Plasma biomarkers, brain amyloid‐beta pathology, and cortical thickness in a non‐Hispanic White and Black/African American middle‐aged community cohort: The HCP‐CoBRA study

Abstract

INTRODUCTION

We evaluated plasma biomarker association with, and classification accuracies for, amyloid beta–positron emission tomography (Aβ‐PET) and cortical thickness in the biracial Human Connectome Project–Connectomics in Brain Aging (HCP‐CoBRA) cohort (53% Black/African American [B/AA] and 47% non‐Hispanic White [NHW]).

METHODS

In n = 218 participants (median age 62, range: 57–71] years, 65% female and 15% Aβ‐PET positive), plasma biomarkers (phosphorylated tau‐181 [p‐tau181], p‐tau217, p‐tau231, glial fibrillary acidic protein [GFAP], neurofilament light chain [NfL], and Aβ42/Aβ40) were compared to Aβ‐PET and magnetic resonance imaging (MRI) neuroimaging indicators.

RESULTS

Plasma p‐tau217 (Johnson & Johnson and ALZpath [areas under the curve [AUCs] = 0.915 vs. 0.919]) had high sensitivity and specificity (>85%) for Aβ‐PET status. All plasma biomarkers except p‐tau231 effectively ruled out Aβ pathology (negative predictive value [NPV] >95%) but only Johnson & Johnson p‐tau217+ was good for confirmation (covariate‐adjusted positive predictive value [PPV] = 0.909). The plasma biomarkers performed poorly for identifying cortical thickness status but were elevated according to joint Aβ‐PET and neurodegeneration profiles. Plasma biomarker accuracies for Aβ‐PET positivity were unaffected by self‐identified race, except for ALZpath p‐tau217 (p = 0.024). However, correlations with Aβ‐PET standardized uptake value ratio varied by self‐identified race.

DISCUSSION

Plasma p‐tau217 is a promising tool for Alzheimer's disease–associated Aβ pathology in older/middle‐aged individuals. However, apparent race‐related performances should be further studied.

Highlights

• Plasma phosphorylated tau‐217 (p‐tau217) and glial fibrillary acidic protein (GFAP) best predicted abnormal brain amyloid beta–positron emission tomography (Aβ‐PET).

• Plasma p‐tau217 accurately identified abnormal Aβ‐PET (Johnson & Johnson p‐tau217: area under the curve [AUC] = 0.9145, 95% confidence interval [CI] = 0.8367–0.9923; ALZpath p‐tau217: AUC = 0.9198, 95% CI = 0.8585–0.981) followed by GFAP and Aβ42/40 ratio (GFAP: AUC = 0.8529, 95% CI = 0.7485–0.9573; Aβ42/40:AUC = 0.7962, 95% CI = 0.6581–0.9346).

• All plasma biomarkers performed poorly in identifying cortical thickness, despite being higher according to combined Aβ‐PET and neurodegeneration profiles.

• Correlations of p‐tau217 (Johnson & Johnson p < 0.001), p‐tau181 (p = 0.005), and Aβ42/40 (p = 0.004) with Aβ‐PET in predicting amyloid burden were stronger in self‐identified non‐Hispanic Whites vs Black/African Americans.

• Biomarker accuracies for Aβ‐PET positivity were unaffected by self‐identified race, except ALZpath p‐tau217 (p = 0.024).

1. BACKGROUND

Despite tremendous advancements in detecting and diagnosing Alzheimer's disease (AD) according to cerebrospinal fluid (CSF) and neuroimaging techniques such as positron emission tomography (PET), these tests remain invasive, time‐consuming, and expensive. ¹ , ² , ³ , ⁴ , ⁵ The development of plasma biomarkers for AD research has substantial potential as screening and confirmatory diagnostic tools for use in intervention trials and clinical care programs. ⁶ , ⁷ , ⁸ , ⁹ The implications are vast including the potential for early detection with pre‐clinical applications. ⁷ , ¹⁰ , ¹¹ There is an abundance of literature regarding the accuracy and robustness of plasma biomarkers to detect neuropathological changes observed with brain imaging. ⁷ , ⁹ , ¹¹ However, studies are limited regarding community‐based cohorts, specifically those comprising diverse groups and backgrounds. ¹¹ , ¹² , ¹³ , ¹⁴ It is imperative to widen the scope of biomarker‐focused investigations to communities and populations that are not routinely included in such studies, toward ensuring that clinically approved tests factor in performance in a more representative group of participants.

The amyloid‐tau‐neurodegeneration (AT(N)) framework proposed by the National Institute on Aging and the Alzheimer's Association (NIA‐AA) workgroup helps to standardize the diagnosis and classification of AD based on the accumulation of amyloid beta plaques (Aβ; A), tau neurofibrillary tangles (T), and neuronal injury/degeneration (N) biomarkers. ¹⁵ , ¹⁶ Specifically, the biofluid‐based biomarkers in the AT(N) scheme include CSF and plasma Aβ42/40 ratio, hyperphosphorylated tau (at various sites, including p‐tau181, p‐tau217, and p‐tau231), and neurofilament light chain (NfL) or total tau (t‐tau), respectively. ¹⁷ Aβ42/40 ratio serves as a marker of amyloid plaque accumulation, typically altered early in preclinical stages. ¹⁸ Phosphorylated tau variants (e.g., p‐tau181, p‐tau217, and p‐tau231) are strongly associated with amyloid status and moderately with tau tangle formation, and they correlate with both AD progression and cognitive decline. ¹⁹ NfL reflects neuronal injury and degeneration, although it is not specific to AD and may be elevated in other neurodegenerative disorders. ²⁰ The recently updated framework recognizes the multifaceted pathophysiological processes in AD, expanding key processes of interest to include inflammation (I) and co‐pathologies such as vascular dysfunction (V) and concomitant neuropathologies such as synucleinopathy (S). ²¹ The inflammation marker GFAP is being increasingly recognized as a marker of astrocytic activation, which often co‐occurs with amyloid pathology. ²² Together, these markers provide a multidimensional view of AD pathophysiology and enable classification and staging. However, the impact of common social determinants of health, such as age, sex, apolipoprotein E (APOE) genotype, and, importantly, self‐identified racial identity, must be critically assessed for real‐world application of these frameworks. ²³ These factors might influence the classification accuracies of plasma biomarkers. ²⁴

This study had a two‐fold aim. Our first aim was to evaluate the clinical performance of plasma biomarkers among a non‐Hispanic White and Black/African American middle‐aged community cohort. We examined classification accuracies of AT(N) plasma biomarkers (including p‐tau181, p‐tau217, p‐tau231, Aβ40, Aβ42, GFAP, and NfL) to identify abnormal brain Aβ‐PET scans as well as to identify abnormal cortical thickness. The second aim was to examine the potential self‐identified race‐dependent association between these plasma biomarkers and Aβ‐PET in the same cohort.

