{
  "id": "34",
  "title": "Microbiome-derived exposures",
  "summary": "Abnormal gut-derived metabolite or LPS output may alter local nerve exposure and contribute to terminal loss.",
  "source_title": "Altered gut-derived chemical exposure contributes to terminal loss",
  "page": "34_microbiome_metabolites.html",
  "source_markdown": "sources/34_microbiome_metabolites.md",
  "structured_data": "data/34_microbiome_metabolites.json",
  "snapshot_date": "2026-09-14",
  "source_review_date": "2026-09-14",
  "scope_label": "Early type 2 diabetes",
  "scope_exception": false,
  "source_headline_markdown": "**Best judgment: about 8%; uncertainty range (sensitivity): 0.1–59%.** Gut interventions can alter neuropathy symptoms and animal nerve structure. The strongest human trial does not measure terminal preservation, and its mouse structural contrast weakens when diabetes without neuropathy supplies the donor comparator. The probabilities below are literature-informed subjective judgments, not measured treatment success rates. Reviewed 14 September 2026. [Probability method](probability_method.md) · [Collection index](README.md).",
  "probability_percent": {
    "skeptical": 0.118125,
    "base": 7.98525,
    "favorable": 58.86675
  },
  "probability_meaning": "Subjective belief in the complete causal proposition. Scenario endpoints are sensitivity products, not confidence limits.",
  "causal_proposition_markdown": "In adults with type 2 diabetes diagnosed within five years and no clinical DPN at entry, an abnormal, specified flow of microbial products from gut to host precedes loss of the connected human peptidergic C-fiber terminals. Restoring that flow to a nondiabetic reference profile 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**.\n\nThis is a bounded chemical-output proposition. It covers excess products and deficient protective products. It does not require a particular bacterial taxon, loss of diversity, or a permanently engrafted transplanted community. Altered microbial production and altered passage across the gut barrier are alternative sources of the exposure. A vagal signaling mechanism without the specified chemical exposure, or a benefit from an unlisted metabolite, does not establish this exact claim.\n\n**Output panel.** The gut-to-host species are **acetate, propionate, butyrate, indole-3-propionate (IPA), indole, p-cresol, trimethylamine (TMA), and lipid-A-bearing bacterial lipopolysaccharides (LPS)**. The LPS profile retains lipid A acylation/phosphorylation identities and corresponding biological activity; different LPS structures are not assigned equal toxicity. Source tracing distinguishes microbial production from food, host synthesis and recycling.\n\n**R1's exposure boundary.** At least 10% of people in a qualifying group have a panel member's microbe-derived net outward flux across the intestinal host interface outside its matched nondiabetic 2.5th–97.5th percentile range and at least 50% from its positive reference median, on at least 30 cumulative days in the first year. For a zero median, twice the validated detection limit replaces the positive-median boundary. Flux is amount entering host tissue per day, not stool concentration or taxonomic abundance. The deviation precedes the connected arbor's first ≥5% net physical deficit lasting 30 days.\n\n**Q — reference gut-output substitution.** For five years, an ideal controller intercepts the listed microbe-derived output at the intestinal host interface and substitutes the joint time-varying profile of a nondiabetic reference matched for age, sex, BMI and habitual diet. Assignment of reference profiles is independent of outcomes and preserves their covariance. Both excess and deficient output are corrected; useful products are not removed indiscriminately.\n\nQ does not transplant an entire microbiome, change unlisted compounds directly, damage the gut barrier, or directly replace nerves, vessels or glia. Diet, medication and other initiating exposures are comparable at assignment. Downstream host metabolism, renal handling, inflammation, gut feedback and repair can change; their benefits and costs count. Thus a downstream glucose improvement is allowed, but a separate direct glucose-lowering treatment is not part of Q.\n\n**R2's host-exposure boundary.** In the same R1-exposed people, at least 10% of mapped connected neuronal units must have a local extracellular panel deviation for ≥24 cumulative hours before their sustained deficit during year one. The local panel is the output species above, their host products **indoxyl sulfate, p-cresyl sulfate and TMAO**, and bioactive **TNF or IL-6** as possible host inflammatory relays. Free/bioavailable concentration, including extracellular particle-bound LPS activity, must lie outside the matched nondiabetic reference interval and at least 50% from its positive median; the zero-median rule applies. Compartments are distal skin interfaces, axon/Remak interfaces and DRG neuron/satellite-glial interfaces.