Plasma Alzheimer's biomarkers and brain amyloid in Hispanic and non‐Hispanic older adults

Abstract INTRODUCTION Alzheimer's disease studies often lack ethnic diversity. METHODS We evaluated associations between plasma biomarkers commonly studied in Alzheimer's (p‐tau181, GFAP, and NfL), clinical diagnosis (clinically normal, amnestic MCI, amnestic dementia, or non‐amnestic MCI/dementia), and Aβ‐PET in Hispanic and non‐Hispanic older adults. Hispanics were predominantly of Cuban or South American ancestry. RESULTS Three‐hundred seventy nine participants underwent blood draw (71.9 ± 7.8 years old, 60.2% female, 57% Hispanic of which 88% were Cuban or South American) and 240 completed Aβ‐PET. P‐tau181 was higher in amnestic MCI (p = 0.004, d = 0.53) and dementia (p < 0.001, d = 0.97) than in clinically normal participants and discriminated Aβ‐PET[+] and Aβ‐PET[‐] (AUC = 0.86). P‐tau181 outperformed GFAP and NfL. There were no significant interactions with ethnicity. Among amnestic MCI, Hispanics had lower odds of elevated p‐tau181 than non‐Hispanic (OR = 0.41, p = 0.006). DISCUSSION Plasma p‐tau181 informs etiological diagnosis of cognitively impaired Hispanic and non‐Hispanic older adults. Hispanic ethnicity may relate to greater likelihood of non‐Alzheimer's contributions to memory loss. Highlights Alzheimer's biomarkers were measured in Hispanic and non‐Hispanic older adults. Plasma p‐tau181 related to amnestic cognitive decline and brain amyloid burden. AD biomarker associations did not differ between Hispanic and non‐Hispanic ethnicity. Hispanic individuals may be more likely to have non‐Alzheimer causes of memory loss.

• AD biomarker associations did not differ between Hispanic and non-Hispanic ethnicity.
• Hispanic individuals may be more likely to have non-Alzheimer causes of memory loss.

BACKGROUND
Dementia diagnoses are more common among Hispanic than non-Hispanic older adults. 1,2Alzheimer's disease (AD) is the most common neuropathological finding in patients with dementia. 3Identifying AD during life has become increasingly precise with fluid and neuroimaging biomarkers including cerebrospinal fluid (CSF) and positron emission tomography (PET) measurement of beta-amyloid (Aβ) and phosphorylated tau (p-tau).Several diagnostic AD biomarker tests are now FDA approved, but widespread access remains limited by cost, time, and invasiveness of test procedures.The rapid development of both sensitive and specific plasma-based AD biomarkers has made it likely that many of these issues can be circumvented, thus enabling their use for diagnosis, prognosis, and tracking treatment response.
Plasma tau phosphorylated at residues threonine 181 (p-tau181) and threonine 217 (p-tau217) are the most studied blood-based AD biomarkers.4][15] However, the lack of ethnic diversity is a common criticism of prior plasma biomarker studies, hindering understanding of generalizability and broad applicability.
Although existing data are sparse, at least one study has demonstrated elevated plasma p-tau181 and p-tau217 levels in participants who were Aβ-PET positive, as well as in those with neuropathologically confirmed AD in a multiethnic older cohort. 16There were no clear differences in biomarker performance across race or ethnicity subgroups, though their sample sizes were relatively small especially for studying plasma biomarkers in comparison to amyloid PET.Additionally, symptoms presumed to be related to underlying AD pathology may differ by ethnicity.One study showed lower rates of positive Aβ-PET scans among Hispanic older adults suspected of having AD clinically. 17inical translation of plasma AD biomarkers requires continued evaluation in study cohorts that reflect the growing sociocultural and ethnic diversity of the population.
In the present study, we evaluated how plasma p-tau181, GFAP,

Participants
All study participants were enrolled in the 1FLADRC, which includes older adults spanning the continuum of normal cognition, MCI, and dementia.The 1FLADRC cohort is unique due to its ethnic diversity.Over 50% of participants self-identify as Hispanic (mostly Hispanic/Latino, which we will refer to as "Hispanic" collectively throughout), primarily of Cuban (32%) or South American (18%) origin.Participants are recruited from outpatient memory disorders clinics, free memory screening programs, and community outreach.Currently, only a small number of participants self-identifying as Black/African American have completed both blood draw and PET scan (N = 6).Therefore, this study focused on Hispanic (HW) and non-Hispanic (NHW) older adults self-reporting as "White" with regards to race.

