Prospective study on clinical utility of plasma p-Tau217 and other biomarkers in Japanese memory clinics using the LUMIPULSE platform

Abstract

Background

In recent years, plasma levels of phosphorylated tau species, particularly p-tau217, have emerged as reliable indicators of amyloid-β (Aβ) pathology in the brain. However, real-world data on plasma biomarkers across diverse populations remain limited. We conducted a prospective multicenter study under real-world clinical settings to evaluate diagnostic performance of plasma biomarkers, including p-tau217, in discriminating amyloid status among a Japanese population.

Methods

A total of 332 participants were recruited from seven memory clinics across Japan. Participants were categorized into four clinical subgroups: cognitively unimpaired (CU), mild cognitive impairment (MCI), Alzheimer’s disease dementia (ADD), and non-ADD. We measured Aβ40, Aβ42, p-tau181, total-tau, and neurofilament light chain (NfL) in CSF and Aβ40, Aβ42, p-tau217, p-tau181, glial fibrillary acidic protein (GFAP) and NfL in plasma using the LUMIPULSE platform. Amyloid status was determined by amyloid PET imaging and/or CSF Aβ42/40 ratio.

Results

Significant differences were observed in plasma biomarker levels, including Aβ42/40, p-tau217, p-tau181, GFAP, and NfL across clinical categories. Plasma p-tau217 and p-tau217/Aβ42 achieved high diagnostic accuracy, with areas under the curve (AUC) exceeding 0.9 with PET amyloid status as the reference, demonstrating comparable performance to the in vitro diagnostic (IVD)-approved CSF Aβ42/40 ratio. The two-cutoff approach using plasma p-tau217 and p-tau217/Aβ42 to achieve 90% sensitivity and 90% specificity provided high negative and positive predictive values. The intermediate range defined by these two-cutoff points was narrower for p-tau217/Aβ42 than for p-tau217. The predefined U.S. Food and Drug Administration (FDA)-approved two-cutoff points were applicable to this cohort with good accuracy. Concordance of plasma p-tau217 or p-tau217/Aβ42 with PET or CSF Aβ42/40 status ranged from 87% to 93%, although larger discordant results were observed in the non-ADD group.

Conclusions

This study demonstrates the clinical utility of plasma p-tau217 and p-tau217/Aβ42 ratio for real-world diagnostic evaluations of dementia. However, careful interpretation of plasma biomarker is warranted in cases showing discordant results with PET or CSF findings.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13195-026-01997-7.

Introduction

In routine clinical practice, accurate diagnosis of dementia subtype including Alzheimer’s disease (AD), in patients exhibiting cognitive decline remains challenging. Biomarkers reflecting neuropathological changes in the brain have been shown to improve the diagnostic accuracy of AD [1], which has become increasingly important in the era of disease-modifying therapies. Confirmation of amyloid-β (Aβ) pathology is required prior to initiating anti-Aβ monoclonal antibody therapy, using reliable tools such as cerebrospinal fluid (CSF) biomarkers or Aβ positron emission tomography (PET) imaging [2, 3]. Although these modalities showed high accuracy in detecting pathological changes of AD, their accessibility remains limited. Blood biomarkers offer a less invasive, more cost-effective, and scalable alternative to CSF and PET tests. Accordingly, they are expected to serve as diagnostic tools in clinical practice. Notably, the revised Alzheimer’s Association criteria for diagnosis and staging of AD include plasma biomarkers as one of core diagnostic markers for AD alongside CSF and PET measures [4].

Several plasma biomarkers have been shown to discriminate between AD patients and controls, including the Aβ42/40 ratio, phosphorylated tau species (p-tau), neurofilament light chain (NfL) and glial fibrillary acidic protein (GFAP) [5]. The plasma Aβ42/40 ratio has been reported to detect Aβ pathology in the brain; however, unlike CSF measurements, plasma Aβ42/40 ratio shows limited performance and poor robustness in detecting Aβ pathology in the brain [6]. Moreover, the fold change in plasma Aβ42/40 levels between Aβ-positive and Aβ-negative subjects is relatively small [7]. Plasma NfL serves as a nonspecific biomarker of neurodegeneration and axonal injury that may be elevated in several neurodegenerative conditions [8]. The levels of plasma NfL are significantly increased in patients with AD, and longitudinal NfL levels increase in association with longitudinal clinical progression of AD [9]. GFAP reflects astrocytic activation and dysfunction and is considered a potential diagnostic biomarker for both neurodegenerative and psychiatric disorders [5, 10]. Plasma GFAP levels rise in parallel with increasing Aβ-PET burden, although this association diminishes at higher Aβ loads [11].

