# 15. Inadequate microvascular oxygen or fuel delivery

**Best judgment: about 15% (15.4% by multiplication); interpretation sensitivity: 0.5–66%.** These are subjective probabilities of the proposition, not effect sizes or sampling intervals. Reviewed 14 September 2026. [Probability method](probability_method.md) · [Collection index](README.md).

## The causal proposition

In adults with type 2 diabetes diagnosed within five years and no clinical DPN, deficient microvascular exchange reduces oxygen or metabolic-fuel availability along the connected peptidergic sensory neuron before terminal loss. Selectively normalizing that exchange would prevent **at least 20% of five-year net distal terminal loss** in a baseline-describable group accounting for **at least 10% of reference-care incident DPN cases**. Severe large-artery ischemia and vasculitis are separate causes; exclude known vasculitis, ankle systolic pressure below 50 mmHg or toe systolic pressure below 30 mmHg from this proposition. These are scope choices.

Reference care continues without the selective correction. Incident DPN means new bilateral length-dependent clinical signs with objective small-fiber loss or abnormal nerve conduction, excluding another sufficient explanation. The structural target is the distal-leg/foot ending of a peptidergic C-fiber, identified independently of changes in peptide expression. For each person, `L = (terminal disappearances − successful new terminals)/baseline terminal number` over five years; a new ending subsequently lost contributes to both counts. Average sampled territories within each person. For a group, `B = (mean L_reference − mean L_corrected)/mean L_reference`, requiring positive reference loss. `C = P(group | incident DPN under reference care)`.

The finite baseline groups are the whole eligible population; HbA1c ≥7.5%; foot transcutaneous oxygen below 40 mmHg; or directly measured local vascular exchange capacity below the matched nondiabetic fifth percentile. The last group requires ideal measurements. These groups may overlap but are not unions selected after outcomes. Membership cannot depend on later hypoxia, degeneration or treatment response. The transcutaneous cutoff defines a candidate group, not a proven neuronal ischemic threshold.

### Exposure and selective correction

Candidate compartments are the cutaneous terminal environment, the connected axon/Remak unit, and its DRG soma. Candidate delivered species are oxygen, glucose, lactate, pyruvate and β-hydroxybutyrate. A single species/compartment or their joint correction can qualify. Alternative fuels, glycolysis and extraction reserve remain available; failure of all energy routes is not assumed.

**Exchange capacity** means the microvascular blood-to-interstitial transfer function: delivered amount per tissue volume and time as a function of arterial content, perfusion pressure and tissue concentration. Measure the function under the same boundary conditions as age-, sex-, site- and subtype-matched nondiabetic tissue. This distinguishes low vascular capacity from low arterial oxygen content, low demand and poor intracellular use. Units are molar flux per tissue volume; oxygen availability is local partial pressure, fuels are free interstitial concentrations.

R1's deficit is below-fifth-percentile exchange capacity for ≥90 cumulative days in year one while the relevant tracked neuronal territory remains structurally intact. It must occur in **at least 10% of baseline mapped neurons, averaged first within each person and then across the group**. One exceptional vessel or cell cannot satisfy this population claim. The molecule and connected compartment must remain the same in later requirements. Capillary counts, basement-membrane thickness and endothelial transcripts alone do not meet this definition.

**Q** restores only a deficient candidate blood-to-interstitial exchange function to the matched nondiabetic median profile, retaining actual arterial content and pressure. An ideal blood-connected exchange interface could supply missing transfer without independently manipulating endothelial growth factors, neuronal receptors or other barrier permeability. Their endogenous responses to restored supply remain free to change. Material comes from circulating blood; Q does not create oxygen, force uptake into a neuron or clamp ATP. Normal exchange is not driven above its reference profile. Single-candidate and joint correction form a bounded OR family, not a search for whatever intervention improves the outcome.

R2 requires a corresponding local availability deficit, below the matched nondiabetic fifth percentile for ≥90 cumulative pre-loss days in at least the same 10% person-weighted neuronal fraction. Q must remove **at least half** of that group's cumulative below-threshold neuronal exposure time over equal observation windows. This is a measured vascular contribution to local supply, not an assumption that the oxygen level is injurious. Normal vascular exchange cannot by definition correct arterial hypoxemia or defective mitochondrial utilization.

All percentiles, durations, burden cutoffs, groups and materiality thresholds specify the proposition; they are not estimated biological constants. This assessment concerns persistent supply deficits. Very brief recurrent ischemia requires a different exposure definition.