Statistical analyses
Data were analyzed using SPSS v28.Plasma biomarker data were log-transformed.We used analysis of covariance (ANCOVA) adjusting For all analyses, we used an a priori alpha of p < 0.05 (two-sided) and report effect size estimates and/or 95% confidence intervals.
For both GFAP and NfL, only AMN-DEM significantly differed (higher) from the other diagnostic groups.There were no statistically significant interactions between clinical diagnosis and ethnicity for any plasma biomarker.

Plasma biomarkers and Aβ-PET
A total of 240 participants underwent Aβ-PET (63% of the overall sample).The cohort with Aβ-PET did not differ significantly in demographics or disease severity (CDR) from the overall sample.Higher concentrations of all three plasma biomarkers were significantly associated with greater Aβ burden (Figure 2).Youden's Index = 2.39 pg/mL, Sens = 75%, Spec = 85%; Figure 3).

F I G U R E 1
Plasma biomarker concentration differences between clinical diagnostic groups for p-tau181, GFAP, and NfL.For p-tau181, participants diagnosed with AMN-DEM had significantly higher concentrations than all other groups (black lines), and those with AMN-MCI had higher concentrations than CN and IMPnonMCI.For both GFAP and NfL, only participants with AMN-DEM had significantly higher concentrations than other diagnostic groups.was more frequently present among misclassified (54%) than correctly classified participants (36%; p = 0.04).We further evaluated whether Aβ burden was associated with classification accuracy.Among Aβ-PET[−] participants, there were no significant differences in CLs between true negative and false positive participants.Among Aβ-PET[+], there was also no significant difference in CLs between true positive and false negative participants.

