Diagnostic and Prognostic Utility of Plasma p‐tau217 for Alzheimer's Disease in Chinese Elderly: Insights From the SILCODE Study With a Derived Threshold

ABSTRACT

Background

Plasma phosphorylated tau 217 (p‐tau217) has emerged as a promising Alzheimer's disease (AD) biomarker, yet its longitudinal associations with neurodegeneration and cognitive decline remain inadequately characterized in Chinese populations, and ethnicity‐specific diagnostic thresholds are lacking for optimal clinical application.

Methods

A total of 541 participants (402 cognitively unimpaired [CU]; 139 cognitively impaired [CI]) from the Sino Longitudinal Study on Cognitive Decline (SILCODE) cohort were enrolled. Cross‐sectional and longitudinal associations of plasma p‐tau217 with amyloid‐β (Aβ) pathology, neurodegeneration, and cognition were evaluated. Diagnostic thresholds were derived using receiver operating characteristic analysis, and Cox regression assessed prognostic value for clinical progression.

Results

Cross‐sectionally, baseline p‐tau217 was associated with greater Aβ burden, neurodegeneration, and poorer cognition in the whole cohort and CI group; in the CU group, associations were confined to amyloid measures. Longitudinally, accelerated p‐tau217 accumulation was associated with faster neurodegeneration and cognitive decline in the whole cohort, with stage‐dependent patterns: nominal associations with neurodegenerative markers in CU and prominent cognitive associations in CI. Plasma p‐tau217 demonstrated high diagnostic accuracy for amyloid positivity (AUC = 0.891; cutoff: 0.529 pg/mL). Threshold‐based stratification effectively differentiated individuals by Aβ burden, neurodegeneration, and cognitive trajectories. Elevated baseline p‐tau217 predicted higher progression risk (Whole cohort: HR = 2.66 [1.28–5.53], p = 0.009; CU: HR = 2.44 [1.07–5.59], p = 0.034).

Conclusion

Plasma p‐tau217 serves as a valuable diagnostic and prognostic biomarker for AD, even among CU individuals, and the ethnicity‐specific threshold of 0.529 pg/mL enhances its clinical applicability for early detection and risk stratification in Chinese populations.

This study evaluated plasma p‐tau217 as a diagnostic, progression‐tracking, and prognostic biomarker for Alzheimer's disease in the SILCODE Chinese aging cohort (n = 541) with up to 8 years of follow‐up. Plasma p‐tau217 demonstrated high diagnostic accuracy for amyloid positivity (AUC = 0.891) with an ethnicity‐specific threshold of 0.529 pg/mL, and exhibited stage‐dependent longitudinal associations with neurodegeneration and cognitive decline. Elevated baseline p‐tau217 independently predicted clinical cognitive progression, even among cognitively unimpaired individuals.

1. Introduction

Alzheimer's disease (AD), the most prevalent neurodegenerative dementia worldwide, is estimated to affect 416 million individuals across the AD continuum [1], a figure substantially exceeding previous estimates of 57.4 million dementia cases globally [2]. This discrepancy may be attributed to the fact that the majority of affected individuals remain in the preclinical stage, characterized by asymptomatic neuropathological changes that may never progress to clinical manifestation [1]. Recent epidemiological data from China reveal concerning upward trends in AD incidence, prevalence, and mortality [3], exerting substantial economic pressures on healthcare systems, social support networks, and families. This escalating public health crisis highlights the pressing need for accessible early detection strategies, particularly given the prolonged preclinical window that precedes symptom onset [4].

Recent diagnostic criteria [5] formally recognize blood‐based biomarkers (BBMs) as a key component, offering distinct advantages over cerebrospinal fluid (CSF) biomarkers and positron emission tomography (PET) in terms of enhanced accessibility, cost‐effectiveness, and reduced invasiveness. Among BBMs, plasma phosphorylated tau (p‐tau) species have emerged as valuable tools for AD diagnosis and prognostication [6], with p‐tau217 standing out as the most promising candidate in current biomarker research. Converging evidence from longitudinal cohort studies has established plasma p‐tau217 abnormality as an early event in AD pathogenesis, closely associated with amyloid‐β (Aβ) accumulation [7, 8, 9]. Furthermore, plasma p‐tau217 also demonstrates high specificity in detecting subsequent tau tangle formation across disease stages [10, 11, 12], while exhibiting strong correlations with key clinical outcomes including accelerated cognitive decline [13], progressive cerebral atrophy [14], and longitudinal disease progression trajectories [15, 16]. With specific regard to Chinese aging populations, a previous study highlighted p‐tau217's diagnostic equivalence to CSF biomarkers for Aβ pathology detection and its superior performance over other plasma analytes in predicting cognitive trajectories [17]. Moreover, elevated p‐tau217 levels have been associated with accelerated cognitive deterioration, progressive neocortical Aβ deposition patterns [18], and increased dementia incidence [19, 20].

Despite these advancements, critical knowledge gaps remain in older Chinese populations. Specifically, current evidence is limited by insufficient longitudinal characterization of plasma biomarker performance across the cognitive spectrum from cognitively unimpaired to impaired states, the absence of ethnically calibrated p‐tau217 thresholds accommodating population‐specific Aβ metabolic variations, and undetermined predictive utility of plasma biomarkers for neurodegeneration trajectories and domain‐specific cognitive decline patterns. To address these gaps, the present study aims: (1) to systematically evaluate the association and prognostic accuracy of plasma p‐tau217 with respect to both neurodegeneration progression and cognitive domain‐specific deterioration patterns across cognitively unimpaired and impaired subgroups; (2) to establish ethnicity‐specific p‐tau217 cutoff values for optimized Aβ pathology detection in Chinese populations; and (3) to validate the clinical utility of these derived thresholds.