Among plasma biomarkers, p-tau217 has emerged as the most robust indicator of AD pathology, demonstrating diagnostic performance comparable to that of in vitro diagnostic (IVD)-approved CSF assays [12]. Head-to-head studies have shown that p-tau217 outperforms p-tau181, Aβ42/40, GFAP, and NfL in detecting cerebral Aβ pathology [13, 14]. Several studies have reported that plasma p-tau217 detects Aβ pathology in the brain with areas under the curve (AUC) exceeding 0.9 when compared with amyloid PET or CSF results as reference [15–17]. This indicates that plasma p-tau217 is equivalent in estimating brain Aβ deposition to IVD-approved CSF Aβ42/40 or p-tau181/Aβ42. Recently, the U.S. Food and Drug Administration (FDA) approved the plasma p-tau217/Aβ42 assay on the LUMIPULSE platform as the first IVD test for aiding in the diagnosis of AD. However, the application of the FDA-approved cutoff points of p-tau217/Aβ42 has not been validated in different populations. Although recent study in heterogeneous memory clinic cohorts suggests that the diagnostic accuracy of plasma p-tau217 was not significantly influenced by ethnicity [18], the generalizability of these FDA-approved cutoff points across different populations and real-world clinical settings remains to be established. Therefore, further studies are needed to validate the clinical utility of plasma p-tau217 and related biomarkers in real-world memory clinic settings across different populations.

The aims of this cross-sectional study were: (1) to evaluate the clinical performance of plasma biomarkers including p-tau217, p-tau181, Aβ42/40, NfL and GFAP on the LUMIPULSE platform in Japanese memory clinics; (2) to establish clinically applicable cutoff values for these biomarkers; and (3) to assess the applicability of FDA-approved cutoff points in a Japanese population; and (4) to characterize patients exhibiting discordant results among blood biomarkers, CSF and PET findings.

Methods

Study design and participants

This multicenter observational study was conducted at seven medical institutions in Japan between October 2022 and June 2024. A total of 332 participants were prospectively recruited. Participants were classified into four clinical categories: cognitively unimpaired (CU), mild cognitive impairment (MCI), Alzheimer’s disease dementia (ADD), and non-ADD. Diagnostic classification of MCI due to AD and ADD was made according to the NIA-AA 2011 criteria [19, 20]. We recruited MCI subjects including MCI due to AD and non-AD etiologies. Each subtype of non-ADD was diagnosed based on its respective established diagnostic criteria. Clinical diagnoses were made based on clinical information without knowledges of CSF and/or amyloid PET results. Among the 332 participants, 330 underwent lumbar puncture for CSF collection, and 102 underwent amyloid PET imaging. Of these, 100 underwent both amyloid PET imaging and CSF collection. In addition to CSF and/or PET assessments, clinical parameters such as body mass index (BMI) and Mini-Mental State Examination (MMSE) scores were obtained. Written informed consent was obtained from all participants or their legal representatives. The study was conducted in accordance with the Declaration of Helsinki and approved by Review Board of Niigata University (IRB approval number: 2023-0087).

Sample collection and processing

Blood and CSF samples were collected according to standardized protocols [21, 22]. Blood samples were drawn into EDTA-2 K tubes (BD Vacutainer) and centrifuged within 2 h at 1,500×g for 10 min at room temperature. Plasma was aliquoted into 0.5-mL portions and stored at − 80 °C until analysis. CSF samples were collected in polypropylene tubes (SARSTEDT, CSF false-bottom tubes) and stored at − 80 °C without centrifugation.

Fluid biomarker measurements and ApoE phenotyping

Plasma concentrations of Aβ40, Aβ42, p-tau181, p-tau217, GFAP, NfL, and CSF concentrations of Aβ40, Aβ42, p-tau181, total-tau (t-tau) and NfL were measured using the LUMIPULSE G1200 analyzer, a fully automated chemiluminescent enzyme immunoassay system (LUMIPULSE system, FUJIREBIO INC., Tokyo, Japan), following the manufacturer’s instructions. FDA-approved assays were used for CSF Aβ42/40 and plasma p-tau217/Aβ42 measurements. All biomarker measurements were conducted at the Research & Development Center of FUJIREBIO INC. (Tokyo, Japan) by experienced researchers who were blinded to clinical diagnosis and imaging results.

In addition, plasma ApoE4 and total ApoE (Pan-ApoE) levels were measured on the LUMIPULSE system. The APOE ε4 allele dosage (0, 1, or 2 alleles) was inferred from the ApoE4/Pan-ApoE ratio using predefined thresholds [23]. This protein-based assay provided an indirect method to estimate APOE ε4 copy number.

Amyloid PET

Amyloid PET imaging was performed in 102 participants using either ¹¹C-Pittsburgh compound B (PiB; n = 52) or ¹⁸F-labeled tracers (n = 50), including flutemetamol or florbetapir, depending on site availability. Amyloid status was determined at each site by board-certified nuclear medicine physicians through visual assessment, supported by quantitative indices such as standardized uptake value ratios (SUVRs) and/or Centiloid values.

Cutoff determination

For each fluid biomarker, optimal cutoff values were determined by maximizing the Youden Index (sensitivity + specificity − 1). A dual-cutoff approach was additionally applied, defining two thresholds to achieve both ≥ 90% sensitivity and ≥ 90% specificity. In the concordance and discordance analyses, cutoff values defined by the Youden index were used for plasma p-tau217 (PET-referenced cutoff = 0.190, CSF-referenced cutoff = 0.212) and p-tau217/Aβ42 (PET-referenced cutoff = 0.007, CSF-referenced cutoff = 0.007). Amyloid PET positivity was determined by visual assessment. CSF amyloid status was defined using a predefined cutoff of the CSF Aβ42/40 ratio (< 0.067) on the LUMIPULSE platform (LUMIPULSE G1200) [24]. Discordant cases were classified into plasma positive / CSF negative, plasma positive / PET negative, plasma negative / CSF positive, plasma negative / PET positive, CSF positive / PET negative, and CSF negative / PET positive.

