Insulin and IGF support

In adults diagnosed with type 2 diabetes within five years and without clinical DPN, deficient input from neuronal insulin/IGF receptors contributes to loss of peptidergic sensory terminals. Restoring that input to its matched nondiabetic physiological level, before the attributed injury and without directly changing systemic glucose, would reduce five-year net terminal loss by at least 20% in a baseline-describable group covering at least 10% of reference-care incident DPN cases.

Insulin and insulin-like growth factors IGF1 and IGF2 bind partly overlapping receptor assemblies. The claim concerns productive signaling from those receptors into the neuron, whether reduced by insufficient ligand delivery, binding proteins that sequester ligand, receptor loss, or impaired proximal coupling. It is not a claim that circulating insulin must be low in type 2 diabetes. It also does not equate reduced response to an experimental insulin pulse with less ongoing support.

The molecular boundary is productive tyrosine-phosphorylation flux from native INSR homodimers, IGF1R homodimers and INSR–IGF1R hybrids into their IRS1/IRS2 and SHC1 adaptor complexes. These adaptors convey receptor activity to growth and metabolic pathways. Resolve INSR isoforms and receptor assemblies; do not count a hybrid twice. Measure mol of productive phosphate transfer per second per identified neuronal volume, together with adaptor occupancy and receptor location. Receptor mRNA, total protein, blood ligand concentration and total phospho-Akt are insufficient substitutes. Akt is a shared downstream kinase; its activation by other growth factors is not attributed to these receptors.

This defines a proximal receptor-input deficit. Resistance confined further downstream, despite normal productive receptor/adaptor input, is adjacent evidence about growth regulation, not this exact initiating defect. Insulin and IGF need not both fail: candidate corrections target INSR homodimers, IGF1R homodimers, hybrids, or the whole receptor family. These are four alternatives, not independent hypotheses whose probabilities are added. A correction retains physiological relative signaling through the named adaptors; it does not activate a chosen downstream branch selected for its growth effect.

For R1, a candidate's mean productive input in a connected peptidergic neuronal compartment must be below the age-, sex- and BMI-matched nondiabetic fifth percentile for at least 90 cumulative days during year one, preceding the losses attributed to it. The compartment is the soma, peripheral axon or terminal arbor of the neurons supplying the tracked skin field. Follow all baseline participants and count early losses; do not select future decliners or surviving axons. Permitted baseline groups are the whole eligible population, HbA1c ≥7.5%, fasting insulin above the matched nondiabetic 95th percentile, IGF1 below its fifth percentile, or directly measured local receptor input below its fifth percentile. Plasma-defined groups need local confirmation under R1. The same group, receptor correction and connected neuronal route must satisfy every row.

The ideal correction Q replaces deficient local receptor activation/coupling with its matched nondiabetic median time profile, including feeding-related variation, only when input is below the normal fifth percentile. It adds no activity above that median and leaves already-normal input unchanged. This requires ideal adult-onset control of ligand access and native receptor/adaptor coupling within the selected neurons. Physiological ATP and substrate costs remain; Q is not an energy source. It does not directly manipulate receptors in liver, muscle, vascular cells, Schwann cells or skin cells. Other initiating exposures remain comparable, including glucose, lipids, oxygen delivery and systemic hormones. Downstream growth signaling, metabolism, axonal transport, firing and neuron-to-glia communication may change. Artificially freezing these consequences would prevent the proposed mechanism from acting.

For each person, L is terminal disappearances minus newly established terminals over five years, divided by baseline terminal number. A new terminal that later disappears contributes to both counts. Identify peptidergic neurons stably, independently of changing CGRP or substance-P expression. With positive reference-care mean loss L0 and corrected mean loss L1, relative preservation is B=(L0−L1)/L0. A smaller contribution is B≥0.05; material contribution is B≥0.20. Coverage C is the qualifying group's fraction of reference-care incident DPN cases, with C≥0.10. Incident DPN means new bilateral length-dependent sensory signs with objective small-fiber loss or abnormal nerve conduction, after excluding other causes. These thresholds and the specified signal boundary define the proposition; they are not measured biological constants.

Source assessment

Best judgment: about 15%. Uncertainty range (sensitivity): 0.7–59%. The conditional product is 15.2%. The central uncertainty is whether an early human neuronal signaling deficit exists that normal physiological support would correct. Growth after extra insulin or insulin-like growth factor does not establish that deficit. These are subjective causal probabilities, not treatment effects or confidence intervals. Reviewed 14 September 2026. Probability method · Collection index.

