# Adversarial review log — bpc-157-tissue-repair (revised edition)

Authoring model: **Sol 6.1**. Adversarial reviewer: **Claude Opus 5.5 (Anthropic)**. Date: **2026-10-14**.
Round 2; group 7 of `NEW_REVIEW_INVENTORY_2026-10-14.json`. Canonical copy is the 9 October 2026 revised edition.

## Summary

| Item | Value |
|---|---|
| Level before | Γ — Mechanistically plausible, partial system support / Low |
| Level after | Γ — Mechanistically plausible, partial system support / Low (unchanged) |
| Findings: Critical | 0 |
| Findings: Major | 1 |
| Findings: Minor | 2 |
| Findings: Editorial | 0 |
| Rejected findings | 7 |
| PMIDs checked / failed | 18 / 0 |
| Load-bearing claims spot-checked | 12 (every human report and every quantitative statement) |
| Validator | **PASS** — 0 fail, 0 warn, 34 pass, before and after (this is a genuine v5.4 package) |
| Approval | **Approved as revised.** |

## Findings

| ID | Severity | Location | Original → corrected | Basis |
|---|---|---|---|---|
| AR-1 | Major | Consensus provenance reconciliation block; `consensus-addendum.md` | "The earlier adversarial approval refers to its original edition" → "No earlier adversarial approval is attached to this case; the 2026-10-06 adversarial pass did not cover it" | False provenance inserted unconditionally by the 9 October revision script. Verified: 56 review files carry a 2026-10-06 approval banner and this case is not among them, and no earlier edition carries an approval section. |
| AR-2 | Minor | LIMITATIONS OF THIS REVIEW, second bullet | "(claim C19, unverified)" → statement that human PK remains unresolved and that non-scoring claim C19 was retired from the scored register on 2026-10-09 | Revision-introduced inconsistency: the claim-register repair retired C19, the register now runs C1–C18, and the provenance header reports 0 unverified. Fixed by scoping, not deleting; the substantive limitation is preserved. |
| AR-3 | Minor | Claim-register repair note | "The original approval predates this repair." → "No earlier adversarial approval is attached to this case." | Same false-approval artifact in a second location. |

## Rejected findings (checked, no change)

1. **Level and aggregation.** Weakest-link pre-registered and justified by the conjunctive claim form; all four human gates at Γ; SC5 correctly scored at rat scope and marked non-gating under the quoted cap-precedence rule; all three rules return Γ. Δ rejected (no controlled human perturbation with a structural endpoint); Β rejected (controlled rat perturbations from outside the originating programme). SC5 at Δ not Ε is correct — the boundary is untested, not mapped.
2. **The "epiphenomenon" role.** Reads harshly but is the framework's step-2 label at the claimed scope, glossed in both AT A GLANCE and the lay summary, with "reinforcing / stabilizing factor" recorded separately at rat scope.
3. **Renderer bug (nominal printed as best-supported rival).** Unaffected: `models_evaluated[0]` is M2 (`is_nominal: false`, rank 1); the nominal M1 ranks third and the prose names M2.
4. **Back-dated identifier.** The provenance block's eleven matched identifiers include PMID 42542926, described elsewhere as newly added. Checked the 5 October enrichment artifacts: all eleven, including 42542926, are present. Accurate.
5. **Confidence band.** Low is correct, derived by quoting the framework's Low text and naming the single connected programme — the only case in this group whose band rests on density and replication rather than access depth.
6. **Reviewer-notes contradiction.** Absent: no preserved sentence makes any claim about Consensus use, so the spliced block contradicts nothing.
7. **Under-reporting of a positive.** The bladder pilot reports complete resolution in 10/12 and 5/5 GRA in all 12; the review gives only design and absence of adverse events. Not an error — the review's point is that an uncontrolled, local, symptomatic study of a different organ cannot bear on the registered claim, and model M3 says so explicitly.

## Benefit-and-risk audit (treatment-relevant review)

Nothing needed correcting. Each human report is described at its real design and endpoint: 2 participants given 10 then 20 mg IV with vitals and blood chemistry and no plasma concentrations, no injury or comparator (40131143); a retrospective chart review reaching 16 of 17 patients for recalled pain, 11/12 and 3/4 reporting significant improvement, no control, no validated instrument, no post-treatment imaging (34324435); 12 pentosan-refractory women given peri-lesional injections at a single cystoscopy, GRA endpoint, no adverse events (39325560). Rat PK is exact — IV half-life under 30 minutes, IM bioavailability ~14–19% in rats and 45–51% in dogs — with the species difference stated (36588717). The independent rat tendon test is called inconclusive at n=4 per group rather than a failed replication (42542926). The 2026 sports-medicine review is quoted only for "remain unsubstantiated by current human trials"; its cardiovascular and metabolic risk findings belong to other peptides and are correctly not attributed to BPC-157 (42578445). The lay summary carries no dosing, sourcing or practical advice and no safety claim the review did not assess.

## Residual limitations

Three human reports and several rat studies were abstract-depth only, so Methods/Results figures (Biçer's medians and p values, the knee review's instrument statement, the FDA reconstruction of the 2005 enema abstract) were confirmable here only against the review's own receipts and the FDA briefing document. The 2005 ulcerative-colitis trial remains a meeting abstract known through an administrative synthesis. Human PK remains unresolved rather than negative.

## Strings for the lead to propagate to package-level files

| Old | New |
|---|---|
| `The earlier adversarial approval refers to its original edition; this provenance supplement is not a new independent or blinded adjudication.` | `No earlier adversarial approval is attached to this case; the 2026-10-06 adversarial pass did not cover it, and this provenance supplement is not a new independent or blinded adjudication.` |
| `The original approval predates this repair.` | `No earlier adversarial approval is attached to this case.` |

No level, confidence, role, source count or claim count changed. `review_metadata.reviewer` was left as house provenance per the lead's instruction. This case has no `propagate_to` copies and no divergent copies.

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