Prefrontal hemodynamic alterations during resting state and stroop interference in cognitively healthy older adults with high plasma pTau-217 levels Authors: Abdulhakim Al-Ezzi, Choo Phei Wee, Xiaomeng Wu, Natalie Astraea, Michael T Kleinman, Wendy Mack, Robert A Kloner, Xianghong Arakaki Published: 2026 ABSTRACT AbstractPlasma phosphorylated tau-217 (pTau-217) has emerged as a sensitive biomarker for early Alzheimer's disease (AD) pathology, including the cognitively healthy (CH) stage when compensatory frontal activity has been reported. This study focused on hemodynamic changes measured using functional near-infrared spectroscopy (fNIRS) in the prefrontal cortex (PFC) in CH older adults. The fNIRS data from 41 CH older adults during resting state, as well as during a color-word Stroop interference task were analyzed. Primary measures included oxyhemoglobin (HbO) responses at left lateral, left medium, right medium, and right lateral PFC regions, exploratory measures included heart rate, Weighted Phase Lag Index (wPLI) for functional connectivity (FC), and correlation between HbO and heart rate (HR) to assess heart–brain coupling. The study revealed that compared to low tertile group, individuals with high tertile pTau-217 exhibited slower response times during the task and region-specific hemodynamic changes - reduced HbO during eyes-open rest (FDR-adjusted p = 0.037) but increased activation during Stroop congruent trials (False Discovery Rate (FDR)-adjusted p < 0.0001), which are supported by sensitivity analysis. Additionally, high tertile pTau-217 group showed strong positive correlation between post-task HR and right-lateral prefrontal HbO (ρ = 0.85, p < 0.001) versus a negative correlation in the low tertile group (ρ = −0.55, p = 0.043) and attenuated interhemispheric synchronization by FC (measured by wPLI), particularly during eyes-open resting state. Overall, the primary findings demonstrate pTau-217-associated differences in prefrontal hemodynamic responses, while the exploratory heart–brain coupling and FC findings are hypothesis-generating and require independent validation. Larger and more diverse cohorts are needed to confirm and generalize these findings.Keywords: AD, pTau-217, fNIRS, Hemodynamics, Functional connectivity, Stroop task, Heart–brain coupling Introduction There is increasing attention to detection of early AD pathology, which has been associated with elevated plasma pTau-217 levels (Barthelemy et al., 2024; Cai et al., 2023; De Meyer et al., 2024). Intriguingly, cognitively healthy (CH) older adults with early AD pathology exhibited deficits in executive function, which may precede memory decline (Harrington et al., 2013). Executive functions are implemented in prefrontal cortex (PFC) and encompass three functions: inhibitory control, working memory, and cognitive flexibility (Diamond, 2013). Inhibitory control, as a common element among core executive functions, is usually tested by the Stroop interference task (Miyake and Friedman, 2012). Several studies have shown that individuals with amyloid and tau burden in body fluid or brain tissue demonstrate altered behavioral performance and atypical brain activities during such tasks (Al-Ezzi et al., 2024a; Al-Ezzi et al., 2024b; Arakaki et al., 2020; Arakaki et al., 2022; Bélanger et al., 2010; Coderre and van Heuven, 2013; de Leeuw et al., 2025). Brain activity can also be measured by functional near-infrared spectroscopy (fNIRS) by detecting changes in oxyhemoglobin (HbO) (Friedman and Robbins, 2022; Xie et al., 2024). The ability to capture subtle changes in cortical activation patterns, often interpreted as reflecting neural efficiency, along with vascular responsiveness and task-evoked hemodynamic responses, makes fNIRS a valuable tool for studying brain function in populations at risk for neurodegeneration (Herrmann et al., 2017; Katzorke et al., 2018; Mei et al., 2023; Wang et al., 2024). Beyond localized activation, functional connectivity (FC) offers insight into the distributed nature of neural dysfunction in preclinical AD (Al-Ezzi et al., 2024a; Wang et al., 2013). FNIRS studies have shown that mild cognitive impairment (MCI) is associated with disrupted functional connectivity (FC), particularly reduced resting-state connectivity in prefrontal networks, connectivity patterns from left-to right-dominant prefrontal regions, and altered interhemispheric interactions (Li et al., 2019; Mizrak et al., 2024; Wang et al., 2015). Task-based studies further reveal condition-dependent changes in FC, including compensatory increases and shifts in network organization, such as rightward redistribution of prefrontal activity in AD patients (Ateş et al., 2017; Bu et al., 2019; Vermeij et al., 2017; Wang et al., 2022). Among multiple FC analysis methods, wPLI is a robust measure of FC that focuses on non-zero phase lags while minimizing the confounding effects of volume conduction (Shafiei et al., 2024). It remains unclear whether there are alterations in PFC HbO or FC during resting or inhibitory control processing in CH older adults with elevated plasma pTau-217. Emerging evidence suggests that autonomic regulation, cerebral perfusion, and brain activity are compromised in the early stages of AD (Nair et al., 2023). Alterations in heart rate and its coupling with brain activity may reflect broader disruptions in neurovisceral central autonomic network (Molloy et al., 2023; Thayer and Lane, 2000). The heart-brain interactions may be altered with early AD pathology, especially during cognitively demanding tasks (Sinha et al., 2025). It remains unclear whether the link between HbO and HR differs by early AD pathology, e.g., between CH older adults with or without elevated plasma pTau-217. The present study has several aims: 1) To investigate prefrontal hemodynamic activity and heart–brain interactions during resting and task conditions in individuals stratified by plasma pTau-217 levels. 