2. METHODS

2.1. Participants

We included participants from the Human Connectome Project–Connectomics in Brain Aging (HCP‐CoBRA), which recruited both cognitively normal and impaired participants, from Pittsburgh, PA, USA, 50–89 years of age. ²⁵ The primary source of recruitment for all participants was the Pitt + Me web portal, with less than 20% of participants being recruited from the University of Pittsburgh Alzheimer's Disease Research Center (Pitt‐ADRC) and by word of mouth. The participants underwent both clinical and cognitive assessments over 3 days, beginning with an informed consent visit. Participants completed functional magnetic resonance imaging (fMRI) tasks both with fasting and then without fasting, as well as underwent magnetoencephalography and ¹¹C Pittsburgh compound B (PiB) PET imaging for brain Aβ plaques a week after the last MRI scanning. Each participant also underwent a brief neuropsychological assessment based on the Pittsburgh ADRC protocol, including the Montreal Cognitive Assessment (MoCA), verbal fluency test, a 30‐item visual naming test, Trail Making Tests, verbal free recall, and the Rey–Osterreith Complex Figure. ²⁵ , ²⁶ Demographic information, including self‐identified racial identity, sex, and education, were also collected. The inclusion criteria and study design have been described in previous publications. ²⁵

All participants completed the informed consent process and intake forms. The HCP‐CoBRA study and its protocols were reviewed and approved by the University of Pittsburgh Institutional Review Board.

2.2. Neuroimaging procedures

Imaging was conducted and processed as described previously. ²⁵ Briefly, Aβ‐PET (A) status was based on a global [¹¹C] PiB standardized uptake value ratio (SUVr) calculated by volume‐weighted averaging of nine composite regional outcomes (anterior cingulate, posterior cingulate, insula, superior frontal cortex, orbitofrontal cortex, lateral temporal cortex, parietal, precuneus, and ventral striatum). ²⁷ Participants were classified as A+ or A– based on a pre‐defined cutoff, with >1.346 as A+. ²⁸ , ²⁹ We classified neurodegeneration (N) status based on MRI scans for cortical thickness (CT). ³⁰ , ³¹ , ³² Specifically, N status was determined by an AD‐signature composite CT index obtained from a surface‐area weighted average of the mean CT of four FreeSurfer regions of interest (ROIs)—entorhinal cortex, inferior temporal and middle temporal gyri, and fusiform—all of which are most predictive of AD‐specific pathology and diagnosis, with <2.7 as N+. ²⁸ , ³³

RESEARCH IN CONTEXT

1. Systematic review: Although plasma biomarkers show high accuracy for detecting Alzheimer's pathophysiology, our search in PubMed identified that studies performed in community/population‐based cohorts were lacking. We examined the clinical performance of several plasma biomarkers covering various amyloid‐tau‐neurodegeneration (AT(N)) categories and their consistency across self‐identified racial groups.

2. Interpretation: Plasma phosphorylated tau‐217 (p‐tau217) and glial fibrillary acidic protein (GFAP) best predicted abnormal brain amyloid beta–positron emission tomography (Aβ‐PET). Yet, all plasma biomarkers performed poorly for N status, despite being higher according to combined A/N statuses. Plasma biomarker correlations with Aβ‐PET tended to vary between self‐identified non‐Hispanic White and Black/African American groups. Biomarker accuracies for Aβ‐PET positivity were unaffected by self‐identified race, except ALZpath p‐tau217 (p = 0.024).

3. Future directions: This study highlights high performance of plasma p‐tau217 and GFAP across communities, whereas demonstrating potential self‐identified racial differences in the association between plasma biomarkers and brain Aβ‐PET. Underlying reasons and the widespread nature of these population‐specific findings remain unknown.

2.3. Plasma collection procedure

Blood collection and processing followed established protocols. ³⁴ Blood samples were collected in 10 mL K2‐EDTA ethylenediamine tetra‐acetic acid tubes (BD Biosciences Cat# 366643) and centrifuged within 2 h at 2000 x g for 10 min at 4°C to obtain the plasma portion of the whole blood. The plasma samples were aliquoted and stored at −80 °C until use. Buffy coats were also collected to determine the APOE genotype according to the published procedure. ³⁵

2.4. Procedures for Single molecule array (Simoa) assays

Simoa assays for the analytes p‐tau181, p‐tau217, p‐tau231, Aβ40, Aβ42, GFAP, and NfL were performed on an HD‐X (Quanterix, Billerica, MA, USA) as described by Zeng and colleagues. ³⁰ Briefly, plasma samples were thawed at room temperature and centrifuged at 4000 x g for 10 min at 4°C to remove particulates. Plasma Aβ40, Aβ42, GFAP, and NfL were measured using the Neurology 4‐Plex E (#103670) kit. We measured p‐tau217 with both the ALZpath Simoa p‐Tau 217 V2 Assay (#104371) and the Johnson & Johnson p‐tau217+ assay, an in‐house assay developed at Johnson & Johnson Research & Development using Johnson & Johnson's proprietary antibodies. ³⁶ We measured p‐tau181 using the Karikari et al., method, ³⁷ using p‐tau181 antibody AT270 for capture and anti‐Tau6‐18 antibody (Tau12; BioLegend, # 806502) for detection. We measured p‐tau231 in the same fashion but with p‐tau231 antibody AT180 for capture and Tau12 for detection. ³⁷ Quality control (QC) samples were analyzed at the start and the end of each run in two to three different concentrations per assay to assess the reproducibility of each assay. All assays for cohort samples were run in singlets, whereas QC samples were run in duplicates at the start and at the end of the plate. These QC procedures remained consistent throughout the study. For p‐tau181 and ALZpath p‐tau217, the within and between‐plate coefficients of variation (CVs) were <5% and <6%, respectively. For the Neurology 4‐plex assays, the within and between‐plate CVs were mostly <10% and <16%, respectively. The Johnson & Johnson p‐tau217+ assay had within and between‐plate CVs of 4.0% and 7.3%, respectively.

2.5. Statistical analysis

Continuous variables were reported as median (interquartile range [IQR]) and categorical variables as counts (%). Statistical testing procedures to compare demographic characteristics and blood‐based biomarkers according to A or N status were done via the Wilcoxon rank‐sum test for continuous variables and the Pearson's chi‐square test for categorical variables. In addition, Spearman's correlations between biomarkers and continuous A or N status were reported. Logistic regression models were fit to predict either A or N status. Univariable models for each biomarker were fit, as well as multivariable models with demographic and genetic characteristics, sex, age (at enrollment), and APOE ε4 carrier status. This allowed for the assessment of each biomarker's predictive capabilities before and after adjustment for demographic variables. Any participants from the HCP cohort that had missing demographic characteristics were excluded from this study. A complete case analysis of all p‐tau biomarkers was conducted first. Then, a reduced dataset was utilized for complete case analysis of NfL, GFAP, and Aβ42/40 ratio measured in a subset of samples with available volume for the Neurology 4‐plex assay.

