Skip to content

Fluid biomarkers

TL;DR — Plasma p-tau217 has moved AD diagnosis out of specialist centres. In a meta-analysis of 113 studies and 29,625 individuals, p-tau217 outperformed every other phosphorylated-tau species against biological reference standards: pooled sensitivity 88.1% (95% CI 86.7–89.5), specificity 88.7% (87.4–89.9), AUROC 91.1% (88.9–92.4), diagnostic odds ratio 50.7 (40.6–63.4), versus AUROC 81.5% for p-tau181 (Therriault 2025, PMID 40818474). Prospectively in Swedish primary care, a mass-spectrometry %p-tau217 plus Aβ42:Aβ40 score achieved AUC 0.96–0.97 with PPV and NPV of 88–92%, against primary-care physicians' 61% (95% CI 53–69) diagnostic accuracy after examination, cognitive testing and CT — the specialists' own accuracy was 73% (68–79) (Palmqvist 2024, PMID 39068545). Assays are not interchangeable: in a 998-participant head-to-head, mass-spectrometry %p-tau217 reached 0.93 accuracy for amyloid-PET status versus 0.83–0.88 for immunoassays, and the authors propose %p-tau217 as a stand-alone confirmatory test while some immunoassays are better used as triage with confirmation (Warmenhoven 2025, PMID 39468767). The 2024 criteria promoted "accurate plasma biomarkers, especially p-tau217" to Core 1 status, sufficient to establish a diagnosis (Jack 2024, PMID 38934362), while the 2022 appropriate-use recommendations — which predate the newest data — advised confirming blood results with CSF or PET and explicitly cautioned against primary-care use (Hansson 2022, PMID 35908251). That gap between what the assays can do and what guidance permits is the live implementation question.

What the fluid markers measure

Marker Compartment What it indexes Direction in AD
Aβ42/Aβ40 ratio CSF, plasma Amyloid deposition (Aβ42 sequestered into plaque) Decreased
p-tau217, p-tau181, p-tau231, p-tau205, p-tau212 CSF, plasma AD-type tau phosphorylation; rises early, driven by amyloid Increased
Total tau, brain-derived tau CSF, plasma Neuronal injury (t-tau non-specific) Increased
Neurofilament light (NfL) CSF, plasma, serum Axonal degeneration, disease-non-specific Increased
GFAP Plasma, CSF Astrocytic reactivity Increased
YKL-40 CSF Glial inflammation Increased
sPDGFRβ CSF Pericyte injury / BBB breakdown Increased (see vascular contributions)

The 2024 criteria's split matters operationally: Core 1 — amyloid PET, approved CSF assays and accurate plasma assays, especially p-tau217 — is diagnostic; Core 2 — tau PET and later-changing biofluid markers — is prognostic (Jack 2024, PMID 38934362). The reason p-tau217 counts as Core 1 despite being a tau marker is that it rises as a consequence of amyloid pathology, tracking Alzheimer's neuropathologic change generally rather than tangle burden specifically.

The temporal ordering comes from dominantly inherited disease: CSF Aβ42 falls about 25 years before expected symptom onset and CSF tau rises about 15 years before (Bateman 2012, PMID 22784036) — see genetics.

CSF: use the ratio, not the single analyte

The diagnostic performance of the core CSF triad was already large before plasma assays existed. A meta-analysis of 231 articles (15,699 AD patients, 13,018 controls) found fold-changes versus controls of 2.54 (95% CI 2.44–2.64) for CSF t-tau, 1.88 (1.79–1.97) for p-tau and 0.56 (0.55–0.58) for Aβ42, with similar separation of MCI-due-to-AD from stable MCI (Aβ42 0.67, p-tau 1.72, t-tau 1.76); CSF NfL (2.35, 1.90–2.91) and plasma t-tau (1.95, 1.12–3.38) also discriminated, while plasma Aβ42 and Aβ40 did not (Olsson 2016, PMID 27068280). That is the CSF reference the blood tests are now matching.

