# Adversarial review — mitochondrial-cascade (ROUND 2)

Authoring model: Sol 6.1. Reviewer: Claude Opus 5.5 (Anthropic). Date: 2026-10-14.
Level before: Γ / Moderate / reinforcing / stabilizing factor. Level after: unchanged.
Findings: 1 Major, 6 Minor, 1 Editorial. PMIDs verified 70/70. Claims spot-checked: 10.
Validator: FAIL on `v5.lay_summary` only (pre-existing package gap, also absent in alpha-synuclein and amyloid-cascade); 31 PASS, 2 WARN.

## ADVERSARIAL REVIEW AND APPROVAL


**Authoring model:** Sol 6.1. **Adversarial reviewer:** Claude Opus 5.5 (Anthropic). **Date:** 2026-10-14.

**Scope of checks.** All 70 reference PMIDs verified against PubMed records (70/70 resolve to the stated title, first author, journal and year; no retractions or expressions of concern). Ten load-bearing claims spot-checked against source abstracts or original bodies (PMID 23486940, 32227247, 27913695, 25668262, 28522181, 34732887, 26182419, 37246815, 35235774, 30768179), plus the UDCA randomisation and subset counts against the original Results. Scoring logic re-derived against `hsef-hypothesis-review-v5/SKILL.md` Step 3 and `references/hsef-framework.md` (caps, scope-matched aggregation, confidence band, momentum artifact, role derivation). No level or confidence change.

| ID | Severity | Location | Original → corrected | Evidence |
|---|---|---|---|---|
| M-1 | Major | Consensus provenance reconciliation (9 October 2026) | "The earlier adversarial approval refers to its original edition" → "No earlier adversarial approval exists for this case" | No adversarial approval section, banner or approval line exists in any edition of this case, including the pre-revision copy in `hsef_v53_popularity_ranks_26_30_2026-10-07`. The sentence asserted an approval that was never issued. Introduced by the 9 October revision — a regression relative to the earlier edition. |
| M-2 | Minor | REFERENCES [31] | ", et al." (author dropped) → "HD iPSC Consortium." | PubMed CollectiveName for PMID 30768179 is "HD iPSC Consortium". Reference-builder collective-author defect. |
| M-3 | Minor | REFERENCES [58] | ", , et al." → "Writing Group, Edaravone (MCI-186) ALS 19 Study Group." | PubMed CollectiveName fields for PMID 28522181 (two collective authors, no individual authors). |
| M-4 | Minor | REFERENCES [64] | ", Beal MF, Oakes D et al." → "Parkinson Study Group QE3 Investigators, Beal MF, Oakes D et al." | PubMed CollectiveName for PMID 24664227. |
| M-5 | Minor | REFERENCES [65] | ", Kieburtz K, Tilley BC et al." → "Writing Group for the NINDS Exploratory Trials in Parkinson Disease (NET-PD) Investigators, Kieburtz K, Tilley BC et al." | PubMed CollectiveName for PMID 25668262. |
| M-6 | Minor | REFERENCES [66] | ", et al." → "Parkinson Study Group STEADY-PD III Investigators." | PubMed CollectiveName for PMID 32227247 (sole author). |
| M-7 | Minor | `final_score.confidence_rationale` (scorecard, 3 occurrences) | "density, heterogeneous controls and incomplete access set the confidence" → rationale restated in terms of evidence density and reproducibility, with access depth explicitly excluded | Framework: "Confidence reflects evidence density and reproducibility." Lead policy on confidence bands (2026-10-14): access depth is never a valid determinant. Re-derived band after striking it: the review documents substantive reproducibility concerns (complex-I subtype heterogeneity, cybrid nuclear-background and preparation differences, unreconciled CypD survival discrepancies), so **Moderate stands**. |
| M-8 | Editorial | REFERENCES [44]; evidence card [60] | "DOI: ." (empty field) removed; "ErratumIn: PMIDnot supplied." → "ErratumIn: PMID not supplied." | PubMed holds no DOI for PMID 10939564 (the journal DOI is a legacy Wiley identifier not indexed by NLM), so the field is dropped rather than left empty. The ErratumIn entry for PMID 24067398 carries a RefSource but no PMID; the glued string was a renderer defect. |

**Rejected findings (checked, no change).**
1. **Headline level Γ and scope-matched derivation.** Re-derived: the central claim is broadly scoped (etiological apex across sporadic AD, PD, HD, ALS); SC5 is the only scope-matched subclaim and is Γ. SC2 Ζ is explicitly scope-restricted to prespecified genetic lesions and toxins and therefore cannot set the headline. Alternatives are reported as required (weakest-link Β, primary-subclaim Γ). No change.
2. **"Best-supported rival" renderer artifact.** Checked `models_evaluated[0]` / rank order: rank 1 is "Multi-hit convergent loop", which is not this review's nominal model (the nominal "Mitochondrial-first cascade" ranks 4). The prose is correct as written.
3. **Momentum "Mechanistic only".** Correct value and a naming artifact is present (2025 rat complex-IV editing, 2026 neuronal MCU loss and encapsulated-mitochondrial transfer). Per lead policy, completed publications could not have supported "Decisive work underway" in any case; the review does not claim it.
4. **REVIEWER NOTES versus the spliced Consensus section.** The notes state two Consensus Pro queries and zero API/Deep calls, which is consistent with the provenance section and with the METHODS statement; there is no "no Consensus work was performed" claim to reconcile. No change.
5. **Citation years.** Reference [16] prints 2023 against a PubMed e-pub year of 2022; within the ±1 print/e-pub convention. Rejected.
6. **PubMed title glueing.** PubMed renders the title of PMID 31316206 as "Pink1mice"; the review's "Pink1-/- mice" is the correct form. Rejected per convention.
7. **UDCA participant count.** The abstract says 30 participants; the review says 31 randomised with 25 contributing 31P-MRS and a smaller pandemic-limited gait subset. Verified against the original Results (PMC10527073): 31 randomised (20 UDCA, 11 placebo), 25 with paired 31P-MRS, 18 in the gait subset. The review is right and the abstract imprecise. No change.

**Residual limitations.** Six of 70 sources had partial-body examination; the remainder rest on abstract and linked-status metadata, so claims about figure-level detail in the other 64 could not be independently confirmed. The erratum content for PMID 34732887, 40069360 and 41856111 remains unexamined, as the review states. This pass is a source, claim and scoring-logic audit, not a blinded recalibration.

**Approval:** Following adversarial review and the corrections listed above, the revised document is approved. — Claude Opus 5.5 (Anthropic), 2026-10-14.