{
  "slug": "weight-loss-mash-reversal",
  "file": "weight-loss-mash-reversal.review.md (shipped review; second independent adversarial pass)",
  "authoring_model": "Sol 6.1",
  "adversarial_reviewer": "Claude Opus 5.5 (Anthropic)",
  "date": "2026-11-02",
  "level_before": "Δ",
  "level_after": "Δ",
  "confidence_before": "Moderate",
  "confidence_after": "Moderate",
  "pmids_checked": 60,
  "pmids_failed": 0,
  "claims_spot_checked": 23,
  "corrections_applied": false,
  "findings": [
    {
      "id": "WL-1",
      "severity": "Major",
      "location": "Evidence and causal interpretation para. 3; SC2 rationale and field-valid line; inventory/card [46]; falsification; steelman; Stability; KNOWLEDGE GAPS; Research gaps; REVIEWER NOTES; Consensus section; 2024 timeline; scorecard final_score.confidence_rationale",
      "original": "Results state nine treatment resolutions but only eight ballooning-zero participants, an internally inconsistent exact count that is left unresolved. We retain the reported qualitative clinical positive with Moderate confidence and seek clarification before using a precise resolution fraction.",
      "corrected": "Results report resolution of steatohepatitis in 9 treatment and 0 control participants (final sample 16 and 8; the Discussion gives 56%). Eight treated participants reduced ballooning to 0; this is compatible with nine resolutions because inclusion required NAS of at least 4 rather than ballooning of at least 1. The counts are reconcilable, not inconsistent. Component reductions are reported within the treatment arm, no between-group test for resolution or components is printed, and controls entered with higher MRS liver fat (24% versus 15%), which the authors say may have contributed.",
      "evidence": "PMC12007730 (doi:10.1016/j.jhep.2024.06.017): inclusion 'NAFLD activity score (NAS) >=4/8'; 'Resolution of steatohepatitis was achieved in 9 treatment subjects and 0 control subjects ... 8 reduced hepatocellular ballooning to a 0'; Discussion '56% of participants ... achieved MASH improvement with treatment (0% of controls)'; Limitations on baseline IHTG imbalance. Overturns the 2026-10-14 pass's rejected finding that called the inconsistency real. Level unchanged (9/16 vs 0/8; Fisher two-sided p about 0.01, reviewer arithmetic). Confidence unchanged."
    },
    {
      "id": "WL-2",
      "severity": "Major",
      "location": "SC2 rationale; decisive-perturbation inventory; tally; M1 tournament entry",
      "original": "(absent) Inventory lists only the two positive randomized paired-biopsy lifestyle trials ([1], [46]).",
      "corrected": "Add Eckard 2013 (PMID 23814606): 56 randomized to standard care or three diet/exercise arms for 6 months with repeat biopsy; 41 completers (88% NASH) improved as a whole (p<0.001) with no between-arm difference including standard care (p=0.31). Add to SC2 rationale that the assigned-package effect is not yet consistently reproduced. Screened discriminating contrasts 33 -> 34.",
      "evidence": "PubMed PMID 23814606, doi:10.1177/1756283X13484078, abstract verified. Same design class and time window as the trials holding SC2; result contrary. Level unchanged (Δ needs one valid positive); supports Moderate."
    },
    {
      "id": "WL-3",
      "severity": "Minor",
      "location": "Tally sentence; scorecard causal_tally_audit",
      "original": "Tally: 1 positive qualifying study unit across scoped subclaims; 33 screened discriminating contrast units.",
      "corrected": "Tally: 2 positive qualifying study units across scoped subclaims ([1] for SC1 biopsy steatosis; [46] for SC2 inflammation/ballooning resolution); 33 screened discriminating contrast units (34 with WL-2). Alternatively keep 1 and label it 'for the primary SC2'.",
      "evidence": "SC1 crosses Δ on PMID19827166 (boundary_check_BGD). Promrat full text PMC2799538: steatosis -1.1 vs -0.3, p=0.02."
    },
    {
      "id": "WL-4",
      "severity": "Minor",
      "location": "COMPETING MECHANISTIC MODELS; scorecard competing_models",
      "original": "Four models, all broadly compatible with the claim; no null/non-specific-effects rival.",
      "corrected": "Add model: 'Non-specific trial effects, regression to the mean and biopsy sampling variability explain small-trial histologic change (Γ).' It does not explain the randomized between-arm differences in [1] and [46], but it does explain the null contrast in Eckard 2013.",
      "evidence": "Promrat full text: control-arm histologic improvement and 'regression to the mean' discussed; Mucinski: both arms reduced liver fat; Eckard 2013: all arms improved."