## Biological reasoning

Net oxygen uptake equals blood flow multiplied by the arterial–venous oxygen-content difference. More flow can coexist with less extraction, but lower uptake can also reflect lower demand. Illustratively, flow 1 with extraction fraction 0.4 gives uptake 0.4; flow 1.5 with extraction 0.2 gives 0.3. Neither this accounting example nor a low venous saturation identifies diabetic ischemia.

Vessel-wall changes could impede diffusion, redirect transit or reduce reserve. They can also coexist with adequate supply. Denervation itself changes vasomotor regulation and demand. Skin oxygen, epineurial laser-Doppler flow, capillary hemoglobin saturation and oxygen near an axon's mitochondria describe different quantities and compartments.

Low delivery could impair ATP-dependent maintenance, alter redox chemistry or invoke hypoxic signaling. ATP depletion is an optional mediator, not a mandatory discount. HIF induction can be compensatory. Correction permits downstream responses, trophic feedback, replacement and activity to change; the outcome is their net structural balance.

## Probabilities of the necessary claims

Every row after R1 assumes all predecessors and retains the same qualifying groups, species and connected compartments. The rows are conditional judgments, not independent marginal probabilities. Their conjunction states the complete operational proposition.

| Claim | Exact additional requirement | Skeptical / best / favorable | Best cumulative | Evidence and uncertainty |
|---|---|---:|---:|---|
| R1 | At least one specified baseline group has the defined pre-loss microvascular exchange-capacity deficit. | 45 / **75** / 95% | 75.0% | Human microangiopathy can precede diagnosed neuropathy, but functional capacity and pre-structural timing are incompletely measured. |
| R2 | In an R1 alternative, the matching availability deficit meets the burden boundary and Q removes ≥50% of that burden. | 25 / **60** / 90% | 45.0% | Human oxygen/transit findings support this possibility. Discordant early flow, extraction reserve and uncertain microregional exposure prevent equating pathology with inadequate supply. |
| R3 | In at least one retained R2 alternative, Q yields B ≥0.05. | 40 / **75** / 95% | 33.8% | Severe supply deficits can injure nerve, and diabetic animal rescue is plausible. Physiological correction, human subtype transfer and structural causation are less established than functional changes. |
| R4 | At least one retained R3 alternative yields B ≥0.20. | 30 / **65** / 90% | 21.9% | Human structural angiogenic results are limited or negative; treatments also have direct trophic actions and mostly concern established disease. |
| R5 | At least one retained R4 group has C ≥0.10. | 35 / **70** / 90% | 15.4% | Microvascular abnormalities are common, but prevalence of this connected, correctable pre-loss deficit among incident cases is unknown. |

R2 can hold with normal ATP and no structural benefit; then R3 fails. B=0.12 passes R3 and fails R4. B=0.30 in a group covering 4% of reference cases passes R4 and fails R5. Merging R1 and R2 gives 45% without changing the final probability.

## Evidence bearing on the judgment

### Human timing, structure and actual oxygen

[Malik et al. (1992)](https://doi.org/10.1136/jnnp.55.7.557) compared 15 people with mild diabetic neuropathy with eight controls. Endoneurial capillary density was lower and basement-membrane/endothelial profiles larger, but luminal size and closed-vessel frequency did not differ. [Giannini and Dyck (1995)](https://doi.org/10.1002/ana.410370412) found wall/debris abnormalities even without diagnosed polyneuropathy; their preceding analysis had not found different mean luminal area or vessel density. These primary abstracts establish pathology, not pre-loss delivery failure. The latter does not supply participant counts.

[Malik et al. (2005)](https://doi.org/10.1007/s00125-004-1663-5) clinically followed 12 people with minimal neuropathy for about 8.7 years after baseline biopsy. Vascular abnormalities accompanied early unmyelinated-fiber changes despite relatively normal myelinated-fiber measures. This is not biopsy before all nerve injury or serial demonstration that falling oxygen preceded terminal loss.

[Thrainsdottir et al. (2003)](https://doi.org/10.2337/diabetes.52.10.2615) related baseline sural morphology to six-year glucose tolerance. Highlighted progression contrasts contain only four versus four people, or five versus ten when glucose-tolerance transitions are combined. Smaller lumens preceded worsening glucose tolerance; greater basement-membrane area accompanied neuropathy. Progression to diabetes is not incident molecularly profiled DPN, and vessel findings do not directly measure supply. Full numerical tables were not recovered for these historical reports.