Statistical analysis

All statistical analyses were conducted using R software (version 4.4.2). Group comparisons for continuous variables across more than two groups were performed using the Kruskal–Wallis rank–sum test, and pairwise comparisons between two groups were assessed using the Wilcoxon rank sum test. Categorical variables were compared using Pearson’s chi-squared test or Fisher’s exact test as appropriate. Pearson’s correlation coefficients were calculated to assess associations between plasma and CSF biomarkers and between biomarkers and clinical variables (age, BMI, and MMSE). These associations were visualized using pair plots and scatterplots displaying univariate distributions and fitted regression lines. To evaluate the diagnostic performance of the fluid biomarkers, receiver operating characteristic (ROC) analyses were performed using two reference standards for amyloid status: amyloid PET and CSF biomarkers. Two-tailed p-values < 0.05 were considered statistically significant. When applicable, multiple comparisons were corrected using the false discovery rate (FDR).

Results

Demographics of participants

Baseline demographics of participants are summarized in Table 1. The median age of participants was 73 years, and 50% were female. APOE ε4 positivity was frequently observed in MCI (46%) and ADD (44%). MMSE scores declined progressively along the clinical spectrum. Amyloid positivity varied markedly across diagnostic groups: CU (20% by PET status and 13% by CSF Aβ42/40), MCI (88% and 66%), ADD (89% and 87%), and non-ADD (7.7% and 27%).

Table 1.

Demographics, baseline characteristics, and biomarker data in clinical categories

	ALL	CU	MCI	ADD	non-ADD
	N = 332	n = 98	n = 61	n = 61	n = 112	p-value*
Age	73 [63, 79]	74 [62, 79]	73 [68, 79]	66 [56, 76]	74 [68, 80]	< 0.01
Female, n (%)	167 (50%)	41 (42%)	37 (61%)	34 (56%)	55 (49%)	0.100
MMSE	25 [21, 28]	28 [26, 30]	27 [24, 28]	21 [17, 23]	23 [17, 27]	< 0.001
BMI	21.6 [19.7, 23.7]	22.0 [20.1, 24.2]	21.6 [19.7, 23.2]	21.3 [19.3, 23.1]	21.6 [19.9, 24.0]	0.600
APOE ε4 number of alleles, n (%)						< 0.001
0	227 (68%)	74 (76%)	33 (54%)	34 (56%)	86 (77%)
1	90 (27%)	24 (24%)	24 (39%)	19 (31%)	23 (21%)
2	15 (4.5%)	0 (0%)	4 (6.6%)	8 (13%)	3 (2.7%)
Amyloid positivity (%)
PET A+	58%	20%	88%	89%	7.7%	< 0.001
CSF A+	41%	13%	66%	87%	27%	< 0.001
CSF
Aβ40	11,254 [8,609, 14,652]	11,046 [8,651, 13,719]	13,060 [10,908, 15,976]	12,689 [9,153, 15,419]	9,225 [7,405, 12,736]	< 0.001
Aβ42	785 [565, 1,105]	967 [772, 1,307]	772 [615, 1,043]	625 [495, 763]	737 [509, 1,169]	< 0.001
Aβ42/40	0.08 [0.05, 0.10]	0.10 [0.09, 0.11]	0.06 [0.05, 0.09]	0.05 [0.04, 0.06]	0.10 [0.06, 0.11]	< 0.001
p-tau181	42.4 [25.9, 78.1]	29.6 [22.9, 41.4]	70.7 [42.2, 100.2]	91.4 [64.1, 138.9]	37.5 [24.7, 49.7]	< 0.001
p-tau181/Aβ42	0.04 [0.03, 0.12]	0.03 [0.02, 0.04]	0.09 [0.03, 0.14]	0.15 [0.11, 0.22]	0.04 [0.03, 0.08]	< 0.001
t-tau	344 [239, 513]	267 [218, 353]	453 [323, 615]	586 [477, 886]	291 [221, 399]	< 0.001
NfL	1,210 [812, 2,155]	1,011 [670, 2,119]	962 [762, 1,258]	1,200 [879, 1,702]	1,914 [1,070, 3,081]	< 0.001
Plasma
Aβ40	334 [297, 378]	335 [303, 379]	332 [298, 362]	305 [277, 352]	350 [304, 400]	0.002
Aβ42	25.7 [22.3, 29.5]	26.9 [23.8, 30.9]	25.1 [21.8, 27.8]	22.3 [20.5, 25.7]	26.2 [23.3, 30.6]	< 0.001
Aβ42/40	0.08 [0.07, 0.08]	0.08 [0.08, 0.09]	0.07 [0.07, 0.08]	0.07 [0.07, 0.08]	0.08 [0.07, 0.08]	< 0.001
p-tau217	0.15 [0.08, 0.40]	0.08 [0.05, 0.12]	0.30 [0.14, 0.44]	0.50 [0.32, 0.72]	0.14 [0.08, 0.30]	< 0.001
p-tau217/Aβ42	0.005 [0.003, 0.017]	0.003 [0.002, 0.004]	0.012 [0.005, 0.017]	0.022 [0.013, 0.031]	0.005 [0.003, 0.013]	< 0.001
p-tau181	1.2 [0.8, 1.7]	0.8 [0.6, 1.2]	1.5 [1.0, 1.8]	1.8 [1.2, 2.3]	1.1 [0.8, 1.8]	< 0.001
p-tau181/Aβ42	0.04 [0.023, 0.07]	0.03 [0.02, 0.04]	0.07 [0.04, 0.08]	0.08 [0.06, 0.10]	0.04 [0.03, 0.07]	< 0.001
GFAP	65.6 [43.5, 94.9]	47.9 [33.0, 68.5]	72.5 [49.9, 86.6]	86.2 [59.5, 117.3]	68.9 [46.9, 113.2]	< 0.001
NfL	24 [16, 41]	22 [13, 34]	20 [16, 26]	19 [15, 32]	38 [24, 66]	< 0.001