2) To identify changes of FC in those with higher plasma pTau-217. In this study, we hypothesized that CH older adults with higher plasma pTau-217 would show altered prefrontal hemodynamic responses compared with those with lower pTau-217, including lower resting HbO and greater task-related HbO change during Stroop performance. In the exploratory analyses, we further hypothesized that higher pTau-217 would be associated with altered HR-HbO relationships and reduced prefrontal functional connectivity, particularly across interhemispheric prefrontal regions. Methodology Participants Participants were enrolled through the Brain Aging Study at Huntington Medical Research Institute (HMRI) [9, 12]. Written informed consent was obtained from all participants under an Institutional Review Board-approved protocol (HMRI Pro00038520). Participants were eligible if they were older than 60 years and classified as cognitively healthy following an approximately 4-h neuropsychological assessment using age-, sex-, and education-adjusted norms. The battery included the Wechsler Adult Intelligence Scale-Third Edition (WAIS-III), Clinical Dementia Rating (CDR), Montreal Cognitive Assessment (MoCA), and Mini-Mental State Examination (MMSE). Exclusion criteria comprised untreated medical conditions, anticoagulant use, or other contraindications to lumbar puncture. Forty-one CH participants were classified into tertiles based on plasma pTau-217 levels (1.9- 18.9 pg/mL). The analysis included low (mean ± SD 2.63 ± 0.37 pg/mL, n = 14) and high tertile (10.09 ± 3.6 pg/mL, n = 14) pTau-217 groups; the mean pTau-217 in the high tertile group is similar to the Youden-defined cutpoint (10 pg/ml) to discriminate AD and control individuals(Kivisakk et al., 2024). There is no established cutoff for pTau-217 to estimate risk of cognitive decline in CH older adults, and quartiles have been used to study potential changes in HRV with cardiac mortality (de Bruyne et al., 1999; Nantsupawat et al., 2022; Singh et al., 2018; Tegegne et al., 2023). Therefore, we ranked the participants based on their plasma pTau-217 levels and compared first tertile and third tertile groups for fNIRS analysis, to maximize sample sizes and pTau-217 differences by excluding the middle tertile samples (n = 13). To address the limitations of categorization and retain information from the full cohort, we also performed sensitivity analyses in all 41 participants with plasma pTau-217 modeled as a continuous predictor. Comprehensive assessments included demographic and clinical evaluations, physical examinations, and blood biomarkers. The tertile analysis was followed up by a sensitivity analysis including all participants and plasma pTau-217 levels as a continuous variable. We measured phosphorylated tau protein 217 (pTau-217) in plasma using the S-PLEX Human Tau (pTau217) kit (MSD, Rockville, MD, USA) and the MSD SQ 120 MM electrochemiluminescence platform (Meso Scale Discovery, Rockville, MD, USA) at HMRI's Analytical Biochemistry Core Laboratory. All blood samples were collected, fractionated, and stored following approved laboratory standardized operating procedures prior to analysis. Analyte and quality control data were processed using the Discovery Workbench Software (MSD, Rockville, MD, USA). All batches included an 8-point standard curve and QC verified acceptance criteria, to ensure method conformance. Samples were run in duplicate with a replicate CV < 20% and values reported as mean concentration. Experimental paradigm and behavioral analysis The color-word Stroop task included congruent and incongruent trials. Each trial began with a fixation cross presented at the center of the screen for 250 ms, followed by a blank screen for another 250 ms. Subsequently, colored words were displayed one at a time, for either congruent trials (e.g., word "Red" in red font) or incongruent trials (e.g., word "Green" in red font). Participants were instructed to respond based on the color of the font, rather than the word itself, by pressing designated keys on a keyboard: ‘1’ for red, ‘2’ for blue, and '3' for green (Fig. 1). Participants practiced before the task trials. The task included three blocks with 110 trials for each block. Task-related epochs for the Stroop congruent and incongruent conditions were defined relative to stimulus onset (−200 to 1500 ms). Time 0 was defined at word onset. Baseline correction was applied using the pre-stimulus interval (−200 to −100 ms). Considering the slower hemodynamic changes, the fNIRS analysis here is based on the averaged post-stimulus activity from 150 to 1500 ms (Arakaki et al., 2022; Hung et al., 2022). Stroop interference testing paradigms. Participants were requested to respond to the color of the word and ignore the meaning of it by press a button (‘1’ for red, ‘2’ for blue, and '3' for green). Tasks include random mixture of low load Congruent trials (Word and color match) or high load Incongruent trials (Word and Color does not match). For behavioral performance, accuracy (ACC) and reaction time (RT) were calculated as previously reported (Al-Ezzi et al., 2024b; Arakaki et al., 2022). Briefly, ACC was defined as the percentage of correctly responded trials among total trials and RT was