Ferroptotic injury

In adults with type 2 diabetes diagnosed within five years and no clinical DPN at entry, iron-dependent phospholipid peroxidation produces irreversible membrane injury in the connected peptidergic sensory unit. Selectively interrupting that chemistry from entry would avert at least 20% of five-year net terminal loss in a baseline-defined group accounting for at least 10% of reference-care incident DPN.

The target is physical cutaneous terminals of adult peptidergic unmyelinated neurons. The finite cellular scope is those neurons, their directly ensheathing nonmyelinating Schwann cells and their somal satellite glia. Spinal microglia, myelinating cells of unrelated axons and circulating markers alone do not qualify.

Here ferroptotic membrane injury requires demonstrated iron-dependent phospholipid oxidation followed by irreversible loss of cellular membrane integrity. In a neuron, this may destroy a bounded peripheral axonal domain without immediate death of its soma. This is a compartmental injury proposition, not a claim that every affected neuron dies as a whole. Reversible oxidative signaling, reduced metabolic-dye conversion or lower GPX4 expression alone fails the fatal-injury requirement.

Identify subtype through combined anatomical and molecular classification. Track actual arbors independently of changing CGRP, PGP9.5 or reporter intensity. For each person, five-year net terminal loss is destruction minus successful replacement, divided by baseline terminal amount. Average people equally:

B = (mean reference net loss − mean net loss with Q) / mean reference net loss.

The reference mean must be positive. Reference care is contemporary diabetes care without Q. Count persistent structures, disappearance and replacement separately; pain, conduction and dye intensity are not structural substitutes.

Finite groups. Evaluate the entire eligible population, males, females, baseline BMI ≥30 kg/m², baseline HbA1c ≥7.5%, and a baseline phospholipid-oxidation-positive group. The last requires the R1 rate below for at least one cumulative hour during a 30-day entry observation in at least 10% of mapped sensory units. Additional intersections and groups defined by subsequent response do not qualify.

Incident DPN means new bilateral length-dependent peripheral nerve signs with objective small-fiber loss or abnormal conduction, excluding another cause. Coverage is C = P(group membership | incident DPN under reference care). All requirements must hold in the same group and connected units.

R1 exposure definition. Measure the formation rate of esterified membrane-phospholipid hydroperoxides per amount of membrane phospholipid per hour, resolving formation separately from reduction, cleavage and membrane replacement. A qualifying rate is at least twice the positive age-, sex-, cell-class- and site-matched nondiabetic median and above its 95th percentile.

At least 10% of mapped units, with equal person weighting, must have one same cell class accumulating at least 24 hours of qualifying activity during year one. Activity must precede the connected arbor's first ≥5% net physical terminal deficit maintained for 30 days. These boundaries define the assessment; they are not established biological thresholds.

Q — selective interruption. From entry for five years, Q suppresses at least 95% of iron-catalyzed initiation and amplification of membrane-phospholipid peroxidation in the three named cell classes. It changes those chemical reactions, not iron abundance generally. It leaves direct oxygen transport, iron-dependent energy enzymes, systemic iron supply, glucose, lipid delivery and other initiating exposures intact. These are ideal molecular capabilities, not claims about available chelators.

Q includes enzymatic and nonenzymatic iron-dependent membrane oxidation; no particular ALOX enzyme, GPX4 deficiency, iron overload or glutathione deficiency is mandatory. Normal protective signaling may also depend on the interrupted reactions. All downstream changes in inflammation, mitochondrial function, membrane repair and replacement count, including adverse consequences.

The priced structural effect is the total effect of interrupting this chemistry in units where a lethal membrane-injury branch exists. It does not claim that every saved terminal was rescued only at the final rupture step. Earlier dysfunction within that route may also contribute.

Source assessment

Best judgment: about 8%; skeptical–favorable sensitivity: 0.09%–47%. These are subjective probabilities of the complete human proposition, not measured treatment effects. The chemistry and capacity for neuronal ferroptosis are established more strongly than their occurrence and structural importance in early human diabetes. Reviewed 14 September 2026. Probability method · Collection index.

Reasoning from the biology

A simplified membrane balance is:

hydroperoxide accumulation = initiation + chain propagation − enzymatic reduction − removal/replacement.