DISCUSSION
We evaluated how plasma biomarkers of AD (p-tau181), astrocyte reactivity (GFAP), and neurodegeneration (NfL) related to clinical diagnosis and Aβ-PET in a large cohort of HW and NHW older adults.Plasma p-tau181 was significantly elevated in both AMN-MCI and AMN-DEM and showed the strongest association with Aβ-PET.
GFAP and NfL differed only in the AMN-DEM group, were modestly associated with Aβ-PET, and did not improve the discriminability of Aβ-PET[+] and Aβ-PET[−] participants beyond plasma p-tau181.We found that misclassifying Aβ-PET status based on plasma biomarkers was more likely in APOE ɛ4 carriers (commonly false negatives) but not associated with age, sex, ethnicity, or Aβ burden.Importantly, ethnicity did not significantly influence biomarker performance, consistent with our group's prior work focused on plasma NfL. 19Our findings are among the strongest evidence to date that plasma AD biomarkers have diagnostic utility and adequately reflect AD pathology among both HW and NHW older adults.However, among participants with AMN-MCI, we saw that HW participants had a lower odds of positive AD biomarkers based on a p-tau181, suggesting higher rates of non-AD etiology(ies) underlying memory loss.
Plasma biomarkers were highest in amnestic-predominant clinical phenotypes, but there was significant variability within groups.
This likely reflects an enrichment for, but not universal presence of, AD as a primary etiology among the amnestic phenotypes. 21,22uropathological heterogeneity among patients with memory loss underscores the importance of AD biomarkers, such as plasma p-tau181, for improving accuracy of the etiological differential.Our understanding of clinical syndromes across neurodegenerative diseases like AD inadequately accounts for the possibility that symptoms may manifest differently across ethnocultural groups.This may partially explain recent findings that Hispanic older adults diagnosed with MCI or dementia due to suspected AD were less likely than NHW to have a positive Aβ-PET scan. 17Our findings are similar when using plasma p-tau181.We contend that integrating AD biomarker measurement into ethnically diverse populations is especially important for better characterizing the spectrum of symptoms potentially related (or unrelated) to AD and understanding whether clinicopathological associations differ across underrepresented patient populations.Further, higher rates of medical comorbidities, such as renal and cardiovascular disease, in underrepresented populations may have important implications for the interpretation and utility of plasma biomarkers. 23,24We did not observe a higher rate of vascular risk factors or diagnoses in our AMN-MCI participants with negative AD biomarkers, but further research with more direct measures of cerebrovascular disease burden is required.
Our data add to the growing evidence for plasma p-tau181 as a proxy for amyloid-mediated brain pathology across ethnicities, 16 with the unique feature in this study of having a predominantly Hispanic population.The AUC for differentiating Aβ-PET positive and negative older adults was 0.85, relatively similar to a prior study of p-tau181 (0.77) and p-tau217 (0.84) in a multiethnic cohort with autopsy validation 16 and other investigations comparing to Aβ-PET but with less ethnically representative cohorts. 6,7,25Despite the cost and time burden, Aβ-PET has been used routinely in recent years to screen patients for clinical trials testing of disease-modifying (eg, amyloidlowering) therapies.7][28] This study importantly demonstrates that continued validation of blood-based tools in multiethnic cohorts may considerably facilitate and reduce the cost of screening representative populations for the purpose of identifying amyloid-positive participants who could fulfill selection criteria for AD clinical trials.
We did not identify obvious factors that predicted misclassification of Aβ-PET status using plasma p-tau181.Most misclassified participants were false negatives, which suggests that the derived threshold for plasma p-tau181 may have been too conservative in relation to Aβ-PET positivity in this cohort.APOE ɛ4 carriers were slightly overrepresented in the misclassified participants.Genome-wide association studies demonstrate links between APOE genotype and plasma p-tau181 regulation, 29,30 but these were based on NHW participants.
Associations between APOE status and AD risk are not universal across racial and ethnic groups. 31,32Our sample comprised a higher percentage of APOE ɛ4 carriers within Hispanic participants of varying ancestries than other studies, 33,34  for incorporating plasma AD biomarkers into routine clinical care for diagnostic and prognostic value. 4,35,36Efforts to fully automate and standardize plasma assays will facilitate clinical translation and interpretability of proposed cutoff values in different contexts of use. 35ile our data provide some assurances that stratification or separate cutoffs related to Hispanic ethnicity alone are unnecessary, it is critical to continue prioritizing sociodemographic representation during biomarker development and validation stages.[39] Neither plasma GFAP nor NfL improved upon p-tau181 for identifying significant levels of cortical Aβ burden but may still represent important and independent aspects of AD or other neurodegenerative pathophysiology.Plasma GFAP is considered a marker of astrocyte reactivity and appears closely linked to early Aβ-related pathology, but not tau pathology, in AD. 40 Plasma NfL is a nonspecific marker of neuronal degeneration or injury that only modestly correlates with AD pathology, but could provide complementary information about disease severity, rates of atrophy, or presence of significant co-pathology. 7,19,41Ultimately, a blood-based biomarker panel that informs multiple components of neurodegenerative disease(s) will optimize the in vivo characterization of older adults experiencing cognitive and/or behavioral decline as they age.

Limitations
Our study characterized plasma p-tau181, GFAP, and NfL in relation to clinical diagnosis, Aβ-PET, and atrophy in a unique sample comprised of over 50% Hispanic older adults.However, our cohort had very limited racial diversity, so we focused instead on Hispanic and non-Hispanic older adults self-identifying as White in terms of race.Data were from one multi-institution ADRC and were cross-sectional.[44][45] While the plasma biomarker assays used to test our samples are common and commercially available, raw values and cutoffs do not necessarily translate to other studies and are not directly comparable since these are not fully standardized, automated, and Clinical Laboratory Improvements Amendments-approved tests.We did not have neuropathological data for this cohort, which limits conclusions about specificity of our findings to AD pathology.We also do not know details about the expected high likelihood of mixed neuropathology across our sample including those with positive Aβ-PET, nor did we have detailed information about renal function (eg, estimated glomerular filtration rate), factors which may influence interpretation of plasma biomarker values 23,46 but also represent risk factors for cognitive decline.Angiotensin II inhibitors can be used in treating patients with chronic kidney disease, and reported use was more common in HW than NHW individuals in our sample, but we cannot say with certainty the reasons for self-reported prescribed medications in our sample.The Hispanic older adults in our sample mostly reported family origins in Cuba or South America and results may not generalize to other regions of Hispanic origin that may have different genetic or social determinants of health considerations.Other cohorts such as the Washington Heights-Inwood Columbia Aging Project (WHICAP), 16 the Health & Aging Brain Among Latino Elders (HABLE), 47 and Health and Aging Brain Study-Health Disparities (HABS-HD) 41 will contribute significantly to advancing AD-related research in Hispanic populations.

