Blood markers p-tau217, NfL, and GFAP levels and their ability to diagnose Alzheimer's, frontotemporal dementia, and psychiatric disorders in a general memory clinic

Blood‐based biomarkers show great potential to transform diagnosis and clinical care of patients presenting with cognitive, psychiatric, and neurological symptoms. Their potential includes improving early and accurate diagnosis of Alzheimer's disease (AD) and resolving real‐world diagnostic challenges, such as distinguishing AD from other neurodegenerative disorders (NDs) like behavioral variant frontotemporal dementia (bvFTD) and from common mimics such as primary psychiatric disorders (PPDs).1, 2, 3, 4, 5, 6 Early, accurate diagnosis is increasingly important with the emergence of disease‐specific treatments, such as monoclonal antibodies targeting amyloid beta.

Two of the most promising biomarkers are plasma phosphorylated tau 217 (p‐tau217) and neurofilament light chain (NfL) protein. P‐tau217 shows strong diagnostic performance and specificity to distinguish AD from non‐AD disorders (including non‐AD dementias and PPD).7, 8, 9, 10 Within AD and control cohorts, plasma levels of p‐tau217 correlate with tau positron emission tomography (PET) pathology11 and tangle load at autopsy12 and are elevated in disorders with tangles but without amyloid plaques,13 supporting its reflection of tau phosphorylation and pathology. NfL, a non‐specific marker of neurodegeneration and acute neuronal injury, particularly in long myelinated axons, distinguishes neurodegenerative disorders (NDs) from ND mimics, including PPDs and non‐NDs,3, 4, 14, 15 and has utility in clinical settings.14, 16, 17, 18, 19 Glial fibrillary acidic protein (GFAP), a marker of astrocytic activation and neuroinflammation, is an additional biomarker of interest in a range of disorders.14, 20, 21, 22, 23, 24, 25, 26 However, the utility of GFAP for the clinical differentiation between AD, non‐AD NDs, and PPDs, especially when compared to p‐tau217 and NfL is not yet clear.

We previously published diagnostic performance, cut‐offs, and reference ranges for plasma p‐tau217 and NfL,10, 14, 18, 27, 28, 29 but we have not investigated their combined use. Furthermore, while we and others have published data showing the superiority of NfL compared to GFAP to distinguish bvFTD from PPDs,30, 31 we have not yet investigated all three, p‐tau217, NfL, and GFAP, in a younger real‐world clinical neuropsychiatric cohort with diverse NDs and PPDs. Some studies have investigated combinations of these markers. Benussi et al. investigated NfL, p‐tau217, NfL/p‐tau217 ratios, and other biomarkers in 374 participants (97 AD, 278 FTD).32 They found strong diagnostic performance of p‐tau217 and slightly superior performance of NfL/p‐tau217 to distinguish between AD and FTD. However, they lacked a PPD group, which is critical to appreciate the use of biomarkers for bvFTD diagnosis. Rousset et al. assessed p‐tau217 and NfL in an unselected memory clinic cohort, showing strong utility of p‐tau217 for AD diagnosis and describing biomarker profiles (high/low p‐tau217 and NfL) across clinical groups.17 To our knowledge, no study to date has jointly investigated p‐tau217, NfL, GFAP, and biomarker profiles, with a specific focus on AD, bvFTD, PPD, in younger people, where differential diagnoses, diagnostic uncertainty, misdiagnosis, and delay are greater.1, 33, 34, 35 Further research is needed to properly establish the roles of single and combination biomarkers in real‐world clinical settings with diverse neuropsychiatric cohorts.