2. Methods

2.1. Study Design and Participants

The Sino Longitudinal Study on Cognitive Decline (SILCODE) study (ClinicalTrials.gov NCT03370744) is an ongoing registered multicenter initiative investigating Alzheimer's disease trajectories in Han Chinese populations across mainland China. Conducted in accordance with the Declaration of Helsinki, this protocol received ethical approval from Xuanwu Hospital's Institutional Review Board at Capital Medical University. Full inclusion/exclusion criteria are detailed in the study protocol (https://www.clinicaltrials.gov/ct2/show/study/NCT03370744). All participants provided written informed consent encompassing clinical data publication rights. Enrolled individuals underwent comprehensive baseline and longitudinal assessments, including standardized neuropsychological evaluations, multimodal neuroimaging (MRI and Aβ PET), and plasma biomarker profiling. Following baseline assessments, participants were followed under a structured longitudinal protocol, with assessment‐specific timelines detailed below. Plasma biomarker measurements were obtained from all 541 participants at baseline; of these, 314 completed repeat sampling at approximately 2‐year intervals (median: 1.79 years, IQR: 1.16–3.59 years). Neuropsychological assessments were administered to all 541 participants at baseline, with 314 completing annual follow‐up evaluations (median interval: 1.36 years, IQR: 0.99–2.65 years). Aβ PET imaging was acquired in a subset of 285 participants at baseline, of whom 65 underwent repeat imaging at a median interval of 2.58 years (IQR: 1.38–5.59 years). Structural MRI was obtained for 309 participants at baseline, with 154 completing repeat scans at a median interval of 1.30 years (IQR: 0.88–2.49 years).

2.2. Plasma Biomarkers and APOE Genotyping

Following a standardized ≥ 6‐h overnight fast, venous blood samples were collected in EDTA tubes during morning phlebotomy sessions. Plasma isolation was achieved through differential centrifugation, with aliquots immediately cryopreserved at −80°C until batch analysis. Plasma p‐tau217 concentrations were determined via single‐molecule array (Simoa) technology on the HD‐X Analyzer (Quanterix Corp). APOE genotyping employed TaqMan SNP genotyping targeting two single nucleotide polymorphisms (rs429358, rs7412), with genomic DNA extracted using a commercially validated isolation kit.

2.3. Clinical and Cognitive Assessments

Demographic parameters, including age, sex, and education years were systematically recorded. All participants completed a multidomain neuropsychological battery comprising standardized metrics across cognitive domains: (1) memory function: Auditory Verbal Learning Test‐Huashan version (AVLT‐H) subcomponents: long‐delayed recall (AVLT‐N5) and recognition (AVLT‐N7); (2) executive function: Shape Trails Test parts A & B (STT‐A/B); (3) language function: 30‐item Boston Naming Test (BNT) and Verbal Fluency Test (VFT); (4) global cognition: Mini‐Mental State Examination (MMSE), Montreal Cognitive Assessment‐Basic (MoCA‐B), and Memory and Executive Screening (MES). Participants were comprised of four clinical groups: cognitively normal (CN), subjective cognitive decline (SCD), mild cognitive impairment (MCI), and AD. For analytical purposes, they are categorized into two primary groups: cognitively unimpaired (CU), encompassing CN and SCD populations, and cognitively impaired (CI), comprising MCI and AD dementia cases. Clinical progression was defined as conversion from CU to CI, or from MCI to dementia.

2.4. Structural MRI and Amyloid PET Image Acquisition and Processing

All neuroimaging data were acquired on a 3.0T time‐of‐flight PET/MR system (SIGNA PET/MR, GE Healthcare, Milwaukee, WI, USA) at Xuanwu Hospital of Capital Medical University. Structural MRI sequences underwent automated parcellation via FreeSurfer v7.2.0. The residual hippocampal volume (rHCV) was determined by calculating the volume of both hippocampal hemispheres, subsequently adjusting for estimated total intracranial volume [21]. Additionally, the cortical thickness of AD‐signature atrophy brain regions was obtained. This was achieved by computing the surface area‐weighted average thickness of the bilateral entorhinal, fusiform, inferior temporal, and middle temporal cortices [22].

For Aβ PET imaging, participants received intravenous administration of 7–10 mCi [¹⁸F]florbetapir (AV45) or [¹⁸F]D3FSP [23], followed by a 20‐min static acquisition 50 min post‐injection. Using in‐house Matlab algorithms, the MRI and PET data were further processed. PET images and corresponding MRI structural images were co‐registered in Statistical Parameter Mapping 12 (SPM12). Regional PET measurements were extracted from the co‐registered images, and 68 Freesurfer‐defined cortical ROI were obtained by MRI segmentation. The standardized uptake value ratio (SUVR) of AD summary cortical regions, including the posterior cingulate cortex, precuneus, frontal lobe, parietal lobe, and lateral temporal, was calculated by normalizing the radiotracer uptake values of these typical AD brain areas against those observed in the cerebellum [24]. Participants were classified as Aβ PET positive if their global average SUVR voxels were ≥ 1.11 for AV45 and ≥ 1.04 for D3FSP [23].