\n\nQ must reduce at least 50% of the cumulative magnitude of that local deviation over the same observation period. This distinguishes an effective chemical connection from a gut-only change. Direct metabolite delivery and a host inflammatory relay are alternatives, not two mandatory paths. This engagement requirement does not claim that every structural benefit is mediated exclusively by the particular assayed local member.\n\n**Groups and endpoints.** Candidate groups are the whole eligible population, males, females, baseline BMI ≥30 kg/m², baseline HbA1c ≥7.5%, and people satisfying the R1 flux criterion during a 30-day baseline observation. No additional intersections or future responder definitions are permitted.\n\nIn fixed distal-leg territories, follow stable peptidergic identity and physical anatomy through five years. Net loss equals cumulative disappearance minus newly established terminal entries, counting any later disappearance of replacements. Let L₀ and L_Q be the equally person-weighted mean losses under contemporary diabetes care and Q, normalized to baseline terminal amount. With L₀ > 0, **B = (L₀ − L_Q)/L₀**. Let **C = P(group | incident DPN under reference care)**. Incident DPN means new bilateral length-dependent nerve signs with objective small-fiber loss or abnormal conduction, excluding another cause.\n\nEach successive claim asks whether at least one candidate group meets every requirement so far. A group that supplies exposure cannot be replaced by a different, exposure-negative group supplying benefit. A complete NO must exclude all candidates satisfying the preceding requirements.",
  "claims": [
    {
      "id": "R1",
      "source_id": "R1",
      "claim": "The specified gut-output deviation occurs before terminal loss.",
      "claim_markdown": "The specified gut-output deviation occurs before terminal loss.",
      "conditional_percent": {
        "skeptical": 30.0,
        "base": 65.0,
        "favorable": 90.0
      },
      "cumulative_base_percent": 65.0,
      "source_cumulative_display": null,
      "reason_markdown": "Diabetic microbial changes are plausible, but early absolute flux, host-source separation and timing are unmeasured.",
      "question": "Does an abnormal flow of microbial short-chain fatty acids, indoles, p-cresol, trimethylamine, or bacterial LPS from gut to host precede terminal loss?"
    },
    {
      "id": "R2",
      "source_id": "R2",
      "claim": "The compatible local host exposure exists and Q corrects ≥50% of it.",
      "claim_markdown": "The compatible local host exposure exists and Q corrects ≥50% of it.",
      "conditional_percent": {
        "skeptical": 35.0,
        "base": 70.0,
        "favorable": 95.0
      },
      "cumulative_base_percent": 45.5,
      "source_cumulative_display": null,
      "reason_markdown": "Transfer/IPA studies support communication; clearance, compartment dose and inflammatory relays limit human delivery inference.",
      "question": "Does restoring the assessed gut-derived metabolite and LPS output to normal correct at least half of the related chemical or inflammatory exposure around the nerve?"
    },
    {
      "id": "R3",
      "source_id": "R3",
      "claim": "Q produces B ≥0.05 in at least one preceding-qualifying group.",
      "claim_markdown": "Q produces B ≥0.05 in at least one preceding-qualifying group.",
      "conditional_percent": {
        "skeptical": 15.0,
        "base": 45.0,
        "favorable": 85.0
      },
      "cumulative_base_percent": 20.475,
      "source_cumulative_display": null,
      "reason_markdown": "Human clinical benefit and non-diabetic repair positives versus weak DPN-specific mouse comparison and absent human structure.",
      "question": "Would restoring the assessed gut-derived metabolite and LPS output to normal avert at least 5% of five-year net terminal loss?"
    },
    {
      "id": "R4",
      "source_id": "R4",
      "claim": "At least one such group has B ≥0.20.",
      "claim_markdown": "At least one such group has B ≥0.20.",
      "conditional_percent": {
        "skeptical": 30.0,
        "base": 60.0,
        "favorable": 90.0
      },
      "cumulative_base_percent": 12.285,
      "source_cumulative_display": null,
      "reason_markdown": "Given some structural contribution, whether correcting this panel supplies a large fraction despite parallel injury and compensation.",
      "question": "Would restoring that gut-derived chemical output to normal avert at least 20% of five-year net terminal loss?"
