# Adversarial review — long-covid-viral-persistence (ROUND 2)

Authoring model: **Sol 6.1**. Adversarial reviewer: **Claude Opus 5.5 (Anthropic)**. Date: **2026-10-14**.
Canonical file reviewed: `consensus_consistency_audit_2026-10-09/revised/long-covid-viral-persistence/long-covid-viral-persistence.review.md`
(the revised edition; reviewed as it now stands).

## Summary

| Item | Result |
|---|---|
| Level before → after | Γ → Γ (no change) |
| Confidence before → after | Moderate → Moderate (no change) |
| Causal role | epiphenomenon (unchanged) |
| Findings | Critical 0 · Major 0 · Minor 3 · Editorial 0 |
| REFERENCE PMIDs verified | 56/56 resolve; 0 failures; 0 retraction/EoC flags; no collective-author, leading-comma or empty-DOI defects |
| Load-bearing claims spot-checked | 10 |
| Validator | package-stored report (v2.3) ok=True; re-run with the bundled v5.4 copy gives 1 FAIL (`v5.lay_summary`, v5.4-only) and 1 WARN (`v5.claim_form` disjunction, rejected by convention), 32 pass — `validator.txt` |
| Regression vs 2026-10-09 edition | one, Minor: see finding 1 |

This was the cleanest of the five reviews in my group on citation integrity: every reference resolved and
none of the four known upstream defect classes fired in the reference list.

## Findings

| ID | Severity | Location | Original → corrected |
|---|---|---|---|
| 1 | Minor (regression) | REVIEWER NOTES | "…Consensus… was performed." → the claim scoped to the original review plus a clause identifying the 9 October 2026 Consensus consistency validation as a later provenance supplement |
| 2 | Minor | Momentum line | "(PMIDs38848477,40188838,42673984)" → "(PMIDs 38848477, 40188838, 42673984)" |
| 3 | Minor | Momentum line | "(PMIDs41794369,41963730)" → "(PMIDs 41794369, 41963730)" |

Finding 1 is the only regression and is the same defect the metformin review in this group carries: the
revision added a Consensus consistency validation section while the preserved REVIEWER NOTES still asserted
that no Consensus work had been done. The fix scopes the original sentence rather than deleting it.

## Verification of the block the revision added

Every quantitative claim is correct against the primary records: the tissue cohort's 225 patients, residual
clinical specimens, 4-month telephone questionnaire and association (odds ratio 5.17, 95% CI 2.64–10.13),
together with the authors' own call for work to verify a mechanistic link; STOP-PASC's 155 participants,
15-week blinded placebo-controlled design, pooled six-symptom primary endpoint at week 10 and null result in
a nearly fully vaccinated cohort; PAX LC's phase 2 design, 100 enrolled participants and day-28 PROMIS-29
physical-health difference of −0.55 (95% CI −2.32 to 1.21, p = 0.54); and RECOVER-VITAL's 959 mITT
participants across three phenotypes with no significant primary benefit under either regimen. The block's
refusal to convert those nulls into proof of reservoir absence is the framework-faithful reading.

## Rejected findings (checked, no change)

1. Aggregation: primary-subclaim with SC3 named; headline = SC3 Γ. Alternatives correct (weakest-link
   min(Β, Γ, Γ) = Β; scope-matched Γ, since SC3 is the only subclaim scoped to defined Long COVID).
2. Causal role "epiphenomenon" correctly derived at step 2.
3. Momentum "Mechanistic only" carries its artifact and correctly declines "Decisive work underway".
4. Causal tally: 7 claimed, 7 bracketed items printed, score card agrees.
5. `v5.claim_form` disjunction warning — the "antigen or RNA" wording is two measurements of one parent
   reservoir claim, decomposed in the subclaim matrix; rejected per the registered convention.
6. `v5.lay_summary` FAIL is a v5.4-only check on a v5.3 review.
7. `review_metadata.reviewer` reads "Codex" in the score card. The same string appears in every score card
   in this corpus, so it is house provenance rather than a defect in this review; flagged for the lead
   rather than changed.

## Copies

- `propagate_to`: none.
- **Divergent copy left untouched:** `hsef_v53_remaining_priority_01_05_2026-10-09/long-covid-viral-persistence/long-covid-viral-persistence.review.md`
  (the pre-audit edition). It carries findings 2 and 3 but **not** finding 1, which exists only because of
  the revision.

## Approval

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