Biological logic and conditional judgments

Receptor input can support protein production, transport, mitochondrial function and renewal. Loss of that input need not injure an axon if other growth factors compensate. Conversely, its restoration can preserve structure through any of those downstream routes; ATP depletion, apoptosis and impaired regeneration are not all mandatory intermediates. The structural comparison directly tests whether compensation leaves a deficit with consequences.

The complete claim is R1 AND R2 AND R3 AND R4 for at least one common candidate/group/compartment combination. Each probability after R1 assumes all its predecessors. R2 and R3 concern different magnitudes of the same causal effect: a 10% preservation effect satisfies R2 and fails R3.

See the claims and probabilities table above.

R1

Rodent ligand/sequestration evidence supports a deficit, but human tissue evidence is small, established-disease and confounded; a blunted Akt pulse need not locate the defect here.

R2

Depletion and structural rescue support a causal trophic role; compensation, cellular targeting and physiological normalization remain unresolved.

R3

Rodent effects can be substantial, but restoration after established injury and broad growth stimulation do not measure five-year prevention.

R4

Metabolic signaling abnormalities can be common; the frequency of the particular causal neuronal deficit is unknown.

The best product is 0.60×0.65×0.60×0.65=0.1521. The evidence cannot reliably distinguish a 15% judgment from nearby values such as 20%. Splitting compensation from its structural consequence would not justify a further automatic probability penalty.

Evidence supporting and challenging the claim

Human nerve findings support a possible IGF-availability defect, with substantial comparator problems. Simon and colleagues measured IGF-binding protein 5, which can sequester IGF1, in sural nerve. Their protein series included six DPN patients, five with type 2 diabetes, versus five non-diseased controls. Median ages were 60.5 versus 33 years; controls included biopsy and autopsy tissue. Protein was reported about 50-fold higher than non-diseased controls and fivefold higher than CIDP controls. The supplementary transcript comparison was one DPN patient versus one control: IGFBP5 7.48-fold higher, IGF1 reported 4.65-fold lower and IGF1R 5.28-fold higher. Neither the large fold changes nor repeated blots expand donor n. There was no diabetic-without-DPN comparison or direct human receptor-input measurement. Simon et al., 2015.

In a separate study of eight diabetic patients, five people with other axonal neuropathies and four diagnostic controls, sural IGF1 mRNA did not differ overall between diabetes and controls. Insulin-treated patients had higher expression than untreated patients, with similar changes in receptor expression. Degeneration and treatment can both alter the signal. This opposes a universal absolute ligand deficiency while leaving a local availability or signaling defect possible. Grandis et al., 2001.

Insulin-pulse resistance is not the same measurement as low trophic input. In adult rodent DRG cultures, 24-hour insulin pretreatment at 2–200 nM blunted the response to a subsequent 20 nM insulin pulse. At 20 nM pretreatment, receptor phosphorylation was not significantly reduced; the authors located much of the change downstream of IRS. Basal Akt phosphorylation tended upward, and IGF1-evoked signaling remained responsive. Thus one hormone's response can decline while another support route remains available. At least three experimental repetitions were reported, not a human donor cohort. Kim et al., 2011.

A simple arithmetic check shows the interpretive problem: a signal rising from 1 to 3 has a threefold pulse response; one rising from 2 to 3 has a 1.5-fold response despite twice the baseline and the same peak. This is an illustration, not reconstructed study data. A reduced challenge fold-change alone cannot establish lower integrated input before injury.

Direct insulin depletion and replacement influence axon properties. Brussee and colleagues gave STZ-diabetic rats intrathecal insulin at 0.1–0.2 IU/day or equimolar IGF1 for four weeks. They reported improved conduction and reversal of sural myelinated-axon atrophy without glycemic correction; the same low insulin dose given subcutaneously did not improve conduction. In nondiabetic rats, intrathecal anti-insulin antibody reduced mean sciatic axon area from 32.6 to 24.8 µm², a 23.9% reduction, compared with anti-albumin antibody. This is an adult depletion contrast, but concerns caliber rather than axon disappearance and is not neuron-exclusive. Arm allocation is insufficiently explicit to reconstruct an animal-level interval from the printed area SEMs. Brussee et al., 2004.

Toth and colleagues added a more relevant structural endpoint: after two months of STZ diabetes, one month of intrathecal insulin or equimolar IGF1 increased epidermal fiber density and length, with the largest improvement under IGF1. The accessible primary abstract does not supply independent n or numerical group effects. This is treatment during established experimental disease, not human pre-injury prevention. Toth et al., 2006.

