CoChat

Comprehensive cross-sectional and longitudinal comparisons of plasma glial fibrillary acidic protein and neurofilament light across FTD spectrum disorders

Open access

The DOI matches its Crossref record; 73 of the 75 references it cites were found in a registry or an index; 28 of the 36 sources its claims cite were checked for support.

Abstract

BACKGROUND: Therapeutic development for frontotemporal dementia (FTD) is hindered by the lack of biomarkers that inform susceptibility/risk, prognosis, and the underlying causative pathology. Blood glial fibrillary acidic protein (GFAP) has garnered attention as a FTD biomarker. However, investigations of GFAP in FTD have been hampered by symptomatic and histopathologic heterogeneity and small cohort sizes contributing to inconsistent findings. Therefore, we evaluated plasma GFAP as a FTD biomarker and compared its performance to that of neurofilament light (NfL) protein, a leading FTD biomarker. METHODS: We availed ARTFL LEFFTDS Longitudinal Frontotemporal Lobar Degeneration (ALLFTD) study resources to conduct a comprehensive cross-sectional and longitudinal examination of the susceptibility/risk, prognostic, and predictive performance of GFAP and NfL in the largest series of well-characterized presymptomatic FTD mutation carriers and participants with sporadic or familial FTD syndromes. Utilizing single molecule array technology, we measured GFAP and NfL in plasma from 161 controls, 127 presymptomatic mutation carriers, 702 participants with a FTD syndrome, and 67 participants with mild behavioral and/or cognitive changes. We used multivariable linear regression and Cox proportional hazard models adjusted for co-variates to examine the biomarker utility of baseline GFAP and NfL concentrations or their rates of change. RESULTS: Compared to controls, GFAP and NfL were elevated in each FTD syndrome but GFAP, unlike NfL, poorly discriminated controls from participants with mild symptoms. Similarly, both baseline GFAP and NfL were higher in presymptomatic mutation carriers who later phenoconverted, but NfL better distinguished non-converters from phenoconverters. We additionally observed that GFAP and NfL were associated with disease severity indicators and survival, but NfL far outperformed GFAP. Nevertheless, we validated findings that the GFAP/NfL ratio may discriminate frontotemporal lobar degeneration with tau versus TDP-43 pathology. CONCLUSIONS: Our head-to-head comparison of plasma GFAP and NfL as biomarkers for FTD indicate that NfL consistently outmatched GFAP as a prognostic and predictive biomarker for participants with a FTD syndrome, and as a susceptibility/risk biomarker for people at genetic risk of FTD. Our findings underscore the need to include leading biomarkers in investigations evaluating new biomarkers if the field is to fully ascertain their performance and clinical value.

Overview

CoChat Summary

## TL;DR

In this ALLFTD study of 1,057 participants, plasma NfL consistently outperformed plasma GFAP across FTD spectrum disorders as a diagnostic, prognostic, and susceptibility/risk biomarker. GFAP was elevated in FTD but generally had poor discriminatory power, while the GFAP/NfL ratio had some value for distinguishing FTLD-tau from FTLD-TDP, largely driven by NfL.

## Key findings

  • In 1,057 participants (161 controls, 127 presymptomatic mutation carriers, 308 bvFTD, 76 nfvPPA, 83 svPPA, 92 CBS, 143 PSP-RS, 67 MBCI), both GFAP and NfL were elevated in FTD spectrum disorders, but NfL was consistently stronger.
  • Baseline GFAP showed poor to moderate discrimination of controls from symptomatic groups, with age/sex-adjusted AUCs of 0.53 for MBCI and 0.56–0.68 for other symptomatic groups; NfL showed better discrimination, with AUCs of 0.69 for MBCI and 0.78–0.87 for other symptomatic groups.
  • In presymptomatic mutation carriers, both biomarkers were higher in phenoconverters, but NfL better separated phenoconverters from non-converters: baseline AUC 0.77 vs 0.61 for GFAP, and rate-of-change AUC 0.78 vs 0.65.
  • Higher GFAP and NfL were associated with worse disease severity and shorter survival after symptom onset, but NfL was stronger: for death after symptom onset, HRs were 2.97 for NfL vs 2.01 for GFAP in bvFTD, and 2.39 vs 1.36 in all FTD syndromes combined.
  • For pathology classification, GFAP alone had little value (AUC 0.55 for FTLD-tau vs FTLD-TDP), while NfL and the GFAP/NfL ratio discriminated better (AUCs 0.78 and 0.73), with the paper stating this was largely driven by NfL.