defined as the duration from stimulus onset to response on correct trials. All participants’ Demographic and clinical variables, including age, years of education, MMSE, plasma pTau217 levels, and MoCA scores are measured. fNIRS data collection Functional near-infrared spectroscopy (fNIRS) data were recorded during resting state with eyes open for 5 min, eyes closed for 5 min, and during the Stroop task (congruent and incongruent trials), from the prefrontal cortex (PFC) using a fNIR100 device and COBI Studio software (Biopac, Inc., San Diego, CA, USA). The sampling frequency was set at 10Hz. Hemodynamic activation was quantified using HbO signals, given their higher contrast-to-noise ratio and greater sensitivity to task-evoked cortical responses compared to HbR. Experimental studies combining NIRS with blood-oxygenation-level-dependent (BOLD) fMRI have demonstrated that HbO exhibits stronger and more consistent correlations with BOLD signals than HbR after accounting for measurement-related confounds, suggesting superior sensitivity of HbO to functional activation (Dhamo et al., 2024; Steinbrink et al., 2006; Strangman et al., 2002). This enhanced performance is attributed, in part, to the higher signal-to-noise characteristics of HbO measurements in optical imaging. For each condition (eyes open, eyes closed, Stroop congruent, Stroop incongruent), HbO time series averaged across the task period to obtain a representative activation value per region of interest (ROI), consistent with standard block-averaging approaches in fNIRS analysis (Yücel et al., 2025). HbO and heart rate (HR) correlation The fNIRS signals over the PFC were preprocessed on a channel-by-channel basis using the fNIRSOFT toolbox (Biopac, Inc., San Diego, CA, USA). fNIRS signals were acquired from 18 prefrontal channels (16 long and 2 short) and grouped into four regions of interest (ROIs): Left Lateral (LL), Left Medial (LM), Right Medial (RM), and Right Lateral (RL), each comprising four channels (LL: 1-4; LM: 5-8; RM: 9-12; RL: 13-16), and short-separation channels (17,18). HbO and HbR concentration changes were calculated using the modified Beer-Lambert law with the Drexel Adult 2.5 cm preset. Two wavelengths were used: 730 nm and 850 nm. The extinction coefficients were: Hb at 730 nm = 1.1022, Hb at 850 nm = 0.69132, HbO2 at 730 nm = 0.390, and HbO2 at 850 nm = 1.058. The fixed DP/pathlength parameter was 0.015. Age- and sex-specific Differential Pathlength Factor (DPF) corrections were not applied; therefore, HbO values were interpreted as relative concentration changes rather than absolute chromophore concentrations. For illustration in figures, HbO signals were averaged within each ROI to obtain a representative measure of regional activation. Heart rate was measured twice before and twice after the task using an Omron device, and the averaged HR was used as pre-task and post-task HR, respectively. Continuous task-synchronized HR was not available; therefore, HR-HbO analyses were designed to examine associations between regional HbO and tonic pre/post-task autonomic state rather than beat-to-beat physiological coupling during individual Stroop trials. To improve data quality, several refinement steps were applied (Fig. 2). Briefly, raw fNIRS signals were preprocessed on a channel-by-channel basis using the fNIRSoft toolbox (Biopac Inc., USA). Ambient light artifacts were first removed to reduce background noise contamination. Raw light intensity signals were then converted to optical density using a logarithmic transformation. Motion artifacts were mitigated using a sliding-window motion artifact rejection (MAR) algorithm, with a window size of 10 samples and upper and lower thresholds set at 25 and 3, respectively, to remove transient disturbances caused by participant movement. To isolate task-relevant hemodynamic fluctuations, slow baseline drift below 0.01 Hz was removed, and a low-pass finite impulse response (FIR) filter (Hamming window, order = 100) with an absolute cutoff frequency of 0.1 Hz was applied. This preprocessing yielded an effective analysis band of approximately 0.01–0.1 Hz, attenuating high-frequency physiological noise, including cardiac (∼0.4-2 Hz) and respiratory (∼0.2–0.3 Hz) components, while preserving slow neurovascular signals relevant to cortical activation(Reddy et al., 2021; Yucel et al., 2021). Hemoglobin concentration changes, including oxyhemoglobin (HbO) and deoxyhemoglobin (HbR), were subsequently estimated using the modified Beer–Lambert law (MBLL), as given in equation (1) and equation (2). The resulting signals were used in subsequent analyses of task-evoked cortical activation and functional connectivity. Where ΔA is the change in optical density absorbance, I is a received light intensity, I0 represents the incident light intensity, εi is the extinction coefficient of chromophore, ci is the concentration change of chromophore (μM), d represents the physical distance between light source and detector (cm), DPf is the differential pathlength factor, and G is the geometry-dependent scattering loss. Overview of the fNIRS data acquisition, preprocessing, feature extraction, and statistical analysis pipeline. Raw light intensity signals at left lateral (LL), left medial (LM), right medial (RM), and right lateral (RL) regions were converted to optical density, followed by motion artifact rejection and temporal filtering (drift removal <0.01 Hz and low-pass filtering at 0.1 Hz), resulting in an effective analysis band of 0.01–0.1 Hz. HbO and HbR were estimated using the modified Beer–Lambert law (MBLL). Feature