    },
    {
      "id": "R5",
      "source_id": "R5",
      "claim": "At least one group satisfying R1–R4 has C ≥0.10.",
      "claim_markdown": "At least one group satisfying R1–R4 has C ≥0.10.",
      "conditional_percent": {
        "skeptical": 25.0,
        "base": 65.0,
        "favorable": 90.0
      },
      "cumulative_base_percent": 7.98525,
      "source_cumulative_display": null,
      "reason_markdown": "Broad metabolic exposure is possible; large benefits could be confined to uncommon output states.",
      "question": "Does the benefiting subgroup account for at least 10% of incident DPN cases under reference care?"
    }
  ],
  "sections": [
    {
      "heading": "The causal claim and its scope",
      "markdown": "In adults with type 2 diabetes diagnosed within five years and no clinical DPN at entry, an abnormal, specified flow of microbial products from gut to host precedes loss of the connected human peptidergic C-fiber terminals. Restoring that flow to a nondiabetic reference profile 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**.\n\nThis is a bounded chemical-output proposition. It covers excess products and deficient protective products. It does not require a particular bacterial taxon, loss of diversity, or a permanently engrafted transplanted community. Altered microbial production and altered passage across the gut barrier are alternative sources of the exposure. A vagal signaling mechanism without the specified chemical exposure, or a benefit from an unlisted metabolite, does not establish this exact claim.\n\n**Output panel.** The gut-to-host species are **acetate, propionate, butyrate, indole-3-propionate (IPA), indole, p-cresol, trimethylamine (TMA), and lipid-A-bearing bacterial lipopolysaccharides (LPS)**. The LPS profile retains lipid A acylation/phosphorylation identities and corresponding biological activity; different LPS structures are not assigned equal toxicity. Source tracing distinguishes microbial production from food, host synthesis and recycling.\n\n**R1's exposure boundary.** At least 10% of people in a qualifying group have a panel member's microbe-derived net outward flux across the intestinal host interface outside its matched nondiabetic 2.5th–97.5th percentile range and at least 50% from its positive reference median, on at least 30 cumulative days in the first year. For a zero median, twice the validated detection limit replaces the positive-median boundary. Flux is amount entering host tissue per day, not stool concentration or taxonomic abundance. The deviation precedes the connected arbor's first ≥5% net physical deficit lasting 30 days.\n\n**Q — reference gut-output substitution.** For five years, an ideal controller intercepts the listed microbe-derived output at the intestinal host interface and substitutes the joint time-varying profile of a nondiabetic reference matched for age, sex, BMI and habitual diet. Assignment of reference profiles is independent of outcomes and preserves their covariance. Both excess and deficient output are corrected; useful products are not removed indiscriminately.\n\nQ does not transplant an entire microbiome, change unlisted compounds directly, damage the gut barrier, or directly replace nerves, vessels or glia. Diet, medication and other initiating exposures are comparable at assignment. Downstream host metabolism, renal handling, inflammation, gut feedback and repair can change; their benefits and costs count. Thus a downstream glucose improvement is allowed, but a separate direct glucose-lowering treatment is not part of Q.\n\n**R2's host-exposure boundary.** In the same R1-exposed people, at least 10% of mapped connected neuronal units must have a local extracellular panel deviation for ≥24 cumulative hours before their sustained deficit during year one. The local panel is the output species above, their host products **indoxyl sulfate, p-cresyl sulfate and TMAO**, and bioactive **TNF or IL-6** as possible host inflammatory relays. Free/bioavailable concentration, including extracellular particle-bound LPS activity, must lie outside the matched nondiabetic reference interval and at least 50% from its positive median; the zero-median rule applies. Compartments are distal skin interfaces, axon/Remak interfaces and DRG neuron/satellite-glial interfaces.