Iron can help initiate or amplify lipid oxidation without increased total tissue iron. Conversely, iron accumulation in lysosomes need not mean more iron available to the relevant membrane. Ferritin, ferroportin and iron-sensitive probes measure different quantities.

GPX4 reduces phospholipid hydroperoxides, but its concentration is not its local protective rate. Substrate access, membrane anchoring, glutathione and competing defenses matter. FSP1/coenzyme-Q chemistry, lipid replacement and membrane composition can compensate. A reduced GPX4 signal therefore does not identify death, while normal protein abundance does not exclude vulnerability.

The chain separates three questions:

excess membrane oxidation → substantial iron dependence → irreversible membrane injury → durable terminal preservation when interrupted.

A peroxide signal may remain sublethal and participate in repair. A metabolic viability assay may change because cells proliferate more slowly. A radical-trapping antioxidant can improve several oxidation-dependent processes. Neither a marker panel nor rescue by one broadly acting drug establishes the full chain.

Cell-to-cell propagation is possible in other systems, but is not required here. The route can injure individual neurons or their support cells without spreading between units. Similarly, caspase activation or another death pathway can coexist; the question is whether the defined iron-dependent membrane chemistry makes a causal contribution, not whether every competing death label is absent.

Evidence that moves the judgment

The principal human association has an inadequate comparator. Wu compared 65 DPN patients with 23 healthy controls. Mean ages were 62.25 versus 26.30 years, and mean diabetes duration was 9.6 years. Serum GPX4/GSH were lower and MDA higher in DPN, but there was no diabetes-without-DPN comparator. The 35.95-year age difference, glycemia and disease status prevent attribution to nerve ferroptosis.

The cell experiments used the rat RSC96 line at 100 mM glucose for 48 hours, with 25 mM as control and 10 μM ferrostatin-1. Readouts emphasized CCK-8, Ki67 and protein expression; the published methods do not supply a direct iron-dependence test or physical death assay. Rescue of proliferation/metabolic activity is weaker than rescue of a demonstrated ferroptotic event. Independent culture n is not clearly reported in the figure captions. Wu et al., 2023.

ALOX15 perturbation is informative, but its discovery analysis is mischaracterized. The 2025 study reports lipid-oxidation and iron-probe changes plus improved CCK-8 after ALOX15 siRNA in undifferentiated SH-SY5Y neuroblastoma cells at 50 mM glucose for 72 hours. Most biochemical assays had three independent preparations; some viability experiments had 3–6. The diabetic mouse comparison had six per group and measured DRG markers and pain, without an in-vivo ALOX15 structural-rescue experiment.

The paper describes GSE34000 as three diabetic and three control rats and reports DESeq2 analysis. The original record instead has one five-rat RNA pool per condition at each of three time points, measured on Affymetrix arrays processed with MAS5. These are neither individual-rat RNA-seq counts nor three exchangeable animals. That invalidates the stated basis of its gene-selection analysis; it does not erase the separate siRNA measurements. Feng et al., 2025, original GEO study.

A rare human disorder supplies stronger mechanistic evidence. Lorenz studied a GPX4 R152H variant in three affected patients; the functional human-cell comparison used one patient's cells, his heterozygous father's cells and corrected clones. Cortical neurons and forebrain organoids were rescued by liproxstatin-1 or genetic correction. The variant impaired membrane anchoring despite retained catalytic activity in biochemical assays.

This demonstrates why protein abundance and soluble-enzyme activity can miss membrane vulnerability. The cortical differentiation used liproxstatin during preparation and tested withdrawal; it was not a diabetic sensory-neuron exposure. Three or more repeated experiments do not turn the two cell donors into a population estimate. The rare genotype establishes biological possibility, not 10% incident-DPN coverage. Lorenz et al., 2026.

A newer neural PARP1 study remains a pain and marker experiment. Guo's nerve-constriction work used PJ34, olaparib, intrathecal ferrostatin-1 and GPX4 inhibitor RSL3. Drug arms had six mice. PARP1 deletion reduced pain/excitability and changed GPX4, iron and lipid-oxidation markers. The methods specify Nestin-Cre, and do not demonstrate an adult DRG-exclusive deletion. Direct death tracking and terminal preservation were not reported. These findings support an adjacent pathway without proving ferroptotic loss of diabetic terminals. Guo et al., 2026.

