# Tau extent outperforms tau load as a predictor of neurodegeneration in Alzheimer's disease.

- **DOI:** [10.1186/s13024-026-00945-1](https://doi.org/10.1186/s13024-026-00945-1)
- **Rating:** Sound (3.6 out of 5)
- **Verdict:** Tau extent shows a credible imaging advantage over tau load, but unresolved longitudinal sample accounting limits confidence in the predictive comparison.
- **Framework version:** v5.9
- **Pipeline version:** v0.4
- **Evaluation date:** 2026-10-08
- **Evaluation page:** https://www.nabu.science/paper/10.1186%2Fs13024-026-00945-1

## Key claims

1. Across cross-sectional and longitudinal analyses, tau extent provided superior predictive value for imaging-based neurodegeneration compared with tau load.
   - Support: Supported
   - Type: Primary finding
   - Basis: consistent imaging-model advantage, limited by in-sample comparisons and unresolved longitudinal sample accounting
   - Location: Abstract, Conclusions
2. Full-cortex SEOT provided the best model fit for cortical thinning.
   - Support: Well-supported
   - Type: Secondary
   - Basis: lowest AIC and higher adjusted R² across the reported cortical-thickness models
   - Location: Abstract, Results; Results, Fig. 1E–F
3. Longitudinally, baseline full-cortex SEOT best predicted future cortical thinning, while temporal SEOT best predicted future hippocampal atrophy.
   - Support: Supported
   - Type: Secondary
   - Basis: mixed-model AIC advantages with sparse repeated observations and an unresolved follow-up denominator
   - Location: Abstract, Results; Results, Figs. 5–6
4. Tau extent was equal or superior to tau load in its correlation with local cortical thickness across all cortical regions.
   - Support: Supported
   - Type: Secondary
   - Basis: region-wise correlations favour extent, though multiplicity handling for the regional tests is unclear
   - Location: Abstract, Results; Results, Fig. 3
5. SEOT outperformed tau load for associations with HCV and for predicting MTA.
   - Support: Tentative
   - Type: Secondary
   - Basis: modest correlation differences and only one significant pairwise MTA AUC comparison
   - Location: Abstract, Results; Results, Figs. 2A and 4C

## Strengths

- **Direct multi-outcome metric comparison** Tau extent and SUVR are compared head-to-head across cortical thickness, hippocampal volume, MTA ratings, fluid biomarkers, regional associations, and longitudinal trajectories. This addresses a specific uncertainty left by earlier SEOT studies.
  - Dimension: contribution
  - Location: Abstract; Results, Figs. 1–6
- **Extensive parameter sensitivity testing** The analyses vary SEOT thresholds and reference groups, exclude temporal ceiling cases, and adjust longitudinal models for baseline neurodegeneration. The principal direction of the imaging findings is largely preserved.
  - Dimension: methodological rigour
  - Location: Results, Sensitivity Analyses; Supplementary Figs. 5–13
- **Imaging and fluid results distinguished** The paper reports better comparative performance for several imaging outcomes while explicitly finding no significant advantage for plasma NfL or CSF total tau. This distinction prevents the conclusion from being generalized to every neurodegeneration marker.
  - Dimension: reporting
  - Location: Abstract, Results and Conclusions; Fig. 2B–C

## Limitations

- **Longitudinal sample count remains unresolved** The paper reports 227 participants with follow-up and 100 without follow-up from a baseline cohort of 367, leaving 40 participants unaccounted for in the attrition comparison. This directly complicates interpretation of selection into the longitudinal analysis.
  - Dimension: methodological rigour
  - Location: Results, opening paragraph; Results, Baseline Differences According to Follow-up Status
- **Superiority language exceeds several comparisons** The claimed imaging advantage relies substantially on in-sample AIC, while some metric differences are small or not significant; only cortical SEOT versus cortical SUVR differed significantly in the MTA AUC comparison. The title and broad predictive wording therefore require tighter calibration.
  - Dimension: reporting
  - Location: Title; Abstract; Results, Figs. 1, 4–6
- **External applicability remains narrowly demonstrated** The evidence comes from a single predominantly White Canadian cohort using one tracer and scanner platform. Threshold, reference-group, smoothing, and cross-site implementation conventions remain to be standardized and independently validated.
  - Dimension: application pathway
  - Location: Methods, Study Population and Neuroimaging Acquisition; Discussion, limitations

## Reliability checks

Caveats (4 of 4 checks).

- **Results check:** Noted
  - SEOT outperformed tau load for associations with HCV and for predicting MTA: "Furthermore, all tau-PET metrics were significantly negatively correlated with hippocampal volume (all p < 0.0001)." (Results — Cross-sectional imaging markers of neurodegeneration) vs "Both temporal (rho = −0.34) and cortical (rho = −0.32) SEOT exhibited stronger correlations with hippocampal volume than temporal (rho = −0.27) and cortical (rho = 0.24) SUVR (Fig. 2A)." (Results — Cross-sectional imaging markers of neurodegeneration). The reported cortical-SUVR rho of positive 0.24 contradicts the statement that every hippocampal-volume correlation is negative, but the supplied text does not establish which report is correct. → Effect on result unclear
  - Minor reporting notes (1):
    - Descriptive: "By comparing baseline characteristics of participants with (n = 227) and without (n = 100) follow-up visits, we observed no differences in age, sex, years of education, or APOE status" (Results — Baseline differences according to follow-up status) vs "In our sample, 38% (n = 140) of participants had only one time point and 34% (n = 125) had exactly two time points;" (Methods — Statistical analysis). Subtracting the reported 227 participants with follow-up from the baseline cohort of 367 gives 140 without follow-up, matching the Methods count rather than the comparison paragraph's 100. → Conclusion holds
- **Reporting and transparency:** Noted. Longitudinal interaction coefficients are labelled standardized without enough information to resolve their scaling across cortical-thickness and hippocampal-volume outcomes.
  - Ethics approval: declared · Data availability: declared
- **Reference integrity:** Clean
  - 68 references in manuscript
  - 68 of 68 checkable references found in an index
  - 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