Data are shown in median [Q1, Q3]; n (%)

*Kruskal-Wallis rank sum test; Pearson’s Chi-squared test

Plasma p-tau217 levels were highest in ADD, followed by MCI, non-ADD and CU. GFAP showed a similar trend, whereas NfL was highest in non-ADD. Among 102 participants who underwent amyloid PET, 58% were amyloid positive (Supplemental Table 1). In 330 participants determined by CSF Aβ42/40, 41% were positive (Supplemental Table 2). Amyloid-positive individuals were more frequently APOE ε4 carriers and cognitively impaired. The non-ADD group included multiple diagnostic categories (Supplemental Table 3). Plasma p-tau217 levels were generally low among non-ADD subtypes, although some Parkinson’s disease with dementia (PDD)/dementia with Lewy bodies (DLB) cases exhibited high plasma p-tau217 levels, suggesting possible Aβ co-pathology. GFAP tended to be elevated in PDD/DLB, frontotemporal dementia (FTD), and normal pressure hydrocephalus (NPH), whereas NfL was highest in NPH.

Performance of plasma biomarkers

To evaluate the diagnostic performance of plasma biomarkers for detecting amyloid pathology, we analyzed their discriminative ability using PET amyloid status (n = 102) and CSF Aβ42/40 status (n = 330) as reference standards (Table 2; Fig. 1). Both plasma p-tau217 and the p-tau217/Aβ42 ratio showed consistently high diagnostic accuracy, with AUC exceeding 0.90 with both reference standards. Plasma p-tau181 also reached an AUC > 0.9 against PET status, but decreased to 0.816 when compared with CSF Aβ42/40 as the reference. There was no significant difference in AUC between CSF Aβ42/40 and plasma p-tau217 or p-tau217/Aβ42 by Delong test when compared with PET amyloid status as the reference. Using PET as the reference, the AUCs were 0.74 for GFAP and 0.59 for NfL. Using CSF Aβ42/40 as the reference, the AUCs were 0.77 for GFAP and 0.52 for NfL. The AUCs for plasma Aβ42/40 were 0.77 (PET reference) and 0.85 (CSF reference), while those for plasma Aβ42 were 0.67 (PET reference) and 0.76 (CSF reference). Plasma p-tau217, p-tau217/Aβ42, p-tau181, and p-tau181/Aβ42 demonstrated significantly higher AUC than plasma Aβ42/40, GFAP, and NfL after FDR correction when compared with PET amyloid status or CSF Aβ42/40 as the reference.

Table 2.

Performance of CSF and plasma biomarkers for differentiating A+ from A- status

Assay	Reference: Amyloid PET				Reference: CSF Aβ42/40
	AUC	95%CI	Sen (%)	Spe (%)	AUC	95%CI	Sen (%)	Spe (%)
CSF
Aβ40	0.56	0.44–0.68	50	29	0.65	0.59–0.70	50	74
Aβ42	0.82	0.72–0.92	95	68	0.83	0.78–0.87	82	71
Aβ42/40	0.93	0.85–1.00	97	90	1.00	1.00–1.00	100	100
p-tau181	0.86	0.77–0.95	85	85	0.93	0.90–0.96	86	88
p-tau181/Aβ42	0.91	0.83–0.99	97	93	0.99	0.98–0.99	95	95
t-tau	0.80	0.70–0.89	80	78	0.88	0.85–0.92	87	76
NfL	0.63	0.51–0.76	93	46	0.51	0.44–0.58	86	29
Plasma
Aβ40	0.54	0.43–0.66	76	42	0.56	0.50–0.63	85	29
Aβ42	0.67	0.56–0.78	76	58	0.76	0.71–0.81	72	69
Aβ42/40	0.77	0.67–0.86	86	61	0.85	0.80–0.89	77	82
p-tau217	0.93	0.87–0.99	93	93	0.92	0.89–0.95	84	90
p-tau217/Aβ42	0.93	0.86–0.99	95	93	0.95	0.92–0.97	90	88
p-tau181	0.94	0.89–0.99	95	91	0.82	0.77–0.86	89	67
p-tau181/Aβ42	0.93	0.87–0.99	95	86	0.88	0.84–0.92	90	78
GFAP	0.74	0.63–0.84	86	81	0.77	0.72–0.82	85	60
NfL	0.59	0.46–0.71	93	35	0.52	0.46–0.59	54	61

Fig. 1.