CONCLUSIONS
Plasma biomarkers reflecting AD pathology (p-tau181, GFAP, NfL) may aid etiological diagnosis of cognitively impaired older adults from Hispanic and non-Hispanic ethnic origins.Hispanic ethnicity alone does not significantly influence the interpretation of how plasma p-tau181, GFAP, and NfL relate to Aβ-PET, but may be linked to greater likelihood of non-AD causes of memory loss.Blood-based biomarkers could help reduce barriers to clinical diagnosis and research participation that disproportionately impact underrepresented sociodemographic groups.
and a marker of neurodegeneration (neurofilament light chain; NfL) differed by clinical diagnosis and related to Aβ-PET in Hispanic and non-Hispanic older adults from the 1Florida Alzheimer's Disease Research Center (1FLADRC).Building off emerging data based on Aβ-PET, we then derived a p-tau181 cutoff for "amyloid positivity" based on participants with a corresponding Aβ-PET to evaluate whether the likelihood of a "positive" p-tau181 test differed between Hispanic and non-Hispanic older adults with a diagnosis of amnestic mild cognitive impairment (MCI).
for age with Tukey-Kramer post hoc analyses to compare biomarker concentrations among five clinical diagnostic groups (CN, AMN-MCI, AMN-DEM, NONAMN, IMPnonMCI).To analyze the association between plasma markers and Aβ-PET, we evaluated (a) how biomarker concentrations related to Aβ burden measured in CLs (Spearman's rho), (b) group differences between Aβ-PET[+] and Aβ-PET[−] (ANOVA), and (c) discriminability (area under the receiver operative characteristics curve [AUC/ROC]).Combined performance of plasma biomarkers for discriminating Aβ-PET[+] and Aβ-PET[−] was evaluated using binary logistic regression and odds ratios (ORs) with and without adjustment for a base model including participant age, sex, and APOE ɛ4 carrier status.We investigated factors potentially influencing the accuracy of Aβ-PET prediction by comparing correctly and incorrectly classified participants on age, sex, ethnicity, and APOE ɛ4 carrier status (ANOVA and chi-square).Interactions between independent variables of interest and ethnicity were evaluated for all analyses to inform whether the observed relationships were significantly different between HW and NHW participants.Lastly, we aimed to build on prior work that used Aβ-PET17 to determine whether plasma p-tau181 differed between HW and NHW older adults diagnosed with amnestic-predominant cognitive decline, the most common clinical phenotype of AD.We focused on individuals diagnosed with AMN-MCI (the largest study group).A Youden's Index cutoff that optimized balance of sensitivity (Sens) and specificity (Spec) to Aβ-PET[+] participants was derived from the cohort that underwent both blood draw and Aβ-PET.Using this cutoff, we created putative AD[+] and AD[−] groups based on plasma p-tau181 (plasmaAD[+]   and plasmaAD[−]) among study participants who contributed blood regardless of whether they had a PET scan.Among those diagnosed with AMN-MCI, we used logistic regression to analyze differences in the likelihood of being plasmaAD[+] between ethnic groups.
Findings did not differ between ethnicity groups.Four participants with very high plasma NfL (> 100 pg/mL, N = 3 AMN-DEM, N = 1 NONAMN) are not shown visually in the figure to avoid extensive y-axis distortion.AMN-DEM, amnestic dementia; AMN-MCI, amnestic MCI; CN, normal controls; GFAP, glial fibrillary acidic protein; IMPnonMCI, cognitively impaired but not MCI; MCI, mild cognitive impairment; NfL, neurofilament light; NONAMN, non-amnestic MCI or dementia F I G U R E 2 Association between plasma biomarker concentrations and cortical Aβ burden.Data are shown in relation to a continuous measure of Aβ burden quantified on the Centiloid scale (top row), and between participants classified as having a positive or negative scan (bottom row).For the associations with the Centiloid scale, statistically significant associations were observed in the overall sample (black line) for all