The primary aim of this study was to compare levels and diagnostic performances of plasma p‐tau217, NfL, NfL/p‐tau217 ratio, and GFAP for distinguishing between AD, bvFTD, and PPDs. We hypothesized that combining p‐tau217 and NfL would improve diagnostic accuracy compared to either alone. We also aimed to describe p‐tau217/NfL biomarker profiles of AD pathology and neuronal injury (e.g., low AD/low neuronal injury, high AD/high neuronal injury) in an unselected population of patients from a neuropsychiatry memory clinic. We focused on p‐tau217 and NfL using our previously described cut‐off for plasma p‐tau21710 and age‐adjusted z‐score reference range models for plasma NfL.27, 30 We wanted to explore whether GFAP could improve any distinctions, for example, between bvFTD and PPDs. Although some studies reported age‐binned reference ranges for GFAP in serum36 and plasma,37 we are unaware of any studies that have described GFAP reference ranges using continuous modeling to derive precise age‐adjusted percentiles and z‐scores, as we did for NfL. As an exploratory aim, we applied a novel z‐score model for plasma GFAP to define biomarker profiles reflecting AD pathology, neuronal injury, and neuroinflammation (p‐tau217/NfL/GFAP).

The current study included participants prospectively recruited between June 2019 and April 2023 who had provided a blood sample for biomarker analysis. This is a follow‐up to our previous study cohort in which we investigated plasma and cerebrospinal fluid (CSF) NfL.14 In the current study, we included participants who had plasma p‐tau217, NfL, and GFAP biomarker results available. In the current study the focus was on AD, bvFTD, and PPD diagnostic groups, with comparator groups being control participants and other NDs (both described in the previous study). Data were available in the current study for additional comparator groups (not previously described): people with mild cognitive impairment (MCI) and people with presymptomatic genetic NDs. The bvFTD group included patients meeting diagnostic criteria for possible bvFTD. The participants in this study did not overlap with our previous study on NfL and GFAP in bvFTD, mood, and psychotic disorders.30

Participants were recruited from the Neuropsychiatry Centre at The Royal Melbourne Hospital, a quaternary service receiving referrals for diagnostically complex cases from primary care and other specialist services within Australia. Patients, as part of routine clinical care through the Neuropsychiatry Centre, received comprehensive multidisciplinary assessments and multimodal investigations, including CSF AD biomarker analysis, with gold standard consensus diagnosis based on established diagnostic criteria, as previously described in detail.14, 18, 28 Diagnostic group categorization was determined based on the most recent diagnosis, at longitudinal follow‐up, based on established diagnostic criteria, blinded to plasma biomarker levels, as previously described.14, 18, 28 Control participants were people recruited from the community. Cognitive screening data are still being collected and not yet available for this study; however, no control participants had symptoms or diagnoses of neurological or NDs and no active psychiatric symptoms or conditions.

EDTA plasma samples were collected during patients’ diagnostic work‐up and at first visit for community controls. Samples were stored at −80°C. Plasma NfL and GFAP were measured using N2PB kits on a Quanterix Single molecule array (Simoa) HD‐X analyzer, according to the manufacturer's instructions (Quanterix Corp., Billerica, MA, USA). Plasma p‐tau217 was measured using an in‐house University of Gothenburg (UGOT) p‐tau217 assay, as previously described in detail.38 The measurements were performed in one round of experiments using one batch of reagents by analysts blinded to clinical data and diagnoses, thereby reducing potential batch effects.

This study, part of the Markers in Neuropsychiatric Disorders Study (The MiND Study, https://themindstudy.org↗), was approved by the human research ethics committee at Melbourne Health (2016.038, 2017.090, 2018.371, 2020.142).

The final cohort consisted of 341 participants: 40 with AD (median age 62 years, 53% female), 15 with bvFTD (median age 57 years, 27% female), and 69 with PPDs (median age 55 years, 48% female). There were 119 controls (median age 63, 74% female), and in the additional comparator groups: 67 with other NDs, 13 with MCI, and 18 with presymptomatic genetic NDs (see Table 1 for full details).

PPDs consisted of major depressive disorder (MDD, n = 21), bipolar disorder (n = 6), functional neurological/cognitive disorder (n = 8), schizophrenia spectrum disorder (n = 16), bvFTD “phenocopy” syndrome (n = 2), and other PPDs (n = 16, which included anxiety, personality, obsessive‐compulsive, post‐traumatic stress, and undifferentiated psychiatric disorders).