In 103 memory-clinic patients with known amyloid-PET status using automated immunoassays, sensitivity/specificity versus amyloid PET were 0.93/0.57 for Aβ42 alone, 0.96/0.69 for p-tau181/Aβ42, 0.92/0.69 for t-tau/Aβ42 and 0.94/0.82 for Aβ42/40; AUCs were 0.78 (95% CI 0.68–0.88) for Aβ42 alone versus 0.90 (0.83–0.97) for Aβ42/40 (Amft 2022, PMID 35473631). The mechanism is straightforward — the ratio normalises for individual differences in total Aβ production and for pre-analytical peptide adsorption — and it is why every current criteria set specifies the ratio. Limitations of that study: n=103, no pre-defined Aβ42/40 cut-off, and a pre-analytical protocol departing from the manufacturer's method sheet.

Plasma p-tau217: the accuracy record

Study Design Key result
Palmqvist 2020 (PMID 32722745) Three cohorts, 1,402 participants: Arizona neuropathology (n=81), BioFINDER-2 (n=699), Colombian PSEN1 E280A kindred (n=622) AUC 0.89 (95% CI 0.81–0.97) for neuropathologically defined AD vs non-AD, exceeding p-tau181 and NfL (AUC 0.50–0.72, P<0.05); AUC 0.96 (0.93–0.98) for AD dementia vs other neurodegenerative disease, not significantly different from CSF p-tau217, CSF p-tau181 or tau-PET; AUC 0.93 (0.91–0.96) for abnormal vs normal tau-PET; elevated in PSEN1 carriers from ~age 25, about 20 years before expected MCI
Therriault 2025 (PMID 40818474) Meta-analysis, 113 studies, 29,625 individuals p-tau217: sensitivity 88.1% (86.7–89.5), specificity 88.7% (87.4–89.9), AUROC 91.1% (88.9–92.4), DOR 50.7 (40.6–63.4). p-tau212: AUROC 90.3%; p-tau205: 85.1%; p-tau181: 81.5%; p-tau231: 80.2%. ~90% of studies rated high risk of bias for not using predefined or externally derived thresholds
Wang 2025 (PMID 40156286) Lumipulse plasma p-tau217/Aβ42 ratio; clinic cohort n=391, community cohort n=121 AUC 0.963–0.966 vs amyloid PET and 0.947–0.974 vs tau PET; clinically equivalent to CSF p-tau181/Aβ42 and Aβ42/40 and better than plasma p-tau217, Aβ42/40, p-tau181 or p-tau181/Aβ42 alone; two-cutoff approach improved specificity without reducing sensitivity; intermediate zone smaller for the ratio than for p-tau217 alone in both clinic (10.6% vs 13.0%) and community (16.5% vs 31.4%) settings
Palmqvist 2024 (PMID 39068545) 1,213 patients with cognitive symptoms, Sweden 2020–2024; predefined cutoffs from an independent cohort; batch and prospective (biweekly) analysis in primary and secondary care APS2 (mass-spec %p-tau217 + plasma Aβ42:Aβ40): primary care batched AUC 0.97 (0.95–0.99), PPV 91% (87–96), NPV 92% (87–96); secondary care batched AUC 0.96 (0.94–0.98); prospective primary care AUC 0.96 (0.94–0.98), PPV 88%, NPV 90%; prospective secondary care AUC 0.97, PPV 91%, NPV 91%; diagnostic accuracy 88–92% across all four cohorts; %p-tau217 alone was equivalent to APS2 (both 90%)
Warmenhoven 2025 (PMID 39468767) 998 BioFINDER-2 participants, five p-tau217 tests head-to-head, external replication n=219 All tests: AUC 0.91–0.96 for abnormal Aβ-PET, 0.94–0.97 for tau-PET. Mass-spec %p-tau217: accuracy 0.93, sensitivity 0.91, specificity 0.94 for Aβ-PET status vs immunoassay accuracy 0.83–0.88, sensitivity 0.84–0.87, specificity 0.85–0.89 (P<0.007). Among immunoassays, Lilly and ALZpath > Janssen for Aβ-PET (P<0.006); Lilly > ALZpath for tau-PET (P=0.025)
Barthélemy 2024 (PMID 38382645) Mass-spec plasma %p-tau217 vs FDA-approved CSF immunoassays; BioFINDER-2 n=1,422, Knight ADRC n=337 Plasma %p-tau217 AUC 0.95–0.97 for Aβ-PET, equivalent to CSF Aβ42/40 and p-tau181/Aβ42; generally superior to CSF for tau-PET (AUC 0.95–0.98). In cognitively impaired subcohorts, accuracy/PPV/NPV 89–90% for Aβ-PET and 87–88% for tau-PET, rising to 95% with two cutoffs
Brum 2023 (PMID 37653254) Two-step workflow, n=348 MCI from BioFINDER-1/2; plasma p-tau217 + age + APOE ε4, then CSF only for the intermediate band Step-1 AUC 89.3% (derivation) / 94.3% (validation). Overall accuracy for Aβ-PET 88.2–92.0% depending on threshold stringency, reducing necessary CSF tests by 61–86%
Karikari 2020 (PMID 32333900) Ultrasensitive plasma p-tau181 immunoassay, 1,131 individuals across four cohorts Distinguished AD dementia from Aβ-negative older controls (AUC 90.21–98.24) and from FTD, vascular dementia, PSP/CBS and PD/MSA; associated with tau-PET (AUC 83–93) and amyloid-PET (76–88) and with 1-year cognitive decline and hippocampal atrophy. Weaker in primary-care MCI vs AD (AUC 55%) — the assay that p-tau217 subsequently outperformed
Salvadó 2026 (PMID 42189519) Plasma %p-tau217 + eMTBR-tau243 staging vs PET-based AA stages; BioFINDER-2 n=872, Knight ADRC n=156 C-index 0.91 (0.90–0.92) vs PET stages and 0.84 vs clinical stage; replicated (C-index 0.91 and 0.86); autopsy ADNC AUC 0.96 (0.91–1.00). Adding eMTBR-tau243 improved intermediate (A+TMOD+) classification versus %p-tau217 alone