    },
    {
      "id": "WL-5",
      "severity": "Minor",
      "location": "Cap audit field-valid definition lines SC1-SC4; upward ladder SC1/SC2 Δ lines",
      "original": "SC2 Δ: “Requires at least one intervention changing lifespan or healthspan in a way consistent with the hypothesis.” Satisfied: yes.",
      "corrected": "SC2 Δ: “Requires at least one intervention changing lifespan or healthspan in a way consistent with the hypothesis.” Literal lifespan/healthspan wording does not apply to this clinical claim; satisfied under the declared clinical adaptation (assigned treatment changing paired biopsy pathology): yes.",
      "evidence": "Rubric 06 aging text quoted verbatim, including 'Must separate developmental effects from true aging effects'. The adaptation is declared, but the printed 'Satisfied: yes' misleads lay readers. Same change for SC1."
    },
    {
      "id": "WL-6",
      "severity": "Minor",
      "location": "reader-summary.json limitations",
      "original": "60 primary reports with 3 selected partial-body examinations; density, heterogeneous controls and incomplete access set the confidence, not correctness.",
      "corrected": "Moderate confidence rests on a single small randomized trial of current MASH resolution that has not been replicated, a second randomized trial with mixed component results, and heterogeneous packages and comparators; it bounds the evidence base, not correctness.",
      "evidence": "WM-2 struck access depth from the scorecard rationale; the lay-facing summary was not updated. The same string appears in at least 30 other slugs' reader summaries (batch propagation item)."
    },
    {
      "id": "WL-7",
      "severity": "Minor",
      "location": "Evidence cards [33], [38], [50] Access/status",
      "original": "ErratumIn: PMID41880634; ErratumIn: PMID32738952; ErratumIn: PMID34147150; ErratumIn: PMID31386114",
      "corrected": "Erratum In: PMID41880634; Erratum In: PMID32738952; Erratum In: PMID34147150; Erratum In: PMID31386114",
      "evidence": "Known renderer space-dropping defect."
    },
    {
      "id": "WL-8",
      "severity": "Editorial",
      "location": "PROPOSED DECISIVE EXPERIMENTS Readout/Controls/Supporting/Falsifying line",
      "original": "separate endpoints.. Controls: / multiplicity addressed.. Supporting: / their separate claims.. Falsifying: / achieved-loss responders..",
      "corrected": "Single full stops.",
      "evidence": "Doubled punctuation from field concatenation."
    }
  ],
  "rejected_findings": [
    "SC2 Δ rather than Γ: randomized package changed paired biopsy resolution (9/16 vs 0/8); no cap or literal Γ hard rule binds measured pathology.",
    "SC1 Δ: Promrat between-group steatosis p=0.02 and responder 13/18 vs 3/10 p=0.03 confirmed in full text; 28/31 paired biopsies as stated.",
    "Ε not reached: correct; WL-2 strengthens the rejection.",
    "Primary-subclaim aggregation applied as declared; weakest-link Γ and scope-matched Δ correctly labelled as sensitivities.",
    "Best-supported rival label correct after WM-1 (M2, rank 1, is_nominal false).",
    "Momentum 'Mechanistic only' correct; ESSENCE 240-week phase is drug evidence.",
    "Drug/surgery/cohort cards (ESSENCE, F4 semaglutide, BRAVES, SYNERGY-NASH, Lassailly, LEAN, NATIVE, pioglitazone, Koutouratsas, Takahashi, Dallio) match abstracts.",
    "Orlistat pilot [15] 'uncontrolled' is correct; PubMed's RCT publication-type tag is an indexing error.",
    "Harrison 2009 orlistat RCT (PMID 19053049) omission not a defect: randomizes orlistat on shared diet; adds only achieved-loss association.",
    "Role of mechanism 'reinforcing / stabilizing factor' consistent with step 5 and explicitly declines endogenous primacy.",
    "Decisive-perturbation (12) and discriminating-contrast (33) counts match scorecard evidence_inventory."
  ],
  "scorecard_changes_proposed": [
    "final_score.confidence_rationale: remove nine/eight discrepancy; add randomized null (WL-2)",
    "causal_tally_audit: tally 1 -> 2; screened_discriminating_contrasts 33 -> 34",
    "competing_models: add non-specific-effects null model (WL-4)",
    "evidence_inventory: add PMID 23814606"
  ],
  "scorecard_changed": false,
  "validator": "Not re-run (no file edited); last recorded PASS - 0 fail, 0 warn, 31 pass (validation.txt)",
  "related_files": {
    "prior_pass": "In-document ADVERSARIAL REVIEW AND APPROVAL dated 2026-10-14 (WM-1..WM-5); its rejected finding on the nine/eight count is overturned by WL-1.",
    "provenance_note": "Review banner says Sol 6.1; scorecard provenance_header.generated_by says 'Codex; HSEF v5.3'. Unresolved; carried forward."
  },
  "residual_limitations": [
    "Full text read for 2 of 60 records (19827166, 38914313); others at abstract depth.",
    "Mucinski figures/supplement not available; between-group component statistics neither confirmed nor excluded.",
    "Omitted-literature search targeted, not systematic."
  ]
}