2.5. Statistical Analysis

All the statistical analyses were conducted using R (v4.3.0). Two‐sided p values < 0.05 were deemed to be statistically significant. Normality of data distribution was assessed using the Shapiro–Wilk test, and variables exhibiting non‐normal distribution were subjected to logarithmic transformation. Inter‐group differences were examined using the Chi‐square test for categorical variables and one‐way analysis of variance (ANOVA) or the Kruskal–Wallis test for continuous variables. Data are presented as mean ± standard deviation (SD) or No. (%) unless otherwise noted. Outliers exceeding 3 SDs from the mean were excluded.

Plasma p‐tau217 levels were first compared between groups categorized by cognitive diagnosis (CU/CI) and Aβ PET status (negative/positive). Linear mixed‐effects models were then employed to characterize longitudinal trajectories of plasma, neurodegenerative, and cognitive markers, while generalized linear models (GLM) were used to evaluate cross‐sectional and longitudinal associations between plasma p‐tau217 and Aβ PET SUVR, neurodegeneration, and cognition across population strata (whole cohort, CU, and CI subgroups). The Benjamini‐Hochberg false discovery rate (FDR) procedure was applied to control for Type I error inflation due to multiple comparisons. For clinical threshold derivation, receiver operating characteristic (ROC) analysis was performed to determine optimal plasma p‐tau217 cutoffs (using Youden index) against Aβ PET classification, differentiating amyloid‐negative from amyloid‐positive participants. Consistent with the methodology established for Chinese aging cohorts, an additional analysis differentiated amyloid‐negative CU participants from amyloid‐positive CI participants [25]. To further assess diagnostic performance across the cognitive spectrum, subgroup ROC analyses were conducted within the CU and CI groups separately. Bootstrap resampling (n = 2000 iterations) was performed to estimate 95% confidence intervals for the optimal threshold, and sensitivity analyses evaluated diagnostic performance across a range of candidate thresholds (0.35–0.70 pg/mL). Participants were subsequently stratified into p‐tau217‐high and p‐tau217‐low subgroups based on the derived threshold. Between‐subgroup comparisons were then performed for baseline amyloid burden, neurodegeneration severity, and cognitive profiles, as well as longitudinal trajectories of neurodegeneration and cognitive markers, with FDR correction applied. To examine the association between p‐tau217 and disease progression risk, Cox proportional hazards regression was performed using two complementary approaches: (a) dichotomized p‐tau217 (high vs. low subgroups), and (b) continuous p‐tau217 (per 1‐SD increase). For each model, we report hazard ratios (HRs), 95% confidence intervals, and p‐values. Kaplan–Meier survival curves stratified by p‐tau217 subgroup (high vs. low) were generated for each cohort, with log‐rank tests used to compare between‐group survival distributions. All analyses were adjusted for age, sex, education, and APOE ε4 status.

3. Results

3.1. Demographic Characteristics of Included Participants

As presented in Table 1, a total of 541 participants were included in the analysis, comprising 402 (74.3%) CU individuals and 139 (25.7%) CI individuals. The mean age of the overall cohort was 66.80 ± 7.20 years, with 330 (61.00%) being females. Compared to the CU group, the CI group was older, had fewer years of education, and had a higher proportion of APOE ε4 carriers. No significant difference in sex distribution was observed between the two groups. Plasma p‐tau217 levels were significantly higher in the CI group compared to the CU group. Furthermore, the CI group consistently exhibited significantly poorer performance across all cognitive domains, lower rHCV and cortical thickness, alongside greater amyloid burden (SUVR‐AV45 and SUVR‐D3FSP) compared to the CU group. Of the 285 participants with available baseline Aβ PET data, 48 CU participants (22.4% of 214) and 47 CI participants (66.2% of 71) were amyloid‐positive at baseline. Follow‐up duration ranged from 0.0 to 8.93 years, with a median of 4.76 years (IQR: 2.62–5.58 years). The number of participants with available data decreased over time due to attrition; 60.5% of participants completed more than 4 years of follow‐up, while 15.6% completed more than 6 years.

TABLE 1.

Demographic, plasma, cognitive and neuroimaging characteristics of population included.

	Total (N = 541)	CU (N = 402)	CI (N = 139)	P
Demographic
Age (years)	66.80 (7.20)	66.04 (6.30)	69.00 (9.00)	< 0.001
Female (%)	330 (61.00)	251 (62.44)	79 (50.36)	0.286
Education (years)	12.79 (3.57)	13.13 (3.50)	11.82 (3.62)	< 0.001
APOE ε4 (%)	162 (30.86)	100 (25.45)	62 (46.97)	< 0.001
Plasma biomarker
Plasma p‐tau217 (pg/mL)	0.45 (0.41)	0.34 (0.22)	0.78 (0.62)	< 0.001
Cognitive Scores
MMSE	26.98 (4.41)	28.54 (1.67)	22.72 (6.38)	< 0.001
MoCA‐B	23.64 (5.13)	25.55 (2.47)	17.82 (6.55)	< 0.001
MES	84.02 (16.28)	89.76 (7.29)	65.72 (22.43)	< 0.001
AVLT‐N5	5.56 (3.12)	6.67 (2.43)	2.12 (2.44)	< 0.001
AVLT‐N7	20.62 (3.81)	21.91 (2.04)	16.66 (5.07)	< 0.001
STT‐A	72.08 (40.60)	58.89 (18.96)	114.91 (58.80)	< 0.001
STT‐B	159.92 (69.79)	138.66 (41.89)	242.97 (91.83)	< 0.001
BNT	24.03 (4.51)	25.48 (2.78)	19.65 (5.77)	< 0.001
VFT	17.94 (5.60)	19.61 (4.72)	12.88 (4.98)	< 0.001
Neuroimaging
rHCV	0.37 (0.86)	0.53 (0.65)	−0.23 (1.23)	< 0.001
Cortical thickness (mm3)	2.71 (0.12)	2.72 (0.09)	2.65 (0.17)	< 0.001
SUVR‐AV45	1.10 (0.27)	1.04 (0.21)	1.30 (0.33)	< 0.001
SUVR‐D3FSP	1.09 (0.17)	1.06 (0.15)	1.14 (0.20)	< 0.001

Note: Data are presented as n (%) for categorical variables and mean (SD) for continuous variables. p values were derived from the Chi‐square test for categorical variables and the ANOVA or the Kruskal–Wallis test for continuous variables.