\n\nQ must reduce at least 50% of the cumulative magnitude of that local deviation over the same observation period. This distinguishes an effective chemical connection from a gut-only change. Direct metabolite delivery and a host inflammatory relay are alternatives, not two mandatory paths. This engagement requirement does not claim that every structural benefit is mediated exclusively by the particular assayed local member.\n\n**Groups and endpoints.** Candidate groups are the whole eligible population, males, females, baseline BMI ≥30 kg/m², baseline HbA1c ≥7.5%, and people satisfying the R1 flux criterion during a 30-day baseline observation. No additional intersections or future responder definitions are permitted.\n\nIn fixed distal-leg territories, follow stable peptidergic identity and physical anatomy through five years. Net loss equals cumulative disappearance minus newly established terminal entries, counting any later disappearance of replacements. Let L₀ and L_Q be the equally person-weighted mean losses under contemporary diabetes care and Q, normalized to baseline terminal amount. With L₀ > 0, **B = (L₀ − L_Q)/L₀**. Let **C = P(group | incident DPN under reference care)**. Incident DPN means new bilateral length-dependent nerve signs with objective small-fiber loss or abnormal conduction, excluding another cause.\n\nEach successive claim asks whether at least one candidate group meets every requirement so far. A group that supplies exposure cannot be replaced by a different, exposure-negative group supplying benefit. A complete NO must exclude all candidates satisfying the preceding requirements."
    },
    {
      "heading": "Reasoning from the biology",
      "markdown": "A microbial species count is several steps from a neuronal dose. Substrate availability changes production per organism. Other microbes consume products; the intestinal barrier, liver and kidney then determine host exposure. For a simple steady-state illustration,\n\n`circulating concentration = (gut input + other input) / clearance`.\n\nEqual plasma concentration can accompany different production and clearance. Conversely, high concentration can reflect retention with unchanged microbial production. Stool abundance is a residual after production, consumption and absorption, not an estimate of portal input. Total protein-bound toxin concentration also differs from the free fraction reaching a nerve.\n\nA local action need not involve direct entry into the neuron. LPS-responsive host cells could relay cytokines; IPA could alter immune recruitment; short-chain fatty acids can change host metabolism and signaling. These paths may increase or decrease regeneration independently of sensory firing. Restoring a reference mixture can remove a compensatory product, so Q is not defined as “remove whichever microbial effects are harmful.”\n\nThe target is net physical preservation. Increased dermal profiles, decreased withdrawal latency, better nerve conduction and a lower clinical score are different observations. Regrowth after nerve crush creates a large demand for repair that early intact diabetes may not create. In the other direction, established denervation or autonomic gut dysfunction can change diet, transit and the microbiome. Randomized intervention can break some of that ambiguity; a microbiome-response association within its treated arm cannot inherit the randomization.\n\nA reference five-year loss of 30% would require six percentage points of retained terminal preservation to meet B ≥0.20. Neither a 20% reduction in a symptom score nor a 20% rise in a metabolite establishes that amount. Those numerical boundaries define the hypothesis; they are not measured biological thresholds."