extraction included task-evoked HbO activation and functional connectivity (measured via wPLI), and statistical analyses were performed using mixed-effects linear models and Spearman correlations. In our case, we used two wavelengths (i.e., 730 nm and 850 nm): Functional connectivity estimation HbO signals were first band-pass filtered at 0.01–0.1 Hz to isolate slow hemodynamic fluctuations commonly examined in fNIRS connectivity analyses. Functional connectivity was estimated using the wPLI, computed on these HbO signals. Phase-based metrics quantify the consistency of non-zero phase relationships between signals over time, capturing temporal coordination of slow hemodynamic oscillations beyond amplitude correlations. HbO signals from the 16 long-separation channels were averaged into four prefrontal ROIs: LL, LM, RM, and RL. For each participant and trial, the ROI-level HbO segment corresponding to the 0.15-1.5s post-stimulus window was extracted. Instantaneous phase was estimated using the Hilbert transform, and pairwise phase-lag connectivity values were calculated between ROIs from the phase-difference time series (16 connections). Connectivity matrices were averaged across trials for each participant and then compared between low and high pTau-217 groups. wPLI is derived from the imaginary component of the cross-spectrum, weighting phase differences according to their magnitude rather than treating all phase lags equally (Schrooten et al., 2017; Vinck et al., 2011). Although originally developed for electrophysiological data, phase-based approaches have been increasingly explored in fNIRS to characterize neurovascular dynamics. Prior work has demonstrated that phase relationships between HbO and HbR signals within the 0.01-0.1 Hz range reflect physiologically meaningful processes, including cerebral autoregulation and blood flow dynamics, and vary systematically across the human lifespan (Liang et al., 2021). These findings support the use of phase-based connectivity measures to probe functional brain organization using fNIRS signals. Higher wPLI values indicate a stable, directionally consistent phase lead or lag between signals over time, reflecting putative functional interactions rather than instantaneous signal mixing (Shafiei et al., 2024). In this study, wPLI was selected as a robust phase-based metric to quantify inter-regional synchronization of slow hemodynamic fluctuations. The wPLI is defined as: Where Axy,t represents the cross-spectral density over time, img(Axy,t) respresnts the imaginary part of the complex cross-spectrum at time/frequency bin t, and sgn(img(Axy,t) represents the sign function, which returns +, −1, and 0 if the imaginary parts are positive, negative, and zero, respectively. Statistical analysis Mixed-effects linear regression models were used to examine group differences in HbO between low- and high-tertile plasma pTau-217 levels across regions and task conditions. To evaluate this, a three-way interaction term (pTau-217 group x brain region x task) was included in the model (and the component 2-way interactions), along with fixed effect variables for pTau-217 group, brain regions (left lateral, left medial, right medial, right lateral PFC), and task conditions. The model incorporated two random effects, one for participant and one for brain region nested within participant, to account for the clustering of repeated HbO measurements across brain regions within individuals and the correlation of multiple measurements within participants. Both unadjusted analyses, with no covariates, and adjusted analyses controlling for age, sex, and years of education were performed. Task- and brain region-specific HbO estimates, as well as HbO estimates for task contrasts (congruent minus eyes open, and eyes closed minus eyes open) at each brain region for the pTau-217 group comparison, along with their corresponding 95% confidence intervals were obtained using contrasts and linear hypothesis testing of the relevant regression model parameters. To include all participants, the tertile analysis was followed by a sensitivity analysis using the same linear mixed model considering pTau-217 as a continuous variable including interaction terms. The false discovery rate (FDR) using the Simes method was applied to adjust for multiple-comparisons, and statistical significance was set at 0.05. Additionally, a mixed-design ANOVA was conducted to examine the effects of Group (Low vs. High pTau-217; between-subject factor) and condition (congruent vs. incongruent; within-subject factor) on both ACC and RT. The analysis was implemented in MATLAB using fitrm and ranova functions, followed by post-hoc comparisons and visualization of mean ± SEM values for each condition. Differences in wPLI connectivity between the low pTau-217 and high pTau-217 cohorts were assessed using a two-tailed two-sample t-test. Spearman's rank correlation was applied to estimate the correlation between HbO concentrations across prefrontal regions and HR (pre- and post-task HR), stratified by plasma pTau-217 group. Therefore, the HbO mixed-effects models were considered the primary analyses and were FDR-corrected. The behavioral, wPLI connectivity, or correlation analyses were exploratory, thus there was no adjustment for multiplicity, and effect sizes (Cohen's d) were calculated. These findings are intended to guide hypothesis generation, and results should be interpreted within this exploratory framework. The statistical analyses were performed using StataNow/SE 19.5 (StataCorp, College Station, TX) and MATLAB (Version, 2022b) (The