We trained and tested the models using a 50–50 training–testing data split, allowing the logistic regression models to be trained with 50% of the data and validated on the remaining 50%. This procedure was done for the prediction of both A and N status models. The area under the receiver‐operating characteristic (ROC) curve (AUC) was calculated from the results of the testing dataset and reported with corresponding 95% confidence intervals (CIs). ³⁸ We compared the biomarker‐only models with the models that incorporated demographic variables via DeLong's test for differences in AUC. ³⁸

For analysis where participants were grouped according to both A and N statuses, to create four possible groups—A–N–, A+N–, A–N+, and A+N+–all continuous variables were first compared using a Kruskal–Wallis rank‐sum test. For biomarkers with a significant global test, pairwise comparisons were made using Wilcoxon rank–sum tests, with a Bonferroni correction for multiple comparisons adjustment. Demographic categorical variables were compared via either a Pearson's chi‐square or Fisher's exact test.

Finally, participants were grouped by self‐identified race, Black/African American (B/AA) or non‐Hispanic White (NHW), and by Aβ‐PET status (A status). Biomarker assays were first compared between self‐identified racial groups via a Wilcoxon rank‐sum test. Statistical testing was also performed within each self‐identified racial group, with continuous variables compared via a Wilcoxon rank‐sum test and categorical variables with either Fisher's exact test or Pearson's chi‐square test. Participants were also grouped by self‐identified race (B/AA or NHW), and by APOE ε4 carrier status (carrier or non‐carrier). Biomarker assays were compared according to APOE ε4 carrier status within self‐identified racial groups by the Wilcoxon rank‐sum test, and Fisher's exact test was used to compare Aβ‐PET status. Logistic regression models were fit using binary Aβ‐PET status as the outcome variable. These models included one of the plasma biomarkers, self‐identified race, and a term for their interaction to assess if the relationship between each biomarker and Aβ‐PET status is altered by self‐identified racial group. Similarly, linear regression models were fit to predict global PiB SUVr using the same predictors to assess if self‐identified racial group alters the relationship between biomarker and global PiB SUVr. The plasma biomarker assays were standardized with the Z‐transformation to make the interaction coefficients more comparable across the linear regression models.

3. RESULTS

3.1. Participant characteristics

We included 218 participants who had Aβ‐PET (A status) data, including 185 (85%) and 33 (15%) A– and A+ individuals, respectively (see Table 1 for participant characteristics according to Aβ‐PET abnormality). Among them, 116 (53%) were self‐identified as B/AA and 102 (47%) were identified as NHW. The A+ participants had a median (IQR) age of 72 (68, 78) years, 14 (42%) female, 13 (39%) APOE ε4 carriers, and 8 (24%) self‐identified as B/AA. The A– participants had a median (IQR) age of 60 (56, 67) years, 127 (69%) were female, 54 (29%) were APOE ε4 carriers, and 108 (58%) self‐identified as B/AA (Table 1). Compared to the A– group, the A+ participants were older (p < 0.001), had fewer females (p = 0.004), and fewer participants who self‐identified as B/AA (p < 0.001), but no statistically significant difference in the frequency of APOE ε4 carriers (p = 0.242).

TABLE 1.

Participant characteristics of the HCP cohort according to Aβ‐PET abnormality and plasma biomarker information.

	Overall, n = 218 a	Aβ PET negative, n = 185 a	Aβ PET positive, n = 33 a	p‐value b
Sex				0.004
Female	141 (65%)	127 (69%)	14 (42%)
Male	77 (35%)	58 (31%)	19 (58%)
Age at enrollment, years	62 (57, 70)	60 (56, 67)	72 (68, 78)	<0.001
APOE ε4				0.242
Non‐carrier	151 (69%)	131 (71%)	20 (61%)
Carrier	67 (31%)	54 (29%)	13 (39%)
Race				<0.001
Black/African American	116 (53%)	108 (58%)	8 (24%)
Non‐Hispanic White	102 (47%)	77 (42%)	25 (76%)
ALZpath p‐tau217, pg/mL	0.25 (0.18, 0.36)	0.24 (0.17, 0.31)	0.51 (0.40,0.80)	<0.001
(Missing)	4	3	1
Johnson & Johnson p‐tau217+, pg/mL	0.035 (0.026, 0.050)	0.033 (0.025, 0.044)	0.070 (0.049, 0.105)	<0.001
(Missing)	22	19	3
p‐tau181, pg/mL	13 (9, 18)	12 (9, 18)	16 (12, 23)	0.014
(Missing)	10	7	3
p‐tau231, pg/mL	9 (6, 14)	9 (6, 13)	10 (6, 17)	0.472
(Missing)	10	7	3
Aß40, pg/mL	68 (53, 79)	69 (53, 79)	64 (54, 78)	0.862
(Missing)	63	58	5
Aß42, pg/mL	4.82 (3.71, 5.61)	4.89 (3.90, 5.67)	3.74 (3.30, 5.14)	0.013
(Missing)	63	58	5
Aß42/40 ratio	0.072 (0.061, 0.080)	0.074 (0.065, 0.083)	0.060 (0.056, 0.069)	<0.001
(Missing)	63	58	5
GFAP, pg/mL	70 (49, 102)	63 (46, 95)	92 (79, 155)	<0.001
(Missing)	63	58	5
NfL, pg/mL	13 (9, 18)	12 (8, 17)	18 (11, 22)	0.003
(Missing)	63	58	5
Global PiB SUVr	1.11 (1.06, 1.19)	1.09 (1.05, 1.14)	1.66 (1.52, 1.91)	<0.001

Abbreviations: Aβ, amyloid beta; GFAP, glial fibrillary acidic protein; HCP, Human Connectome Project; NfL, neurofilament light chain; PET, positron emission tomography; p‐tau; phosphorylated tau; PiB, Pittsburgh Compound B; SUVr, standardized uptake value ratio.

^a n (%) or median (IQR).

^b Pearson's chi‐square test for categorical and Wilcoxon rank‐sum test for numerical variables.

The neurodegeneration status (N status) analysis included 220 participants with complete CT‐based MRI data, with 142 (65%) and 78 (35%) being N– and N+, respectively (see Table 2 for demographic characteristics according to neurodegeneration status as assessed by cortical thickness). There was no significant difference between the N+ and N– groups in terms of age (p = 0.814), sex (p = 0.434), APOE ε4 carrier status (p = 0.056), or self‐identified race (p = 0.119). The N+ participants had a median (IQR) age of 62 (56, 71) years, 53 (68%) were female, 30 (38%) were APOE ε4 carriers, and 47 (60%) self‐identified as B/AA. The N– participants had a median (IQR) age of 63 (57, 70) years, 89 (63%) were female, 37 (26%) were APOE ε4 carriers, and 70 (49%) self‐identified as B/AA.

TABLE 2.

Participant characteristics of the HCP cohort according to neurodegeneration status as assessed by cortical thickness.