The clinician-versus-test comparison

The most policy-relevant number in this literature is not an AUC. In Palmqvist's prospective cohorts, primary-care physicians achieved 61% (95% CI 53–69) diagnostic accuracy for clinical AD after clinical examination, cognitive testing and a CT scan, rising to 91% (86–96) with the APS2. Dementia specialists achieved 73% (68–79) versus 91% (88–95) with the APS2 (PMID 39068545). Both figures are from a research protocol in a single country with predefined cutoffs and a single laboratory, so they bound what is achievable rather than what will be achieved.

Where the numbers break

Threat Evidence Practical implication
Assay heterogeneity Mass-spec %p-tau217 accuracy 0.93 vs immunoassays 0.83–0.88 for Aβ-PET status (PMID 39468767) A "p-tau217 result" is not a portable quantity; cutoffs are assay-specific
Threshold derivation ~90% of studies in the meta-analysis were high risk of bias for not using predefined/external thresholds (PMID 40818474) Published accuracy is optimistic; prospective external-cutoff studies (PMID 39068545) are the relevant evidence
Intermediate zone 10.6–16.5% (ratio) and 13.0–31.4% (p-tau217 alone) fell in the intermediate range, larger in the community cohort (PMID 40156286) A meaningful minority need confirmatory CSF/PET regardless of assay quality
Borderline results in treatment decisions 10% of DMT candidates had borderline p-tau217 (0.273–0.399 pg/mL) requiring confirmation; externally derived cutoffs had lower specificity, risking treatment of Aβ-negative patients (Howe 2024, PMID 38971815) Two-cutoff strategies are needed where the consequence is an infusion
Kidney function Biomarker concentrations were highest at CKD stage 3; associations with eGFR persisted after adjustment for Aβ40, Aβ42, NfL and GFAP but not for p-tau217; adding eGFR did not improve prediction of Aβ positivity for any biomarker (Arslan 2025, PMID 39972340) Within normal-to-mild renal impairment, eGFR adjustment is not required for p-tau217
Population representativeness In 1,170 memory-clinic patients from 91 countries, p-tau217 concentrations and diagnostic accuracy did not differ by ethnicity, though the authors note comorbidities affecting plasma markers may disproportionately affect minoritised groups (Kjaergaard 2025, PMID 40501109) Reassuring for the assay; not reassuring for the comorbidity burden around it
Implementation infrastructure Pre-analytical and analytical standardisation, confounder characterisation, clinician education and clear guidelines are all identified as unmet prerequisites, especially for primary care and any population screening (Schöll 2024, PMID 39369727) Accuracy is necessary but not sufficient for deployment