## Methods

  • Cross-sectional and longitudinal analysis using ALLFTD resources, including clinically normal mutation-negative controls, presymptomatic mutation carriers, symptomatic FTD spectrum groups, and an MBCI comparison group.
  • Plasma GFAP and NfL were measured by single molecule array (Quanterix); the paper reports baseline and longitudinal measures across 1,691 samples from the 1,057 participants.
  • Main analyses used multivariable linear regression, Cox proportional hazards models, Spearman correlations, and AUC-based discrimination analyses, with Bonferroni correction and sensitivity analyses using GEE/marginal models and BMI adjustment.

## Limitations

  • The authors state that participants “may not represent the general FTD population,” and diagnoses were clinical rather than neuropathologic for most participants.
  • The paper states BMI was missing for 18% of participants, so it was not included in primary multivariable models; this is an evident limitation because BMI can confound biomarker levels.
  • The authors note medication/supplement use was not accounted for, and give minocycline as a possible confounder of NfL.
  • The paper also notes some phenotype groups were small, which raises the possibility of false-negative findings.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1186/s13024-025-00821-4

Cites

73 of 73 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Associations of plasma GFAP and NfL concentrations with age and sex in some phenotype groups. Associations of baseline concentrations of GFAP (a) or NfL (b) with age and sex assessed using linear regression models adjusted for age, sex and, for bvFTD, also for symptom duration. The number of participants (n), β coefficients (β), 95% confidence intervals (95% CI) and p values are shown. p < 0.025 (controls and presymptomatic mutation carriers) or p < 0.017 (bvFTD) are considered statistically significant. Grey circles represent seven bvFTD participants with an unknown mutation status, two bvFTD participants with a TARDBP mutation, and one bvFTD participant with a likely VCP pathogenic variant. GFAP and NfL concentrations are shown on the base 2 logarithm scale. Black horizontal bars represent median GFAP or NfL concentrations. See also Tables S6 and S7 for associations of baseline GFAP and NfL with age, sex and disease duration in other disease groups
Fig. 1 Associations of plasma GFAP and NfL concentrations with age and sex in some phenotype groups. Associations of baseline concentrations of GFAP (a) or NfL (b) with age and sex assessed using linear regression models adjusted for age, sex and, for bvFTD, also for symptom duration. The number of participants (n), β coefficients (β), 95% confidence intervals (95% CI) and p values are shown. p < 0.025 (controls and presymptomatic mutation carriers) or p < 0.017 (bvFTD) are considered statistically significant. Grey circles represent seven bvFTD participants with an unknown mutation status, two bvFTD participants with a TARDBP mutation, and one bvFTD participant with a likely VCP pathogenic variant. GFAP and NfL concentrations are shown on the base 2 logarithm scale. Black horizontal bars represent median GFAP or NfL concentrations. See also Tables S6 and S7 for associations of baseline GFAP and NfL with age, sex and disease duration in other disease groupsCC BY · PubMed Central
