Bornhorst, Joshua

Abstract INTRODUCTION Plasma p‐tau217 and p‐tau217/Aβ42 reliably identify non‐Hispanic White individuals with symptomatic and pre‐clinical AD. However, performance across ethnoracially diverse groups remains unknown. METHODS Cognitively normal Black, Hispanic, and non‐Hispanic White participants completed cognitive evaluations, brain magnetic resonance imaging (MRI), amyloid and tau positron emission tomography (PET), and measures of kidney function. Plasma p‐tau217 and Aβ42 were measured using Fujirebio Lumipulse assays. Pre‐clinical AD detection was evaluated using symptomatic and cohort‐derived optimized cutoffs. RESULTS Among 191 participants (57 Black, 46 Hispanic White, 88 non‐Hispanic White), age‐adjusted plasma p‐tau217 and p‐tau217/Aβ42 performed similarly in detecting amyloid‐positive individuals (≥ 25 Centiloids), regardless of race/ethnicity. Both measures independently correlated with amyloid deposition, whereas estimated glomerular filtration rate (eGFR) correlated solely with p‐tau217. Pre‐clinical AD optimized cutoffs (p‐tau217 ≥ 0.132 pg/ml; p‐tau217/Aβ42 ≥ 0.0059) outperformed cutoffs established in symptomatic individuals (sensitivity: 83% vs. 57%; 77% vs. 63%; negative predictive value [NPV] > 93%). DISCUSSION Plasma biomarkers performed similarly across ethnoracially diverse cohorts. Optimized cutoffs may improve early detection and trial enrollment.

This cross-sectional study evaluates the performance of the US Food and Drug Administration–cleared phosphorylated tau 217 (p-tau217) to β-amyloid 42 ratio cut points used to detect amyloid pathology associated with Alzheimer disease.

Abstract Background Decreased kidney function, indicated by a lower estimated glomerular filtration rate (eGFR), may affect interpretation of Alzheimer's disease (AD) blood biomarkers, including phosphorylated tau ( p ‐tau) and amyloid‐beta (Aβ) peptides, due to altered clearance. However, there is limited data available to determine the influence of eGFR on these biomarkers and whether it differs by biomarker or assay type, especially in racially diverse studies of community‐based older adults. Methods We used data from 557 participants in the Mayo Clinic Study of Aging (Rochester, MN n = 182; age 79±8 years; 36.3% Female, White participants) and the UMMC MIND Center‐Mayo Clinic Study of Aging (Jackson, MS n = 375; age 69±8 years; 73% Female 27% Black participants). eGFR was determined using the 2021 CKD‐EPI Creatinine equation. Plasma biomarkers were measured using the Lumipulse immunoassays for p ‐tau217, Aβ42, and Aβ40 (Fujirebio Diagnostics; n = 420); or the mass spectrometry‐based PrecivityAD2 assay (C2N Diagnostics; n = 253) for p ‐tau217, np‐tau217, Aβ42, and Aβ40. Common ratio values were calculated. Aβ42/40 measures were inverted so higher values are worse for all biomarkers. Log base‐2 transformed regression analysis related eGFR to each single analyte and ratio; coefficients (95% CI) reflect estimated doubling in biomarker per doubling in eGFR. Group specific [MN White ( n = 182), MS White ( n = 224), MS Black ( n = 151) participants] estimates are reported, as the distribution of social determinants of health can vary across populations and locations. Results Lower eGFR was associated with higher (worse) AD‐plasma biomarkers, and group stratified estimates were mostly consistent (Figure 1). Ratios of biomarkers diminished but did not entirely alleviate associations with eGFR. For example, MN White participants’ C2N p ‐tau217 association of ‐1.47 (‐1.94, ‐1.00), was attenuated to ‐0.67 (‐1.11, ‐0.22), when using p ‐tau217/np‐tau217. Conclusions Kidney function may be associated with AD‐plasma biomarker values, even with common ratio corrections. There was not strong evidence for highly differential associations of eGFR by race/site categories, though sample sizes are currently limited. The implications for diagnosis and risk classification need to be explored further to support health equity. Optimal consideration of comorbidities on blood biomarker concentrations must continue to be informed by investigation in large, racially and ethnically diverse, community‐based studies.