[Newrick et al. (1986)](https://doi.org/10.1136/bmj.293.6554.1053) directly measured endoneurial oxygen in **11 people with chronic DPN**: mean 39.7 mmHg, SD 10.2. Ten had nerve oxygen below dorsal-foot venous oxygen, compared with none of four non-neuropathic comparators. The accessible primary abstract does not give the comparator mean. This is meaningful local evidence, but 39.7 mmHg is not itself a demonstrated terminal-injury threshold.

[Ibrahim et al. (1999)](https://doi.org/10.1007/s001250051222) studied **10 with mild/moderate DPN, nine diabetic participants without DPN and nine healthy volunteers**. Reported DPN-versus-healthy endoneurial hemoglobin saturation was 67.1 versus 76.7%, with fluorescein rise times 48.5 versus 14 seconds. Three hundred microvascular measurements do not make n=300. Lower hemoglobin saturation may involve extraction as well as delivery; it is not interstitial neuronal oxygen. The accessible primary abstract does not resolve the full diabetic-without-DPN comparison or structural mediation.

[Zochodne and Ho (1997)](https://doi.org/10.1093/brain/120.7.1131) found no simple flow deficit in **26 people with mild DPN versus 17 with other neuropathies**. Flow did not track fiber density; ten had contralateral repeat measurements after one year without lower flow despite a downward fiber-density trend. Other neuropathies are not healthy controls, laser-Doppler flow is not neuronal oxygen, and a small null is not equivalence. This nevertheless weakens a universal low-flow account.

A [2024 thigh-skin study](https://doi.org/10.3389/fpain.2024.1485420) found increased vWF staining in painful DPN alongside lower epidermal innervation. It measured a vascular marker, not oxygen or flow. The report states 41 participants, with ten missing vascular immunohistochemistry, and inconsistently assigns six versus seven participants to painless/no-DPN groups across text and tables. Exact paired subgroup denominators cannot be reconstructed reliably. A nerve-to-vWF ratio can fall because its numerator falls or denominator rises; it does not isolate failed delivery. Clinical and biopsy compartments also differ.

### Public data and its sampling limit

The public [EDDIE dataset](https://doi.org/10.5258/SOTON/D1581) contains **92 independent people with T2D**. Foot transcutaneous oxygen and toe–brachial index jointly explain about **0.04%** of log(1+sural amplitude) variance. With age, height, sex, duration, HbA1c, temperature, smoking and centre adjustment, the oxygen coefficient is **−0.071 per SD, HC3 95% interval −0.259 to +0.117**. This gives no positive linear oxygen–amplitude relationship in this cross-section.

However, oxygen has median **67.2 mmHg**, range **20.2–147**: only **one person is below 30 and three below 40**. The null scarcely samples very low oxygen. It cannot reject a nonlinear low-tail mechanism, historical/transient hypoxia, or a peptidergic structural effect. Eleven amplitudes are coded zero and retained using log(1+x); this check does not reinterpret them as missing. These participants do not establish coverage among early-T2 incident cases.

### Interventions: rescue and attribution

[Low et al. (1984)](https://doi.org/10.1073/pnas.81.21.6894) reported at least one-third lower nerve blood flow after four months of STZ diabetes in rats. Oxygen measurements below 25 mmHg were more frequent, and four weeks of oxygen supplementation improved conduction and resistance to ischemic block without changing plasma glucose. Local polyol metabolites also changed. The primary abstract does not supply animal-level structural counts; oxygen sites are not independent animals. Functional normalization is not a measured terminal-prevention fraction.

[Schratzberger et al. (2001)](https://doi.org/10.1172/JCI12188) reported vascular/functional rescue with VEGF gene transfer in diabetic rats and rabbits. But [Murakami et al. (2006)](https://doi.org/10.1002/jgm.893) found behavioral rescue without increased endoneurial vessel counts, including contralateral effects. Counts need not track flow, but VEGF is not a selective exchange correction. These primary abstracts do not give a subtype-specific terminal rescue contrast.