    },
    {
      "heading": "Evidence that moves the judgment",
      "markdown": "**The human FMT trial supports clinical benefit in established disease.** Yang randomized **37 participants, 25 to FMT and 12 to placebo**; **22 and 10** completed 84-day follow-up. The primary outcome was TCSS, a composite of symptoms, reflexes and sensory examination. Public paired observations give mean improvement **3.09 versus 0.60 points**, difference **2.49 points (approximate 95% interval 0.76–4.22)**. Adjusting day-84 score for baseline preserves an FMT advantage of **2.19 points (0.55–3.83)**.\n\nThis is a meaningful favorable anchor for an established-neuropathy clinical effect. It is not a structural trial. The completers include **16/6 type 2/type 1 participants in FMT and 8/2 in placebo**; diabetes duration averages approximately nine and seven years. All glucose-lowering drugs except insulin and conventional DSPN treatments were withdrawn during a 14-day run-in. FMT consisted of two 150-mL mid-gut infusions on consecutive days. That treatment context differs from prevention during ordinary contemporary care. [Yang et al., 2023](https://doi.org/10.1016/j.cmet.2023.06.010); [raw-data calculations](../../../findings/damage_calibration_microbiome_quantitative.md).\n\nCompleter analysis excludes five randomized people. Assigning zero change to all missing outcomes gives a 2.22-point advantage. Assigning the scale's most adverse possible missing changes gives bounds **−3.23 to +7.67 points**; those are identification bounds, not plausible imputations or confidence limits. Equal and opposite missing changes of about 7.74 points would erase the observed assigned-arm advantage. The data support benefit, while leaving attrition sensitivity unresolved.\n\n**Objective functional support is uneven and does not demonstrate new terminals.** In the same source workbook, the sural sensory-velocity change is **+5.25 m/s (−1.12 to +11.62)** versus placebo, with **20/9** evaluable participants. Ulnar sensory velocity is **+3.76 m/s (0.45–7.07)** with **21/9**. The remaining five conduction contrasts are smaller or imprecise. An exploratory correction across 16 extracted clinical, conduction and circulating-marker comparisons retains only the anxiety-score contrast below 0.05; this does not retroactively replace TCSS as the paper's named primary outcome.\n\nFive-Hz current-perception testing is functional, not a count of surviving C fibers. The trial measured no skin IENFD. LBP and IL-6 changes support a host-response effect, but **LBP is a host binding protein, not measured circulating LPS**, and neither identifies local neural exposure. Fecal short-chain fatty acids and bile acids did not show significant treatment changes; their concentrations do not measure production or absorbed flux.\n\n**The mouse transfer is structural, but its comparator and donor unit matter.** The same study pooled stool from **five people per donor group** before transfer to antibiotic-treated db/db mice. The first experiment's epidermal outcome has **seven mice per arm**, not the larger behavioral cohort n. Mean IENFD was **24.71 profiles/mm with nondiabetic donor material, 16.57 with diabetes-without-DSPN material, and 14.14 with DSPN material**.\n\nThe DSPN–nondiabetic contrast is **−10.57 (−18.76 to −2.38)**. The more specific DSPN–diabetes-without-DSPN contrast is **−2.43 (−8.06 to +3.20)**. The latter does not establish equivalence, but it does not show a clear extra structural effect of the neuropathy donor state. A second donor set yielded **28.5 versus 16.0 profiles/mm**, difference **−12.5 (−22.4 to −2.6)**, with four mice per arm; it again compared DSPN with nondiabetic donors.\n\nThese intervals describe mice receiving the particular pooled inocula. They cannot establish between-human-donor reproducibility. The source's broader human donor cohorts also differ in diabetes duration: roughly one year without DSPN versus nine years with DSPN. The two transfer experiments support a gut-material effect under a diabetic recipient state, but do not isolate early DPN-specific microbial chemistry, a single panel member or pre-loss timing. [Primary report and full supplement](https://ars.els-cdn.com/content/image/1-s2.0-S1550413123002206-mmc8.pdf); [calculations](../../../findings/damage_calibration_microbiome_quantitative.md).\n\n**Another FMT experiment separates dermal from epidermal benefit.** Bonomo treated Western-diet obese mice after 12 weeks, using antibiotics, an initial five-day diet switch and two weeks of lean-donor FMT versus corresponding controls. FMT increased dermal nerve density, but the reported epidermal comparison was not significant; the text and caption disagree on its P value (0.7618 versus 0.07618). Structural n is reported as 6–10 per arm, preventing a precise interval without additional values. Antibiotics alone reduced skin nerve density in obese mice. The human component comprised 29 people, only four with diabetes; it concerned distal pain and circulating short-chain fatty acids.