Abstract Background Plasma p ‐tau217 represents a promising alternative to PET‐ and CSF‐based biomarkers of AD in well‐characterized research cohorts. However, the performance of this accessible biomarker in outpatient clinics comprised of patients with multiple causes of cognitive impairment has yet to be established. Method Plasma Aβ42/40 and p ‐tau217 concentrations were measured using Fujirebio Lumipulse assays in patients evaluated within a subspecialty memory clinic at Mayo Clinic in Florida. Syndromic and etiologic diagnoses were established by consensus, integrating clinical data and findings from MRI ( n = 484), FDG‐PET ( n = 117), and CSF AD biomarkers ( n = 440). Plasma biomarker performance was evaluated using established diagnostic cutpoints to identify patients with “positive/elevated”, “intermediate”, or “negligible” levels of cerebral amyloid pathology on amyloid‐PET. Result Plasma AD biomarkers were measured in 509 patients (mean 68.6±9.3 years; 47% female; 91.3% non‐Hispanic White), including patients with typical amnestic AD ( n = 237, 46.6%), non‐amnestic presentations of AD ( n = 48, 9.4%), and non‐AD causes of cognitive concerns ( n = 224, 44.0%). “Intermediate” plasma p ‐tau217 and p ‐tau217/Aβ42 concentrations were reported in <15% of patients (43% of patients with Aβ42/40), implying the need for further testing to inform the likelihood of AD neuropathologic change in these patients. After excluding these patients, plasma p ‐tau217 reliably distinguished patients with symptomatic AD with 95% sensitivity and 82% specificity. Integration of Aβ42 measures (ptau217/Aβ42) incrementally improved specificity (86%). Diagnostic performance was unaffected by age (> vs < 65‐years). Plasma p ‐tau217 (AUC: 0.94; 95%CI: 0.91‐0.96) and p ‐tau217/Aβ42 performance (AUC: 0.96; 95%CI: 0.94‐0.98) were closely aligned with results of established CSF AD biomarkers ( p ‐tau181/Aβ42), outperforming Aβ42/40 (AUC: 0.78; 95%CI: 0.74‐0.83). Reduced kidney function was associated with elevated plasma p ‐tau217 and p ‐tau217/Aβ42 concentrations in patients without AD (referencing CSF AD biomarkers). Conclusion Concentrations of plasma p ‐tau217, but not Aβ42/40, strongly associated with clinical diagnoses of symptomatic AD and CSF AD biomarker results. Integration of Aβ42 measures yielded marginal improvements in performance, with increased complexity and cost. These findings support the use of plasma p ‐tau217 in heterogeneous clinical cohorts, including patients with multiple causes of cognitive impairment. Caution is advised when interpreting p ‐tau217 concentrations in patients with decreased kidney function, due to potential elevations in plasma biomarker concentrations.

Abstract Background Black and Latino Americans experience a disparate dementia burden yet remain underrepresented in dementia research, including clinical trials designed to slow or prevent symptomatic Alzheimer's disease (AD). Accessible and non‐invasive AD blood biomarkers may improve identification and recruitment of underrepresented participants into AD clinical trials. There is a clear need to assess the performance of emergent plasma AD biomarkers for the detection of AD neuropathology in cognitively normal Black and Latino Americans. Method Cognitively normal Black and Latino participants underwent cognitive evaluations and PET (amyloid‐PiB, tau‐flortaucipir) imaging at Mayo Clinic in Florida. Plasma p ‐tau217 was measured using Fujirebio Lumipulse assays. We evaluated the diagnostic performance of a p ‐tau217 threshold (≥0.186 pg/mL) known to identify non‐Hispanic White participants with cognitive impairment and “positive” amyloid PET scans (>25 Centiloids). Next, we derived an optimized threshold to identify participants with “positive” amyloid PET using Youden's method. Result The cohort included 58 Black (55%) and 47 Latino‐White (45%) well‐educated (mean 16.1±2.3 years) individuals, with mean age 64.7±8.8 years, and slight female bias (63%). Plasma p ‐tau217 concentrations increased with age (rho=0.269, p = 0.004) and declined with MoCA scores (rho=‐0.225, p = 0.027). Sixty‐nine participants completed amyloid‐ and tau‐PET imaging. Plasma p ‐tau217 concentrations increased with amyloid‐ (rho=0.254, p = 0.044) but not with tau‐PET standardized uptake value ratio (rho=0.127; p = 0.334). p ‐tau217 concentrations were ≥0.186 pg/mL in 10/69 participants (14%), including 3/3 amyloid‐ and tau‐PET positive (A+T+) participants, 3/8 A+T‐ participants (sensitivity: 55%), and 4/58 A‐T‐ participants (specificity 93%). An optimized p ‐tau217 threshold of ≥0.118 pg/mL identified 10/11 A+T± participants (sensitivity 91%), and 16/58 A‐T‐ participants (specificity 73%). In A‐T‐ participants, estimated glomerular filtration rate was lower in participants with ‘false positive’ elevations in plasma p ‐tau217 (mean difference: 67 vs 81 ml/min/1.73m 2 , p = 0.004), emphasizing the need for caution when interpreting plasma p ‐tau217 concentrations in patients with decreased kidney function. Conclusion Plasma p ‐tau217 thresholds validated in participants with symptomatic AD (≥0.186 pg/mL) failed to identify early AD neuropathology in 5/8 cognitively normal Black and Latino participants. Optimized thresholds are required to improve detection of preclinical AD in Black/Latino individuals.

Abstract Background Artificial intelligence (AI) augmented laboratory tests can improve quality, efficiency, cost-effectiveness, and staff satisfaction. However, the clinical success of these tests can introduce unforeseen challenges for model retraining. This study describes the discovery of an “AI label feedback loop” in a clinically implemented AI-augmented kidney stone composition test. Methods An AI-augmented kidney stone composition test (V1) has been previously deployed for clinical kidney stone characterization. After several years of clinical use, a retrained model (V2) was developed using 6 times more data. Model performance of both V1 and V2 were evaluated across 3 datasets: a recent production validation (hold-out) set (mostly AI-influenced labels), the original V1 validation set (pre-AI, entirely human-labeled), and a subset of recent cases with exclusively human-generated or human-corrected labels. Results V2 demonstrated a 10% lower concordance rate than V1 when evaluated on the recent production hold-out set, despite a much larger training dataset. Performance between V1 and V2 was similar when applied to the pre-AI validation set. Notably, V2 outperformed V1 on the recent subset of cases with human-only or human-corrected labels, particularly for less-common stone types. These findings revealed an AI label feedback loop, confounding retraining and evaluation. Conclusions The integration of AI into clinical practice can potentially influence reported test results, complicating the development and evaluation of future models. To mitigate AI label feedback loops, ongoing human annotation and careful validation set construction are essential. These strategies can ensure reliable performance assessment and support the safe evolution of clinical AI systems.