biomarkers, but effect sizes were notably larger for plasma p-tau181 (ρ = 0.59 [0.50 to 0.67]) and GFAP (ρ = 0.42 [0.30 to 0.52]) than for NfL (ρ = 0.23 [0.10 to 0.35]).Results were similar when comparing participants based on dichotomous (positive or negative) Aβ-PET status.The strength of association between plasma markers and Aβ-PET outcomes did not significantly differ between Hispanic (orange) and Non-Hispanic (purple) ethnicity groups.Aβ, beta-amyloid; GFAP, glial fibrillary acidic protein; NfL, neurofilament light; PET, positron emission tomography F I G U R E 3 Area under the curve analysis demonstrating discriminability between Aβ-PET positive and negative participants for each plasma biomarker.Data are shown for the overall cohort, Non-Hispanic participants, and Hispanic participants.Plasma p-tau181 concentrations most accurately discriminated Aβ-PET positive and negative participants (maroon) followed by GFAP (blue) and NfL (green).Discrimination accuracy was similar for both ethnicity groups.Aβ, beta-amyloid; GFAP, glial fibrillary acidic protein; NfL, neurofilament light; PET, positron emission tomography with available APOE genotyping (∼90% of sample), an APOE ɛ4 allele There were 161 participants diagnosed with AMN-MCI (N = 65 NHW, N = 96 HW).Applying the Youden's Index cutoff of 2.39 pg/mL derived from the Aβ-PET cohort, 73 (45.3%) were classified as plasmaAD[+] (ie, plasma p-tau181 > 2.39pg/mL) and 88 (54.7%) as plasma AD[−].HW older adults with AMN-MCI had significantly lower odds of being plasmaAD[+] than NHW older adults with AMN-MCI (36.5% vs 58.5%; OR = 0.41 [0.21 to 0.78], p = 0.006).Results remained significant after controlling for age and sex (OR = 0.45 [0.23 to 0.87], p = 0.02), and were slightly attenuated when further including APOE ɛ4 carrier status in the model (OR = 0.52 [0.25 to 1.1], p = 0.08).When restricting the analysis to the subset of 61 participants with plasma only who did not contribute to the plasmaAD cutoff derivation (N = 25 NHW, N = 36 HW), there remained lower odds of HW being plasmaAD[+], but the effect was not statistically significant due to the smaller sample (OR = 0.39 [0.14 to 1.1], p = 0.08), and again slightly attenuated when further adjusting for age and sex (OR = 0.49 [0.16 to 1.5], p = 0.2) and APOE ɛ4 (OR = 0.53 [0.15 to 1.8], p = 0.3).To explore whether vascular disease risk contributed to discrepancies between plasmaAD status and AMN-MCI diagnoses (ie, cerebrovascular drivers of memory loss), we compared overall VBS scores between plasmaAD[+] and plasmaAD[−] participants with AMN-MCI (Mann-Whitney U).No significant differences were observed for the overall cohort or within the HW or NHW groups separately (all p > 0.5).
Sample characteristics for the overall study sample (N = 379) with plasma biomarker data and for the sub-cohort who underwent both blood draw and positron emission tomography imaging (N = 240, 63% of overall sample) alone misclassified Aβ-PET status for 45 (22%) participants with a tendency towards false negatives (ie, 33/45 were observed Aβ-PET[+] but predicted to be Aβ-PET[−] based on P-tau181).Misclassified participants did not significantly differ from correctly classified participants in age, sex, or ethnicity.Among those TA B L E 1 factors linked to variability in ancestral origins influence agreement between AD biomarker modalities (eg, PET vs CSF vs blood) and associations with AD pathology, and whether and how such relationships are consistent across racial and ethnic groups.Valid blood tests for AD offer the potential of removing barriers to diagnosis associated with more invasive and costly tests, including CSF and PET biomarkers.Persuasive arguments have been made which may reflect differences in recruitment sources and a bias towards presence of cognitive impairment in tertiary clinics typical of ADRC populations.Findings may not generalize to patients evaluated in general medicine clinics or at a general population level.Future work should aim to understand whether polygenetic