MathWorks, Inc., Natick, MA). Results Behavioral analysis A mixed-design ANOVA revealed a significant main effect of task condition on accuracy (ACC), F (1,26) = 15.31, p = 0.0006, with participants demonstrating higher accuracy during congruent trials (0.96 ± 0.04) than incongruent trials (0.82 ± 0.19) (Fig. 3A). The group main effect was not significant, F (1,26) = 2.26, p = 0.145, and the group × condition interaction did not reach significance, F (1,26) = 3.22, p = 0.085. For RT, there was a strong main effect of condition, F (1,26) = 51.18, p < 0.001, with faster responses during congruent (716.88 ± 90.15 ms) compared to incongruent trials (834.15 ± 145.38 ms) (Fig. 3). Neither the group main effect (F (1,26) = 2.18, p = 0.152) nor the group × condition interaction (F (1,26) = 0.001, p = 0.974) was significant. Group differences in behavioral performance during the Stroop interference task stratified by Plasma pTau217 levels. Mean accuracy (%) (A) and response time (ms) (B) are shown for congruent and incongruent conditions in participants with low and high pTau217. Additionally, a mixed-design ANOVA was conducted to examine the effects of Group (Low vs. High pTau-217; between-subject factor) and Condition (Congruent vs. Incongruent; within-subject factor) on both ACC and RT. Error bars represent the Standard Error of the Mean (SEM). * Indicates p < 0.05; ***p < 0.001; ns = not significant. Direct between-group comparisons showed that individuals in the high pTau-217 group exhibited slower RTs than the low pTau-217 group during both incongruent (0.87 ± 0.18 s vs. 0.80 ± 0.11 s, p = 0.273, Cohen's d = 0.43) and congruent trials (0.75 ± 0.09 s vs. 0.69 ± 0.08 s, p = 0.066, Cohen's d = 0.73) with medium effect size. Accuracy did not differ significantly between groups for congruent trials (0.96 ± 0.04 vs. 0.95 ± 0.05, p = 0.765, Cohen's d = 0.11), however high pTau-217 group had lower accuracy than low pTau-217 group for incongruent trials with medium effect size (0.76 ± 0.24 vs 0.88 ± 0.09, p = 0.103, Cohen's d = −0.64). After the study session, participants completed a brief questionnaire assessing their perceived physical discomfort, mental effort, task difficulty, and fatigue (Table 1). Responses were recorded on a 0-3 severity scale, with 0 indicating no difficulty/discomfort and 3 indicating severe difficulty or discomfort. Intermediate ratings, such as 1.5, were permitted when participants felt their experience fell between two scale points. Table 1Demographics and post-study subjective effort report.Table 1VariableLow pTau-217 (M ± SD)High pTau-217 (M ± SD)Sample (N)1414Age (Year)71.07 ± 9.276.6 ± 7.7Sex (F/M)10/412/2 †Education Years17.46 ± 1.816.2 ± 1.7MMSE29.67 ± 0.728.9 ± 1.4pTAu-217 (Pg/ml)2.63 ± 0.3710.09 ± 3.6 **MoCA28.1 ± 1.328.2 ± 1.86Handedness (Right/Left/mixed)11/1/211/2/1 †Post-study questionnaires 0: Not at all – 3: Severe; (Median (IQR))Do you have any physical troubles during tasks0 (0-0)0 (0-0)Difficulty: Color-Word Matches1 (1-1)1 (1-2)Difficulty: Color-Word Mismatches1 (1-1)2 (1-2) **How much effort did you use for the task2 (1-2)2 (1-2)Do you feel tired after the task1 (1-1)1 (1-2)Demographic and clinical variables, including age, years of education, MMSE, plasma pTau217 levels, and MoCA scores, were compared between groups using two-sample t-tests. Categorical variables, including handedness and sex, were compared using Fisher's exact test †. Post Questionnaire data are reported as median and interquartile range (IQR). Group comparisons for post-study questionnaire measures were performed using a nonparametric approach using the Mann–Whitney U test (Wilcoxon rank-sum test). **: p < 0.01. Demographics and post-study subjective effort report. Demographic and clinical variables, including age, years of education, MMSE, plasma pTau217 levels, and MoCA scores, were compared between groups using two-sample t-tests. Categorical variables, including handedness and sex, were compared using Fisher's exact test †. Post Questionnaire data are reported as median and interquartile range (IQR). Group comparisons for post-study questionnaire measures were performed using a nonparametric approach using the Mann–Whitney U test (Wilcoxon rank-sum test). **: p < 0.01. Oxygenated hemoglobin concentration analysis During Stroop congruent trials, HbO in the right medial region was statistically significantly higher in the high-tertile pTau-217 group compared with the low-tertile group after adjusting for age, sex, and years of education (FDR-adjusted p = 0.032), whereas HbO levels in the right lateral region were higher but did not reach statistical significance (FDR-adjusted p = 0.052; Table 2a) (Fig. 4A). HbO was also statistically significantly higher in the left medial region in the high-tertile pTau-217 group after adjusting for age, sex, and years of education (covariate-adjusted mean difference = 0.006, 95% CI = 0.001 to 0.01, FDR-adjusted p = 0.04). During eyes open, HbO in the right lateral region was significantly lower in the high-tertile pTau-217 group in the adjusted model (FDR-adjusted p = 0.037) (Fig. 4B). Scalp topographies of Cohen's d further illustrate these region-specific HbO differences between groups for both the Stroop congruent (Fig. 4C) and eyes open (Fig. 4D) conditions. Further analysis of HbO with plasma pTau-217 levels in all participants revealed a negative correlation (ρ = −0.41, p = 0.008): higher plasma pTau-217 related to lower levels of HbO in right lateral PFC (Fig. 4E). Additionally, the HbO changes from eyes open to congruent trials, which reflected HbO response with