Fig. 2 Baseline plasma GFAP and NfL are elevated in FTD syndromes. a, b Comparison of baseline plasma GFAP (a) and NfL (b) between healthy controls or presymptomatic mutation carriers and all participants for a given symptomatic group. c, d Comparison of baseline GFAP (c) and NfL (d) between controls or presymptomatic carriers and participants in symptomatic groups with a CDR® + NACC-FTLD global score of 0 or 0.5. a-d Heat maps show AUCs comparing controls to the indicated groups that either include all individuals (All participants) or only those with an CDR® + NACC-FTLD global score of 0 or 0.5 (Mildly impaired participants) from unadjusted or age and sex-adjusted analyses. The number of participants (n) is shown. p values are from analysis adjusted for age and sex; p < 0.0071 (comparisons to controls) or p < 0.0083 (comparisons to presymptomatics) were considered statistically significant. ****p < 0.0001 and **p < 0.01 (comparison to controls); ####p < 0.0001, ###p < 0.001 and ##p < 0.01 (comparison to presymptomatic mutation carriers). Horizontal bars represent median GFAP or NfL concentrations, which are shown on the base 2 logarithm scale. See Tables S8 and S9 relating to panels a and b, and Tables S14 and S15 relating to panels c and d
Fig. 2 Baseline plasma GFAP and NfL are elevated in FTD syndromes. a, b Comparison of baseline plasma GFAP (a) and NfL (b) between healthy controls or presymptomatic mutation carriers and all participants for a given symptomatic group. c, d Comparison of baseline GFAP (c) and NfL (d) between controls or presymptomatic carriers and participants in symptomatic groups with a CDR® + NACC-FTLD global score of 0 or 0.5. a-d Heat maps show AUCs comparing controls to the indicated groups that either include all individuals (All participants) or only those with an CDR® + NACC-FTLD global score of 0 or 0.5 (Mildly impaired participants) from unadjusted or age and sex-adjusted analyses. The number of participants (n) is shown. p values are from analysis adjusted for age and sex; p < 0.0071 (comparisons to controls) or p < 0.0083 (comparisons to presymptomatics) were considered statistically significant. ****p < 0.0001 and **p < 0.01 (comparison to controls); ####p < 0.0001, ###p < 0.001 and ##p < 0.01 (comparison to presymptomatic mutation carriers). Horizontal bars represent median GFAP or NfL concentrations, which are shown on the base 2 logarithm scale. See Tables S8 and S9 relating to panels a and b, and Tables S14 and S15 relating to panels c and dCC BY · PubMed Central
Fig. 3 GFAP demonstrates less utility as a susceptibility/risk biomarker compared to NfL. a, b Comparison of baseline GFAP (a) or NfL (b) concentrations in presymptomatic carriers who phenoconverted to controls or to presymptomatic carriers who remained asymptomatic for at least one year. c, d Comparison of baseline GFAP (c) or NfL (d) concentrations in controls to presymptomatic individuals with either a C9orf72, GRN or MAPT mutation. e, f Comparison of GFAP (e) or NfL (f) rates of change in presymptomatic carriers who phenoconverted to controls or to presymptomatic carriers who remained asymptomatic for at least one year. g, h Comparison of GFAP (g) or NfL (h) rates of change in controls to presymptomatic individuals with either a C9orf72, GRN or MAPT mutation. The number of participants (n) is shown. p values are from analysis adjusted for age and sex. p < 0.0167 is considered statistically significant. ****p < 0.0001, **p < 0.01 and *p = 0.013 (comparison to controls); ##p < 0.01 (comparison of phenoconverters to non-converters). In panels a-d, horizontal bars represent median GFAP or NfL concentrations, which are shown on the base 2 logarithm scale. In panels e–h, rates of GFAP or NfL change are shown with box and whiskers plots representing minimum, first quartile, median, third quartile, and maximum values. See Table S16 relating to panels a, b, e and f, and Table S18 relating to panels c, d, g and h