\n\nThis provides an animal structural effect and an explicit boundary on the target terminal endpoint. It does not justify claiming FMT restored epidermal fibers or that butyrate uniquely caused the structural result. [Bonomo et al., 2020](https://doi.org/10.1073/pnas.2006065117).\n\n**IPA supplies a more specific repair mechanism, with dose and compartment limits.** Serger's nerve-crush experiments connect microbial IPA production, immune recruitment and regeneration. Reanalysis of the released skin observations gives **5.45 versus 15.65 epidermal profiles/mm** at day 16, four mice per arm. Whole-animal profile analysis supports reduced fasting benefit under vancomycin; separate PXR and CXCR2 perturbations constrain the IPA mechanism but do not all survive a combined exploratory multiplicity correction. The skin outcome and behavioral endpoints do not share a released animal key. [Serger et al., 2022](https://doi.org/10.1038/s41586-022-04884-x); [animal-level reanalysis](../../../model_effort/integrative_model_v1/review/new_ipa_repair_findings.md).\n\nA subsequent study used oral IPA **20 mg/kg/day for three days after rat sciatic crush**, reporting longer GAP43-positive axons and increased Schwann-cell proliferation. Its cultured RSC96 optimum was **500 µM IPA**; the stated 12 rats per treatment group does not identify every panel's independent n. Systemic PI3K inhibition is not Schwann-specific mediation. [2025 study](https://doi.org/10.1016/j.neurot.2025.e00578).\n\nIn adult mouse DRG cultures, **100–1000 µM IPA reduced outgrowth even after neuronal AhR deletion**; the released measurements came from only two mice per genotype without animal-linked fields. Thus a positive IPA effect cannot simply be labeled neuronal AhR activation, and a Schwann-cell optimum does not establish a safe or effective neuronal exposure. [Halawani et al., 2026](https://doi.org/10.1038/s41586-026-10295-z); [corrected-source analysis](../../../model_effort/integrative_model_v1/review/new_ahr_ipa_findings.md).\n\nDirect diabetic-rat IPA treatment, **10 or 20 mg/kg for two weeks starting six weeks after STZ**, reports improved pain and stress/mitochondrial measures. The accessible primary abstract does not give animal n, complete cell doses, glycemia under treatment or a terminal-count result. It supports a diabetic benefit lead without supplying a structural rescued fraction. [Gundu et al., online 2022](https://doi.org/10.1080/13813455.2021.2024577).\n\n**Public within-diabetes plasma data limit simple circulating-marker stories.** Among **48 DPN and 49 diabetes-without-DPN donors** in ST001411, indoxyl sulfate had a DPN/reference ratio **0.902 (0.724–1.123)**; a seven-toxin family gave permutation P = 0.907. A separate seven-indole/kynurenine analysis found no strong discriminator in this cohort or an 88-person obese-neuropathy comparison. A fixed seven-metabolite model transferred with **AUROC 0.525 and 0.574** in the two directions, with broad donor-bootstrap intervals spanning 0.5.\n\nIndolepropionate's estimate was especially sensitive to imputation: **0.951 (0.598–1.513)** in processed data versus **1.128 (0.795–1.601)** among detected observations. These are established-disease relative plasma measurements, not local free concentrations, microbial flux or incident DPN. Individual renal function and other covariates are unavailable; absence of an unadjusted signal is not proof that adjustment could never reveal one. [Toxin analysis](../../../findings/s30_s31_d33_d44_lipid_toxin_gate.md); [original-data and transfer correction](../../../model_effort/integrative_model_v1/review/new_dpn_indole_public_findings.md).\n\nProspective sampling provides a different constraint. In **1,947 DPPOS participants**, baseline quinolinic acid predicted later monofilament-defined neuropathy in the lifestyle arm, **OR 1.64 (1.24–2.19)**. Participants entered with prediabetes, and neuropathy was assessed as prevalent status after about 15 years. Quinolinic acid is a host kynurenine-pathway metabolite; this result does not identify microbial origin, skin terminals or incidence after an established diabetes-without-DPN baseline. [Perng et al., 2025](https://doi.org/10.1007/s00125-025-06571-8).\n\n**Neither more symptom trials nor taxon associations close the mechanism.** FADIGAS randomized 20 people with longstanding type 1 diabetes and severe gastrointestinal symptoms; FMT improved gastrointestinal scores at four weeks, without measuring peripheral terminal repair. A 2025 db/db study's microbial/metabolite “mediation” was derived from cross-sectional associations, not randomized mediator manipulation. Older microbiome MR studies also used population controls rather than diabetes-conditioned controls, allowing effects on diabetes itself to enter the complication association. [FADIGAS](https://doi.org/10.1016/j.eclinm.2024.103000); [2025 mouse study](https://doi.org/10.3389/fcimb.2025.1599954); [endpoint audit](../../../findings/idea11_endpoint_family_adversarial.md)."