Abbreviations: AVLT‐N5, Auditory Verbal Learning Test‐Huashan version‐long delayed recall; AVLT‐N7, Auditory Verbal Learning Test‐Huashan version‐recognition; BNT, Boston Naming Test; CI, cognitively impaired; CU, cognitively unimpaired; MES, Memory and Executive Screening; MMSE, Mini‐Mental State Examination; MoCA‐B, Montreal Cognitive Assessment‐Basic; N, number; rHCV, residual hippocampal volume; STT, Shape Trail Test; SUVR, standardized uptake value ratio; VFT, Verbal Fluency Test.

To further elucidate the influence of clinical and biological factors on plasma p‐tau217, comparative analyses were conducted on the distribution of baseline plasma p‐tau217 levels and their longitudinal slopes across different subgroups. As shown in Figure 1, participants were categorized into four groups based on their cognitive diagnosis (CU/CI) and (negative [N]/positive [P]): CU_N, CU_P, CI_N, and CI_P. Both baseline levels and longitudinal slopes of plasma p‐tau217 demonstrated a pattern primarily driven by amyloid status. Levels were comparable between the CU_N and CI_N groups, whereas amyloid‐positive groups (CU_P and CI_P) exhibited markedly elevated levels (all pairwise comparisons p < 0.05), with the CI_P group displaying the highest values overall. Notably, no significant difference was observed between the CI_N and CU_N groups. Further subgroup analyses (see Figure S1) consistently demonstrated that baseline plasma p‐tau217 levels were significantly higher in CI individuals, those aged ≥ 65 years, females, and APOE ε4 carriers. Similarly, the longitudinal increase in plasma p‐tau217 (see Figure S2) was significantly greater in CI individuals, the older age group (≥ 65 years), and APOE ε4 carriers. However, no significant sex‐related difference was observed in the longitudinal rate of p‐tau217 change.

FIGURE 1.

Distribution of baseline plasma p‐tau217 levels and longitudinal slopes across cognitive and amyloid status subgroups. Box plots showing (A) baseline plasma p‐tau217 concentrations and (B) longitudinal change (Δ) in plasma p‐tau217 levels across four subgroups categorized by cognitive diagnosis (CU, Cognitively unimpaired; CI, Cognitively impaired) and amyloid‐PET status (N: Amyloid‐negative; P: Amyloid‐positive). The central line in each box represents the median, the box bounds denote the interquartile range (IQR), and whiskers extend to 1.5 times the IQR. The horizontal dashed blue line represents a reference level.

3.2. Cross‐Sectional and Longitudinal Associations of Plasma p‐tau217 With Amyloid Deposition, Neurodegenerative Markers, and Cognitive Function

As depicted in Figure 2A, Table S1, and Figure S3, higher baseline plasma p‐tau217 levels were significantly associated with greater amyloid deposition, lower hippocampal volume, and poorer cognitive function across multiple domains at baseline in the whole cohort. Within the CI group, these baseline associations were broadly consistent with those observed in the whole cohort, with the notable exceptions of cortical thickness, AVLT‐N7, and STT‐B, which did not reach statistical significance. In the CU group, significant correlations were confined to elevated SUVR‐AV45 and SUVR‐D3FSP. The significant highlighted above remained robust after FDR correction for multiple comparisons.

FIGURE 2.

Associations of baseline plasma p‐tau217 with baseline biomarkers and longitudinal change rates across the Whole cohort, CU, and CI groups. Forest plots illustrating standardized beta (β) coefficients and 95% confidence intervals (CIs) derived from generalized linear models. Each model assesses the association of baseline plasma p‐tau217 levels with: (A) Baseline levels of each variable; (B) Longitudinal change rates (Δ) of each variable. All associations were adjusted for age, sex, education, and APOE‐ε4 status. Solid red circles indicate FDR‐corrected significance (P_FDR < 0.05); open red circles indicate nominal significance (uncorrected p < 0.05 only); gray triangles indicate non‐significant associations (uncorrected p ≥ 0.05). Error bars represent 95% CIs. AV45, [¹⁸F]florbetapir; AVLT, the Auditory Verbal Learning Test‐HuaShan version; BNT, the 30‐item Boston Naming Test; CI, cognitively impaired; CU, cognitively unimpaired; D3FSP, [¹⁸F]D3FSP; MES, the Memory and Executive Screening; MMSE, Mini‐Mental State Examination; MoCA‐B, the Montreal Cognitive Assessment‐Basic Version; p‐tau, phosphorylated tau; rHCV, residual hippocampal volume; STT‐A&B, the Shape Trails Test Part A and B; SUVR, standardized uptake value ratio; VFT, the Verbal Fluency Test.