The accessible diabetic structural positive is not selective. The sinomenine-hydrogel primary abstract reports improved mouse IENFD, axonal diameter and myelin, together with lower glucose and higher insulin. The proposed cell mechanism is microglial and the intervention changes Nrf2/Hmox1 broadly. Full methods, structural n and effect values were not recovered; no quantitative preservation fraction can be assigned. This is a structural lead under altered systemic exposure, not a selective local ferroptosis experiment. Chen et al., 2026.

Lipid oxidation can also support repair. A mouse transection/resuture study applied the GPX4 inhibitor ML162 locally at 0.25 μM for one minute and reported improved early recovery, alone or with PEG. Behavioral arms were 5/3/10/9, with additional endpoint-specific exclusions. Sensory-neuron Gpx4 deletion produced some improved early functional scores at n=4/4 but was followed by death within about three weeks. These observations do not recommend chronic GPX4 inhibition; they show why short-term signaling, regeneration and fatal oxidation must be distinguished. The source does not demonstrate membrane fusion directly in mice. Ko et al., 2025.

Reverse direction and pathway labels are credible alternatives. In a corneal-denervation study, loss of TRPV1-positive sensory input preceded epithelial ferroptosis-associated changes and delayed healing. The epithelial phenotype can therefore be downstream of lost innervation. Corneal study, 2025.

A 2026 omega-3 analysis explicitly assessed ferroptosis-associated stress rather than demonstrated cell death. Its human component reused a three-month supplementation cohort of 40 adults with T2D; serum pathway predictions were not local nerve-death measurements. Omega-3 study, 2026.

No accessed human study directly measured this sequence before incident DPN or randomized an engaged, selective interruption with structural follow-up. Human bulk nerve/DRG expression and enrichment cannot substitute for those measurements. Failure of an expression screen to discriminate ferroptosis is an assay limitation, not evidence of zero biological involvement.

Probabilities of the logical claims

Each row assumes all preceding rows are true in the same group. These are conditional judgments; the chain rule requires no independence.

See the claims and probabilities table above.

R1 credits diabetic oxidation evidence while retaining the large compartment, timing and formation-versus-disposal gaps. R2 is relatively high conditional on genuine excess membrane oxidation: iron-dependent chemistry is plausible, although neither high total iron nor an ALOX expression change establishes the required share.

R3 is the main unresolved biological transition. Genetic and human-cell evidence establish that this mode of injury exists, but diabetic studies often substitute metabolic activity or marker panels for physical failure. Its 20–80% range represents the disagreement between those interpretations.

R4 is relatively high once actual injury in connected units is established; it does not discount the existence of ferroptotic damage again. Replacement and glial redundancy can still prevent a durable terminal difference. R5 and R6 separately price effect magnitude and population coverage, neither measured by the rare human genetic example.

The absolute 0.2-percentage-point event reduction in R3 defines a small lethal-injury contribution; it is distinct from the later terminal-loss materiality threshold. A B of 10% passes R4 and fails R5. A large effect in a group covering 3% of incident DPN fails R6.

Ideal experiments that would resolve each claim

Assume noninjurious human membrane-chemistry measurements, continuous structural tracking and selective control of the named reactions. Count independent humans. Unlimited precision defines YES/NO; finite intervals crossing a boundary, incomplete engagement or ambiguous classification remain unresolved.

R1 — formation before loss. In representative eligible adults and matched nondiabetic references, quantify formation and disposal of esterified phospholipid hydroperoxides in the mapped neuronal, Schwann and satellite-glial membranes during entry and year one. Establish the two-fold/95th-percentile boundary, 24-hour duration, ≥10% unit fraction and pre-deficit timing. YES means one complete candidate-group/cell-class condition holds; NO means none does. Concentration alone, blood markers or post-degeneration samples are insufficient. Existing DPN studies chiefly supply abundance and reporter measurements, not this longitudinal flux.

R2 — iron dependence. Assuming R1, apply randomized rapid local Q on/off comparisons to the same domains at their observed exposure, before downstream changes in membrane composition or supply obscure the reaction contrast. Define excess as the untreated rate minus the matched reference median. Average excess rates across qualifying domains with equal person weighting and measure the proportion removed by Q, verify ≥95% inhibition of the intended iron-dependent reactions, and check that non-target iron functions remain directly comparable. YES requires ≥50% reduction of that excess in the qualifying domains; NO is <50% for every still-qualifying group. An orthogonal membrane radical-trapping manipulation and restoration of the iron-dependent reaction distinguish peroxidation from probe artifacts. Existing ALOX15 siRNA and iron-probe evidence approach parts of this test but do not resolve it.