Other NDs consisted of CJD (n = 2), dementia with Lewy bodies (n = 5), dementia not otherwise specified (n = 6), Huntington's disease (HD, n = 15), mixed AD/vascular (n = 3), substance‐related cognitive impairment/dementia (n = 2), vascular dementia (n = 5), and a range of other NDs (n = 29, which included autoimmune encephalitis, cerebral amyloid angiopathy, corticobasal syndrome, central nervous system vasculitis, Down syndrome, Fahr disease, metabolic disorders, Niemann–Pick Type C, Parkinson's disease, and cerebellar degenerative disorder). The presymptomatic genetic ND group consisted of AD (PSEN1, n = 2), genetic CJD (n = 6), HD (n = 6), CADASIL (n = 1), bvFTD (C9orf72n = 1 and GRNn = 2).

CSF biomarker analysis, including AD proteins CSF AB42 and p‐tau181 (Supporting Information Table 1), was conducted on 80 patients. Briefly, 27/40 (68%) of AD patients had CSF AD biomarkers. Most (18/27, 67%) had a CSF AD biomarker profile consistent with AD, as defined by an amyloid‐positive, p‐tau‐positive (A+T+) profile, categorized using established cut‐offs, as described in our previous study.10 Other profiles in the AD group were 8/27 with A+T− and 1/27 with A−T+. Two patients with AD had amyloid PET (but not CSF), and both were positive for amyloid plaques. Further information, including phenotype information for AD, are in the Supporting Information Tables 1 and 2. Forty‐seven percent (7/15) of bvFTD patients had CSF AD biomarkers (none with an A+T+ profile, 4/7 with A−T−, and 3/7 with A+T−). None of the patients in the other groups who had CSF AD biomarkers (20 in the PPD group, 21 in other NDs, and five MCI) had A+T+ profiles (further details in Supporting Information Tables 1 and 2).

We investigated multiple biomarkers in an unselected real‐world clinical cohort of younger participants seen in a neuropsychiatry memory clinic, focusing on AD, bvFTD, and PPDs. The main findings were as follows: (1) very strong diagnostic performance of plasma p‐tau217 to distinguish AD from bvFTD, and AD from PPDs; (2) superior performance of p‐tau217 for AD diagnosis on its own, compared to NfL/p‐tau217 ratio, NfL, and GFAP; (3) strongest performance of NfL for distinguishing bvFTD from PPDs, and all NDs from PPDs, consistent with previous studies.30, 31 This adds important further evidence to the limited literature so far on roles for different biomarkers in a real‐world clinical cohort, with p‐tau217 being a very sensitive and specific test for AD and NfL having the strongest diagnostic performance to distinguish neurodegeneration from PPDs/non‐NDs, with little role for GFAP in any of these distinctions. The key strengths of this study included a real‐world, unselected diverse, clinical, and younger cohort – a relatively underinvestigated group. Establishing the role for these biomarkers in younger people, who face greater rates of misdiagnosis and diagnostic uncertainty, is important for increasing clinical and research access, where arguably the clinical, treatment, and broader psychosocial benefits could be greater.

This study builds on our previous work on p‐tau217,10 providing further evidence of very strong diagnostic performance for some of the most common diagnostic distinctions, distinguishing AD from bvFTD, and AD from PPDs. We saw very high specificity and sensitivity (91% and 95%), establishing a role for such an accurate test (93%) in younger people. We found no benefit of NfL/p‐tau217 ratio over p‐tau217 alone in terms of improving the distinction between AD and bvFTD, and bvFTD from PPD, in contrast to Benussi et al.32 This difference in their study was slight and may reflect their older cohort and greater FTD sample sizes. Although there has been interest in GFAP assisting with AD versus non‐AD and ND versus PPD/non‐ND distinctions, we found little to no utility for GFAP. Importantly, while different p‐tau217 assays correlate well with each other, they have not been standardized with each other and thus differ in absolute concentrations.9 Assay standardization projects are under way; however, at present cut points must be regarded as assay‐specific and, on the very safe side, as lab‐specific.