Figure 2B, Table S2, and Figure S4 further illustrated the associations between baseline plasma p‐tau217 and faster longitudinal decline in hippocampal volume and cognition across multiple domains in the whole cohort. The association with the slope of AVLT‐N5 was nominally significant but did not survive FDR correction. In the CI group, significant associations were observed for the slopes of MMSE, MoCA‐B, MES, AVLT‐N7, and cortical thickness, with effect sizes substantially larger than those in the whole cohort. However, associations with slopes of rHCV, VFT, and BNT in the CI group were nominally significant but did not survive FDR correction. In the CU group, no longitudinal associations reached significance after FDR correction, though nominal associations were observed for slope of STT‐A and BNT.

The dynamic associations between plasma p‐tau217 slopes and the rates of change in neurodegenerative markers and cognitive function are delineated in Figure 3, Table S3, and Figure S5. In the whole cohort, a steeper increase in plasma p‐tau217 was significantly associated with a more pronounced longitudinal decline in rHCV, cortical thickness, and nearly all assessed cognitive domains, including MMSE, MoCA‐B, MES, AVLT‐N5, AVLT‐N7, STT‐A, and VFT after FDR correction. The association with the slope of STT‐B was nominally significant but did not survive FDR correction. In the CI group, significant associations after FDR correction were observed for the slopes of MMSE, MoCA‐B, MES, and AVLT‐N7, with markedly larger effect sizes compared to the whole cohort. In the CU group, no associations survived FDR correction, though nominal significance was observed for the slopes of rHCV and cortical thickness.

FIGURE 3.

Associations of longitudinal plasma p‐tau217 slopes with rates of change in neurodegenerative and cognitive markers across the whole cohort, CU, and CI groups. Forest plots illustrating the associations of the longitudinal plasma p‐tau217 slope with the longitudinal change rates (Δ) of neurodegeneration markers and cognitive functions.

3.3. Clinical Utility of Plasma p‐tau217 Threshold for Identifying Amyloid Positivity and Stratifying Neurodegeneration and Cognitive Decline

The diagnostic performance of plasma p‐tau217 for identifying brain amyloid positivity was evaluated across multiple analytical frameworks (Figures 4 and S6). In the primary analysis including all amyloid‐negative and amyloid‐positive participants regardless of cognitive status (N = 285), plasma p‐tau217 achieved an AUC of 0.891 (95% CI: 0.843–0.933) with an optimal cutoff of 0.529 pg/mL, yielding a sensitivity of 0.737 and specificity of 0.958 (Figure 4A). When the extreme‐group approach was applied to differentiate CU amyloid‐negative from CI amyloid‐positive participants, the AUC increased to 0.977 (95% CI: 0.955–0.994; Figure S6A). Subgroup analyses further demonstrated an AUC of 0.819 (95% CI: 0.736–0.895) within the CU group and 0.944 (95% CI: 0.890–0.986) within the CI group (Figure S6B,C). Bootstrap resampling confirmed the stability of the derived threshold, with the 95% CI of the optimal cutoff ranging from 0.377 to 0.548 pg/mL (Figure 4D). Sensitivity analysis across candidate thresholds from 0.35 to 0.70 pg/mL demonstrated that values within 0.50–0.55 pg/mL consistently yielded the highest Youden indices (0.669–0.696; Table S4), further supporting the robustness of the established cutoff. The distribution of plasma p‐tau217 levels across the cohort, with the identified threshold indicated, is presented in Figure 4B, illustrating that the majority of values cluster below the threshold while higher values extend into the positive range. Based on this cutoff, participants were stratified into p‐tau217‐high and p‐tau217‐low subgroups for subsequent analyses.

FIGURE 4.

Diagnostic performance and threshold stability of plasma p‐tau217 for identifying brain amyloid positivity. (A) ROC curve evaluating the diagnostic performance of plasma p‐tau217 for identifying brain amyloid positivity across all participants (amyloid‐negative vs. amyloid‐positive, N = 285). The optimal cutoff value (0.529 pg/mL) was determined using the Youden index. (B) Histogram and density plot illustrating the distribution of baseline plasma p‐tau217 levels across the cohort. The vertical red line indicates the optimal cutoff derived from the ROC analysis. (C) Overlay comparison of ROC curves from four analytical approaches: CU amyloid‐negative vs. CI amyloid‐positive, all amyloid‐negative vs. all amyloid‐positive, within CU only, and within CI only. (D) Bootstrap analysis of the optimal threshold from the primary ROC analysis. The solid red line indicates the optimal threshold. AUC, area under the curve; CI, cognitively impaired; CU, cognitively unimpaired.

Figure 5A, Table S5, and Figure S7 present the cross‐sectional comparisons between the p‐tau217‐high and p‐tau217‐low groups across amyloid burden, neurodegenerative markers, and cognitive functions. In the whole cohort and CI group, the p‐tau217‐high group showed significantly higher SUVR‐AV45 and SUVR‐D3FSP, lower rHCV and cortical thickness, and worse cognitive performance across all cognitive domains after FDR correction. In the CU group, significant differences after FDR correction were limited to amyloid PET measures (SUVR‐AV45 and SUVR‐D3FSP), while AVLT‐N5 and STT‐B showed nominal significance only.

FIGURE 5.

Differences in baseline biomarkers and longitudinal change rates between plasma p‐tau217‐high and p‐tau217‐low groups. Forest plots illustrating standardized beta (β) coefficients and 95% CIs for between‐group comparisons. (A) Cross‐sectional differences between the p‐tau217‐high and p‐tau217‐low groups in Aβ pathology, neurodegeneration, and various cognitive functions. (B) Longitudinal change rates (Δ) of neurodegeneration markers and cognitive functions between the two groups. Analyses were conducted across the whole cohort, CU group, and CI group. Symbol conventions, covariates, and abbreviations are as defined in Figure 2.

