# Impact of zervimesine on the neuroinflammatory biomarker GFAP and related proteomic molecular correlates in plasma of participants from a phase 2 clinical trial in Alzheimer's disease.

- **DOI:** [10.1186/s13195-026-02025-4](https://doi.org/10.1186/s13195-026-02025-4)
- **Rating:** Sound (3.0 out of 5)
- **Verdict:** Randomized trial biomarker analyses generate plausible enrichment and pharmacodynamic hypotheses, but uncontrolled multiplicity makes the central signals unreliable without replication.
- **Framework version:** v5.9
- **Pipeline version:** v0.4
- **Evaluation date:** 2026-10-08
- **Evaluation page:** https://www.nabu.science/paper/10.1186%2Fs13195-026-02025-4

## Key claims

1. Collectively, plasma biomarker findings in the low p-tau217 subgroup were more robust than in the mITT population. Levels of GFAP were significantly decreased and NfL, Aβ42 and Aβ40 levels trended towards a decrease with zervimesine compared to placebo.
   - Support: Tentative
   - Type: Primary finding
   - Basis: multiple nominal subgroup comparisons without multiplicity control or a treatment-by-subgroup interaction test
   - Location: Abstract; Results, Figure 1B
2. Proteomics analyses identified candidate pharmacodynamic biomarkers of zervimesine, and gene ontology and pathway analyses pointed to an impact on amyloid biology, trafficking, lipid metabolism, and immune response.
   - Support: Tentative
   - Type: Secondary
   - Basis: small selected proteomic subsets and thousands of unadjusted protein-level tests make individual candidates highly uncertain
   - Location: Abstract; Results, Figures 2–4
3. Exploratory plasma biomarker findings align with the degree of clinical benefit in SHINE mITT and low p-tau217 populations, and support future trial enrichment with patients with lower levels of pathology as defined by lower baseline plasma levels of p-tau217.
   - Support: Tentative
   - Type: Secondary
   - Basis: a single sample-specific subgroup without formal interaction testing cannot validate a responder-enrichment strategy
   - Location: Abstract; Conclusion

## Strengths

- **Randomized placebo-controlled biomarker source trial** SHINE used a randomized, double-blind, placebo-controlled design, allowing biomarker changes to be compared against a concurrent control rather than an untreated historical series.
  - Dimension: methodological rigour
  - Location: Methods, Clinical Trial Design
- **New plasma biomarker dataset** The study extends prior CSF work with targeted plasma markers and untargeted TMT-MS proteomics, creating a useful hypothesis set for later studies.
  - Dimension: contribution
  - Location: Background, final paragraphs; Results, Figures 1–6
- **Limitations are explicitly acknowledged** The Limitations section identifies small subgroup samples, unadjusted p-values, false-positive risk, peripheral biomarker ambiguity, and the need for replication.
  - Dimension: positioning
  - Location: Limitations

## Limitations

- **Multiplicity threatens central biomarker signals** More than 2,000 proteins were tested using unadjusted p-values, while the resulting hit counts are near chance expectations and the Discussion misstates what a 5% test-wise error rate implies for identified proteins.
  - Dimension: methodological rigour
  - Location: Methods, Differential Abundance Analyses; Results, Figures 2–4; Discussion, GFAP-correlate paragraph
- **Responder inference lacks interaction testing** The claim that low-p-tau217 findings are more robust relies on nominal subgroup results rather than a reported treatment-by-subgroup interaction, limiting the enrichment inference.
  - Dimension: methodological rigour
  - Location: Abstract; Results, Figure 1B; Conclusion
- **Conclusions exceed exploratory evidence** Phrases such as “leads to positive changes” and “identify a patient population more likely to respond” give small, multiplicity-exposed subgroup findings greater certainty than the analysis supports.
  - Dimension: reporting
  - Location: Conclusion

## Reliability checks

Caveats (4 of 4 checks).

- **Results check:** Clean. The reported numbers and statistics are consistent with the stated results.
- **Reporting and transparency:** Noted. The 15 proteins overlapping the GFAP correlates and differential-abundance results are described using different significance thresholds; their individual p-values are not supplied, so the number meeting the stricter threshold cannot be determined.
  - Conflict of interest: not stated · Ethics approval: not stated · Preregistration: declared · Data availability: not stated · Code availability: not stated
- **Reference integrity:** Noted
  - 53 references in manuscript
  - 45 of 45 checkable references found in an index
  - 2 are books, websites or datasets — counted, but not index-checkable
  - 6 could not be checked — not the same as no match found
  - References not verified automatically are resolved manually within 24 hours.

## Post-publication record

- **Status:** Clean
- **Summary:** No retraction or expression of concern found in Crossref (includes Retraction Watch).
- **Last checked:** 2026-10-17