	Overall, n = 220 a	Neurodegeneration‐ negative, n = 142 a	Neurodegeneration‐positive, n = 78 a	p‐value b
Sex				0.434
Female	142 (65%)	89 (63%)	53 (68%)
Male	78 (35%)	53 (37%)	25 (32%)
Age at enrollment, years	62 (57, 71)	63 (57, 70)	62 (56, 71)	0.814
APOE ε4				0.056
Not Carrier	153 (70%)	105 (74%)	48 (62%)
Carrier	67 (30%)	37 (26%)	30 (38%)
Race				0.119
Black/African American	117 (53%)	70 (49%)	47 (60%)
Non‐Hispanic White	103 (47%)	72 (51%)	31 (40%)
ALZpath p‐tau217, pg/mL	0.25 (0.18, 0.37)	0.24 (0.18, 0.34)	0.27 (0.19,0.43)	0.109
(Missing)	4	2	2
Johnson & Johnson p‐tau217+, pg/mL	0.035 (0.026, 0.050)	0.034 (0.026, 0.046)	0.038 (0.025, 0.065)	0.161
(Missing)	22	16	6
p‐tau181, pg/mL	13 (9, 18)	12 (8, 18)	14 (11, 19)	0.011
(Missing)	10	6	4
p‐tau231, pg/mL	9 (6, 14)	8 (6, 13)	10 (7, 15)	0.029
(Missing)	10	7	3
Aß40, pg/mL	68 (53, 79)	68 (53, 82)	68 (53, 78)	>0.900
(Missing)	63	39	24
Aß42, pg/mL	4.82 (3.71, 5.61)	4.82 (3.73, 5.71)	4.79 (3.60, 5.43)	0.630
(Missing)	63	39	24
Aß42/40 ratio	0.072 (0.061, 0.080)	0.072 (0.062, 0.083)	0.072 (0.061, 0.079)	0.591
(Missing)	63	39	24
GFAP, pg/mL	70 (49, 102)	68 (49, 99)	76 (55, 107)	0.517
(Missing)	63	39	24
NfL, pg/mL	13 (9, 18)	13 (9, 18)	15 (10, 21)	0.281
(Missing)	63	39	24
CT Composite Region	2.74 (2.66, 2.79)	2.78 (2.74, 2.82)	2.63 (2.55, 2.67)	<0.001

Abbreviations: Aβ, amyloid beta; GFAP, glial fibrillary acidic protein; HCP, Human Connectome Project; NfL, neurofilament light chain; p‐tau, phosphorylated tau.

^a n (%) or median (IQR).

^b Pearson's chi‐square test for categorical and Wilcoxon rank‐sum test for numerical variables.

A total of 191 individuals had available data on plasma p‐tau measurements and Aβ‐PET, whereas a subset of 142 had measurements from the Neurology 4‐Plex E assay (NfL, GFAP, Aβ42, Aβ40) and for Aβ‐PET. Of these, 96 individuals were used for training and 95 for testing for the p‐tau assays. For the 4‐plex assays, 72 individuals were used for training and 70 for testing. Likewise, 193 individuals had available data on plasma p‐tau measurements and CT‐based MRI, 97 of which were used for training and 96 for testing. A subset of 144 participants had information for the 4‐plex assays and CT‐based MRI data, 72 of which were used for training and 72 for testing.

3.2. Association of plasma biomarkers with amyloid pathology (A status)

Several plasma biomarkers showed statistically significant differences between the A+ and A– groups (Table 1 and Figure 1). The ALZpath p‐tau217 assay had higher median levels (p < 0.001) in the A+ (0.51 pg/mL, IQR: 0.40, 0.80 pg/mL) versus the A– group (0.24 pg/mL, IQR: 0.17, 0.31 pg/mL). The same trend was seen for the Johnson & Johnson p‐tau217+ levels (A+: 0.070 pg/mL, IQR: 0.049, 0.105 pg/mL vs A–: 0.033 pg/mL, IQR: 0.025, 0.044 pg/mL, p < 0.001). Significant differences between the A+ and A– groups were also observed for p‐tau181 (p = 0.014), with median levels at 16 (IQR: 12, 23) pg/mL and 12 (IQR: 9, 18) pg/mL, respectively. Similarly, plasma GFAP levels were higher in the A+ (92 pg/mL, IQR: 79, 155 pg/mL) versus the A– (63 pg/mL, IQR: 46, 95 pg/mL) groups (p < 0.001). The NfL assay had higher median levels in the A+ group (18 pg/mL; IQR: 11, 22 pg/mL) than the A– group (12 pg/mL, IQR: 8, 17 pg/mL) (p = 0.003). The median plasma Aβ42/40 ratio was lower (p < 0.001) in the A+ group (0.060, IQR: 0.056, 0.069) compared with the A– group (0.074, IQR: 0.065, 0.083), as expected. There was no significant median level difference for p‐tau231 (p = 0.472) between A+ (10 pg/mL, IQR: 6, 17 pg/mL) and A– (9, IQR pg/mL: 6, 13 pg/mL) individuals.

FIGURE 1.

Plasma biomarker profiles according to dichotomized amyloid beta–positron emission tomography (Aβ‐PET) status. The boxplots with overlaid data points showed the distribution of plasma biomarkers in Aβ‐PET positive versus negative groups: (A) ALZpath p‐tau217, (B) Johnson & Johnson p‐tau217+, (C) p‐tau181, (D) p‐tau231, (E) Aβ42/40, (F) glial fibrillary acidic protein (GFAP), and (G) neurofilament light chain (NfL). The box represents the interquartile range, with the end points as the 25th and 75th percentiles, and the median line within the box. The whiskers are the most extreme non‐outlier points and any points beyond the whiskers are more than 1.5*inter‐quartile range (IQR) lower than quartile 1 (Q1) or higher than quartile 3 (Q3). p‐values displayed are from the Wilcoxon rank‐sum test. N = 218.

Among all plasma biomarkers, the p‐tau217 assays had the largest fold changes when comparing the A+ and A– groups, with Johnson & Johnson having a fold change of 2.20 and ALZpath of 2.14. These were followed by NfL, GFAP, p‐tau181, and p‐tau231, with fold changes of 1.52, 1.50, 1.33, and 1.19, respectively. Aβ42/40 was lower in A+, at 0.81 times that of A–.

We next examined the strength of the correlation between plasma biomarkers and brain Aβ plaque burden, as assessed by PiB SUVr (Figure 2). All biomarkers showed significant association except for p‐tau231 (p = 0.125). Positive correlations were in the following decreasing order: Johnson & Johnson p‐tau217+ (R = 0.38, p < 0.001), GFAP (R = 0.35, p < 0.001), ALZpath p‐tau217 (R = 0.33, p < 0.001), NfL (R = 0.21, p = 0.009), and p‐tau181 (R = 0.16, p = 0.024). Plasma Aβ42/40 showed an inverse correlation (R of −0.38, p < 0.001).

FIGURE 2.