Discriminative performance of plasma and CSF biomarkers. A AUC and 95% CI for each biomarker using amyloid PET status (n = 102) as the reference are shown. AUCs for plasma p-tau217, p-tau217/Aβ42, p-tau181, and p-tau181/Aβ42 were comparable to that of the IVD-approved CSF Aβ42/40 ratio. B AUC and 95% CI for each biomarker using CSF Aβ42/40 (n = 330) as the reference are shown. The vertical dashed line indicates AUC = 0.90. Abbreviations: Aβ40, amyloid-β 1-40; Aβ42, amyloid-β 1-42; AUC, area under the curve; CI, confidence interval; CSF, cerebrospinal fluid; GFAP, glial fibrillary acidic protein; NfL, neurofilament light chain; PET, positron emission tomography; p-tau181, phosphorylated tau at threonine 181; p-tau217, phosphorylated tau at threonine 217; t-tau, total tau

CSF and Plasma biomarkers in clinical subgroups

To characterize CSF biomarker profiles across clinical subgroups, we first compared CSF biomarker levels across diagnostic categories (Supplemental Fig. 1). Comparison of CSF biomarker levels across diagnostic categories revealed the expected AD pattern: lower Aβ42/40 and higher p-tau181, t-tau, and p-tau181/Aβ42 ratios from CU to ADD. Non-ADD cases had low tau markers but the highest NfL levels. These findings highlight distinct biomarker profiles between ADD and non-ADD, with tau elevations specific to ADD and NfL elevations characteristic of non-ADD dementias.

We next evaluated plasma biomarker levels across the same diagnostic categories to assess whether similar disease-related patterns were observed in blood (Fig. 2). As shown in Fig. 2, p-tau217 and p-tau217/Aβ42 increased stepwise from CU to MCI to ADD (p < 0.001). Similarly, p-Tau181 and p-tau181/Aβ42 also increased significantly across these groups (p < 0.001), although p-tau181 levels did not differ significantly between MCI and ADD. The Aβ42/40 ratio was lower in ADD than in CU and non-ADD, with no difference between ADD and MCI. The non-ADD group showed a slight reduction in the Aβ42/40 ratio compared with CU. GFAP increased from CU to ADD, with comparable elevations in ADD and non-ADD, whereas NfL was highest in non-ADD.

Fig. 2.

Comparison of plasma biomarkers among clinical categories. A Boxplots showing plasma Aβ42/40 ratio across four clinical categories: CU, MCI, ADD, and non-ADD). B p-tau217, C ptau217/Aβ42, D p-tau181, E p-tau181/Aβ42, F GFAP (log10transformed), and G NfL (log10-transformed). Group differences were assessed using Kruskal–Wallis tests with post hoc pairwise comparisons. ***p < 0.001, **p < 0.01, *p < 0.05. Abbreviations: ADD, Alzheimer’s disease dementia; Aβ42/40, amyloid-β 1-42 to 1-40 ratio; CU, cognitively unimpaired; GFAP, glial fibrillary acidic protein; MCI, mild cognitive impairment; NfL, neurofilament light chain; non-ADD, non-AD dementia; p-tau181, phosphorylated tau at threonine 181; p-tau217, phosphorylated tau at threonine 217

When stratified by amyloid status (A+ /A−) based on PET or CSF, both p-tau217 and p-tau217/Aβ42 were significantly higher in A+ individuals across all diagnostic categories (p < 0.0001) (Fig. 3).

Fig. 3.

Comparison of plasma p-tau217 and p-tau217/Aβ42 in clinical categories stratified by Aβ status. A Boxplots of plasma p-tau217 levels stratified by amyloid status (positive [A+] or negative [A−]) are shown. Amyloid status was based on concordant results from amyloid PET and CSF Aβ42/40. Eight subjects with discordant PET and CSF results were excluded from this analysis. B Boxplots showing p-tau217/Aβ42 levels are shown. ****p < 0.0001 for all shown comparisons

Association between plasma biomarkers and clinical parameters

We next examined associations of plasma p-tau217 and p-tau217/Aβ42 with age, BMI, and MMSE (Supplemental Fig. 2). In CU, both plasma markers showed a significant positive association with age, whereas in MCI and ADD, age showed a significant inverse association. BMI demonstrated a significant inverse correlation with plasma p-tau217 in CU, MCI, and non-ADD as well as with p-tau217/Aβ42 in MCI and non-ADD. MMSE scores showed a significant inverse association with plasma p-tau217 and p-tau217/Aβ42 in non-ADD.

Cutoff determination and stratified analyses

Table 3 summarizes the performance of plasma p-tau217 and the p-tau217/Aβ42 ratio for discriminating amyloid PET/CSF-defined positivity using single- and two-point cutoffs. AUC was 0.92 for p-tau217 and 0.95 for p-tau217/Aβ42. Single-point cutoffs derived by maximizing the Youden index yielded sensitivities of 84–90% and specificities of 88–91%. Two-point cutoffs were optimized to achieve ≥ 90% sensitivity and ≥ 90% specificity, categorizing cases as positive, intermediate, or negative. The intermediate range was smaller for the p-tau217/Aβ42 ratio (3%) than for p-tau217 alone (15%). Supplemental Figs. 3 and 4 show that the p-tau217/Aβ42 ratio consistently minimized intermediate zone compared to p-tau217 across clinical diagnostic and age-stratified groups, improving separation between amyloid-positive and -negative groups. We next applied FDA-approved cutoffs points to our datasets, which provided 98% sensitivity, 89% specificity, 98% negative predictive value (NPV), and 88% positive predictive value (PPV), with an intermediate range of approximately 20%.

