# Adversarial review log — saturated-fat-linoleic-coronary

- **Authoring model:** Sol 6.1 (HSEF v5.3 pipeline)
- **Adversarial reviewer:** Claude Opus 5.5 (Anthropic)
- **Date:** 2026-10-14 (round 2)
- **Canonical file reviewed:** `hsef_v53_traffic_ranks_11_15_2026-10-13/saturated-fat-linoleic-coronary/saturated-fat-linoleic-coronary.review.md`
- **Level before → after:** Γ (Moderate, epiphenomenon) → Γ (Moderate, epiphenomenon) — unchanged
- **Corrections applied:** 5 (0 Critical, 0 Major, 1 Minor, 4 Editorial)
- **PMIDs verified:** 60/60 title + first author + journal + year; 60/60 DOI strings; 0 failures; 0 retractions or expressions of concern
- **Load-bearing claims spot-checked:** 15 (all supported)
- **Status:** Approved as revised

## Findings

| ID | Severity | Location | Original → corrected | Evidence |
|---|---|---|---|---|
| A1 | Editorial | Throughout (144 lines) | Lost inter-word spaces restored, e.g. "LDLcholesterol" → "LDL cholesterol", "was384days" → "was 384 days", "The195-person16-week" → "The 195-person 16-week", "UKBiobank" → "UK Biobank" | Known upstream builder defect (round-2 defect list item 4). Whitespace-only repair: with whitespace stripped, the revised file is character-identical to the Sol 6.1 original apart from A4. |
| A2 | Editorial | Evidence cards [2], [5], [43] | "ErratumIn: PMIDnot supplied." → "ErratumIn: PMID not supplied." | Same builder defect. The erratum identifiers are genuinely absent from retrieved metadata, so the substance is retained. |
| A3 | Editorial | REFERENCES 9, 10, 45, 48 | "DOI: ." → "DOI: not indexed in PubMed." | PubMed holds no DOI for PMID 1484945, 1431574, 16483027, 5900208. |
| A4 | Editorial | Evidence cards [10], [23] | "lipoproteina" → "lipoprotein[a]" | Source titles read "serum lipoprotein[a] levels" (PMID 1431574) and "serum lipoprotein(a) concentration" (PMID 38714425). |

| A5 | Minor | scorecard `final_score.confidence_rationale` | Access depth removed as a stated basis for the band | The framework ties confidence to evidence density and reproducibility, not to the reviewer's access. Re-derived on legitimate grounds — 60 reports, but the exact clinical question rests on a handful of conflicting allocated trials and mutually disagreeing cohort and genetic estimates — the band remains **Moderate**; only its basis is corrected. Lead policy of 2026-10-14 (two-step test) applied: access depth struck, band re-derived. Step 2 keeps **Moderate** — the accessible allocated trials conflict outright (Minnesota null, Sydney adverse, the Finnish hospital periods favourable) and the cohort and genetic estimates disagree with each other — which are legitimate determinants; the High band's "extensive independent replication across labs, models, and assay modalities" is not met, and access depth was not the only thing holding the band down. |

## Rejected findings (checked, no change)

1. Aggregation and headline letter — `preregistration.json` declares `rule: primary-subclaim`, `primary: SC2`; SC2's scope matches the central claim, so scope-matched and primary-subclaim coincide at Γ, and SC4's Α is correctly reported as a narrower subclaim sensitivity.
2. Β/Γ boundary — controlled isoenergetic feeding trials give partial experimental system support (above Β); no intervention moves the registered clinical endpoint (below Δ).
3. Causal role `epiphenomenon` — correctly derived at framework Step 2 and scoped to the proposed benefit account.
4. Momentum "Mechanistic only" — correct; named artifacts are mechanistic/genetic and no pending allocated oil-swap event trial exists.
5. Collective-author rendering bug — not present (no consortium first authors).
6. Finnish Mental Hospital Study downgraded from the PubMed "randomized controlled trial" tag to a hospital-period natural experiment — defensible reading of the source abstract, and the tension is stated in the review.
7. PubMed title for reference 36 drops the italicised gene name — standing rejected-finding convention.
8. Print/issue vs e-pub year — no reference differs by more than ±1.
9. **Renderer defect (nominal printed as its own "Best-supported rival") — checked, absent.** `models_evaluated[0]` is rank 1 with `is_nominal` false; the nominal is rank 4.
10. **Spliced-provenance artifacts — checked, absent.** This package carries no consensus-consistency block and no "earlier adversarial approval" clause.

## Validator

`hsef-hypothesis-review-v5/tools/validate_review.py` on the revised scorecard + prose: **32 pass, 0 warn, 1 fail**. The single fail is `v5.lay_summary`, which the validator marks "mandatory from v5.4" and does not apply to this v5.3 package. Output saved to `validator.txt`; the pre-review `validation-report.json` is preserved in `sol61-original/`.

## Files changed

- `saturated-fat-linoleic-coronary.review.md` (edited; original in `sol61-original/`)
- `saturated-fat-linoleic-coronary.scorecard.json` — `final_score.confidence_rationale` only (surgical text edit, formatting preserved). Level, confidence band, role, momentum and counts unchanged.

## Propagation

Finished canonical copied byte-identically to:
- `hsef_v53_topical_all_22_2026-10-13/cycles/hsef_v53_traffic_ranks_11_15_2026-10-13_share/saturated-fat-linoleic-coronary/saturated-fat-linoleic-coronary.review.md`
- `hsef_v53_traffic_ranks_11_15_2026-10-13_share/saturated-fat-linoleic-coronary/saturated-fat-linoleic-coronary.review.md`

No divergent copies recorded for this slug.