separate severe-event coefficients were imprecise and did not establish a depth gradient. Trial participants with recurrent severe hypoglycemia were often excluded. The study supports temporal association, not source-specific energetic mediation. Mellor et al., 2024.

Corneal structure supplies a small exploratory association. Gad studied 20 insulin-treated T1D and 20 insulin-treated T2D participants, plus 28 controls, with four days of glucose monitoring. Median corneal branch density was 25.0 versus 37.5 branches/mm² above versus below 4% time in level-1 hypoglycemia, P=0.045; for >1% versus <1% time below 54 mg/dL, it was 25.0 versus 35.4, P=0.040.

Other corneal endpoints did not show the corresponding significant contrasts. Across all 72 displayed within-diabetes contrasts and correlations in Tables 2–3, Benjamini–Hochberg correction gives minimum q=0.396, with none below 0.05. This is a combined exploratory family, not an adjusted result reported by the authors.

Short exposure sampling, unknown temporality and no joint confounder-adjusted structural model prevent a causal interpretation. Group medians and ranges do not provide the individual data or subgroup n needed to reconstruct a reliable effect interval. Corneal branches are not distal cutaneous peptidergic terminals. Gad et al., 2022.

Schwann-cell experiments support stress, without demonstrating axonal energy failure. Kato compared normal, constant low, constant high, intermittent low and intermittent high glucose over three days; intermittent exposure lasted one hour three times daily. These are separate low and high schedules, not necessarily one alternating low/high sequence. Viability fell and stress/apoptosis markers rose; 4-phenylbutyrate reduced these effects. An immortalized cell line and stress markers cannot determine the human terminal-loss consequence. Kato et al., 2019.

A 2025 study in the same line used 2.5 or 25 mmol/L glucose for one hour, with 100 μM imeglimin given for 24 hours. ATP content decreased slightly, alongside membrane-potential, cell-count and LDH changes; imeglimin improved them. Assay n ranged from 3 to 10. Total ATP normalized to protein is not measured free ATP activity or hydrolysis energy in an axon. The tested drug concentration was about 15 times the cited human plasma peak, and the timing is not delayed terminal rescue. Kato et al., 2025.

Insulinoma cases support lasting human dysfunction, with different exposure and anatomy. A 2026 report described a 22-year-old woman with recurrent symptoms over 6–12 months and fasting glucose 1.7 mmol/L. After tumor removal, median-nerve sensory amplitude improved from 3.3 to 16.7 μV at four months, while some motor deficits persisted. Her sural sensory amplitude was normal before treatment.

This is a non-diabetic case with motor-predominant, partly upper-limb disease and additional treatment, not serial small-fiber histology or a controlled prevention effect. Its partial reversibility supports neither inevitable permanent C-fiber destruction nor a claim that hypoglycemic neuropathy is always transient. Körei et al., 2026.

Conditional probabilities

Each row is conditional on every earlier requirement being true in at least one surviving baseline group. Lower/best/upper values are coordinated sensitivity judgments, not independent measured frequencies.

See the claims and probabilities table above.

R1

The population is selected for substantial systemic lows, but local trajectories and recurrence are not established for every group.

R2

Acute nerve physiology and cell ATP changes support plausibility; alternative substrates and reduced demand can preserve usable energy.

R3

Prolonged animal exposures cause physical injury, while human small-fiber specificity and selective energy rescue remain missing.

R4

Five-year loss may mainly reflect hyperglycemia, vascular injury or repair failure despite a real hypoglycemic contribution.

R5

Enrichment for substantial lows improves plausibility, but the frequency of destructive local deficits is unmeasured.

An electrical failure can satisfy R1 and R2 without satisfying R3. A real 10% structural benefit satisfies R3 but fails R4. A large benefit confined to a baseline group containing 3% of cases fails R5. No extra generic “translation factor” is multiplied after these judgments.

Ideal experiments that would resolve each claim

These comparisons define decisive evidence, not proposed laboratory work. Assume exact systemic/local glucose and chemical-activity measurements, selective U/Q, and nondestructive tracking of connected human terminals. Randomize independent people where manipulation is required. Measure complete structural and vital-status trajectories; disappearance from follow-up cannot be coded as recovery. Finite confidence intervals crossing a boundary, inadequate target engagement or incomplete follow-up mean unresolved.

