# Adversarial review log — nad-boosters-human-function

Authoring model: **Sol 6.1**. Adversarial reviewer: **Claude Opus 5.5 (Anthropic)**. Date: **2026-10-14**.
Round 2 of the HSEF v5.3 adversarial pass. Canonical copy:
`hsef_v53_traffic_ranks_21_22_2026-10-13/nad-boosters-human-function/nad-boosters-human-function.review.md`.

## Summary

| Item | Before | After |
|---|---|---|
| HSEF level | Δ — First field-valid causal perturbation crossed | Δ — First field-valid causal perturbation crossed (unchanged) |
| Confidence | Low | Low (unchanged) |
| Causal role | reinforcing / stabilizing factor | unchanged |
| Momentum / Stability / Divergence | Mechanistic only / Stable / Contested | unchanged (Divergence rationale given its required artifact) |
| Positive qualifying study units | 0 | 1 |
| Validator | not previously emitted | PASS — 0 fail, 0 warn, 31 pass (`validator.txt`) |

Findings by severity: **Critical 0, Major 3, Minor 3, Editorial 1** (7 corrections applied).
PMIDs checked: **70/70 references verified, 0 failed**, plus 2 linked correction notices verified as status flags.
Load-bearing claims spot-checked against source abstracts: **14, all accurately attributed**.

## Findings

| ID | Severity | Location | Correction |
|---|---|---|---|
| A1 | Major | Decisive-perturbation inventory and causal tally; scorecard `causal_tally_audit` | Tally 0 → 1 positive qualifying study unit (the twelve-week randomized gait-time contrast supporting SC2). A zero tally cannot coexist with a Δ headline or with the same paragraph's statement that one positive qualifying source is counted. |
| A2 | Major | Decisive-perturbation inventory, closing paragraph | "No positive field-valid crossing was verified for the registered causal scope" restricted to the SC1, SC3 and SC4 scopes, which is what the evidence supports; as written it contradicted SC2's recorded positive crossing. |
| A3 | Minor | M/S/D section; scorecard `extended_scoring_MSD.divergence` | Divergence "Contested" given the checkable artifact the framework requires: the rationale now names the split (promotional framing above the assessed level, trial literature at or below it). Value unchanged. |
| A4 | Minor | REFERENCE 33 | Empty "DOI: ." field → "DOI: none indexed in PubMed." (PubMed holds no DOI for this record). |
| A5 | Minor | Throughout, ~50 sites | Builder rendering defect: missing spaces around numerals and symbols restored. No numeric value, unit or identifier changed. |
| A6 | Editorial | Throughout, 10 sites | Doubled field-separator full stops reduced to one; "MIB626" normalised to "MIB-626". |
| A7 | Major | AT A GLANCE, "Best-supported rival" | The field named the nominal model (score card `is_nominal: true`, tournament rank 2) as the best-supported rival. Now names the highest-ranked non-nominal model, "Exposure and functional response are partly dissociated", and states that the nominal ranks second. Known renderer defect printing `models_evaluated[0]`; the score card's rank and nominal flags are correct and unchanged. |

## Rejected findings (checked, no change)

1. Citation years differing by one from PubMed e-pub dates (refs 5, 8, 23, 34, 45, 47, 63, 64) — house convention.
2. PubMed titles dropping superscripts/italics (NAD+, β-NMN, H2O2, RYR1) — field artifact.
3. The DOI-only Igarashi 2026 npj Aging source — a PubMed search on 2026-10-14 still returns nothing for it, so the
   separately displayed, non-counted handling with no invented PMID is correct.
4. Momentum "Mechanistic only" — required artifact present and does not claim decisive work underway.
5. Δ for SC2 resting on a secondary endpoint — argued against and retained; the registered gate admits measured gait
   and requiring a trial-primary endpoint would add a rule after seeing the result.
6. Aggregation rule (primary-subclaim, SC2) applied as written; scope-matched would also return Δ.
7. No collective-author defect in this reference list (no consortium first authors).
8. Confidence band Low — verified against the framework's confidence text rather than against access depth. The crossing rests on one secondary endpoint in one trial, so neither Moderate nor High is reachable; the band is correct even though the score card's `confidence_rationale` boilerplate also mentions incomplete access.
9. No false-approval or false-provenance clause is present in this review, and no earlier adversarial approval is asserted.

## Files changed

- `nad-boosters-human-function.review.md` — banner, `## ADVERSARIAL REVIEW AND APPROVAL` section, A1–A7.
- `nad-boosters-human-function.scorecard.json` — `causal_tally_audit` (tally 0 → 1, unit text), `extended_scoring_MSD.divergence.justification`.
- `validator.txt` — new; PASS.
- `sol61-original/` — pre-review copies of the review, scorecard and validation report.

No package-level file (OVERVIEW.md, MANIFEST.json, READER.html, BATCH_COMPLETION.json) was edited.

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