An additional strength of the study was defining and describing biomarker profiles using previously published cut‐offs and reference ranges.10, 27, 40 Although the previous p‐tau217 study and cut‐off had some overlap with this study cohort, a new separate analysis of AD versus controls in this study yielded a very similar cut‐off of (2.18 pg/mL, Supporting Information Table 3). The NfL reference range was from a completely independent, large reference cohort.27, 40 We found higher p‐tau217 diagnostic performance (AUC 0.97 for AD vs PPD, and 0.98 for AD vs controls, sensitivities and specificities, and accuracies > 90%), compared to Rousset et al. (AUCs approximately 0.78 to 0.81 depending on the group being compared to AD),17 and we also found greater proportions of high p‐tau217/high NfL and high p‐tau217/low NfL profiles in our AD group. Our findings may reflect our younger AD group (median 62 years of age vs 66). Although most patients with PPDs had low p‐tau217/low NfL profiles, there were still a number with other profiles, including 25% with high NfL. This builds on numerous other studies that demonstrated that, although very elevated NfL levels are rarely seen in PPDs, slightly elevated levels can be seen in numerous PPDs, and PPDs cannot be assumed to be equivalent to controls/non‐neurological controls.3, 4, 25, 26, 27, 30, 41 Although some interesting patterns could be observed between individual PPDs (Supporting Information Figures 1 and 2), any confident interpretations are limited by the small sample sizes, and deeper investigations in larger cohorts are warranted.

Among AD, bvFTD, and other ND, we found the greatest proportion (47%) of “normal” AD/neuronal injury biomarker profile (Low p‐tau217/Low NfL) in bvFTD. This was reduced slightly when considering “normal” AD/neuronal injury/neuroinflammation (low p‐tau217/low NfL/low GFAP) profile (40%). This could reflect issues previously described with NfL technical/analytical/reference range issues10, 42 or, more likely, highlights greater complexity and clinical and biomarker heterogeneity, seen in bvFTD.43 Although p‐tau217 can accurately distinguish frontal AD from bvFTD, and NfL has strong diagnostic performance for bvFTD versus PPDs at group levels, our findings suggest that many individuals with bvFTD would still have non‐diagnostic biomarker levels and profiles. We found weaker diagnostic performance of NfL for bvFTD versus PPD distinction in this study compared to one of our previous studies that compared bvFTD to PPD27 and other studies.41 Patients in this study were from within the same unselected clinical cohort, whereas those other studies had patients from separate clinics/cohorts. Therefore, differences in diagnostic performance most likely reflect differences in cohorts, with the present study reflecting the performance, and limitations and ongoing challenges, in a real‐world, unselected clinical cohort. Studies are needed and under way to more deeply understand this discordance in larger groups of people with bvFTD in real‐world clinical cohorts, including multimodal neuroimaging, clinical, multi‐omic, and genomic data, to ultimately improve upon single diagnostic biomarker performance for more accurate diagnosis of bvFTD. Describing biomarker profiles were secondary, exploratory aims of this study. A limitation was binary/dichotomous “low”/“high” classification. Binary classification loses important nuance in interpreting an individual level (e.g., an individual with level 1% below the binary cut‐off or 95th percentile, and someone with a level 50% below the cut‐off or 95th percentile, would both be classified as “low”). Unfortunately, small sample sizes prevented more nuanced categorization (e.g., “high,” “low,” and “borderline”).