Guo and colleagues treated opposite paws with insulin or carrier in STZ and db/db mice after three months of diabetes. Two days of local treatment increased PGP9.5-positive epidermal innervation at day five without lowering systemic glucose. The paired design strengthens the local contrast, but insulin receptors also occurred in keratinocytes. Numerical histology and animal counts were not accessible; separate day-14 behavioral measurements cannot be paired to those histology animals. Guo et al., 2011.

Receptor deletion and ligand sequestration give different answers. Constitutive sensory-neuron INSR deletion caused no detected skin-fiber loss through about 29 weeks in otherwise euglycemic mice: IENFD used four mice per genotype, conduction 24 controls and 23 knockouts. Whole-DRG receptor protein fell about 60% and insulin-evoked Akt activation was lost. Small structural n does not establish equivalence, and developmental adaptation or IGF support could compensate. Nevertheless, the result argues against inevitable axon loss from missing INSR alone. Grote et al., 2018.

In Simon's IGFBP5-overexpressing mice, ten-month epidermal PGP9.5 fiber density was approximately 507 versus 348 fibers/mm², about 31% lower, with four mice per genotype. A ±1.5-pixel reading check gives roughly 30–33% for that ratio; this is digitization sensitivity, not sampling uncertainty. Substance P also fell, but expression changes alone would not establish peptidergic axon loss. The model expressed IGFBP5 from development, could affect neighboring cells, and did not test selective adult reversal in diabetes. Neuronal IGF1R deletion produced myelinated/motor axon pathology, strengthening receptor dependence in those systems without identifying the target C-fiber effect. Simon et al., 2015, supplementary Figure A3.

Endogenous IGF can affect sensory-neuron growth without proving the population claim. In DRG cultures, neutralizing IGF1 or reducing its expression suppressed neurite outgrowth; about 85% knockdown at 180 nM siRNA was rescued by exogenous IGF1. IGF1 knockdown also blocked CEBPβ-driven increases in respiration and outgrowth. Tissue and culture contrasts generally used 3–6 animals or culture groups, not hundreds of independent neurons. The mouse cell atlas showed higher IGF1 transcript trends in nonpeptidergic neurons and satellite glia than in peptidergic neurons; it is a source-context observation, not absence of peptidergic receptor signaling. These experiments connect endogenous ligand to growth but do not supply an in-vivo terminal-prevention effect. Aghanoori et al., 2022.

Increasing growth can act beyond restoration of a deficit. Local insulin plus PTEN knockdown improved epidermal innervation and other nerve measures in chronic type 1 diabetic mice without glycemic correction. PTEN regulates a growth pathway used by many receptors; the combination does not isolate insulin deficiency or neuronal action. The accessible primary abstract supports the structural direction, but not a numerical independent insulin contribution. Pham et al., 2023. Sensory-neuron PTEN deletion can also produce extra axonal branches and epidermal innervation in intact nondiabetic nerves. That is an explicit counterexample to interpreting every growth response as correction of a diabetic deficit. Eaton et al., 2026.

Direction also depends on tissue. Adult-induced sensory INSR depletion prevented high-fat-diet-associated airway hyperinnervation in a six-mice-per-arm morphology comparison. More innervation is not universally a beneficial outcome, and this airway experiment cannot decide cutaneous preservation. Calco et al., 2022.

The human intervention evidence does not resolve prevention. Forty patients with established painful idiopathic small-fiber-predominant neuropathy were randomized to IGF1, 0.05 mg/kg twice daily, or placebo for six months. Pain and functional secondary measures showed no consistent benefit. There was no skin structural endpoint in the reported endpoint list and no diabetic prevention population. The trial limits a broad clinical-growth argument but does not estimate B for this claim. Windebank et al., 2004. The 2009 intranasal-insulin experimental paper was retracted and contributes no positive evidence. Publisher retraction, 2014.

Ideal experiments that would resolve each claim

These are definitions of ideal resolution, not proposed studies. Enroll representative independent humans from the stated early type 2 population, fix baseline group membership, identify the connected peptidergic neurons and track actual terminals for five years. Measure local receptor/adaptor input with sufficient temporal resolution to distinguish physiological pulses from a persistent deficit. For each of the four receptor corrections, randomize Q versus reference care before attributed loss. Confirm actual input restoration, assembly/substrate specificity and physiological energy costs throughout. An independent implementation and restoration of the original low-input profile distinguish the intended effect from an intervention artifact. Other initiating exposures remain comparable; downstream consequences remain free to respond.

Retain the same candidate/group/compartment combinations across rows. YES requires at least one retained combination to meet the row's boundary; NO requires every retained combination to fail. Use simultaneous intervals across the defined alternatives and groups. Equality belongs to the ≥ side for duration, preservation and coverage; a finite interval crossing a threshold remains unresolved. Nonpositive reference mean loss, failed signal engagement, loss of subtype identity or missing early terminal losses cannot establish preservation.