Fig. 3 GFAP demonstrates less utility as a susceptibility/risk biomarker compared to NfL. a, b Comparison of baseline GFAP (a) or NfL (b) concentrations in presymptomatic carriers who phenoconverted to controls or to presymptomatic carriers who remained asymptomatic for at least one year. c, d Comparison of baseline GFAP (c) or NfL (d) concentrations in controls to presymptomatic individuals with either a C9orf72, GRN or MAPT mutation. e, f Comparison of GFAP (e) or NfL (f) rates of change in presymptomatic carriers who phenoconverted to controls or to presymptomatic carriers who remained asymptomatic for at least one year. g, h Comparison of GFAP (g) or NfL (h) rates of change in controls to presymptomatic individuals with either a C9orf72, GRN or MAPT mutation. The number of participants (n) is shown. p values are from analysis adjusted for age and sex. p < 0.0167 is considered statistically significant. ****p < 0.0001, **p < 0.01 and *p = 0.013 (comparison to controls); ##p < 0.01 (comparison of phenoconverters to non-converters). In panels a-d, horizontal bars represent median GFAP or NfL concentrations, which are shown on the base 2 logarithm scale. In panels e–h, rates of GFAP or NfL change are shown with box and whiskers plots representing minimum, first quartile, median, third quartile, and maximum values. See Table S16 relating to panels a, b, e and f, and Table S18 relating to panels c, d, g and hCC BY · PubMed Central
Fig. 4 Rates of GFAP and NfL change across prodromal and symptomatic phases. a,b For individuals with one or more serial GFAP (a) or NfL (b) measurements at least one year from baseline, we show comparisons of rate of change in biomarker concentrations per year for controls, all presymptomatic mutation carriers combined (All PreSx), presymptomatic carriers who did not convert (Non-conv), those that did phenoconvert (Phenoconv) and participants with MBCI, bvFTD, PPA or parkinsonian disorders. The number of participants (n) is shown. p values are from analysis adjusted for age and sex when comparing rates of GFAP or NfL change between controls and the indicated groups. p < 0.0071 is considered statistically significant; ****p < 0.0001, *** p < 0.001 and **p < 0.01. See also Table S19
Fig. 4 Rates of GFAP and NfL change across prodromal and symptomatic phases. a,b For individuals with one or more serial GFAP (a) or NfL (b) measurements at least one year from baseline, we show comparisons of rate of change in biomarker concentrations per year for controls, all presymptomatic mutation carriers combined (All PreSx), presymptomatic carriers who did not convert (Non-conv), those that did phenoconvert (Phenoconv) and participants with MBCI, bvFTD, PPA or parkinsonian disorders. The number of participants (n) is shown. p values are from analysis adjusted for age and sex when comparing rates of GFAP or NfL change between controls and the indicated groups. p < 0.0071 is considered statistically significant; ****p < 0.0001, *** p < 0.001 and **p < 0.01. See also Table S19CC BY · PubMed Central
