# Estimating the time course of biomarker changes in Alzheimer's disease.

- **DOI:** [10.1093/brain/awaf413](https://doi.org/10.1093/brain/awaf413)
- **Rating:** Sound (3.6 out of 5)
- **Verdict:** A useful cross-cohort Alzheimer’s biomarker timeline with broad modality coverage, constrained by internal validation that does not establish universal staging superiority.
- **Framework version:** v5.9
- **Pipeline version:** v0.4
- **Evaluation date:** 2026-10-08
- **Evaluation page:** https://www.nabu.science/paper/10.1093%2Fbrain%2Fawaf413

## Key claims

1. The proposed staging approach was found to produce stronger correlations with unseen cognitive measures and biomarkers compared to alternative staging methods, including amyloid and tau PET clocks (all pairwise P < 0.05).
   - Support: Tentative
   - Type: Primary finding
   - Basis: the stated universal significance conflicts with the non-significant GRACE comparison and rests on internal correlation-based validation
   - Location: Abstract; Results, Disease progression modeling results and validation
2. The period from Aβ-PET positivity to end-stage Alzheimer's disease dementia (MMSE = 0) was estimated at 20-25 years, with a presymptomatic phase of 7-11 years.
   - Support: Tentative
   - Type: Secondary
   - Basis: model-based extrapolation is sensitive to sparse late-stage data, increasing variance, and informative dropout
   - Location: Abstract; Results, Disease progression modeling results and validation; Discussion, limitations paragraph
3. CSF Aβ42/Aβ40 became abnormal about a year before Aβ-PET positivity, CSF phosphorylated-tau (p-tau)231, p-tau217 and plasma phosphorylated/neuritic plaque-tau217 1-3 years after, and tau-PET about 8 years after.
   - Support: Supported
   - Type: Secondary
   - Basis: bootstrap-timed model estimates show consistent ordering across cohorts, though abnormality thresholds and Aβ-status rules shape timing
   - Location: Abstract; Results, Biomarker trajectories; Fig. 4
4. Findings related to biomarker trajectories were highly consistent across cohorts.
   - Support: Supported
   - Type: Secondary
   - Basis: trajectory ordering and abnormality timings broadly agree across ADNI and BioFINDER despite different measurement systems
   - Location: Abstract; Results, Biomarker trajectories; Figs 3–4

## Strengths

- **Broad cross-modality biomarker timeline** Plasma, CSF, MRI, amyloid-PET, and tau-PET measures are placed on a shared predicted time-since-Aβ-PET-positivity axis, providing fine-grained relative timing across disease domains.
  - Dimension: contribution
  - Location: Results, Biomarker trajectories; Figs 3–4
- **Replication across two large cohorts** The broad ordering and abnormality timings are reported as highly consistent across ADNI and BioFINDER despite differences in assays, tracers, processing, and cognitive measures.
  - Dimension: methodological rigour
  - Location: Results, Biomarker trajectories; Figs 3–4 and Supplementary Fig. 9
- **Relevant comparator and sensitivity analyses** The model is compared with clinical staging variables, amyloid and tau clocks, GRACE, and LTJMM, while random-effect sub-models and bootstrap timing intervals provide additional checks.
  - Dimension: methodological rigour
  - Location: Methods, Model validation; Results, Tables 2–3 and Supplementary Tables 1–4

## Limitations

- **Universal superiority claim is inconsistent** The Abstract says all pairwise comparisons were significant, but Results report P = 0.1796 against GRACE. The conclusion’s broad superiority claim therefore exceeds the reported comparison evidence.
  - Dimension: reporting
  - Location: Abstract; Results, Disease progression modeling results and validation; Discussion, concluding paragraph
- **Validation does not test prognosis** Validation uses correlations with other measures from the same source cohorts rather than held-out subjects, future outcomes, or operational decisions. This does not establish the claimed prognostic or trial-staging utility.
  - Dimension: methodological rigour
  - Location: Methods, Model validation; Discussion, clinical application claims
- **Absolute durations remain model-dependent** Residual variance increases with disease age, late-stage estimates may be lengthened by informative dropout, and the model assumes one multivariate trajectory. These issues particularly qualify the 20–25-year duration estimate.
  - Dimension: positioning
  - Location: Results, conditional residuals; Discussion, limitations paragraph

## Reliability checks

Caveats (4 of 4 checks).

- **Results check:** Noted
  - The proposed staging approach was found to produce stronger correlations with unseen cognitive measures and biomarkers compared to alternative staging methods, including amyloid and tau PET clocks (all pairwise P < 0.05): "The proposed staging approach was found to produce stronger correlations with unseen cognitive measures and biomarkers compared to alternative staging methods, including amyloid and tau PET clocks (all pairwise P < 0.05)." (Abstract) vs "The predicted disease showed numerically stronger average correlations than GRACE disease time (P = 0.1796) and LTJMM disease time (P = 0.0001)." (Results — Disease progression modeling results and validation). The reported GRACE comparison has P = 0.1796 > 0.05, contradicting the unrestricted statement that all pairwise comparisons with alternative staging methods were significant. → Not supported by the paper's own data
- **Reporting and transparency:** Noted. The diagnosis-imputation sensitivity analysis is explicitly reported as data not shown, leaving its stated result without presented supporting estimates.
  - Conflict of interest: not stated · Ethics approval: declared · Data availability: declared · Code availability: declared
- **Reference integrity:** Clean
  - 69 references in manuscript
  - 65 of 65 checkable references found in an index
  - 4 have no canonical index record — counted, but not index-checkable
  - 1 reference confirmed by manual review
  - 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