    },
    {
      "heading": "Probabilities of the claims",
      "markdown": "The rows are conditional on the whole preceding prefix, not independent discounts. Injury, mitochondrial effects, immune recruitment and altered replacement remain alternative mediators.\n\n| Claim | Exact additional requirement | Best conditional probability | Uncertainty range (sensitivity) | Main uncertainty |\n|---|---|---:|---:|---|\n| R1 | The specified gut-output deviation occurs before terminal loss. | 65% | 30–90% | Diabetic microbial changes are plausible, but early absolute flux, host-source separation and timing are unmeasured. |\n| R2 | The compatible local host exposure exists and Q corrects ≥50% of it. | 70% | 35–95% | Transfer/IPA studies support communication; clearance, compartment dose and inflammatory relays limit human delivery inference. |\n| R3 | Q produces B ≥0.05 in at least one preceding-qualifying group. | 45% | 15–85% | Human clinical benefit and non-diabetic repair positives versus weak DPN-specific mouse comparison and absent human structure. |\n| R4 | At least one such group has B ≥0.20. | 60% | 30–90% | Given some structural contribution, whether correcting this panel supplies a large fraction despite parallel injury and compensation. |\n| R5 | At least one group satisfying R1–R4 has C ≥0.10. | 65% | 25–90% | Broad metabolic exposure is possible; large benefits could be confined to uncommon output states. |\n\nThe preferred product is **0.65 × 0.70 × 0.45 × 0.60 × 0.65 = 0.0798525**, reported as about 8%. Independently considering the whole proposition suggests a credible but low-probability prevention claim: randomized clinical benefit and specific repair evidence support causal potential, while human terminal preservation and the common early exposure remain unshown. That supports a single-digit-to-low-teens neighborhood without identifying an exact percentage. Clinical FMT success is not a resolved prediction of Q's structural outcome."
    },
    {
      "heading": "Ideal experiments that would resolve each claim",
      "markdown": "With unlimited capabilities, map microbial source flux and the connected neural units without measurement injury from the baseline observation through five years. Randomize people to verified Q or reference care, retaining their original group definitions and complete outcomes. Independent n is people. Repeated matched nondiabetic measurements supply reference profiles. These are ideal truth conditions, not a practical experimental proposal.\n\n**R1 — source, amount and timing.** Trace microbe-produced molecules across the intestinal interface, measuring net outward flow rather than stool residuals. Follow the same people's arbors. YES: at least one candidate group meets the ≥10%-of-people, magnitude, duration and pre-deficit boundary. NO: no candidate group does. A taxonomic association alone cannot supply YES.\n\n**R2 — host delivery or relay and engagement.** Among R1-qualified groups, measure the defined local bioavailable exposures and their change under Q over equal pre-loss windows. YES: at least one group meets both local-exposure and ≥50% correction boundaries. NO: all R1-qualified groups fail at least one. Gut correction without a measurable compatible host exposure fails this exact claim; unknown clearance or unmeasured delivery leaves it unresolved.\n\n**R3 — nontrivial structural contribution.** Compare true mean L₀ and L_Q in groups satisfying R1–R2, accounting for all destruction, replacement and indirect costs. YES: at least one has positive L₀ and B ≥0.05. NO: all have B <0.05 or no positive reference loss. Improved pain, conduction or autonomic symptoms without physical preservation cannot satisfy R3.\n\n**R4 — material contribution.** In the same comparison, restrict attention to groups satisfying R1–R3. YES: at least one has B ≥0.20. NO: every remaining group has B <0.20. The percentage is relative to reference net terminal loss, not baseline symptom score or total microbial output.\n\n**R5 — case coverage.** In the complete reference-care cohort, measure incident DPN and count membership of the original baseline groups. YES: at least one R1–R4 group contains ≥10% of cases. NO: every such group contains <10%. Selecting only people with favorable post-treatment microbiomes cannot satisfy coverage.\n\nFailure to implement Q, incomplete follow-up or uncertain physical identity is not biological NO."