Table 3.

Cutoff values for plasma p-tau217 and p-tau217/Aβ42 to detect amyloid status

1-point cutoff
Assay	Cutoff	AUC	Sen (%)	Spe (%)	OPA (%)	NPV (%)	PPV (%)
p-tau217 (pg/mL)	0.212a	0.92 (0.89, 0.95)	84 (77, 89)	91 (86, 94)	88 (84, 91)	89 (84, 93)	87 (80, 92)
p-tau217/Aβ42	0.007a	0.95 (0.93, 0.97)	90 (84, 94)	88 (83, 92)	89 (85, 92)	93 (88, 96)	84 (77, 89)

2-point cutoffs
Assay	Cutoff		Sen (%)	Spe (%)	OPA (%)	NPV (%)	PPV (%)	Intermediate zone (%)
	Low	High
p-tau217 (pg/mL)	0.134b	0.247b	90 (84, 95)	90 (84, 94)	90 (86, 93)	93 (88, 96)	87 (79, 92)	15
p-tau217/Aβ42	0.00613b	0.00700b	90 (83, 94)	90 (85, 94)	90 (86, 93)	93 (88, 96)	87 (80, 91)	3
p-tau217/Aβ42	0.00370c	0.00738c	98 (94, 100)	89 (83, 93)	93 (89, 96)	98 (95,100)	88 (81, 92)	20

^aCutoff values determined by maximizing the Youden index

^bCutoff values determined to achieve both ≥ 90% sensitivity and ≥ 90% specificity

^cFDA-approved cutoff values applied

Stratification based on plasma p-tau217 or the p-tau217/Aβ42 ratio is presented in Supplementary Table 4. Participants were classified into plasma biomarker-positive (51% according to p-tau217 and 58% according to p-tau217/Aβ42), intermediate (9% and 2%, respectively), and biomarker-negative groups (40% for both measures). When plasma biomarkers were used as a rule-out tool for amyloid pathology, the number of PET or CSF examinations could be reduced by approximately 40%. When plasma biomarkers were used as a stand-alone test, with additional PET or CSF examinations required only for participants in the intermediate group, the need for PET or CSF examinations could be reduced by 91–98%.

Concordance and discordant cases for plasma, CSF and PET scan

We examined concordance of plasma p-tau217 or p-tau217/Aβ42 with amyloid PET status or CSF Aβ42/40 across diagnostic groups of CU, MCI, ADD, and non-ADD (Table 4). In PET-based analyses, both plasma measures showed similarly high concordance (93–96%) across CU, MCI and ADD. The largest discordance was observed in non-ADD (11.5%). In the CSF-based dataset, the p-tau217/Aβ42 ratio yielded slightly lower discordance rates than p-tau217 in MCI (p-tau217/Aβ42: 6.6% vs. p-tau217: 9.8%) and ADD (5.0% vs. 8.3%). The largest discordance was observed in non-ADD (19.6%). Overall, both plasma measures demonstrated approximately 90% concordance with amyloid PET or CSF Aβ42/40 with consistently lower concordance in non-ADD.

Table 4.

Concordance of plasma biomarkers with amyloid PET or CSF Aβ42/40 across clinical categories

	Reference: Amyloid PET					Reference: CSF Aβ42/40
	ALL	CU	MCI	ADD	non-ADD	ALL	CU	MCI	ADD	non-ADD
	N = 102	n = 15	n = 26	n = 35	n = 26	N = 330	n = 97	n = 61	n = 60	n = 112
Plasma p-tau217
Concordant, n (%)	95 (93.1)	14 (93.3)	25 (96.2)	33 (94.3)	23 (88.5)	288 (87.3)	88 (90.7)	55 (90.2)	55 (91.7)	90 (80.4)
Discordant, n (%)	7 (6.9)	1 (6.7)	1 (3.8)	2 (5.7)	3 (11.5)	42 (12.7)	9 (9.3)	6 (9.8)	5 (8.3)	22 (19.6)
Plasma+ / Ref.–, n (%)	3 (42.9)	0 (0.0)	0 (0.0)	1 (50.0)	2 (66.7)	21 (50.0)	1 (11.1)	3 (50.0)	2 (40.0)	15 (68.2)
Plasma– / Ref.+, n (%)	4 (57.1)	1 (100.0)	1 (100.0)	1 (50.0)	1 (33.3)	21 (50.0)	8 (88.9)	3 (50.0)	3 (60.0)	7 (31.8)
Plasma p-tau217/Aβ42
Concordant, n (%)	95 (93.1)	14 (93.3)	25 (96.2)	33 (94.3)	23 (88.5)	292 (88.5)	88 (90.7)	57 (93.4)	57 (95.0)	90 (80.4)
Discordant, n (%)	7 (6.9)	1 (6.7)	1 (3.8)	2 (5.7)	3 (11.5)	38 (11.5)	9 (9.3)	4 (6.6)	3 (5.0)	22 (19.6)
Plasma+ / Ref. –, n (%)	3 (42.9)	0 (0.0)	0 (0.0)	1 (50.0)	2 (66.7)	21 (55.3)	1 (11.1)	3 (75.0)	1 (33.3)	16 (72.7)
Plasma– / Ref.+, n (%)	4 (57.1)	1 (100.0)	1 (100.0)	1 (50.0)	1 (33.3)	17 (44.7)	8 (88.9)	1 (25.0)	2 (66.7)	6 (27.3)