The human [Ropper et al. (2009)](https://doi.org/10.1002/ana.21675) trial randomized **39 to VEGF plasmid and 11 to placebo** for established DPN. Symptom change was −1.2 versus −0.9; the reported adjusted difference was −1.40, p=0.01, using post-treatment untreated-leg change as a prespecified covariate. That is the reported analysis, but the contralateral limb is not guaranteed to escape treatment effects. Overall improvement was 12/39 versus 2/11: an unadjusted difference of **12.6 percentage points**, Newcombe 95% interval **−19.4 to +33.0**. This does not replace the trial's continuous adjusted analysis. Nerve conduction potentials did not improve; terminal structure and local exchange engagement were not measured.

The **170-person SB-509 phase IIb** trial did not meet its 180-day conduction, NIS-LL or IENFD endpoints according to the [sponsor's primary results announcement](https://www.fiercebiotech.com/biotech/sangamo-biosciences-announces-phase-2b-trial-of-sb-509-diabetic-neuropathy-did-not-meet-key). A selected earlier sensory-score subgroup result was not a sustained structural positive. This is a sponsor report, not a recovered full peer-reviewed dataset; no numerical IENFD effect or verified selective exchange correction is available. It constrains broad VEGF-induction treatment without settling Q.

For HGF, the [2015 trial](https://doi.org/10.1002/acn3.186) randomized 104 people and treated 103; its principal efficacy set excluded protocol violations and included 84, rather than 84 independently randomized participants. IENFD assessment was described as inconclusive. The later [500-person VM202 analysis](https://doi.org/10.1111/cts.12977) was negative on its main endpoints; the positive 101-person extension was a subset, not independent replication. Neither trial identifies oxygen-mediated terminal preservation; HGF has direct neuronal actions.

In [2025 NGF-R100W/VEGFA mRNA experiments](https://doi.org/10.2337/db24-0989), local treatment of established db/db neuropathy increased PGP9.5 density, particularly with the combination; structural panels report n=6 per group and flow n=10. Numerical density coordinates were not extracted from the accessible full text. NGF monotherapy also helped; the combination cannot establish a solely vascular route or quantified interaction. A [2026 focused-ultrasound report](https://doi.org/10.1109/TUSON.2026.3700212) reports rat skin-perfusion and nerve-density benefit, with n=8 for perfusion and n=5 for density in its primary abstract. Direct neural, thermal and mechanical effects were not separated in the accessible evidence.

HIF activity is not an oxygen meter: sensory-neuron HIF1α deletion worsened later diabetic injury in [Rojas et al. (2018)](https://doi.org/10.1007/s00109-018-1707-9). This supports a protective response under those conditions without demonstrating the human vascular defect. DCE-MRI permeability and fingertip oxygen-saturation associations likewise do not directly establish exchange at the target terminal.

## Ideal experiments that would resolve each claim

These are conceptual truth tests with ideal selective exchange and longitudinal human terminal measurements, not proposed practical studies. People are the inference unit; neurons, vessels and time points estimate within-person quantities. Resolve the finite alternatives with simultaneous precision. An interval crossing a boundary, failed specificity or incomplete engagement leaves a claim unresolved.

**R1 — preceding exchange-capacity deficit; no predecessors.** Enroll a representative eligible early-T2 cohort and matched nondiabetic references. Assign the specified groups at baseline and follow mapped terminal–axon–soma pathways through year one. Quantify candidate transfer functions at matched arterial/tissue boundary conditions, separately from actual availability; track structural integrity independently of marker expression. Keep measurement challenges brief enough not to create the exposure and verify that measurement itself does not reduce flow. YES requires at least one group/candidate to meet the fifth-percentile, 90-day and 10% person-weighted neuronal-fraction boundaries before loss. NO requires all alternatives to fail at least one boundary. Post-loss changes alone are NO. Uncertain transfer estimation or insufficient observation before disappearance is unresolved. This establishes exposure in diabetes, not an additional claim that glucose caused it.

**R2 — vascular contribution to low availability; assume R1.** Within R1 groups, randomize Q alternatives versus matched sham before anticipated loss. Measure interstitial oxygen partial pressure or free fuel concentration at the same connected sites over equal year-one windows. Keep arterial oxygen content, hemoglobin, circulating fuels, pressure, medication, activity and external temperature comparable; allow local consumption, extraction and downstream metabolism to respond. Verify the prescribed transfer function and absence of independent neuronal receptor stimulation, endothelial trophic manipulation or changes to noncandidate transport; endogenous downstream feedback may respond. An independently implemented correction should give the same exposure effect. YES requires the defined local deficit in reference care and Q removing ≥50% of its below-threshold neuronal exposure time in a retained group. NO requires no retained alternative to meet both boundaries despite faithful correction. More bulk flow without restored availability is NO. Changes caused by different arterial content or incomplete Q are unresolved. This does not yet estimate structural benefit.