Fig. 5 Plasma NfL better predicts survival after symptom onset than plasma GFAP. Comparisons of baseline GFAP (a, b) or NfL (c, d) concentrations in predicting survival after symptom onset in participants with bvFTD or in the combined group of participants with a FTD syndrome. For ease of presentation, GFAP and NfL were divided into a 2-level categorical variable based on sample medians. bvFTD, n = 301 (participants who died, n = 42); all FTD syndromes, n = 678 (participants who died, n = 105). In (a): Low ≤ 247.00 pg/ml, High > 247.00 pg/ml. In (b): Low ≤ 237.00 pg/ml, High > 237.00 pg/ml. In (c): Low ≤ 42.00 pg/ml, High > 42.00 pg/ml. In (d) Low ≤ 32.00 pg/ml, High > 32.00 pg/ml. See also Table S26
Fig. 5 Plasma NfL better predicts survival after symptom onset than plasma GFAP. Comparisons of baseline GFAP (a, b) or NfL (c, d) concentrations in predicting survival after symptom onset in participants with bvFTD or in the combined group of participants with a FTD syndrome. For ease of presentation, GFAP and NfL were divided into a 2-level categorical variable based on sample medians. bvFTD, n = 301 (participants who died, n = 42); all FTD syndromes, n = 678 (participants who died, n = 105). In (a): Low ≤ 247.00 pg/ml, High > 247.00 pg/ml. In (b): Low ≤ 237.00 pg/ml, High > 237.00 pg/ml. In (c): Low ≤ 42.00 pg/ml, High > 42.00 pg/ml. In (d) Low ≤ 32.00 pg/ml, High > 32.00 pg/ml. See also Table S26CC BY · PubMed Central
Fig. 6 Baseline NfL and GFAP/NfL ratio can discern FTLD-tau from FTLD-TDP pathology. a-c Comparisons of baseline GFAP (a), NfL (b), or GFAP/NfL (c) in the combined group of participants with a FTD syndrome with no mutation, or either a C9orf72, GRN, or MAPT mutation. p values are from analysis adjusted for age, sex and symptom duration. p < 0.0083 is considered statistically significant; ****p < 0.0001 (comparison to no mutation participants, “None”); ####p < 0.0001 (comparison to GRN mutation carriers). d-f Comparisons of GFAP (d), NfL (e), or GFAP/NfL (f) between neuropathologically confirmed participants with FTLD-tau or with FTLD-TDP. p values are from analysis adjusted for age at biomarker measurement, sex, symptom duration, and age at death. p < 0.025 is considered statistically significant; ****p < 0.0001. In all panels, the number of participants (n) is shown, and horizontal bars represent median GFAP, NfL or GFAP/NfL concentrations, which are shown on the base 2 logarithm scale. See Tables S27 and S28 for panels (a-c), and Table S29 for panels (d-f)
Fig. 6 Baseline NfL and GFAP/NfL ratio can discern FTLD-tau from FTLD-TDP pathology. a-c Comparisons of baseline GFAP (a), NfL (b), or GFAP/NfL (c) in the combined group of participants with a FTD syndrome with no mutation, or either a C9orf72, GRN, or MAPT mutation. p values are from analysis adjusted for age, sex and symptom duration. p < 0.0083 is considered statistically significant; ****p < 0.0001 (comparison to no mutation participants, “None”); ####p < 0.0001 (comparison to GRN mutation carriers). d-f Comparisons of GFAP (d), NfL (e), or GFAP/NfL (f) between neuropathologically confirmed participants with FTLD-tau or with FTLD-TDP. p values are from analysis adjusted for age at biomarker measurement, sex, symptom duration, and age at death. p < 0.025 is considered statistically significant; ****p < 0.0001. In all panels, the number of participants (n) is shown, and horizontal bars represent median GFAP, NfL or GFAP/NfL concentrations, which are shown on the base 2 logarithm scale. See Tables S27 and S28 for panels (a-c), and Table S29 for panels (d-f)CC BY · PubMed Central