Longitudinal comparisons between p‐tau217‐high and p‐tau217‐low groups were presented in Figure 5B, Table S6, and Figure S8. In the whole cohort, the p‐tau217‐high group demonstrated significantly faster declines after FDR correction in rHCV, MMSE, MoCA‐B, MES, AVLT‐N5, AVLT‐N7, and VFT, as well as significantly greater increases in STT‐A compared to the p‐tau217‐low group. In the CI group, accelerated longitudinal declines in rHCV, cortical thickness, MMSE, MoCA‐B, MES, AVLT‐N7, and VFT were similarly observed in the p‐tau217‐high group relative to the p‐tau217‐low group. In the CU group, no longitudinal between‐group differences survived FDR correction, though nominal differences were observed for the slopes of rHCV and STT‐A.

3.4. Prognostic Value of Plasma p‐tau217 for Clinical Cognitive Progression

During the follow‐up, 36 of 262 CU participants (13.7%) subsequently converted to MCI or dementia, and 7 of 32 MCI participants (21.9%) progressed to dementia. As illustrated in Figure 6, in the whole cohort, both dichotomized and continuous p‐tau217 were significantly associated with increased risk of cognitive progression. Participants in the p‐tau217‐high subgroup had a 2.66‐fold increased risk of progression compared to those in the p‐tau217‐low subgroup (HR = 2.66, 95% CI: 1.28–5.53, p = 0.009). When modeled continuously, per SD increase in p‐tau217 was associated with a 2.05‐fold increased risk (HR = 2.05, 95% CI: 1.68–2.50, p < 0.001). In the CU subgroup, the association remained significant for both dichotomized (HR = 2.44, 95% CI: 1.07–5.59, p = 0.034) and continuous p‐tau217 (HR = 1.76, 95% CI: 1.15–2.69, p = 0.009). In the CI subgroup, the p‐tau217‐high subgroup exhibited a 1.96‐fold higher risk of progression relative to the p‐tau217‐low subgroup (HR = 1.96, 95% CI: 0.20–19.12, p = 0.564); the log‐rank test, however, indicated a significant difference between survival curves (p = 0.005). The continuous model yielded significant results (HR = 1.98, 95% CI: 1.45–2.71, p < 0.001).

FIGURE 6.

Clinical cognitive progression in participants stratified by plasma p‐tau217 status. Kaplan–Meier survival curves stratified by baseline plasma p‐tau217 levels for the (A) whole cohort, (B) CU subgroup, and (C) CI subgroup. Participants were dichotomized as p‐tau217‐high (≥ 0.529 pg/mL) or p‐tau217‐low (< 0.529 pg/mL). Shaded areas represent 95% confidence intervals. Numbers at risk are shown below each panel. Hazard ratios and p values are from Cox proportional hazards models adjusted for age, sex, education, and APOE ε4 status. Log‐rank p values compare the survival distributions between the two groups.

4. Discussion

The rapidly evolving field of AD biomarkers has seen significant advancements, particularly with the development of highly accurate blood‐based markers, most notably plasma p‐tau217. The present study comprehensively evaluated the diagnostic, prognostic, and longitudinal utility of plasma p‐tau217 as a multifaceted biomarker for Aβ pathology, neurodegeneration, and cognitive decline in the SILCODE cohort, a well‐characterized Chinese aging population with up to 8 years of follow‐up. Our findings yielded several key insights: (1) plasma p‐tau217 demonstrated strong cross‐sectional and longitudinal associations with Aβ pathology, neurodegeneration, and cognitive decline, with distinct stage‐dependent patterns across CU and CI groups; (2) plasma p‐tau217 exhibited high diagnostic accuracy for identifying brain amyloid positivity, with an AUC of 0.89 and a derived cutoff value of 0.529 pg/mL; (3) the threshold‐based stratification effectively differentiated subgroups with distinct profiles of amyloid burden, neurodegeneration, and cognitive trajectories; (4) elevated baseline plasma p‐tau217 independently predicted future clinical cognitive progression, even among CU individuals, underscoring its prognostic value.