    },
    {
      "heading": "Uncertainty and evidence updates",
      "markdown": "These are complete interpretations of the same bounded proposition, without assigned scenario weights.\n\n| Interpretation | R1–R5 conditional percentages | Whole probability |\n|---|---|---:|\n| Preferred | 65, 70, 45, 60, 65 | 7.99% |\n| Lower: weak early output deviation; host adaptation; effects mainly symptomatic | 30, 35, 15, 30, 25 | 0.12% |\n| Upper: persistent early exposure; effective delivery; substantial structural effects in common groups | 90, 95, 85, 90, 90 | 58.87% |\n| Strong chemical effect in an uncommon subgroup | 70, 90, 80, 85, 25 | 10.71% |\n| Common output disturbance, mostly small structural contributions | 85, 85, 55, 30, 85 | 10.13% |\n\nThe **0.1–59% sensitivity range**, about 59 percentage points wide, describes uncertainty across defensible interpretations, not a confidence interval or an empirical distribution. The upper interpretation is permitted by positive transfer/repair experiments and the lack of a decisive human structural test. The lower interpretation is permitted by donor pooling, comparator differences, missing local dose measurements and symptom–structure separation. Shared study limitations move several rows together.\n\nSampling intervals above concern particular measured effects. They do not resolve validity, dose transport, human subtype, source attribution or subgroup coverage. More repeated bacterial sequences cannot shrink donor uncertainty. No prospective set of comparable resolved prevention predictions supplies calibration weights here.\n\n**Hypothetical upward update:** source-resolved early human exposure plus selective material terminal preservation across a common baseline group could justify [85,90,80,80,80]%, giving **39.2%**. **Hypothetical downward update:** verified host exposure correction with little structural benefit, despite improved symptoms, could justify [60,65,15,35,50]%, giving **1.0%**.\n\nChanging only R3 from 45% to 15% or 80% gives **2.7% or 14.2%**. Even certainty about that step alone gives **17.7%** under frozen other judgments; this is not expected research value. Actual data can change several conditionals simultaneously.\n\nChanging the panel, accepting any vagal route, or lowering the 20% preservation boundary changes the proposition. The available data do not estimate how those choices change early-human coverage. Microbial outputs can interact; splitting this joint controller into separate metabolite rows must not automatically multiply additional marginal discounts."
    },
    {
      "heading": "Overlap and boundary",
      "markdown": "Microbial products can act through [inflammation](19_tnf_tlr4_inflammation.md), [energy support](07_bioenergetic_insufficiency.md) or [repair restriction](44_intrinsic_growth_restriction.md). [Uremic exposure](38_kidney_failure_uremic_exposure.md) shares some retained molecules but concerns clearance, which Q does not directly normalize. The [ER-stress assessment](23_er_stress.md) includes pharmacological TMAO protection, underscoring that metabolite name alone does not fix sign across doses. Different source boundaries, interventions and timing prevent a logical probability ordering among these complete claims. Their probabilities are not independent disease fractions."
    }
  ],
  "source_claim_table_markdown": "| Claim | Exact additional requirement | Best conditional probability | Uncertainty range (sensitivity) | Main uncertainty |\n|---|---|---:|---:|---|\n| R1 | The specified gut-output deviation occurs before terminal loss. | 65% | 30–90% | Diabetic microbial changes are plausible, but early absolute flux, host-source separation and timing are unmeasured. |\n| R2 | The compatible local host exposure exists and Q corrects ≥50% of it. | 70% | 35–95% | Transfer/IPA studies support communication; clearance, compartment dose and inflammatory relays limit human delivery inference. |\n| R3 | Q produces B ≥0.05 in at least one preceding-qualifying group. | 45% | 15–85% | Human clinical benefit and non-diabetic repair positives versus weak DPN-specific mouse comparison and absent human structure. |\n| R4 | At least one such group has B ≥0.20. | 60% | 30–90% | Given some structural contribution, whether correcting this panel supplies a large fraction despite parallel injury and compensation. |\n| R5 | At least one group satisfying R1–R4 has C ≥0.10. | 65% | 25–90% | Broad metabolic exposure is possible; large benefits could be confined to uncommon output states. |"
}