R3 — irreversible membrane injury. Assuming R1–R2, randomize people to Q or reference care from entry and track the same mapped units through year one. Count a unit as affected when at least one named cell or bounded neuronal membrane domain loses barrier integrity, physically disintegrates and fails to restore viable continuity for at least 24 hours. Exclude channel-mediated permeability and transient dye entry. Measure lipid oxidation before rupture, distinguish cell proliferation from survival, and confirm the injury reduction with an independent peroxidation-interruption method while keeping initiating exposures comparable. Estimate the person-weighted difference in the fraction of mapped units affected, using all entry-mapped units in each randomized group. YES requires at least 0.2 percentage points fewer affected units with Q; NO means every qualifying group has a smaller reduction. For example, 2.0% versus 1.5% affected passes, whereas 2.0% versus 1.9% fails. Do not select units using post-treatment oxidation or survival. Sublethal oxidation alone fails this claim. Incomplete reaction control or insufficiently resolved physical fate remains undecided. Existing human genetic models demonstrate rescue of neuronal degeneration under a different initiating defect, while diabetic marker studies lack the complete comparison.

R4 — lasting terminal contribution. Assuming R1–R3, continue the same randomized groups for five years with durable local Q engagement. Measure physical peptidergic terminal loss and successful replacement throughout follow-up; permit downstream signaling, metabolism and repair costs to change. YES requires positive reference loss and B≥5% in a same qualifying group; NO is B<5% for every such group. Glial death without consequential terminal loss can pass R3 and fail R4. Brief pain improvement, low GPX4 or incomplete subtype tracking does not resolve it. The sinomenine structural report supplies a related endpoint but changes systemic exposure.

R5 — material magnitude. Assuming R1–R4, estimate that same human structural contrast and denominator with precision sufficient to distinguish B=20%, including delayed damage and replacement. YES is B≥20% for at least one still-qualifying group; NO means all have 5%≤B<20%. An interval spanning 20%, a shorter observation window or selecting future responders remains unresolved. No accessed selective ferroptosis study provides this five-year human quantity.

R6 — incident-case coverage. Assuming a group passes R1–R5, follow a representative reference-care population and ascertain incident DPN uniformly over five years. Calculate the fraction of all incident cases belonging to each qualifying entry-defined group, resolving outcomes after dropout. YES requires C≥10% for one such group; NO means all have C<10%. Rare mutations, serum-marker prevalence and enriched treatment samples cannot supply the denominator. Existing evidence does not measure it.

Uncertainty, sensitivity and overlap

Best inputs [65,80,50,75,65,60]% give 7.605%. Skeptical [25,50,20,40,35,25]% and favorable [90,95,80,90,90,85]% give 0.0875% and 47.0934%. These are coordinated evidence interpretations, not an effect-size interval or an empirically calibrated posterior.

A model-restricted interpretation, [45,65,30,60,45,40]%, gives 0.95%. A physiological membrane-injury interpretation, [80,90,70,85,80,80]%, gives 27.42%. No empirical weights or variance are assigned to these scenarios.

Hypothetically, direct early-human membrane-fate evidence moving R3 to 80%, with other judgments fixed, raises the headline to 12.17%. Evidence moving R3 to 20% lowers it to 3.04%. Establishing R3 with certainty leaves 15.21%. These numbers describe evidence updates, not the expected value of another study.

R1–R3 carry 26% joint probability. Merging their explanation must preserve that belief; adding unnecessary compulsory GPX4 depletion, iron overload or intercellular propagation must not impose additional discounts. Definition changes to timing, event frequency or materiality require a separate re-elicitation.

This route overlaps oxidative injury, PARP1 and glial support. A broader oxidation claim can hold without ferroptotic membrane failure. Different interventions and coverage definitions prevent casual arithmetic comparisons; the probabilities are not additive contributions or treatment rankings.

Reproducible checks (background note outside this collection) retain the human age-comparator problem and recover the actual GSE34000 sample design without treating pooled arrays as individual rats.