There was a very high p‐tau217 outlier in the bvFTD group. They had a complex history, initially diagnosed with AD based on memory impairment and neuroimaging. They subsequently developed significant personality and behavior change and eventually developed motor neuron disease (MND). At that point, the diagnosis was revised to bvFTD‐MND. Notably, their blood sample and biomarker level was prior to MND onset. Their elevated p‐tau217 level possibly reflected skeletal muscle source of p‐tau217,44 or it could suggest that they were unfortunate to have multiple pathologies, such as AD, as well as MND, and possibly FTD. Unfortunately, this patient did not have genetic testing, CSF AD protein analysis, or amyloid PET. It is possible that blood‐based biomarkers could help improve not only accurate AD diagnosis but also identification of cases of multiple co‐pathologies, especially important for older people, where co‐pathology is more common, and in this era of increasingly available disease‐specific treatments.45

The study's limitations include the single‐center, cross‐sectional design, relatively small FTD subgroup, and small AD variant subgroups preventing deeper analyses, a lack of gold standard reference tests for AD (CSF AD proteins and amyloid PET) for all participants, reliance on diagnostic categorization based on gold standard clinical diagnoses over biomarker profiles (and thus the range of CSF profiles, such as an A−T+ CSF profile for one patient with an AD diagnosis, although their CSF AB42 was borderline), and lack of genetic or post mortem diagnostic confirmation. Additional exploratory analyses indicated that cognitive severity did not differ significantly between AD and bvFTD and was not significantly associated with plasma biomarkers within AD, suggesting that observed biomarker patterns were unlikely to be driven by disease stage; however, these findings need to be interpreted with caution given the small sample sizes. Other NDs, MCI, and presymptomatic genetic NDs were included as comparator groups for secondary/exploratory/descriptive analyses, but their small sample sizes limited further in‐depth analyses or interpretations. While all cases had systematic longitudinal diagnostic data collection, it was beyond the scope of this study to systematically review all cases with discordant biomarkers (e.g., high NfL in PPD), for reasons of discordance and potential misclassification. Such detailed investigations will be required in future studies. Our control group relied on self‐reports since cognitive screening data on controls were not available for this study. While we could not completely rule out the possibility of some control participants with MCI due to this significant limitation, the relatively younger age, narrow range, and mainly normal biomarker levels in this group suggest that this is very unlikely to be a significant issue. We developed a novel GAMLSS plasma GFAP model. While we found a U‐shaped curve (higher levels in younger people, lower in midlife, then increasing again with age), similar to other studies,36, 37 our model was based on a relatively small number of individuals and therefore should be interpreted with caution. Studies are under way aiming to replicate this study's findings and to investigate the incremental validity of multiple biomarkers in a larger sample size, including imaging and other multi‐omic data, and in different clinical settings.

This study adds important evidence for the very strong diagnostic performance of p‐tau217 to distinguish AD from bvFTD and PPDs and identified important future research needs to improve bvFTD diagnosis. The strength and simplicity of single blood‐based biomarkers for AD diagnosis are particularly important in the era of increasing availability of anti‐amyloid and other disease‐specific treatments for AD. Ongoing further research will further establish the role of p‐tau217 and NfL in precision diagnostic algorithms to reduce misdiagnosis and diagnostic delay and improve clinical and research outcomes.

The authors declare no conflicts of interest. Author disclosures are available in the. supporting information

KB has served as a consultant and at advisory boards for Abbvie, AC Immune, ALZPath, AriBio, Beckman‐Coulter, BioArctic, Biogen, Eisai, Lilly, Moleac Pte. Ltd, Neurimmune, Novartis, Ono Pharma, Prothena, Quanterix, Roche Diagnostics, Sunbird Bio, Sanofi, and Siemens Healthineers; has served at data monitoring committees for Julius Clinical and Novartis; has given lectures, produced educational materials and participated in educational programs for AC Immune, Biogen, Celdara Medical, Eisai, and Roche Diagnostics; and is a co‐founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program, outside the work presented in this paper. HZ has served on scientific advisory boards and/or as a consultant for Abbvie, Acumen, Alector, Alzinova, ALZpath, Amylyx, Annexon, Apellis, Artery Therapeutics, AZTherapies, Cognito Therapeutics, CogRx, Denali, Eisai, Enigma, LabCorp, Merck Sharp & Dohme, Merry Life, Nervgen, Novo Nordisk, Optoceutics, Passage Bio, Pinteon Therapeutics, Prothena, Quanterix, Red Abbey Labs, reMYND, Roche, Samumed, ScandiBio Therapeutics AB, Siemens Healthineers, Triplet Therapeutics, and Wave; has given lectures sponsored by Alzecure, BioArctic, Biogen, Cellectricon, Fujirebio, LabCorp, Lilly, Novo Nordisk, Oy Medix Biochemica AB, Roche, and WebMD; is a co‐founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program; and is a shareholder of MicThera (outside submitted work). The other authors have nothing to disclose.