**R3 — small structural causal effect; assume R1–R2.** Sustain retained Q or sham for five years from pre-loss baseline in randomized independent people. Establish specificity as in R2; keep other initiating exposures comparable while allowing energy state, redox, HIF, firing, endogenous trophic responses and repair to change. Track actual terminal disappearances and successful replacements with stable subtype identity. Estimate person-averaged L and B, with absolute counts alongside the fraction. YES requires B ≥0.05 in a retained alternative with positive reference loss; NO requires all retained alternatives to have B <0.05 under faithful Q. Pain, function or staining changes alone are NO. Imprecision across 0.05, uncertain identity or negligible reference loss is unresolved. Restoration of the original exchange deficit, if used as an independent causal check, must reinstate that exposure rather than impose an unrelated ischemic insult.

**R4 — material magnitude; assume R1–R3.** Use the same five-year randomized human contrasts, matched upstream inputs, verified Q and permitted downstream responses. Resolve B for retained alternatives. YES requires at least one B ≥0.20; NO requires every retained B below 0.20 despite an established ≥0.05 benefit. A precise B=0.12 is NO here. An interval spanning 0.20 is unresolved. This is a magnitude question, not a second injury mechanism.

**R5 — case coverage; assume R1–R4.** Follow a representative reference-care cohort from the same population for five-year incident DPN under the stated clinical definition, retaining baseline group membership. Randomized arms establish benefit but do not define the reference-case denominator. Estimate C separately for every retained beneficial group, without adding overlapping groups or selecting responders. YES requires at least one C ≥0.10; NO requires all retained C below 0.10. An interval crossing 0.10 or incomplete case ascertainment is unresolved. No vascular intervention is needed in this descriptive denominator cohort; ordinary disease and care may evolve naturally.

## Uncertainty and sensitivity

The conditional vectors are **[45,25,40,30,35]%**, **[75,60,75,65,70]%** and **[95,90,95,90,90]%**, giving **0.47%, 15.36% and 65.79%**. The skeptical scenario combines largely adequate early exchange with weak structural transfer and restricted coverage. The favorable scenario supposes common microregional deficits missed by available measurements and transferable rescue once delivery is restored. These are coordinated interpretations, not independent draws from invented distributions.

An intermediate mainly adequate-exchange interpretation gives **3.1%**, using [65,35,55,45,55]%; an early microregional-deficit interpretation gives **44.2%**, using [90,85,85,80,85]%. Current sources do not justify probability masses or a variance for these scenarios. The unresolved human exposure-to-structure connection dominates the range.

Hypothetically raising R2 from 60% to 85% with other conditional judgments fixed raises the joint price to **21.8%**; lowering it to 25% gives **6.4%**. Longitudinal demonstration that selective exchange correction removes pre-loss local deficits would be more discriminating than another endothelial-marker association. These are hypothetical updates, not trial-derived estimates.

Changing the 90-day rule, neuronal-fraction boundary, fuel catalog or subgroup family changes the proposition and needs a new elicitation. Relaxing B from 20% to 5% removes R4 as a separate magnitude requirement; it does not retain an automatic discount. Present data cannot numerically calibrate these alternate definitions. A skin oxygen cutoff cannot be calibrated from the near-zero EDDIE regression because the low tail is scarcely sampled.

## Overlap and implication

[PKC](16_pkc_vascular_signaling.md), [RAGE](04_rage_signaling.md) and [complement](22_complement_microvascular_injury.md) can be upstream routes. [Energy insufficiency](07_bioenergetic_insufficiency.md) is one downstream consequence; [NGF](14_ngf_trophic_support.md), [Schwann support](17_schwann_metabolic_support.md) and [skin repair](32_skin_repair_environment.md) can interact with supply. Shared angiogenic experiments are not independent support for each entry; these probabilities must not be added.

The broader claim that severe ischemia can injure peripheral nerve is strong. The priced claim is narrower: a common, persistent microvascular exchange deficit causes preventable peptidergic terminal loss before clinical DPN in early T2 diabetes. Human work makes that plausible but does not resolve it.