task, was significantly higher in the high-tertile pTau-217 group compared with the low-tertile pTau-217 group after adjusting for age, sex, and years of education for the right lateral (covariate-adjusted mean difference = 0.01, 95% CI = 0.007 to 0.02, FDR-adjusted p < 0.0001), left medial, and right medial (both FDR-adjusted p < 0.0001) regions (Fig. 4F). Finally, HbO in the right lateral between eyes close and eyes open, which reflected neural reactivity, was higher in high-tertile pTau-217 group compared with the low tertile group in the adjusted model (covariate-adjusted mean difference = 0.007, 95% CI = 0.002 to 0.01, FDR-adjusted p = 0.032) (Fig. 4G–Table 2b). Table 2aModel-based comparisons of HbO levels between pTau-217 tertile groups across tasks and brain regions.Table 2aUnadjusted mixed-effects modelMixed-effects model adjusted for age, sex, and years of educationModel-estimated mean difference (95% CI)FDR adjusted p-valueModel-estimated mean difference (95% CI)FDR adjusted p-valueCongruent Left lateral•High tertile vs low tertile pTau-2170.003 (−0.001, 0.008)0.2920.003 (−0.001, 0.01)0.365 Left medial•High tertile vs low tertile pTau-2170.004 (−0.0001, 0.009)0.1440.006 (0.001, 0.01)0.04 Right medial•High tertile vs low tertile pTau-2170.006 (0.001, 0.01)0.040.007 (0.002, 0.01)0.032 Right lateral•High tertile vs low tertile pTau-2170.006 (0.002, 0.01)0.0370.006 (0.001, 0.01)0.052Eyes close Left lateral•High tertile vs low tertile pTau-2170.001 (−0.004, 0.005)0.9510.0002 (−0.004, 0.005)0.965 Left medial•High tertile vs low tertile pTau-2170.002 (−0.002, 0.007)0.5490.002 (−0.002, 0.007)0.549 Right medial•High tertile vs low tertile pTau-2170.001 (−0.004, 0.005)0.8640.001 (−0.004, 0.01)0.951 Right lateral•High tertile vs low tertile pTau-2170.001 (−0.003, 0.005)0.7910.001 (−0.004, 0.005)0.900Eyes open Left lateral•High tertile vs low tertile pTau-217−0.002 (−0.006, 0.003)0.593−0.002 (−0.007, 0.002)0.549 Left medial•High tertile vs low tertile pTau-217−0.002 (−0.006, 0.002)0.549−0.003 (−0.007, 0.002)0.549 Right medial•High tertile vs low tertile pTau-217−0.001 (−0.006, 0.003)0.739−0.002 (−0.007, 0.003)0.593 Right lateral•High tertile vs low tertile pTau-217−0.006 (−0.01, −0.001)0.045−0.006 (−0.01, −0.002)0.037Incongruent Left lateral•High tertile vs low tertile pTau-217−0.002 (−0.006, 0.003)0.594−0.002 (−0.007, 0.003)0.556 Left medial•High tertile vs low tertile pTau-2170.0004 (−0.004, 0.005)0.9550.0004 (−0.004, 0.005)0.955 Right medial•High tertile vs low tertile pTau-2170.002 (−0.002, 0.006)0.5560.002 (−0.003, 0.007)0.568 Right lateral•High tertile vs low tertile pTau-2170.0002 (−0.004, 0.005)0.9650.0001 (−0.005, 0.005)0.986 Model-based comparisons of HbO levels between pTau-217 tertile groups across tasks and brain regions. •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 HbO Changes level during Congruent task and eyes open in HC individuals with high plasma pTau-217 levels. HbO values represent the mean concentration change across the task period for each condition. (A–B) Group differences in HbO changes between high (red) and low (blue) pTau-217 tertiles during the congruent (A) and eyes-open resting (eyes open) (B) conditions. Scalp topographies of Cohen's d effect sizes illustrating region-specific HbO differences between groups for the congruent (C) and eyes open (D) conditions. A correlation between eyes-open right lateral HbO and pTau-217 level (E). HbO differences for the task–rest contrast (Congruent–eyes open) (F) and resting contrast (eyes close–eyes open) (G) across lateral and medial PFC subregions. All reported results are based on the averaged post-stimulus activity from 150 to 1500 ms. Error bars represent ± SD. * indicates p < 0.05; **p < 0.01; ns = not significant. Table 2bModel-based comparisons of HbO difference levels between pTau-217 tertile groups for task contrasts by brain region.Table 2bUnadjusted mixed-effects modelMixed-effects model adjusted for age, sex, and years of educationModel-estimated mean difference (95% CI)FDR adjusted p-valueModel-estimated mean difference (95% CI)FDR adjusted p-valueCongruent – Eyesopen Left lateral•High tertile vs low tertile pTau-2170.005 (0.0004, 0.01)0.0970.006 (0.0009, 0.01)0.065 Left medial•High tertile vs low tertile pTau-2170.006 (0.002, 0.01)0.0370.009 (0.004, 0.01)<0.0001 Right medial•High tertile vs low tertile pTau-2170.007 (0.002, 0.01)0.0320.009 (0.004, 0.01)<0.0001 Right lateral•High tertile vs low tertile pTau-2170.01 (0.007, 0.02)<0.00010.01 (0.007, 0.02)<0.0001Eyes Open – Eyes close Left lateralHigh tertile vs low tertile pTau-2170.002 (−0.003, 0.01)0.5560.003 (−0.002, 0.007)0.549 Left medial•High tertile vs low tertile pTau-2170.004 (−0.0003, 0.01)0.1630.005 (0.0001, 0.01)0.126 Right medial•High tertile vs low tertile pTau-2170.002 (−0.003, 0.007)0.5640.002 (−0.002, 0.007)0.549 Right lateral•High tertile vs low tertile pTau-2170.007 (0.002, 0.01)0.0370.007 (0.002, 0.01)0.032 Model-based comparisons of HbO difference levels between pTau-217 tertile groups for task contrasts by brain region. •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 High tertile vs low tertile pTau-217 High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 •High tertile vs low tertile pTau-217 As a sensitivity analysis, the same linear mixed-effects models were used, with pTau-217 modeled as a continuous variable including interaction terms, as reported in Table 3a, Table 3ba and 3b. During eyes open, higher pTau-217 levels was significantly associated with lower HbO levels in the right lateral region in both unadjusted (β = −0.00073, 95% CI -0.00122–-0.00024, FDR-adjusted p = 0.028) and covariate-adjusted analyses (β = −0.00078, 95% CI -0.00128–-0.00027, FDR-adjusted p = 0.028). For the task contrasts (congruent versus eyes-open and eyes-closed and eyes-open), pTau-217 was also significantly associated with HbO levels in the right lateral region in both unadjusted (congruent vs eyes-open: β = 0.00102, 95% CI 0.00051–0.00153, FDR-adjusted p = 0.003; eyes-closed vs eyes-open: β = 0.00072, 95% CI 0.00021–0.00123, FDR-adjusted p = 0.039) and covariate-adjusted analyses (congruent vs eyes-open: β = 0.00104, 95% CI 0.00052–0.00155, FDR-adjusted p = 0.003; eyes-closed vs eyes-open: β = 0.00075, 95% CI 0.00024–0.00127, FDR-adjusted p = 0.034). These results were consistent with the primary analysis for the right lateral region during the eyes-open trial and task contrasts. Table 3aModel-based association between pTau-217 and HbO levels across tasks and brain regions.Table 3aUnadjusted mixed-effects modelMixed-effects model adjusted for age, sex, and years of educationModel-estimated slope (95% CI)FDR adjusted p-valueModel-estimated slope (95% CI)FDR adjusted p-valueCongruentLeft lateral−0.00014 (−0.00062, 0.00035)0.783−0.00016 (−0.00066, 0.00035)0.772Left medial0.00007 (−0.00042, 0.00056)0.9370.00018 (−0.00033, 0.00069)0.734Right medial0.00009 (−0.00039, 0.00058)0.8800.00015 (−0.00035, 0.00066)0.772Right lateral0.00029 (−0.00020, 0.00077)0.5980.00026 (−0.00025, 0.00076)0.643Eyes closeLeft lateral−0.00014 (−0.00063, 0.00034)0.772−0.00016 (−0.00067, 0.00035)0.772Left medial0.00002 (−0.00047, 0.00050)1.0000.00001 (−0.00050, 0.00052)1.000Right medial−0.00006 (−0.00055, 0.00043)0.937−0.00008 (−0.00059, 0.00043)0.918Right lateral−0.000009 (−0.000497, 0.000479)1.000−0.00003 (−0.00053, 0.00048)1.000Eyes openLeft lateral−0.00037 (−0.00085, 0.00012)0.416−0.00040 (−0.00091, 0.00010)0.410Left medial−0.00028 (−0.00076, 0.00021)0.598−0.00030 (−0.00081, 0.00021)0.598Right medial−0.00032 (−0.00081, 0.00017)1.000−0.00036 (−0.00086, 0.00015)1.000Right lateral−0.00073 (−0.00122, −0.00024)0.028−0.00078 (−0.00128, −0.00027)0.028IncongruentLeft lateral0.00003 (−0.00045, 0.00052)1.0000.0000097 (−0.000497, 0.000516)1.000Left medial0.00011 (−0.00045, 0.00052)0.8300.00012 (−0.00038, 0.00063)0.830Right medial0.00020 (−0.00028, 0.00069)0.7010.00021 (−0.00030, 0.00071)0.703Right lateral0.00019 (−0.00030, 0.00068)0.7040.00019 (−0.00032, 0.00070)0.722 Model-based association between pTau-217 and HbO levels across tasks and brain regions. Table 3bModel-based association between pTau-217 and HbO levels for task contrasts by brain region.Table 3bUnadjusted mixed-effects modelMixed-effects model adjusted for age, sex, and years of educationModel-estimated slope (95% CI)FDR adjusted p-valueModel-estimated slope (95% CI)FDR adjusted p-valueCongruent - Eyes openLeft lateral0.00023 (−0.00028, 0.00074)0.6800.00025 (−0.00027, 0.00076)0.658Left medial0.00034 (−0.0017, 0.00085)0.5250.000481 (−0.000034, 0.000996)0.313Right medial0.000414 (−0.000097, 0.000925)0.4100.000508 (−0.000007, 0.001023)0.271Right lateral0.00102 (0.00051, 0.00153)0.0030.00104 (0.00052, 0.00155)0.003Eyes close- Eyes openLeft lateral0.00022 (−0.00029, 0.00073)0.6880.00024 (−0.00027, 0.00076)0.658Left medial0.00029 (−0.00022, 0.00080)0.5980.00031 (−0.00020, 0.00083)0.598Right medial0.00026 (−0.00025, 0.00077)0.6430.00028 (−0.00024, 0.00079)0.646Right lateral0.00072 (0.00021, 0.00123)0.0390.00075 (0.00024, 0.00127)0.034 Model-based association between pTau-217 and HbO levels for task contrasts by brain region. Association between HbO and heart rate (HR) The associations between right lateral PFC HbO levels during Stroop congruent trials and HR (post- and pre-task) differed between those with high and low tertile plasma pTau-217 groups. Specifically, individuals with high pTau-217 concentrations exhibited a strong positive correlation between post-task HR and HbO levels (ρ = 0.85, p < 0.001), such that higher post-task HR related to increased HbO levels in the right lateral PFC. In contrast, those with low pTau-217 showed a moderate negative correlation (ρ = −0.55, p = 0.043). Similarly, participants with high pTau-217 exhibited a positive correlation between pre-task HR and HbO (ρ = 0.72, p < 0.005). Conversely, individuals with low pTau-217 demonstrated a negative correlation (ρ = −0.53, p = 0.053) (Fig. 5). Overall correlations between HbO and HR are listed in Table S1. Heart rate (HR) and right lateral HbO. Correlation between pre-task HR (beats per minute) and right lateralprefrontal cortex (PFC) oxygenated hemoglobin (HbO) levels (A) and post-task HR (beats per minute) and right lateral PFC HbO levels (B) during congruent trials for individuals with low (Blue color) and high (Red color) pTau-217 levels. Shades indicated 95% confidence intervals. Functional connectivity Connectivity analysis between high and low pTau-217 groups revealed distinct patterns between four regions, Left Lateral (LL), Left Medial (LM), Right Medial (RM), and Right Lateral (RL), across experimental conditions. The only observed group difference was FC between LL and RL, with wPLI values peaking at 0.93 (Low pTau-217) and 0.91 (High pTau-217) with (p = 0.03), Cohen's d = −0.89 during eyes open, indicating reduced interhemispheric synchronization in the high plasma pTau-217 group (Fig. 6, Table S2). Group-level descriptive statistics (mean ± SD), inferential statistics (p-values), and standardized effect sizes (Cohen's d) for all wPLI connectivity comparisons are summarized in Table S2. Weighted Phase Lag Index (wPLI) matrices across cognitive and resting-state