The non-specific markers

Neurofilament light indexes axonal degeneration in any disease. Review of the cross-sectional and longitudinal literature positions blood NfL as a strong monitoring biomarker — reflecting current severity of atrophy, hypometabolism and white-matter integrity loss in AD-typical regions — and as a prognostic and susceptibility biomarker predicting subsequent structural and functional change, including in cognitively unimpaired people with elevated amyloid. It is not a diagnostic biomarker for AD, because it is elevated across neurodegenerative and traumatic conditions (Jung 2024, PMID 37540027).

GFAP and YKL-40 are positioned inside the preclinical cascade rather than downstream of it: in 384 cognitively unimpaired ALFA+ participants, plasma GFAP mediated the relationship between CSF Aβ42/40 and amyloid-PET, and CSF YKL-40 partly explained the paths from amyloid to p-tau181 and from p-tau181 to NfL (Pelkmans 2024, PMID 37690071). See neuroinflammation and glia.

Assay identity is part of the result

“Plasma p-tau217” is not a single interchangeable test. In 392 participants, automated Lumipulse and ALZpath Simoa assays had nearly identical discrimination of AD from other neurodegenerative disease (AUC 0.952, 95% CI 0.927–0.978 versus 0.955, 0.928–0.982) and from healthy controls (0.938 versus 0.937) (Pilotto 2025, PMID 39679606). But a ten-assay comparison in 135 people with MCI found mass-spectrometry p-tau217 best for both abnormal amyloid (AUC 0.947) and progression to AD dementia (AUC 0.932); immunoassay AUCs ranged from 0.642 to 0.889, and plasma–CSF correlations from 0.320 to 0.891 (Janelidze 2023, PMID 36087307). A commercial head-to-head study likewise found p-tau217 measures strongest across amyloid PET, tau PET and cortical-thickness outcomes, while plasma Aβ42/40 classified amyloid relatively poorly (Schindler 2024, PMID 39394841).

The evidence base behind the 2025 guideline makes the uncertainty explicit. Across 49 observational studies and 31 tests, pooled sensitivity ranged from 49.3% (95% CI 41.2–57.4) to 91.4% (86.6–94.6), specificity from 61.5% (45.6–75.3) to 96.7% (87.8–99.2), and GRADE certainty from moderate to very low; most studies had high risk of bias in patient selection, index-test conduct or reference standard (Pahlke 2025, PMID 41193403). Reporting only the analyte, without platform, cut-point, setting and pre-test probability, is therefore incomplete.

Markers outside the AT core add biology rather than diagnostic specificity. In AIBL, lower plasma Aβ42/40 and higher p-tau181, GFAP and NfL predicted prospective cognitive decline over 7–10 years, while the combination of Aβ42/40, p-tau181 and GFAP equalled or exceeded any single marker for amyloid status (Chatterjee 2023, PMID 36574591). These associations do not turn GFAP or NfL into AD-specific diagnostic tests.

Guidance versus capability

The Alzheimer's Association appropriate-use recommendations for blood biomarkers (2022) permitted use as pre-screeners for trial enrolment with PET/CSF confirmation; encouraged longitudinal measurement within trials; advised against use as primary endpoints in pivotal trials; recommended cautious use in specialised memory clinics with confirmation wherever possible; and stated that more data were needed before stand-alone diagnostic use or any primary-care use (Hansson 2022, PMID 35908251). The 2024 diagnostic criteria then declared an abnormal Core 1 biomarker — explicitly including accurate plasma assays — sufficient to establish a diagnosis (Jack 2024, PMID 38934362), and the 2024 primary-care validation data appeared after both documents. The 2025 GRADE clinical practice guideline then set performance-based rules for specialised care: ≥90% sensitivity and ≥75% specificity as a triage test; ≥90% sensitivity and ≥90% specificity as a substitute for amyloid PET or CSF in cognitively impaired patients — with the explicit caution that many commercial tests do not meet those thresholds on a single cutoff (Palmqvist 2025, PMID 40729527).