The present study delineates the dynamic associations of plasma p‐tau217 with Aβ pathology, neurodegenerative processes, and cognitive function, reinforcing its value as a longitudinal and prognostic marker. Notably, the majority of associations reported here survived rigorous FDR correction for multiple comparisons, attesting to the robustness of our findings. Across the whole cohort, baseline plasma p‐tau217 levels were broadly associated with neurodegenerative markers (rHCV, cortical thickness) and multiple cognitive domains, consistent with the established role of p‐tau as a marker of AD pathology that drives neurodegeneration. More importantly, longitudinal analyses demonstrated that an accelerated rate of plasma p‐tau217 accumulation was significantly associated with more pronounced declines in rHCV and nearly all assessed cognitive domains in the whole cohort, further supporting its utility as a robust progression marker, in line with previous findings [19, 26]. Interestingly, our results reveal stage‐dependent patterns in the association between plasma p‐tau217 change rates and specific disease hallmarks, suggesting a shifting biological role for this biomarker across the AD continuum. In the CU group, baseline plasma p‐tau217 was exclusively confined to amyloid PET measures after FDR correction. Longitudinally, baseline plasma p‐tau217 showed nominal associations with the rates of change in STT‐A and BNT, while the rate of plasma p‐tau217 change showed nominal associations with the rates of change in rHCV and cortical thickness; however, none of these longitudinal associations survived FDR correction. While these findings may represent true but statistically underpowered associations in the current sample, they should be considered exploratory and warrant validation in larger independent cohorts before definitive conclusions can be drawn. Nevertheless, the convergence of these cross‐sectional and longitudinal findings reveals a coherent pattern: in the preclinical stage, plasma p‐tau217 primarily captures early cerebral Aβ pathology, with only tentative associations with neurodegenerative markers and no detectable associations with cognitive measures. This supports the hypothesis that plasma p‐tau217 elevation precedes neurofibrillary tangle accumulation, significant neurodegeneration, and overt cognitive symptoms, and aligns with the proposed temporal sequence of AD biomarker changes in which amyloid accumulation represents the earliest detectable pathological event. This early detection capability has important implications for timely identification and intervention, particularly given the rapid advancements in Aβ monoclonal antibody therapies. Conversely, in CI patients, an increasing plasma p‐tau217 slope was correlated with greater declines across global, memory, and language cognitive domains, yet did not show significant associations with neurodegenerative markers. This intriguing dissociation warrants consideration. One plausible explanation may relate to the smaller sample size and less frequent longitudinal MRI assessments compared to cognitive evaluations, thereby limiting statistical power to detect subtle ongoing structural changes. Alternatively, this dissociation may reflect a shift in what plasma p‐tau217 levels capture as the disease progresses. Given that significant macroscopic neurodegeneration is likely well‐established in CI individuals, rising p‐tau217 levels at this stage might more closely track the rate of functional cognitive decline, potentially reflecting downstream pathological processes beyond the initial stages of structural atrophy. Further mechanistic investigations, ideally with adequately powered and uniformly sampled longitudinal multimodal imaging data, are warranted to elucidate the underlying biological mechanisms.

Another key contribution of this study is the demonstration of excellent diagnostic performance of plasma p‐tau217 in differentiating amyloid‐positive from amyloid‐negative individuals. The diagnostic performance of plasma p‐tau217 varied across analytical frameworks, providing insights into its clinical applicability. The extreme‐group comparison (CU amyloid‐negative vs. CI amyloid‐positive) yielded an AUC of 0.977, representing the upper bound of diagnostic performance between the two most diagnostically distinct populations. When all participants were included regardless of cognitive status, the AUC was 0.891, reflecting the inclusion of diagnostically challenging intermediate cases. This more conservative estimate aligns with realistic clinical screening scenarios and remains indicative of excellent diagnostic accuracy. The within‐CU AUC of 0.819 highlights the inherent difficulty of detecting preclinical amyloid positivity, suggesting that p‐tau217 alone may benefit from combination with complementary biomarkers such as plasma Aβ42/Aβ40 for screening CU populations. Conversely, the within‐CI AUC of 0.944 confirms robust diagnostic utility in symptomatic individuals. Our optimal p‐tau217 cutoff of 0.529 pg/mL is consistent with published values using the Simoa ALZpath platform, which range from 0.338 to 0.63 pg/mL across studies [9, 27, 28, 29]. Higher cutoffs reported in memory clinic cohorts (0.52–0.55 pg/mL) [9, 28] align closely with our value, while the lower thresholds (0.338 pg/mL) [29] observed in a community‐based cohort likely reflect the lower amyloid prevalence in that population. The high specificity achieved with our cutoff (95.8%) is well‐suited for clinical settings where minimizing false‐positive results remains a priority. Despite cross‐study variation in cutoff values, AUC values remain consistently high (above 0.91), confirming robust diagnostic performance of plasma p‐tau217 across different populations and analytical thresholds. This level of accuracy is comparable to previously reported diagnostic utility for amyloid pathology in large‐scale studies and surpasses that of other plasma AD biomarkers [30, 31, 32]. Notably, its performance in our cohort was comparable to that of CSF markers, consistent with previous findings [17], further supporting the potential of plasma p‐tau217 as a minimally invasive alternative to CSF‐based assessment for Aβ pathology detection.

A further critical contribution of this study is the establishment of a clinically actionable threshold for plasma p‐tau217 within this Chinese aging cohort. While prior research has explored the diagnostic capabilities of plasma p‐tau217 for AD pathology [15, 33], there remains a notable gap in the establishment of definitive, clinically applicable positivity thresholds, especially within large, well‐characterized Chinese cohorts. This scarcity can be attributed to methodological variations across studies, including differences in participant populations and assay platforms [9, 33]. Nevertheless, a single cutoff is needed to identify AD pathology and facilitate widespread, standardized clinical application across diverse situations [34]. Utilizing our derived cutoff of 0.529 pg/mL, we observed significant between‐group differences in baseline Aβ pathology, as well as both baseline levels and longitudinal change rates for neurodegeneration and cognitive decline, between the threshold‐defined high and low groups. This stratification was evident across the whole cohort and, specifically, within the CI subgroup. Even within the CU group, this threshold effectively differentiated subgroups with significant differences in baseline Aβ pathology and longitudinal decline rates of rHCV and STT‐A.