Verification

Claims the paper makes

11 Valid · 0 Not valid · 3 Not checked · 117 Own results and interpretations

62 more claims · 2 with their cited sources checked

Create a free account to see all 131 claims and to verify the 1 reference still open.

See all claims — free account
  • Probably not a substantial claim. 22
  • It may state two separate claims as one. 46

Create a free account to see these claims.

Citations

Cites

Show all 73 references

Cited by

See all 20

Metadata

Authors
Sheth, U., Öijerstedt, L., Heckman, M. G., White, L. J., Heuer, H. W., Lario Lago, A., Forsberg, L. K., Faber, K. M., Foroud, T. M., Rademakers, R., Ramos, E. M., Appleby, B. S., Bozoki, A. C., Darby, R. R., Dickerson, B. C., Domoto-Reilly, K., Galasko, D. R., Ghoshal, N., Graff-Radford, N. R., Grant, I. M., Hales, C. M., Hsiung, G.-Y. R., Huey, E. D., Irwin, D., Kwan, J. Y., Litvan, I., Mackenzie, I. R., Masdeu, J. C., Mendez, M. F., Onyike, C. U., Pascual, B., Pressman, P. S., Roberson, E. D., Snyder, A., Tartaglia, M. C., Seeley, W. W., Dickson, D. W., Rosen, H. J., Boeve, B. F., Boxer, A. L., Petrucelli, L., & Gendron, T. F.
Affiliations
  • Mayo Clinic (US)
  • Mayo Clinic in Florida (US)
  • University of California, San Francisco (US)
  • University of California, Los Angeles (US)
  • Case Western Reserve University (US)
  • University of North Carolina at Chapel Hill (US)
  • Vanderbilt University (US)
  • Harvard University (US)
  • University of California System (US)
  • Washington University in St. Louis (US)
  • Northwestern University (US)
  • Emory University (US)
  • University of British Columbia (CA)
  • Brown University (US)
  • University of Pennsylvania (US)
  • National Institutes of Health (US)
  • Houston Methodist (US)
  • Johns Hopkins University (US)
  • Oregon Health & Science University (US)
  • University of Alabama at Birmingham (US)
  • University of Toronto (CA)
Corresponding author
Gendron, Tania F.
Venue
Molecular Neurodegeneration
Published
2025-03-12
Volume
20
Issue
1
Pages
30-30
ISSN
1750-1326 (ISSN-L)
OpenAlex
W4408362064
Cited by
20
References
75
Citations by year
2026: 8 · 2025: 12
FWCI
13.38
Type
journal article
Language
en
Open access
Licence
Locations
  • Molecular Neurodegeneration (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • IUScholarWorks (Indiana University) (submittedVersion, cc-by)
  • eScholarship (California Digital Library) (submittedVersion)
  • Digital Commons@Becker (Washington University School of Medicine) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Institute on Aging (U19AG063911, U01AG045390, U24AG021886, P30AG062677, P01AG019724, K24AG045333, P30AG066509)
  • National Institute of Neurological Disorders and Stroke (U54NS092089, R35NS097273, P01NS084974, P01NS099114)
  • Association for Frontotemporal Degeneration
Topics
  • Health Sciences › Medicine › Psychiatry and Mental health › Dementia and Cognitive Impairment Research
  • Health Sciences › Medicine › Clinical Neurology › Amyotrophic Lateral Sclerosis Research
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Alzheimer's disease research and treatments
Keywords
frontotemporal dementia, frontotemporal lobar degeneration, glial fibrillary acidic protein, neurofilament light chain, plasma biomarkers, presymptomatic mutation carriers, phenoconversion, tau pathology, TDP-43 pathology, disease progression, survival, prognostic factors, biomarkers of susceptibility, single molecule array, longitudinal biomarker changes
MeSH terms
Adult, Aged, Cross-Sectional Studies, Female, Glial Fibrillary Acidic Protein, Humans, Longitudinal Studies, Male, Middle Aged, Biomarkers, Neurofilament Proteins, Frontotemporal Dementia

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Comprehensive cross-sectional and longitudinal comparisons of plasma glial fibrillary acidic protein and neurofilament light across FTD spectrum disorders. https://app.cochat.ai/papers/10.1186/s13024-025-00821-4
The paper
Sheth, U., Öijerstedt, L., Heckman, M. G., White, L. J., Heuer, H. W., Lario Lago, A., Forsberg, L. K., Faber, K. M., Foroud, T. M., Rademakers, R., Ramos, E. M., Appleby, B. S., Bozoki, A. C., Darby, R. R., Dickerson, B. C., Domoto-Reilly, K., Galasko, D. R., Ghoshal, N., Graff-Radford, N. R., . . . Gendron, T. F. (2025). Comprehensive cross-sectional and longitudinal comparisons of plasma glial fibrillary acidic protein and neurofilament light across FTD spectrum disorders. Molecular Neurodegeneration. https://doi.org/10.1186/s13024-025-00821-4

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.