This knowledge base records that as an unresolved conflict rather than adjudicating it: the 2022 recommendations were written before the prospective primary-care evidence, the 2024 criteria are not a workflow guideline, and the 2025 CPG is scoped to specialised care. What follows from the data is narrower than any of the three documents: a mass-spectrometry %p-tau217-based score with predefined cutoffs, run in a laboratory that established those cutoffs, performs at ~90% accuracy in primary care and is clinically equivalent to FDA-approved CSF tests (PMID 39068545; PMID 38382645); a two-step model can reserve CSF/PET for the intermediate band, but accuracy and test-avoidance trade against each other — lenient step-1 thresholds avoided 85.9% of CSF tests at 88.2% accuracy, stringent ones avoided 61.2% at 92.0% (Brum 2023, PMID 37653254). Whether that transfers to other assays, other laboratories, other health systems and asymptomatic populations is untested. See guidelines.

Open questions

  • Do prospective accuracy results transfer from a single Swedish laboratory with predefined cutoffs to routine multi-laboratory practice in other health systems (Palmqvist 2024, PMID 39068545; Schöll 2024, PMID 39369727)?
  • What should be done with the 10–30% of results in the intermediate zone, and does that fraction shrink with better assays or is it intrinsic to a continuous biology (Wang 2025, PMID 40156286)?
  • Should immunoassay-based p-tau217 be restricted to triage with mandatory confirmation, as the head-to-head data imply (Warmenhoven 2025, PMID 39468767)?
  • What is the performance of plasma p-tau217 in asymptomatic people — the population any screening programme would test — as distinct from people presenting with cognitive symptoms (Therriault 2025, PMID 40818474)?
  • Given that ~90% of published studies used data-derived thresholds, how much of the reported accuracy is optimism, and what would a fully prespecified replication show (PMID 40818474)?
  • Do multimorbidity patterns that differ by population change cutoffs even when ethnicity itself does not affect assay performance (Kjaergaard 2025, PMID 40501109; Arslan 2025, PMID 39972340)?
  • Can plasma markers serve as trial endpoints, or does the 2022 prohibition still hold given target-engagement effects on p-tau217 (Hansson 2022, PMID 35908251)?
  • Can plasma %p-tau217 plus eMTBR-tau243 replace tau PET for biological staging, given C-index 0.91 versus PET stages and 0.96 AUC versus autopsy ADNC in research cohorts (Salvadó 2026, PMID 42189519)?