We identified 10 individuals (10%) with discrepancies among 100 subjects who received plasma p-tau217, CSF Aβ42/40, and amyloid PET tests (Supplemental Table 5). These cases included MCI (n = 2), ADD (n = 3), non-ADD (vascular dementia, suspected non-Alzheimer’s disease pathophysiology (SNAP), and FTD: n = 4), and CU (n = 1). Most were APOE ε4 non-carriers (70%), with MMSE scores ranging from 14 to 30. The results of plasma p-tau217 and the p-tau217/Aβ42 ratio were fully concordant with each other. Patterns of discordance were heterogeneous: six were plasma-positive with CSF-negative/PET-positive (n = 3) and CSF-positive/PET-negative (n = 3), and four were plasma-negative with CSF-negative/PET-positive (n = 2) and CSF-positive/PET-positive (n = 2).

Discussion

In this study, we evaluated several plasma biomarkers as predictors of Aβ pathology in participants recruited from seven memory clinics across Japan. Our major findings were as follows. First, we demonstrated that plasma p-tau217 and the p-tau217/Aβ42 ratio showed high accuracy in distinguishing Aβ-positive from Aβ-negative individuals across the AD continuum including MCI, ADD, and non-ADD. These differences were evident even in CU and non-ADD groups, suggesting that elevated plasma p-tau217 reflects underlying amyloid pathology independent of clinical conditions. Plasma p-tau217 (AUC = 0.93) and p-tau217/Aβ42 (AUC = 0.93) achieved diagnostic performance comparable to that of the IVD-approved CSF Aβ42/40 ratio (AUC = 0.93) when PET amyloid status was used as the reference. These findings suggest that plasma p-tau217 and the p-tau217/Aβ42 ratio are robust blood-based biomarkers for detecting amyloid pathology. Second, applying a two-cutoff approach to meet the 90% sensitivity and specificity thresholds resulted in relatively small intermediate range for p-tau217 (15%) and p-tau217/Aβ42 (3%). The small proportion of intermediate cases for p-tau217/Aβ42 was consistent across diagnostic categories and age groups (Supplemental Figs. 3 and 4). When FDA-approved predefined two-cutoff points were applied to our cohort, the p-tau217/Aβ42 ratio maintained high diagnostic performance with 98% sensitivity and 89% specificity with approximately 20% intermediate range. However, our findings seem to contrast with the finding from another research group presented at CTAD 2025 meeting (https://www.alzforum.org/news/conference-coverage/trouble-fujirebios-fda-cleared-blood-test-or-lousy-lot). More real-world studies are required to validate the FDA-approved two-cutoff points of plasma p-tau217/Aβ42. Third, we identified a subset of participants showing discordant results between plasma and CSF biomarkers, and PET imaging. Taken together, these results highlight the clinical potential of integrated plasma biomarker assessment for detecting amyloid pathology and facilitating accurate diagnosis of AD.

We report that plasma p-tau181 performed as well as p-tau217 when amyloid PET was used as the reference, but not when CSF Aβ42/40 was used as the reference. Previous study demonstrated that changes in CSF Aβ42/40 start prior to detectable amyloid accumulation on PET imaging [25]. In addition, plasma p-tau217 has been reported to become abnormal earlier than plasma p-tau181 in AD continuum [26]. Participants with CSF Aβ42/40 positivity may be likely to include an earlier stage of AD continuum. Within this population, plasma p-tau217 may provide better diagnostic performance than plasma p-tau181.

A previous study reported that a two-step diagnostic workflow incorporating plasma biomarkers improved diagnostic performance and expanded their clinical applicability [27]. Introducing plasma biomarkers into the diagnostic workflow may reduce the number of PET or CSF examinations required to identify amyloid pathology. In the present study, we demonstrated that a substantial proportion of PET or CSF examinations could be reduced through stratification based on plasma p-tau217 and the p-tau217/Aβ42 ratio. Consistent with our findings, a prior study reported that a two-cutoff approach using a plasma biomarker with 90% sensitivity and 90% specificity reduced the need for subsequent CSF testing by 86.5% when the plasma biomarker was used as a stand-alone tool [27]. Overall, implementing a diagnostic workflow that incorporates plasma biomarkers for the detection of amyloid pathology could substantially reduce the use of advanced CSF or PET testing, thereby minimizing the burden on patients and caregivers.