Scatterplot distribution of plasma biomarkers (Y‐axis) and global PiB SUVR (X‐axis). The data points were color‐coded according to the Aβ‐PET status, with blue dots for Aβ‐PET negative and red dots for Aβ‐PET positive. Correlation coefficients and corresponding p‐values were determined using Spearman correlation. N = 218. The blue line indicates the best‐fit linear regression line, and the gray zones represent the 95% confidence interval of the regression lines. (A) ALZ path p‐tau217, (B) Johnson & Johnson p‐tau217+, (C) p‐tau181, (D) p‐tau231, (E) Aβ42/40, (F) GFAP, and (G) NfL.

3.3. Predictive accuracy of plasma biomarkers for Aβ‐PET status

For plasma biomarker‐only models, we found that both p‐tau217 assays showed a high predictive ability to distinguish the A+ from the A– participants (Figure S1A), with AUCs of 0.919 (95% CI = 0.859–0.981) for ALZpath and 0.914 (95% CI = 0.837–0.992) for Johnson & Johnson assay. Plasma p‐tau181 and p‐tau231 were inferior to p‐tau217, with AUCs of 0.676 (95% CI = 0.543–0.808) and 0.544 (95% CI = 0.366–0.722), respectively (Table S1). All sensitivities, specificities, positive predictive values (PPVs), and negative predictive values (NPVs) for Aβ‐PET status using the biomarker‐only models can be found in the supplement (Table S1).

We then utilized multivariable models to adjust demographic and genetic covariates (age, sex, and APOE ε4 carrier status) and compared the multivariable models to the corresponding biomarker‐only models (Figure S1B). The inclusion of the covariates significantly increased the AUCs for p‐tau181 to 0.858 (95% CI = 0.774–0.942) and the p‐tau231 to 0.860 (95% CI = 0.776–0.944), with p values of 0.023 and 0.003, respectively (Table S2). The AUC for Johnson & Johnson p‐tau217+ showed a slight but nonsignificant increase to 0.922 (95% CI = 0.851–0.992), (p = 0.813). ALZpath, on the other hand, showed a slight decrease of AUC to 0.863 (95% CI = 0.779–0.948). However, the difference was also not statistically significant (p = 0.275). All sensitivities, specificities, PPVs, and NPVs for Aβ‐PET status using the covariate‐adjusted models can be found in Table S2.

Within the 4‐plex assays (Figure S1A), GFAP (AUC = 0.853, 95% CI = 0.749–0.957) had the highest AUC for differentiating A+ and A– participants, followed by Aβ42/40 (AUC = 0.796, 95% CI = 0.658–0.934) and NfL (AUC = 0.773, 95% CI = 0.657–0.890) (Table S1). Adding demographic covariates resulted in slightly higher AUCs for GFAP (AUC = 0.856, 95% CI = 0.765–0.946), NfL (AUC = 0.823, 95% CI = 0.719–0.927), and Aβ42/40 (AUC = 0.903, 95% CI = 0.826–0.979)(Figure S1B and Table S2). However, none of the increases were statistically significant (p > 0.05).

3.4. Plasma biomarker association with neurodegeneration (N status)

When dichotomizing participants based on N status, only p‐tau181 and p‐tau231 showed significant differences, with p‐values of 0.011 and 0.029, respectively (Table 2, Figure 3). Levels of p‐tau181 displayed the highest fold increase, with N+ being 1.18 times that of N– individuals. The level of p‐tau231 was 1.11 times higher in N+ compared to N–. Levels of p‐tau217 from both ALZpath and Johnson & Johnson, GFAP, and NfL also showed a slight increase in N+ compared to N–. However, none of these increases were statistically significant (Figure 3). The ALZpath (p = 0.109) and Johnson & Johnson (p = 0.161) p‐tau217 levels were 1.11 times higher in N+ compared to N–. The level of GFAP was 1.16 times higher in N+ compared to N– (p = 0.517), and the level of NfL was 1.15 times higher in N+ compared to N– (p = 0.281). Aβ42/40 ratio was 0.99 times lower in N+ compared to N–, but this difference was not statistically significant (p = 0.591).

FIGURE 3.

Plasma biomarker profiles according to neurodegeneration status. Shown are the boxplots with overlaid data points for the distribution of plasma biomarker levels in N‐positive and N‐negative groups: (A) ALZpath p‐tau217, (B) Johnson & Johnson p‐tau217+, (C) p‐tau181, (D) p‐tau231, (E) Aβ42/40, (F) GFAP, and (G) NfL. The box represents the interquartile range, with the end points as the 25th and 75th percentiles, and the median line within the box. The whiskers are the most extreme non‐outlier points and any points beyond the whiskers are more than 1.5*IQR lower than quartile 1 (Q1) or higher than quartile 3 (Q3). p‐values displayed are from the Wilcoxon rank‐sum test. N = 220.

In addition, we evaluated the association of plasma biomarkers with the continuous measure of cortical thickness (or CT). Both p‐tau181 and p‐tau231 showed a highly associated inverse relationship with CT, with Spearman correlation coefficient of −0.24 (p < 0.001) and −0.19 (p = 0.005), respectively. Johnson & Johnson p‐tau217+ also showed a marginally significant inverse relationship, with R of −0.14 (p = 0.047), (Figure S2). The remaining biomarkers, including ALZpath p‐tau217, GFAP, NfL, and Aβ42/40, showed no significant association with CT (all p’s > 0.05).

3.5. Predictive accuracy of plasma biomarkers for N status

All p‐tau assays displayed low predictive capabilities in differentiating N+ from N– participants (Figure S3A and Table S3). Specifically, p‐tau181 (AUC = 0.615, 95% CI = 0.494–0.737) had the highest AUC followed by ALZpath p‐tau217 (AUC = 0.601, 95% CI = 0.478–0.724), p‐tau231 (AUC = 0.585, 95% CI = 0.460–0.710), and Johnson & Johnson p‐tau217+ (AUC = 0.558, 95% CI = 0.422–0.693). Including demographic and genetic covariates (age, sex, race, and APOE ε4 status) increased the AUCs of p‐tau. However, none of the increases were statistically significant (all p’s > 0.05), (Figure S3B and Table S4). All sensitivities, specificities, PPVs, and NPVs for neurodegeneration status using the biomarker only (Table S3) and covariate‐adjusted models can be found in Table S4.

In terms of the 4‐plex assays, NfL (AUC = 0.587, 95% CI = 0.441–0.734) had a slightly higher AUC compared to GFAP (AUC = 0.568, 95% CI = 0.422–0.713) and Aβ42/40 (AUC = 0.490, 95% CI = 0.344–0.637), (Figure S3B and Table S3). Adding the same covariates to the models resulted in little to no change in their predictive performance for N status, with AUCs of 0.552 for NfL (95% CI = 0.413–0.692), 0.568 for GFAP (95% CI = 0.432–0.703), and 0.600 for Aβ42/40 (95% CI = 0.461–0.739), (Figure S3B and Table S4).