References

  1. Therriault J, et al. Blood phosphorylated tau for the diagnosis of Alzheimer's disease: a systematic review and meta-analysis. Lancet Neurol. 2025;24:740-752. PMID 40818474.
  2. Palmqvist S, et al. Discriminative accuracy of plasma phospho-tau217 for Alzheimer disease vs other neurodegenerative disorders. JAMA. 2020;324:772-781. PMID 32722745.
  3. Palmqvist S, et al. Blood biomarkers to detect Alzheimer disease in primary care and secondary care. JAMA. 2024;332:1245-1257. PMID 39068545.
  4. Wang J, et al. Diagnostic accuracy of plasma p-tau217/Aβ42 for Alzheimer's disease in clinical and community cohorts. Alzheimers Dement. 2025;21:e70038. PMID 40156286.
  5. Warmenhoven N, et al. A comprehensive head-to-head comparison of key plasma phosphorylated tau 217 biomarker tests. Brain. 2025;148:416-431. PMID 39468767.
  6. Hansson O, et al. The Alzheimer's Association appropriate use recommendations for blood biomarkers in Alzheimer's disease. Alzheimers Dement. 2022;18:2669-2686. PMID 35908251.
  7. Amft M, et al. The cerebrospinal fluid biomarker ratio Aβ42/40 identifies amyloid positron emission tomography positivity better than Aβ42 alone in a heterogeneous memory clinic cohort. Alzheimers Res Ther. 2022;14:60. PMID 35473631.
  8. Schöll M, et al. Challenges in the practical implementation of blood biomarkers for Alzheimer's disease. Lancet Healthy Longev. 2024;5:100630. PMID 39369727.
  9. Arslan B, et al. The impact of kidney function on Alzheimer's disease blood biomarkers: implications for predicting amyloid-β positivity. Alzheimers Res Ther. 2025;17:48. PMID 39972340.
  10. Kjaergaard D, et al. Influence of ethnicity on the diagnostic accuracy of plasma p-tau217 for the identification of Alzheimer's disease: a consecutive mixed memory clinic cohort study. Alzheimers Dement. 2025;21:e70315. PMID 40501109.
  11. Jung Y, et al. The potential of blood neurofilament light as a marker of neurodegeneration for Alzheimer's disease. Brain. 2024;147:12-25. PMID 37540027.
  12. Pelkmans W, et al. Astrocyte biomarkers GFAP and YKL-40 mediate early Alzheimer's disease progression. Alzheimers Dement. 2024;20:483-493. PMID 37690071.
  13. Jack CR, et al. Revised criteria for diagnosis and staging of Alzheimer's disease: Alzheimer's Association Workgroup. Alzheimers Dement. 2024;20:5143-5169. PMID 38934362.
  14. Bateman RJ, et al. Clinical and biomarker changes in dominantly inherited Alzheimer's disease. N Engl J Med. 2012;367:795-804. PMID 22784036.
  15. Howe MD, et al. Clinical application of plasma P-tau217 to assess eligibility for amyloid-lowering immunotherapy in memory clinic patients with early Alzheimer's disease. Alzheimers Res Ther. 2024;16:154. PMID 38971815.
  16. Olsson B, et al. CSF and blood biomarkers for the diagnosis of Alzheimer's disease: a systematic review and meta-analysis. Lancet Neurol. 2016;15:673-684. PMID 27068280.
  17. Barthélemy NR, et al. Highly accurate blood test for Alzheimer's disease is similar or superior to clinical cerebrospinal fluid tests. Nat Med. 2024;30:1085-1095. PMID 38382645.
  18. Brum WS, et al. A two-step workflow based on plasma p-tau217 to screen for amyloid β positivity with further confirmatory testing only in uncertain cases. Nat Aging. 2023;3:1079-1090. PMID 37653254.
  19. Karikari TK, et al. Blood phosphorylated tau 181 as a biomarker for Alzheimer's disease: a diagnostic performance and prediction modelling study using data from four prospective cohorts. Lancet Neurol. 2020;19:422-433. PMID 32333900.
  20. Salvadó G, et al. Plasma eMTBR-tau243 and %p-tau217 for biological staging of Alzheimer disease. JAMA Neurol. 2026;83:654-666. PMID 42189519.
  21. Palmqvist S, et al. Alzheimer's Association Clinical Practice Guideline on the use of blood-based biomarkers in the diagnostic workup of suspected Alzheimer's disease within specialized care settings. Alzheimers Dement. 2025;21:e70535. PMID 40729527.
  22. Pilotto A, et al. Plasma p-tau217 in Alzheimer's disease: Lumipulse and ALZpath SIMOA head-to-head comparison. Brain. 2025;148:408-415. PMID 39679606.
  23. Janelidze S, et al. Head-to-head comparison of 10 plasma phospho-tau assays in prodromal Alzheimer's disease. Brain. 2023;146:1592-1601. PMID 36087307.
  24. Schindler SE, et al. Head-to-head comparison of leading blood tests for Alzheimer's disease pathology. Alzheimers Dement. 2024;20:8074-8096. PMID 39394841.
  25. Pahlke S, et al. Blood-based biomarkers for detecting Alzheimer's disease pathology in cognitively impaired individuals within specialized care settings: a systematic review and meta-analysis. Alzheimers Dement. 2025;21:e70828. PMID 41193403.
  26. Chatterjee P, et al. Plasma Aβ42/40 ratio, p-tau181, GFAP, and NfL across the Alzheimer's disease continuum. Alzheimers Dement. 2023;19:1117-1134. PMID 36574591.