Most importantly, our Cox proportional hazards models demonstrated the prognostic value of our established plasma p‐tau217 threshold. Over the up to 8‐year follow‐up period, individuals with baseline plasma p‐tau217 levels above this threshold exhibited a significantly higher risk of clinical cognitive progression compared to those below it. This prognostic capability was notable in the CU group, indicating that elevated baseline plasma p‐tau217 independently and significantly predicts a higher likelihood of future progression to MCI or dementia, even in individuals initially without any apparent cognitive impairment. These findings are highly consistent with previous research [26, 35], reinforcing the potential of plasma p‐tau217 as a valuable tool for identifying individuals at high risk for AD progression and facilitating earlier enrollment into clinical trials. Of note, the survival analyses were conducted using both dichotomized and continuous p‐tau217, providing complementary insights. The dichotomized approach offers clinical interpretability for risk stratification but necessarily discards information, as reflected in the wider confidence intervals compared to continuous models. The continuous approach yielded consistently more robust associations, particularly in the CU subgroup where statistical power was limited. For clinical implementation, a threshold‐based approach may be more appropriate for dichotomous decision‐making, whereas continuous values provide superior statistical efficiency for research and risk modeling purposes.

The present study presents several notable strengths. First, it leverages a large‐scale, longitudinally followed Chinese aging cohort, offering invaluable data for assessing the long‐term dynamics and clinical relevance of plasma p‐tau217 in this specific ethnic population. Second, our findings robustly confirm the excellent diagnostic accuracy for brain amyloid positivity and the prognostic capability of plasma p‐tau217, even among CU individuals. Third, and of particular significance, is the empirical establishment of a clinically actionable plasma p‐tau217 cutoff value (0.529 pg/mL) tailored to this Chinese cohort, addressing a key gap in population‐specific threshold determination.

Several limitations should be acknowledged. First, the use of the Simoa platform for p‐tau217 quantification means that our absolute values and threshold may not be directly comparable to other assay methodologies (e.g., mass spectrometry), necessitating cross‐platform validation. Second, limited tau‐PET data precluded a detailed exploration of the association with tau tangle pathology, warranting future studies with comprehensive tau‐PET imaging. Third, the relatively modest sample sizes of certain subgroups (particularly the CI group in survival analyses) may have limited statistical power to detect significant associations, and our findings in these subgroups should be interpreted with caution. Fourth, as a single‐center Chinese cohort study, the generalizability of our derived threshold to other ethnic populations and clinical settings requires further external validation.

5. Conclusions

In conclusion, the present study provides robust evidence supporting plasma p‐tau217 as a highly accurate diagnostic, progression‐tracking, and prognostic biomarker for AD. Its excellent diagnostic performance for amyloid positivity, strong associations with neurodegeneration and cognitive decline, and capacity to predict future clinical progression, even among cognitively unimpaired individuals, position it as a promising tool for both AD research and clinical practice. Furthermore, the ethnicity‐specific threshold of 0.529 pg/mL established in this Chinese aging cohort enhances its practical clinical applicability, facilitating improved early risk stratification and more timely therapeutic interventions.

Author Contributions

Ruixian Li: conceptualization, methodology, software, formal analysis, visualization, writing‐original draft. Wenhui Chai: validation. Jie Yang: data curation, resources. Hongwei Liu: resources. Xuanqian Wang: resources. Shuyu Zhang: resources. Hongtao Cai: data curation. Ying Han: conceptualization, methodology, writing – review and editing, funding acquisition, supervision, project administration. Jinghua Wang: conceptualization, methodology. Tengfei Guo: conceptualization, methodology, writing – review and editing.

Funding

This work was supported by STI2030‐Major Projects, 2022ZD0211800; National Natural Science Foundation of China, 82327809; the Construction Fund of Key Medical Disciplines of Hangzhou, 2025HZGF02; Initiative Funding of Hainan University, KYQD‐ZR‐21057; Education Funding from Mr Zhenhai Song and Mr Jinbo Yan.

Ethics Statement

Conducted in accordance with the Declaration of Helsinki, this protocol received ethical approval from Xuanwu Hospital's Institutional Review Board at Capital Medical University. All participants provided written informed consent encompassing clinical data publication rights.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Association between plasma p‐tau217 and Aβ pathology, neurodegeneration and cognitive function at baseline.

Table S2: Association between baseline plasma p‐tau217 and change rates of neurodegeneration and cognitive function.

Table S3: Association between change rates of plasma p‐tau217 and neurodegeneration and cognitive function.

Table S4: Sensitivity analysis of plasma p‐tau217 at varying thresholds for identifying brain amyloid positivity.

Table S5: Difference of Aβ pathology, neurodegeneration and cognitive function between plasma p‐tau217 high and low groups at baseline.

Table S6: Difference of change rates of neurodegeneration and cognitive function between plasma p‐tau217 high and low groups.

Figure S1: Differences of baseline plasma p‐tau217 between diagnosis, age, sex, and APOE4 status groups.

Figure S2: Differences of change rates of plasma p‐tau217 between diagnosis, age, sex, and APOE4 status groups.

Figure S3: Cross‐sectional correlations between plasma p‐tau217 with Aβ pathology, neurodegeneration, and cognitive performance in whole/CU/CI cohort.

Figure S4: Longitudinal associations between plasma p‐tau217 and rates of cognitive decline and neurodegenerative progression in whole/CU/CI cohort.

Figure S5: Longitudinal associations of change rates of plasma p‐tau217 with cognitive decline and neurodegenerative progression in whole/CU/CI cohort.

Figure S6: Subgroup ROC analyses for plasma p‐tau217 in differentiating amyloid positive and negative status.

Figure S7: Comparison of baseline amyloid deposition, neurodegeneration, and cognitive function between plasma p‐tau217 positive and negative groups.

Figure S8: Comparison of longitudinal change rates of neurodegeneration, and cognitive function between plasma p‐tau217 positive and negative groups.

Data Availability Statement

The datasets used and analyzed in this study are from the SILCODE cohort, and the derived data may be available for the qualified investigator after contacting the corresponding author upon the terms of a data use agreement.