3.6. Plasma biomarker profiles in combined A and N status groups

We next evaluated whether the observed significance between dichotomized A status depends on N status and vice versa, by categorizing participants into four groups according to the combined A and N statuses. Figure 4 shows the results of the statistical comparisons among groups and Table S5 shows the demographic distribution. All except p‐tau231 showed overall significance among the four groups, with Kruskal–Wallis p‐values < 0.001 for ALZpath p‐tau217, Johnson & Johnson p‐tau217+, GFAP, and Aβ42/40, and p‐values of 0.012 for p‐tau181 and 0.025 for NfL. In addition, p‐tau231 and NfL saw no differences across any pairwise comparisons among the four groups after the Bonferroni correction. Furthermore, none of the biomarkers had statistically significant differences between the N– and N+ groups after controlling for A status (A–N– vs A–N+, and A+N– vs A+N+).

FIGURE 4.

Plasma biomarker profiles according to combined A and N statuses. Shown are boxplot distributions of plasma biomarker levels among the four groups categorized by A and N status. (A) ALZpath p‐tau217, (B) Johnson & Johnson p‐tau217+, (C) p‐tau181, (D) p‐tau231, (E) Aβ42/40, (F) GFAP, and (G) NfL. The box represents the interquartile range (IQR), with the end points as the 25th and 75th percentiles, and the median line within the box. The whiskers are the most extreme non‐outlier points, and any points beyond the whiskers are more than 1.5*IQR lower than quartile 1 (Q1) or higher than quartile 3 (Q3). Significant p‐values are shown from Wilcoxon rank‐sum tests with a Bonferroni correction. N = 218.

ALZpath p‐tau217, Johnson & Johnson p‐tau217+, and Aβ42/40 showed significant differences between A– and A+, regardless of N status. However, all three biomarkers showed larger A+/A– fold change among N+ participants, with fold changes of 3.21 (p < 0.001) for ALZpath p‐tau217, 2.44 (p < 0.001) for Johnson & Johnson p‐tau217+, and 0.79 (p = 0.002) for Aβ42/40 among N+ participants, compared to 1.92 (p < 0.001), 2.05 (p < 0.001), and 0.845 (p = 0.018) in N– participants. In contrast, GFAP only showed significant differences among N+ groups (A–N+ vs A+N+). Its level was 1.76 times higher in A+ compared to A– among N+ participants (p = 0.002), compared to a fold change of 1.41 in N– participants (p = 0.28).

3.7. Impact of self‐identified race on plasma biomarkers and APOE ε4 carrier status

When participants were grouped by self‐identified race (B/AA or NHW), and by APOE ε4 carrier status (carrier or noncarrier), we found that APOE ε4 effects on Aβ and tau biomarkers are more pronounced in NHW participants (Table S6). We found no significant difference between non‐carriers and carriers in B/AA participants for Aβ‐PET positivity or any biomarker. Conversely, for NHW participants, APOE ε4 carriers had higher positivity (44%) compared to noncarriers (18%), suggesting that the APOE ε4 effect is more pronounced in the NHW group (p = 0.015) in detecting Aβ‐PET positivity compared to the B/AA group (p = 0.709). In NHW participants, carriers had higher levels than non‐carriers in ALZpath p‐tau217 (p = 0.005), Johnson & Johnson p‐tau217+ (p = 0.010), and p‐tau231 (p = 0.028). Carriers had lower Aβ42 (p = 0.009) and Aβ42/40 ratio (p = 0.005) in NHW, consistent with amyloid accumulation. Aβ40, GFAP, and NfL showed no significant differences between carriers and noncarriers in NHW or B/AA. This aligns with previous findings using only neuroimaging data. ³⁹

3.8. Impact of self‐identified race on the link between plasma biomarkers and Aβ‐PET status

When grouped by self‐identified race, there were significant differences in several plasma biomarkers. The level of p‐tau231 was 1.33 times higher in those who identified as B/AA compared to NHW (p < 0.001), followed by p‐tau181 with levels 1.15 times higher in B/AA compared to NHW (p = 0.018). The median fold change comparing B/AA to NHW for Johnson & Johnson p‐tau217+ was 0.87 (p = 0.002), and the ALZpath p‐tau217 levels were 0.81 times lower in those who identified as B/AA compared to NHW (p < 0.001). The median level for GFAP was 0.78 times lower for B/AA compared with NHW (p = 0.039). Aβ42/40 levels were 1.03 times higher in B/AA versus NHW, but this difference was not statistically significant (p = 0.354). NfL median levels were 0.913 times lower for those who identified as B/AA compared to NHW (p = 0.066).

Cohort characteristics, separated by self‐identified race and Aβ‐PET status, are shown in Table S6. As shown in the table, several plasma biomarkers exhibited race‐dependent associations with Aβ‐PET status. For example, although plasma p‐tau217 (measured by both Johnson & Johnson and ALZpath) and GFAP showed a statistically significant difference between A+ and A– regardless of self‐identified race, the fold change increases were larger in B/AA participants versus NHW participants. Johnson & Johnson p‐tau217+ had a fold change of 2.29 versus 1.91 (A+/A–) for B/AA and NHW, respectively, but the interaction between race and Johnson & Johnson p‐tau217+ in the logistic regression model was not significant (p = 0.135). ALZpath p‐tau217 showed a fold change of 2.57 for B/AA compared to 1.97 for NHW and the interaction with race was significant (p = 0.024). Similarly, GFAP showed a fold change of 1.89 in B/AA compared to 1.36 in NHW, albeit there was no significant interaction (p = 0.330). Aβ42/40 showed significant differences between A+ and A– regardless of self‐identified race, but the fold change of 0.80 was greater for those who identified as NHW compared to the fold change of 0.82 within the B/AA group. The logistic regression interaction term between race and Aβ42/40 was nonsignificant (p = 0.784).

For p‐tau181 and p‐tau231, significant differences between the A+ and A– groups were recorded only in the NHW group (Figure 5). Specifically, p‐tau181 had a fold change of 1.65 in B/AA versus 1.51 in NHW. For p‐tau231, the fold change was 1.47 for B/AA versus 1.35 for NHW. In contrast, NfL exhibited a significant difference between A+ and A– only in B/AA participants, with a fold change of 2.15 in B/AA and 1.39 in NHW. None of the logistic regression interaction terms between race and p‐tau181, p‐tau231, or NfL were significant when predicting A status, with p‐values of 0.171, 0.269, and 0.322, respectively.

FIGURE 5.