All human subjects provided informed consent.

The authors would like to thank all the patients, participants, and their families for their participation. KB is supported by the Swedish Research Council (2017‐00915 and 2022‐00732), the Swedish Alzheimer Foundation (AF‐930351, AF‐939721, AF‐968270, and AF‐994551), Hjärnfonden, Sweden (ALZ2022‐0006, FO2024‐0048‐TK‐130, and FO2024‐0048‐HK‐24), the Swedish state under the agreement between the Swedish government and the County Councils, the ALF agreement (ALFGBG‐965240 and ALFGBG‐1006418), the European Union Joint Programme for Neurodegenerative Disorders (JPND2019‐466‐236), the Alzheimer's Association 2021 Zenith Award (ZEN‐21‐848495), the Alzheimer's Association 2022‐2025 Grant (SG‐23‐1038904 QC), La Fondation Recherche Alzheimer (FRA), Paris, France, the Kirsten and Freddy Johansen Foundation, Copenhagen, Denmark, Familjen Rönströms Stiftelse, Stockholm, Sweden, and an anonymous philanthropist and donor. HZ is a Wallenberg Scholar and a Distinguished Professor at the Swedish Research Council supported by grants from the Swedish Research Council (2023‐00356, 2022‐01018, and 2019‐02397), the European Union's Horizon Europe research and innovation programme under Grant Agreement 101053962, Swedish State Support for Clinical Research (ALFGBG‐71320), the Alzheimer Drug Discovery Foundation (ADDF), USA (201809‐2016862), the AD Strategic Fund and the Alzheimer's Association (ADSF‐21‐831376‐C, ADSF‐21‐831381‐C, ADSF‐21‐831377‐C, and ADSF‐24‐1284328‐C), the European Partnership on Metrology, co‐financed by the European Union's Horizon Europe Research and Innovation Programme and by the Participating States (NEuroBioStand, 22HLT07), the Bluefield Project, Cure Alzheimer's Fund, the Olav Thon Foundation, the Erling–Persson Family Foundation, Familjen Rönströms Stiftelse, Familjen Beiglers Stiftelse, Stiftelsen för Gamla Tjänarinnor, Hjärnfonden, Sweden (FO2022‐0270), the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska‐Curie Grant Agreement 860197 (MIRIADE), the European Union Joint Programme–Neurodegenerative Disease Research (JPND2021‐00694), the National Institute for Health and Care Research University College London Hospitals Biomedical Research Centre, the UK Dementia Research Institute at UCL (UKDRI‐1003), and an anonymous donor. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. AFS is primarily funded by the Swedish federal government under the ALF agreement (ALF 2022 YF 0017) and The Åke Wiberg Foundation.

Open access publishing facilitated by The University of Melbourne, as part of the Wiley ‐ The University of Melbourne agreement via the Council of Australian University Librarians.

Eratne D, Kang M, Malpas CB, et al. Plasma p‐tau217, NfL, GFAP diagnostic performance and biomarker profiles in Alzheimer's disease, frontotemporal dementia, and psychiatric disorders, in a prospective unselected neuropsychiatry memory clinic. Alzheimer's Dement. 2025;21:e70717. 10.1002/alz.70717