R1 — systemic lows reach the target compartment. Under reference care, follow complete glucose histories and local availability in the mapped units through year one, alongside structural trajectories. Use the local-input comparison to establish the systemic episode's contribution. YES: at least one candidate meets two episodes, ≤2.5 mmol/L, ≥30 minutes, ≥10% of people/units and the pre-further-loss boundary. NO: complete measurements exclude the conjunction in every candidate. Four days of arm-sensor data or a fasting plasma nadir alone leave this local claim unresolved.

R2 — compensation fails to maintain usable energy. In R1-qualified units, compare reference care with U during the same exposure patterns. Measure A, Φ, alternative substrate use and the baseline/reference profiles, without treating ATP per biopsy as axonal activity. YES: a surviving group meets the defined 20% and fifth-percentile deficit for ≥30 minutes in both episodes, and U removes ≥50% of the integrated shortfall. NO: every remaining group fails at least one boundary under complete exposure and chemical measurement. Conduction silence with preserved A/Φ is NO for this energy-deficit claim. Brown measures a related functional reserve; current human studies do not make this comparison.

R3 — energy replacement preserves actual terminals. Randomize Q versus reference care within surviving groups for five years. Verify that Q reproduces the joint U energy trajectory while systemic lows and other incoming exposures remain comparable. Track physical connected structures, including regrowth and its later loss. YES: at least one group meeting R1–R2 has L₀>0 and B≥5%. NO: none does with complete engagement and observation. Reduced pain or better conduction alone cannot establish YES. Insulin-infused rats show structural plausibility but do not isolate this intervention.

R4 — preservation is material. Retain the same intervention, groups and terminal-loss denominator. YES: at least one R1–R3 group has B≥20%. NO: every such group has B<20%. A precise 10% benefit is a positive causal result and an ideal NO for materiality; an interval spanning 0–30% remains unresolved. Neither a rat degeneration incidence nor a human neuropathy-report rate ratio estimates this fraction.

R5 — the benefiting group is sufficiently common. Follow the full eligible reference population and classify incident/progressive neuropathy and baseline groups independently of treatment response. YES: a group satisfying R1–R4 contains ≥10% of cases. NO: every remaining group falls below 10%. Overlapping groups are evaluated separately, not added. Missing the prolonged-low group or excluding people with the heaviest prior exposures leaves a complete NO unresolved.

Calibration and sensitivity

The best product is 0.85×0.55×0.60×0.55×0.65 = 10.027875%, reasonably about one chance in ten. Evidence supports “sufficiently prolonged hypoglycemia can injure peripheral axons” more strongly than this specified human energetic-preservation proposition.

The lower scenario [60,25,25,35,35]% → 0.459% assumes most observed human episodes are locally buffered, with apparent associations dominated by disease severity, ascertainment and nonenergetic pathways. The upper [95,85,90,85,90]% → 55.60% assumes substantial lows repeatedly exhaust distal reserves and that large-fiber animal injury transfers to a common small-fiber energy mechanism. The main uncertainty is not whether zero glucose can stop a nerve; it is which real human exposures cause an energy-dependent structural deficit.

The approximately 55.1-percentage-point width is sensitivity to these coherent interpretations, not a sampling interval or a probability distribution estimated from study n. More participants in an adverse-event analysis cannot substitute for local energetic engagement or the physical endpoint.

A favorable hypothetical update would show episode-linked axonal A/Φ collapse, verified U correction and selective Q preserving connected human terminals across common baseline groups. [95,80,85,80,80]% → 41.34% expresses that coordinated update. Well-measured local energy preservation during severe systemic lows, or complete Q engagement with little physical benefit, would instead support [80,35,25,35,40]% → 0.98%. Exact ideal NO for any necessary claim makes the conjunction false.

Another possible interpretation is structural damage mainly during glucose recovery through inflammation or oxidant production. Q could then change energy measurements without preserving terminals. Conversely, a small group with many hours near 1 mmol/L could have compelling injury while failing the 10% case-coverage requirement. These alternatives are not resolved by the current pooled association or by an assay precision estimate.

Merging R4 and R5 gives 35.75%, preserving the same best joint. Adding a compulsory glycogen, lactate or mitochondrial pathway would incorrectly require one optional explanation. Lowering B or C cannot lower the probability of the weaker conjunction; changing eligibility changes the population and can move probability either way.

This is a source- and episode-specific version of local energy failure, with a different population and intervention, so no direct probability ordering is implied. Rapid glycemic correction need not involve absolute lows. Glucose variability asks about waveform effects beyond matched hypoglycemic burden. These are overlapping causal questions, not independent contributions to add.

Quantitative support (background note outside this collection).