### Consensus consistency validation — 9 October 2026

Scope: retrospective validation of the archived v5.3 dossier, not a blind re-review. The preregistered scope remains **persistent SARS-CoV-2 material as a primary cause maintaining symptoms in people meeting prospectively defined Long-COVID criteria**; aggregation remains primary-subclaim SC3. Prior score: Γ, Moderate. Literature cutoff: 2026-10-09. Source records below were independently checked against PubMed primary reports on 2026-10-09 local (2026-10-10 UTC); this reconciles the archive's explicit UTC examination date.

#### Raw Consensus provenance

The parent's Consensus Pro response is preserved unchanged at `consensus-browser-response.md`, with raw thread record `browser-thread.json`. The direct Consensus keyword search results used in this check are also retained as source leads in that response's reference list. Do not treat the Consensus synthesis as original verified evidence or edit its citation markers.

#### Primary-source checks

| Load-bearing claim | Primary report checked | Source-level observation | Before/after judgement |
|---|---|---|---|
| Tissue persistence exists and is associated with symptoms, but does not establish human causal primacy. | Zuo et al., PMID 38663423; DOI 10.1016/S1473-3099(24)00171-3; PubMed abstract and metadata (`https://pubmed.ncbi.nlm.nih.gov/38663423/`); abstract_metadata. | Single-centre cross-sectional cohort; 225 recovered patients, residual clinical specimens; RNA found in selected solid tissues at 1–4 months; symptom questionnaire at 4 months; association OR 5.17 (95% CI 2.64–10.13). Authors explicitly state further work is needed to verify a mechanistic link. Subgenomic RNA detection is not by itself infectious-virus recovery. Exact source was already in inventory as PMID38663423. | No change to SC1 Β or SC3 Γ. Confirms existing evidence card. No temporal intervention/clearance or symptom rescue.
| Established Long-COVID symptoms did not significantly improve under the tested 15-day nirmatrelvir/ritonavir regimen. | Geng et al. STOP-PASC, PMID 38848477; DOI 10.1001/jamainternmed.2024.2007; PubMed abstract (`https://pubmed.ncbi.nlm.nih.gov/38848477/`); abstract_metadata. | 155 participants; 15-week blinded placebo-controlled RCT; primary pooled six-symptom severity at week 10; no significant between-group difference and no significant differences across listed secondary symptom/function measures. PMID 38848477 is the verified identity. Consensus raw response incorrectly gives PMID 38848147 for this DOI; keep its raw file unchanged and correct only this audit's citation mapping. | No change. This constrains benefit from this regimen in a protracted, mostly vaccinated cohort; no relevant tissue clearance/antigen target-engagement measurement demonstrated.
| A second established-disease trial also had a null primary health outcome with a 15-day regimen. | Sawano et al. PAX LC, PMID 40188838; DOI 10.1016/S1473-3099(25)00073-8; PubMed abstract (`https://pubmed.ncbi.nlm.nih.gov/40188838/`); abstract_metadata. | Double-blind phase 2 RCT; 100 randomized; day-28 PROMIS-29 physical-health difference −0.55 (95% CI −2.32 to 1.21; p=.54), with no significant improvement. It tested a short course and did not establish reservoir removal. It is already in inventory as its own trial (NCT05668091). | No change. Same directional constraint as prior appraisal; small phase 2 and no verified tissue clearance limit mechanistic falsification. It is distinct from STOP-PASC and RECOVER-VITAL.
| A larger phenotype-stratified established-disease trial found no significant primary benefit under 15- or 25-day regimens. | Baden et al. RECOVER-VITAL, PMID 42673984; DOI 10.1016/S1473-3099(26)00406-8; PubMed/primary abstract (also Consensus record `5ee123b1308d5f668b93f70c368bd043`); abstract_metadata. | 959 mITT participants across cognitive, autonomic, and exercise phenotypes. No statistically significant primary endpoint benefit for any phenotype in either treatment arm vs placebo; no secondary endpoint differences. Trial targets protease-dependent replication pharmacologically, but its reported abstract does not demonstrate tissue reservoir clearance or quantify tissue target engagement. The archived primary report is counted once as study NCT05595369. | No change to Γ or Moderate. Additional null evidence lowers the plausibility of a broadly treatment-responsive, accessible-replication model under regimens tested, but cannot rule out non-replicating antigen, unexposed compartments, longer treatment, or subgroup response.

#### Consensus-result reconciliation

Consensus usefully surfaced the Zuo cohort, STOP-PASC, PAX LC, and RECOVER-VITAL. Each already appears in the archived source inventory (Zuo PMID38663423; STOP-PASC PMID38848477; PAX LC PMID40188838; RECOVER-VITAL PMID42673984). The RECOVER-VITAL platform protocol is a protocol, not an outcome report, and is not counted as an independent trial. STOP-PASC secondary reports remain within the STOP-PASC trial family. The acute-treatment prevention study PMID42462744 remains outside the registered maintenance subclaim.

The raw Consensus answer overstates inference when it calls these trials "targeted antiviral eradication," describes a "clear dissociation between antiviral inhibition and clinical improvement," or says clinical intervention confirms that the virus does not independently maintain disease. Primary trial reports support **no significant clinical benefit under the tested regimen and population**. They do not show that tissue reservoirs were cleared or that tissue target engagement occurred. Preserve this as a claim-scope caveat; do not convert the nulls into proof of absence.

#### Per-subclaim score decision

- SC1 (presence after acute infection): Β, Moderate — unchanged. Tissue RNA/antigen presence is observed in selected human samples; the primary source checks corroborate the registered description.
- SC2 (viral material drives sustained immune/physiological disruption): Γ, Moderate — unchanged. The reviewed studies do not add a selective human reservoir perturbation or demonstrate a mechanistic clinical rescue.
- SC3 (primary cause of symptoms in clinically defined Long COVID): Γ, Moderate — unchanged. The four primary reports add no verified reservoir-removal intervention followed by symptom rescue; established-disease nulls are constraints on tested antiviral regimens, not direct clearance experiments.
- Aggregation remains primary-subclaim SC3: overall Γ, Moderate. No confidence-band change proposed: new validation confirms and sharpens source interpretation but does not resolve the causal gap or add independent positive causal lines. This is a retrospective consistency check, not blind calibration.

#### Sources

- Consensus raw response: `consensus-browser-response.md`; raw thread: `browser-thread.json`.
- PubMed: https://pubmed.ncbi.nlm.nih.gov/38663423/ ; https://pubmed.ncbi.nlm.nih.gov/38848477/ ; https://pubmed.ncbi.nlm.nih.gov/40188838/ ; https://pubmed.ncbi.nlm.nih.gov/42673984/

This dated supplement adds source reconciliation to the original review. Its original narrative, registered scope, source inventory and grades remain historical records. No earlier adversarial approval has been verified for this edition at the supplement date. This provenance supplement has no separate approval and is not an independent or blinded adjudication. New sources described here are supplemental and are not counted in the original inventory.