Plasma biomarker levels according to Aβ‐PET status classified based on self‐identified Black/African American (B/AA) versus non‐Hispanic White (NHW) participants. (A) ALZpath p‐tau217, (B) Johnson & Johnson p‐tau217+, (C) p‐tau181, (D) p‐tau231, (E) Aβ42/40, (F) GFAP, and (G) NfL. The box represents the interquartile range (IQR), with the end points as the 25th and 75th percentiles, and the median line within the box. The whiskers are the most extreme non‐outlier points, and any points beyond the whiskers are more than 1.5*IQR lower than quartile 1 (Q1) or higher than quartile 3 (Q3). Statistical comparisons were made using the Wilcoxon rank‐sum test for self‐identified racial groups. p‐values are shown at the top of each plot, underneath the NHW or B/AA title. N = 218. (See Figure 6 and Table S7 for further analysis of these results.)

We evaluated whether there were any racial differences in the association of plasma biomarkers with amyloid burden assessed by global PiB SUVr (Figure 6). Significant racial differences, as determined by the significance of the interaction term between race and biomarkers in the linear regression models were observed for p‐tau217 measured with Johnson & Johnson assay, p‐tau181, and Aβ42/40, with p values of < 0.001, 0.005, and 0.004, respectively. NHW was associated with a steeper incline in global PiB SUVr with increased Johnson & Johnson p‐tau217+ (0.086 for those who self‐identified as B/AA vs 0.208 increase per standard deviation [SD] in NHW). A similar steeper incline was observed for p‐tau181, with an increase from 0.003 for B/AA to 0.134 for NHW in global Aβ PiB SUVr per SD increase in p‐tau181. Aβ42/40, on the other hand, showed a much steeper decrease in NHW, with a 0.027 decrease in PiB SUVr per SD increase in Aβ42/40 in B/AA, compared to a 0.150 decrease for those who identified as NHW. The ALZpath p‐tau217 (p = 0.435), p‐tau231 (p = 0.109), GFAP (p = 0.247), and NfL (p = 0.979) biomarker assays did not show a significant racial impact on their association with global PiB SUVr and the assays.

FIGURE 6.

Correlation of standardized plasma biomarkers with Aβ‐PET uptake according to self‐identified Black/African American (B/AA) versus non‐Hispanic White (NHW) racial groups. p‐values and β estimates (B/AA as reference) are from the interaction term between self‐identified race and biomarkers in a regression model to predict Global Pittsburgh Compound B (PiB) standardized uptake value ratio (SUVr). Blue lines and dots represent self‐identified NHW participants. The red lines and dots represent self‐identified B/AA participants. Only β estimates with a significant p‐value are displayed. N = 218. (A) ALZpath p‐tau217, (B) Johnson & Johnson p‐tau217+, (C) p‐tau181, (D) p‐tau231, (E) Aβ42/40, (F) GFAP, and (G) NfL.

4. DISCUSSION

The use of plasma biomarker tests for detecting Aβ pathology is critical in the effort to independently detect AD preclinically and symptomatically in an efficient and less‐invasive manner than using CSF and neuroimaging methods. ⁴⁰ Although many plasma biomarkers have been shown to be related to Aβ pathology, it is still unclear which amalgamations are superior in predicting A status and N status in a heterogeneous population.

We evaluated the classification accuracies of several plasma biomarkers for brain Aβ pathology and neurodegeneration among a non‐Hispanic White and Black/African American middle‐aged community cohort. Although there have been a number of studies demonstrating high performances of plasma biomarkers, especially p‐tau217, to identify abnormal brain Aβ‐PET, the vast majority of those studies lack diverse representation and are composed heavily of self‐identified NHW participants. ⁴⁰ , ⁴¹ Hence, there is a need for studies focused on widening the participation of diverse populations in biomarker studies. ⁴⁰

Plasma p‐tau217 assays from Johnson & Johnson and ALZpath demonstrated superior classification accuracy in identifying Aβ pathology status, followed by GFAP, Aβ42/40, and p‐tau181. These findings agree with those from several other cohorts/including community/population‐based ones, ⁶ , ³⁰ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ establishing that the value of these biomarkers might be applicable to wider populations. To further emphasize the superior classification accuracy of p‐tau217 in diverse populations, it has been reported that there is no significant difference in the relationship between p‐tau217 and amyloid PET when comparing underrepresented groups in AD to non‐underrepresented groups. ⁴⁴ Furthermore, adjusting the analysis for the demographic covariates of age, sex, and APOE carrier status did not provide much improvement to the AUCs, indicating that the biomarkers themselves can perform well unaided. In a cohort of only 15% A+ individuals and substantially cognitively normal participants, the plasma p‐tau217 assays had the best performance as triaging tests.

Although the aforementioned plasma biomarkers are highly effective at identifying Aβ pathology, current blood‐based assessments are inadequate for characterizing the T and N of the diagnostic framework. For plasma biomarker association with CT, the lack of significant AUC values with N status suggests that these biomarkers individually may not be adequate predictors of N status, thus calling for novel AD‐type N status markers anticipated in future analysis. We did not include diagnostic or cognitive stage as a covariate in CT models, as N status was defined using MRI‐derived measures of AD‐signature CT. Nonetheless, we recognize that biomarker–atrophy associations may evolve across disease stages, and future work should examine whether stage‐specific differences influence the relationship between MRI‐derived CT and AD biomarkers. A potentially useful marker might be brain‐derived tau, which has been shown to be a plasma biomarker that associates with AD‐specific neurodegenerative features. ⁴⁸ , ⁴⁹

Results based on comparing plasma biomarker profiles among groups stratified by A and N statuses indicated that, compared to the A–N– group, the A+N+ group showed the largest differences in p‐tau181, p‐tau217, GFAP, and Aβ42/40 levels, followed closely by the A+N– group. These findings suggest that Aβ pathology is the primary driver of increased plasma biomarker levels in p‐tau181, p‐tau217, GFAP, and Aβ42/40, with neurodegeneration contributing further but with more subtle abnormalities, particularly in the presence of brain Aβ pathology. These results align with published findings that core AD biomarkers, particularly p‐tau181 and p‐tau217, are associated with the earlier phases of AD pathophysiology. ⁴² , ⁴⁵ , ⁵⁰ , ⁵¹ , ⁵²

Possible racial differences in plasma biomarkers for AD have been increasingly recognized, although the underlying causes remain complex and multifactorial. Studies have reported that B/AA individuals tend to show lower levels of p‐tau biomarkers (such as p‐tau181 and p‐tau217) and reduced tau PET binding compared to NHW individuals, even when matched for disease stage. ⁵³ , ⁵⁴ In addition, it has been shown that there are significant differences in plasma eligibility rates between people who self‐identify as B/AA versus NHW. ⁵⁵ B/AA were shown to qualify less frequently for trial eligibility based on the plasma biomarker screen despite showing no racial or ethnic differences in amyloid positivity. Findings regarding Aβ42/40 ratios are more variable, with some studies indicating higher or similar ratios in B/AA individuals, whereas others, including ours, observing the opposite. ⁵⁶ NfL and GFAP levels may also differ by race, but evidence remains limited. ⁵⁷ These differences may not reflect innate biological variation, but rather be influenced by social determinants of health, including access to health care, comorbid conditions (e.g., hypertension, diabetes), chronic stress, and structural racism.