A subset of participants showed discordant results between plasma p-tau217 and PET amyloid status or CSF Aβ42/40, which may affect clinical decision-making for diagnosis and treatment. Discordant rates were 7% between plasma p-tau217 and PET results, and 13% between plasma p-tau217 and CSF results. Similarly, discordance was 7% for PET and 11% for CSF when using the p-tau217/Aβ42 ratio. Discordant cases were most frequent in the non-ADD group. Among ten discordant cases, two were negative for plasma p-tau217 and p-tau217/Aβ42 but positive for CSF and PET tests, suggesting potential false-negative blood biomarker results that could lead to misclassification of AD at a screening process. However, we cannot exclude a possibility that plasma biomarker results were truly negative, and PET and CSF results were false-positive. Cautious interpretations of discordant results are required. Yun et al. reported slightly higher discordant rates of 6% for plasma p-tau217-positive/Aβ PET-negative and 5% for plasma p-tau217-negative/Aβ PET-positive groups than our rates [28]. They concluded that the mechanisms underlying the discordance between plasma p-tau217 and Aβ PET may be heterogeneous. They also found that the p-tau217-negative/Aβ PET-positive group had a higher BMI compared with the p-tau217-positive/Aβ PET-positive group, which could contribute to the discordance rates. Consistent with their observation, among the two discordant participants with plasma biomarker-negative but PET- and CSF-positive results in our study, one individual had a high BMI (≥ 30). Of the four possible biomarker combinations, the p-tau217-positive/Aβ PET-positive group showed the fastest cognitive decline, followed by the p-tau217-positive/Aβ PET-negative group [28]. Longitudinal follow-up of the discordant cases in our study will be necessary to clarify their clinical significance. Understanding the determinants of concordance and discordance among AD plasma and CSF biomarkers, and PET imaging is crucial for accurate interpretation of plasma biomarker results. Such discrepancies may arise from pre-analytical variability in blood sampling, technical issues in measurement, inconsistent cutoff values, borderline or misinterpreted PET findings, and the presence of confounding factors and unknown analytical interference affecting plasma biomarkers. Discordance between fluid biomarkers and PET Aβ imaging may reflect difference between soluble and aggregated Aβ species [29]. Reduction in Aβ42/40 or elevation of p-tau217 can precede PET-detectable fibrillar Aβ deposition and neuritic plaque formation, suggesting that some discordant cases represent early-stage of Aβ accumulation [30]. Additionally, variability in visual assessment of amyloid PET scans may also contribute to discordant findings. Nonetheless, current evidence remains limited, underscoring the need for larger, real-world datasets to understand the implications of biomarker discordance in clinical practice.

The strengths of this study include the recruitment of participants from multiple memory clinics, enhancing the generalizability of our findings to real-world clinical settings. Moreover, we utilized FDA-approved assay kits for CSF Aβ42/40, plasma p-tau217 and Aβ42. However, several limitations should be acknowledged. First, this was a cross-sectional baseline study, and longitudinal data on biomarker changes and clinical outcomes were not available. Second, potential confounding factors that might influence plasma biomarkers levels, such as renal dysfunction, were not considered and may have led to false-positive results. Third, different amyloid PET tracers were used across sites, depending on local availability. Fourth, the number of discordant cases was relatively small, limiting statistical power. Further studies are needed to confirm these findings and explore the mechanisms underlying discordance.

Conclusions

This study supports the clinical utility of plasma p-tau217 and the p-tau217/Aβ42 ratio in the real-world diagnostic evaluation of dementia. The performance of plasma p-tau217 and p-tau217/Aβ42 aligns with current recommendations for blood-based biomarker testing and demonstrate promise as practical tools for detecting Aβ pathology in clinical settings.

Abbreviations

Aβ42

Amyloid-β 1–42

ADD

Alzheimer’s disease dementia

AUC

Area under the curve

CSF

Cerebrospinal fluid

CU

Cognitively unimpaired

MCI

Mild cognitive impairment

MMSE

Mini-Mental State Examination

non-ADD

Non-AD dementia

PET

Positron emission tomography

p-tau217

Phosphorylated tau at threonine 217

ROC

Receiver operating characteristic

Authors’ contributions

TIs: data acquisition, performed analysis, drafted manuscript. MK: data acquisition, performed analysis, revised manuscript. YN: data acquisition, performed analysis, revised manuscript. EK-S: drafted manuscript, data acquisition, performed analysis, revised manuscript. KK: data acquisition, revised manuscript. NT: data acquisition, revised manuscript. SK: data acquisition, revised manuscript. YK; data acquisition, revised manuscript. KI: data acquisition, revised manuscript. MH: data acquisition, revised manuscript. NS: data acquisition, revised manuscript. KI: data acquisition, revised manuscript. HN: data acquisition, revised manuscript. JK: data acquisition, revised manuscript. KA: revised manuscript, supervision. SS: data acquisition, revised manuscript. HM: data acquisition, revised manuscript. TT: data acquisition, revised manuscript. OO: revised manuscript, supervision. TIw: designed study, revised manuscript, supervision. MY: designed study, data acquisition, revised manuscript, supervision. TY: designed study, data acquisition, revised manuscript, supervision. AI: designed study, data acquisition, revised manuscript, supervision. TIk: designed study, data acquisition, revised manuscript, supervision.

Funding

This study was supported by FUJIREBIO INC., Integrated Research Initiative for Living Well with Dementia (IRIDE) of the Tokyo Metropolitan Institute for Geriatrics and Gerontology (MK, KI, AI), and AMED JP25dk0207070 (AI, TIk).

Data availability

The datasets used in this study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participants

This study was conducted in accordance with the tenets of the Declaration of Helsinki and subsequent amendments. All participants provided written informed consent. This study is approved by the ethical review boards of Niigata University (2023-0087), Fujirebio and each of participating institutes. All participants provided written informed consent to use medical data and biomaterials for this study.

Competing interests

E Kikkawa-Saito, H Nojima, and J Kamada are full-time employees of FUJIREBIO INC, and K Aoyagi is a board member of FUJIREBIO INC.