R1 — local input deficit, no predecessors. In unmanipulated reference-care participants and matched nondiabetic comparators, directly measure productive receptor-to-adaptor flux in the soma, axon and terminal compartments from baseline through year one. Relate the measured cells to their tracked terminals. YES is mean input below the matched fifth percentile for ≥90 cumulative days before attributed losses in a permitted group and candidate. NO is adequate measurement showing all candidates fail the level or duration condition; a 20-day reduction fails even if deep. An insulin-pulse response, lower IGF1 mRNA or bulk phospho-Akt alone is unresolved. This establishes exposure, not its harm. Human sural binding-protein findings and rodent ligand studies approximate parts of this comparison.

R2 — structural consequence, assuming R1. Use Q in the same qualified neurons and baseline group. Count each person's terminal disappearances and successful replacements to estimate B over five years. YES is B≥0.05; NO is B<0.05 with sufficiently narrow bounds across retained candidates. For example, L0=0.30 and L1=0.28 gives B=0.067 and meets R2. A precise upper bound below 0.05 establishes NO, even if signaling and neurite growth improve. Receptor input corrected only after loss, supraphysiological stimulation, or a change confined to conduction, pain or peptide staining remains unresolved. Rodent local rescue and depletion support causation without resolving this selective physiological comparison. Compensation is part of the result, not an extra undefined claim.

R3 — material effect, assuming R1–R2. Use that same randomized comparison and candidate, without selecting responders. YES is B≥0.20; NO under the predecessors is 0.05≤B<0.20. L0=0.30 and L1=0.24 meets R3, whereas L1=0.28 meets only R2. Require precision distinguishing the 20% boundary; a confidence interval spanning it remains unresolved. Neither an engineered 31% mouse density deficit nor reversal of axon atrophy estimates this human selective-prevention fraction.

R4 — coverage, assuming R1–R3. In a representative reference-care cohort, measure baseline membership in each independently qualified group and ascertain all incident DPN cases over five years. Estimate C=P(group | reference-care incident DPN). YES is C≥0.10 for at least one retained combination; NO requires C<0.10 for every one. A group with large preservation but 2% case coverage fails. An interval spanning 10%, selected clinical biopsies, or defining a group from its treatment response is unresolved. Blood insulin or IGF abnormality prevalence alone does not settle the prevalence of the structurally qualified group.

Sensitivity and consequential evidence updates

The coordinated skeptical interpretation uses [30,25,30,30]% and gives 0.675%. It treats much of the human association as age, treatment or degeneration, rodent rescue as additional growth stimulation, and normal trophic compensation as sufficient for most early type 2 terminals. The favorable interpretation uses [85,90,85,90]% and gives 58.52%: local ligand sequestration or receptor-input loss precedes injury commonly, adult depletion and structural rescue transfer to human neurons, and a baseline group captures substantial incident disease. These jointly defensible scenarios generate the row ranges and the approximately 58-percentage-point headline span. The span is sensitivity to interpretation, not a calibrated interval or independent row distributions.

Two additional scenarios separate the main ambiguities. Growth stimulation with uncommon true input deficiency, [30,65,60,65]%, gives 7.6%. A common deficit with a reproducible but small structural effect, [80,80,25,65]%, gives 10.4%. Sampling more cells within the same human biopsies would settle neither interpretation.

Hypothetically, a selective physiological human comparison showing preservation could raise R2 from 65% to 90%, moving the whole assessment to 21.1% with other rows fixed. Successful restoration with B<0.05 could lower R2 to 25%, giving 5.9%. Evidence about magnitude or coverage should also update those rows. Merging R1–R2 into their joint 39% probability preserves the same headline; decomposition alone does not change the evidence.

The strongest unresolved issue is direct early human input and its selective structural consequence. The detailed numerical results of Toth 2006, Guo 2011 and the full 2023 combination study were not accessible; this limits extraction, not their biological effects. Changing the receptor boundary or normalization profile changes the proposition and requires reassessment. With Q fixed, relaxing the 20% benefit or 10% coverage thresholds enlarges the truth set, without providing a numerical effect distribution.

This claim can overlap bioenergetic insufficiency or intrinsic growth restriction, but it does not require either as the sole mediator. Schwann-cell receptor deficits belong to glial support. Type 1 insulin deficiency is an adjacent exposure, and C-peptide deficiency is a separate proposition. Numerical support (background note outside this collection).