Furthermore, it remains unclear why the p‐tau217 assays from Johnson & Johnson and ALZpath show different results: in the linear regression model, they differ in predicting amyloid burden based on biomarker and global PiB SUVr, and in the logistic regression model, they diverge in their predictive performance as well. Perhaps the differences can be accounted for by the varying affinities to different tau variants; Johnson & Johnson p‐tau217+ has been shown to have high affinity when there is additional phosphorylation at threonine 212, but further research is needed to explore these varying effects. ⁵⁸ Biologically, both ALZpath p‐tau217 and Johnson & Johnson p‐tau217+ assays specifically recognize tau phosphorylated at Thr217. However, Johnson & Johnson p‐tau217+ has enhanced affinity if Thr212 or Ser214 are also phosphorylated, which perhaps shifts associations with PET and clinical stage. ⁴⁴ The “+” designation reflects this enhanced recognition capability. ⁵⁹ Consequently, biomarker thresholds and interpretations need to be contextualized within diverse populations to ensure accurate and equitable diagnosis and treatment.

We observed lower levels of plasma p‐tau biomarkers in B/AA individuals, in agreement with some previous literature (Figure 5). ⁴³ , ⁶⁰ The fold changes were higher in B/AA for all biomarkers. However, we observed an unexpected lower Aβ42/40 ratio in B/AA compared to NHW individuals, opposing several prior publications that found higher or similar Aβ42/40 ratios in B/AA and NHW individuals. ⁴³ , ⁶¹ Although the reasons underlying our findings remain uncertain, they may be influenced by social determinants of health, particularly comorbidities and related health conditions. We also observed that B/AA participants showed less biomarker differentiation by APOE ε4, especially for p‐tau and Aβ measures. The discrepancies between our findings and the existing literature may reflect underlying social determinants of health such as differential access to care, comorbidity profiles (e.g., vascular risk), educational opportunities, and chronic stress exposure, rather than innate racial differences. ⁴⁷

A significant strength of this study is the biracial nature of the cohort, composed of almost equal numbers of NHW and B/AA participants. Most studies examining the association between plasma biomarkers and AD pathology have included predominantly NHW participants in their cohorts. In contrast, our study included almost equal representation of B/AA and NHW volunteers. This allowed us to examine the racial impact on the AT(N) or AT/N/I/V/S frameworks for their real‐world application. Additional advantages of this study include correlating biomarkers to both PET‐PiB imaging and MRI‐based CT, and the evaluation of multiple biomarkers that reflect different pathophysiological processes in AD. Our study is limited by the lack of imaging information for tau status, and also, only two self‐identified racial groups were included: B/AA and NHW. In addition, our analyses relied on self‐identified race, which has important limitations. Self‐report captures social identity and lived experiences that shape health through structural and environmental exposures, but it does not necessarily reflect underlying genetic ancestry. Recent studies using genetically determined ancestry demonstrate that ancestry‐based approaches can provide complementary insights into biological differences in biomarker expression. ⁵⁶ Thus, although our findings highlight disparities across self‐identified racial groups, future work integrating both social identity and genetic ancestry will be important to disentangle social and biological contributors to plasma biomarker variability. Finally, it is important to consider that the thresholds utilized were developed primarily in cohorts consisting of predominantly NHW individuals and may not perform equally across other racial and ethnic groups.

Together, this study evaluated the clinical performance of the plasma biomarkers in identifying neuroimaging‐based A and N statuses. Our study indicated that p‐tau217, measured by both the Johnson & Johnson and ALZpath assays was the best predictor of Aβ‐PET pathology for triaging purposes. No substantial improvement was observed with the inclusion of covariates (age, sex, and APOE ε4 carrier status). Level of p‐tau181 and p‐tau231, but not p‐tau217, showed significant correlation with neurodegeneration status (based on CT). Self‐identified racial identity significantly influenced the association between p‐tau217 (Johnson & Johnson), p‐tau181, Aβ42/40, and Aβ plaque burden measured by PiB SUVr. However, biomarker accuracies for Aβ‐PET positivity were unaffected by self‐identified race, except for ALZpath p‐tau217 (p = 0.024). Thus, the evaluated Aβ‐associated plasma biomarkers demonstrate effectiveness even in populations underrepresented in the development of such tests. These findings support the broader implementation of these tests in general population settings. This deserves further attention to ascertain the widespread nature of these findings and their root causes.

CONFLICT OF INTEREST STATEMENT

Gallen Triana‐Baltzer and Hartmuth Kolb are employees of Johnson & Johnson Research and Development. The plasma p‐tau217+ measurements were performed at Quanterix and managed by Johnson & Johnson Research and Development, but both parties were blinded to sample ID and were not involved in the data analysis. The co‐authors employed by Johnson & Johnson provided comments on the manuscript and provided approval for submission of the manuscript. Xuemei Zeng is an inventor on University of Pittsburgh provisional patents on anti‐tau antibodies and plasma amyloid beta peptide biomarker assays by immunoprecipitation‐mass spectrometry. Thomas K. Karikari has consulted for Quanterix Corporation, SpearBio Inc., Neurogen Biomarking LLC. and Alzheon and has served on advisory boards for Siemens Healthineers and Neurogen Biomarking LLC., outside the submitted work. He has received in‐kind research support from Johnson & Johnson Research Laboratories, SpearBio Inc., and Alamar Biosciences, as well as meeting travel support from the Alzheimer's Association and Neurogen Biomarking LLC., outside the submitted work. Thomas K. Karikari has received royalties from Bioventix for the transfer of specific antibodies and assays to third party organizations. He has received honoraria for speaker/grant review engagements from the National Institute of Health (NIH), Universtiy of Pennsylvania (UPENN), University of Wisconsin‐Madison (UW‐Madison), the Cherry Blossom symposium, the Health and Aging Brain Study‐Health Disparities (HABS‐HD)/Alzheimer's Disease Neuroimaging Initiative 4 (ADNI4) Health Enhancement Scientific Program, Advent Health Translational Research Institute, Brain Health conference, Barcelona‐Pittsburgh conference, the International Neuropsychological Society, the Icahn School of Medicine at Mount Sinai, and the Quebec Center for Drug Discovery, Canada, all outside of the submitted work. Thomas K. Karikari and Xuemei Zeng are inventors on patents and provisional patents regarding biofluid biomarker methods, targets, and reagents/compositions, which may generate income for the institution and/or self should they be licensed and/or transferred to another organization. The other authors report no conflict of interest. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

All participants provided written consent, and the University of Pittsburgh Institutional Review Board approved the study.

Supporting information

Supporting information

Supporting information

Brodman ST, Heaton N, Triana‐Baltzer G, et al. Plasma biomarkers, brain amyloid‐beta pathology, and cortical thickness in a non‐Hispanic White and Black/African American middle‐aged community cohort: The HCP‐CoBRA study. Alzheimer's Dement. 2025;21:e70985. 10.1002/alz.70985