# Tau368 improves p-tau diagnostic accuracy for FTLD-tau from FTLD-TDP.

- **DOI:** [10.1007/s00401-026-03042-1](https://doi.org/10.1007/s00401-026-03042-1)
- **Rating:** Limited (2.9 out of 5)
- **Verdict:** Autopsy-anchored CSF data identify promising FTLD-tau markers, but post hoc PSP exclusion and inconsistent AUC reporting make tau368’s incremental value provisional.
- **Framework version:** v5.9
- **Pipeline version:** v0.4
- **Evaluation date:** 2026-10-08
- **Evaluation page:** https://www.nabu.science/paper/10.1007%2Fs00401-026-03042-1

## Key claims

1. CSF p-tau181 and p-tau212, as well as p-tau181/tau368 and p-tau212/tau368, were higher in FTLD-tau compared with non-tau groups, and the diagnostic accuracy to discriminate FTLD-tau from FTLD-TDP improved.
   - Support: Tentative
   - Type: Primary finding
   - Basis: small retrospective subgroups, post hoc PSP exclusion, and inconsistent AUC reporting limit the claimed incremental accuracy
   - Location: Abstract; Results, FTLD ROC analysis
2. In AD, CSF p-tau181 and p-tau212 were significantly higher than in all other groups, including FTLD-tau, while tau368/t-tau was significantly lower.
   - Support: Well-supported
   - Type: Secondary
   - Basis: autopsy-defined AD comparisons and reported group medians consistently support the stated biomarker pattern
   - Location: Abstract; Results, AD biomarker comparisons; Table 1
3. Levels of the ratios were increased in behavioral and primary progressive aphasia variants of tauopathies when compared to FTLD-TDP.
   - Support: Tentative
   - Type: Secondary
   - Basis: very small bvFTD groups and supplementary PPA analyses limit clinical-subtype inference
   - Location: Abstract; Results, clinical variants; Fig. 5
4. the biomarkers showed significant correlation with both FTLD-tau and AD tau burden at autopsy across brain regions.
   - Support: Tentative
   - Type: Secondary
   - Basis: small exploratory correlations and incomplete within-FTLD significance limit the breadth of the statement
   - Location: Abstract; Results, brain tau correlation sections

## Strengths

- **Autopsy reference for central comparison** FTLD-tau and FTLD-TDP are defined neuropathologically, providing a strong reference standard for the headline disease comparison.
  - Dimension: methodological rigour
  - Location: Methods, Penn biomarker cohort participants
- **Addresses an unmet biomarker need** The study targets the absence of reliable fluid biomarkers for FTLD-tau and connects the proposed markers to clinical-trial selection and differentiation of overlapping syndromes.
  - Dimension: contribution
  - Location: Introduction, paragraph 1; Conclusions
- **Multiple safeguards in quantitative analysis** The analysis reports assay precision, FDR correction, bootstrap confidence intervals, and blinded case-level image quantification.
  - Dimension: methodological rigour
  - Location: Methods, Biomarker measurements; Microscopy; Statistical analyses

## Limitations

- **Central AUC reporting does not reconcile** The Results and Discussion report materially different AUCs for FTLD-tau versus FTLD-TDP, changing the apparent size of tau368’s incremental benefit.
  - Dimension: reporting
  - Location: Results, FTLD ROC analysis; Discussion, FTLD diagnostic-performance paragraph
- **PSP exclusion is outcome informed** Four PSP cases were removed from the main ADNC-negative analysis after being identified as tau368-ratio outliers, while analyses retaining them appear only in supplementary material.
  - Dimension: methodological rigour
  - Location: Results, “CSF p-tau/tau368 ratios improve the diagnostic accuracy…”, paragraph 1
- **Clinical claims exceed validation stage** The Conclusions predict fewer false results and improved diagnosis despite a small, single-biobank proof-of-concept cohort with no external validation.
  - Dimension: positioning
  - Location: Discussion, limitations paragraph; Conclusions

## Reliability checks

Caveats (4 of 4 checks).

- **Results check:** Noted
  - the biomarkers showed significant correlation with both FTLD-tau and AD tau burden at autopsy across brain regions: "The correlation was significantly negative, possibly reflecting accumulation of this fragment in the brain and proving the association between tau368 and tangle pathology." (Discussion, paragraph beginning “Having in mind the complex physiology”) vs "CSF tau368 was negatively correlated with tau burden, although this did not reach significance (r = − 0.44, 95% CI − 0.75–0.01, p = 0.057)" (Results, Tau368 mediates the correlation of p-tau biomarkers in the FTLD-tau group). The reported tau368 correlation has p = 0.057, above the stated 0.05 threshold, so the Discussion's assertion of statistical significance contradicts the Results. → Not supported by the paper's own data
  - CSF p-tau181 and p-tau212, as well as p-tau181/tau368 and p-tau212/tau368, were higher in FTLD-tau compared with non-tau groups, and the diagnostic accuracy to discriminate FTLD-tau from FTLD-TDP improved: "Thus, analyses in ADNC-negative subcohort excluded four PSP cases (n = 102)." (Results, CSF p-tau/tau368 ratios improve the diagnostic accuracy) vs "The figure shows levels of tau biomarkers in clinically healthy individuals without cognitive impairment (normal; n = 41), frontotemporal lobar degeneration type tau (FTLD-Tau; n = 21), frontotemporal lobar degeneration type TDP (FTLD-TDP; n = 21), and neuronal α-synuclein diseases (αSyn; n = 12)." (Figure 3 caption). Figure 3's group counts sum to 41 + 21 + 21 + 12 = 95, seven fewer than the stated post-exclusion cohort of 102, without an explanation identifying the correct analytic denominator. → Effect on result unclear
  - CSF p-tau181 and p-tau212, as well as p-tau181/tau368 and p-tau212/tau368, were higher in FTLD-tau compared with non-tau groups, and the diagnostic accuracy to discriminate FTLD-tau from FTLD-TDP improved: "p-tau181 (n = 130; 33 AD, 20 FTLD-tau, 17 FTLD-TDP, 18 αSyn, 42 Control)" (Methods, Biomarker measurements) vs "clinically healthy individuals without cognitive impairment (normal; n = 41)" (Figure 2 caption; corroborates Table 1 Normal group count). The Methods report p-tau181 measurements for 42 controls, one more than the 41 controls reported for the cohort, and the supplied text does not establish which count is correct. → Effect on result unclear
- **Reporting and transparency:** Noted. AUC reporting does not consistently label the PSP inclusion basis, logistic-model effects are called hazard rates rather than odds ratios, and the correlation-method descriptions differ between Methods and Figure 4.
  - Ethics approval: declared
- **Reference integrity:** Clean
  - 126 references in manuscript
  - 126 of 126 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