conditions. The heatmaps depict pairwise wPLI values between four cortical regions: Left Lateral (LL), Left Medial (LM), Right Medial (RM), and Right Lateral (RL), during four Stroop interference task and resting state: Incongruent, Congruent, Eyes Closed, and Eyes Open. Top panels represent individuals with low pTau-217 levels, while bottom panels show those with high pTau-217. Warmer colors (red) indicate stronger phase-based connectivity, while cooler colors (blue) reflect weaker synchronization. The size of the circles indicates the strength of the synchronization as well. Discussion These results suggest unique cerebral vascular changes in cognitively healthy older adults with higher plasma pTau-217 levels in comparison with those with lower levels. First, high pTau-217 individuals had lower HbO at the right lateral region during resting state with eyes open, which was significantly greater during low-load congruent Stroop testing. Their HbO at the right lateral PFC was higher with HR after Stroop testing, however the opposite relationship was observed in low pTau-217 participants. Second, the FC of HbO between left and right lateral regions during resting state Stroop testing was lower than those with low plasma pTau-217 levels. This analysis offers potential insights into the much-needed detection and management of early AD risk in cognitively healthy older adults. As there is no published cutoff for estimating cognitive decline risk in cognitively healthy older adults, we separated the participants into two very distinct groups, high-and low-tertile. Although not an ideal separation method, quartile cutpoints have been used to study potential changes in HRV with cardiac mortality(de Bruyne et al., 1999; Nantsupawat et al., 2022; Singh et al., 2018; Tegegne et al., 2023). To maximize sample sizes and avoid middle uncertain samples, we compared first tertile and third tertile groups. Results from this exploratory study need to be confirmed with a larger sample size when a confirmed cutoff point for plasma pTau-17 has been established. Behavioral analysis Our behavioral findings suggest that elevated pTau-217 levels may be associated with early disruptions in cognitive efficiency, particularly evident in slower response times despite preserved task ACC (Fig. 3). Although participants with high and low pTau-217 levels showed comparable ACC on congruent trials and no significant differences on incongruent trials, the high pTau-217 group exhibited significantly longer response times across both conditions. This dissociation between preserved ACC and delayed RT may indicate early inefficiencies in cognitive control and processing speed that preceded overt performance decline (Cabeza et al., 2018; Reuter-Lorenz and Cappell, 2008). Prior research has similarly shown that cognitive slowing can be an early behavioral marker of AD pathology, particularly linked to tau accumulation, even in cognitively unimpaired individuals (Betthauser et al., 2019; Simon et al., 2023). Such slowing may reflect compensatory mechanisms or early functional compromise in neural circuits mediating attentional regulation and executive function (Kaufmann et al., 2008; Reuter-Lorenz and Cappell, 2008; Yoon et al., 2019). Notably, our findings align with a previous study in which individuals with MCI exhibited significantly longer RTs than HC during both congruent and incongruent Stroop trials (Fan et al., 2025). Our results support prior findings that RT is more sensitive than ACC for early AD pathology detection (Hartle et al., 2022; Wong et al., 2017). Oxygenated hemoglobin concentration analysis Our findings reveal region-specific alterations in prefrontal hemodynamic responses associated with elevated pTau-217 levels (Fig. 4). Individuals in the high tertile pTau-217 group exhibited reduced HbO (0.01–0.1 Hz) in the lateral right prefrontal cortex during the eyes-open resting condition, suggesting impaired baseline neural activation or diminished neurovascular responsiveness in this region. Furthermore, our analysis suggested a blunted change in HbO between eyes-closed and eyes-open states, indicating reduced neural adaptability during shifts in arousal or attentional state (Petro et al., 2022a). Notably, participants with elevated pTau-217 demonstrated a greater increase in HbO from rest to the congruent task condition in the same region, which may reflect compensatory overactivation during cognitive engagement. This interpretation aligns with previous findings in individuals with MCI, where increased HbO in the ventrolateral PFC (VLPFC) and parietal lobe (PL) during the late phase of the Stroop task was interpreted as compensatory neural recruitment to preserve task performance (Fan et al., 2025). Interestingly, several fNIRS studies have reported reduced hemodynamic responses in the inferior frontotemporal cortex of MCI and AD patients (Herrmann et al., 2017; Katzorke et al., 2018; Mei et al., 2023), highlighting the impact of AD pathology on frontal cortex. However, no lateralized changes were observed (Herrmann et al., 2017; Katzorke et al., 2018; Mei et al., 2023). In our study, task-dependent HbO changes over the right lateral PFC in those with early AD pathology were observed, specifically a decrease during resting eyes closed, but an increase during low-load congruent Stroop trials. These region-specific and context-dependent effects were supported by a follow up sensitivity analysis using the same linear mixed-effects models considering pTau-217 as a continuous variable. This was in line with previous reports regarding vulnerability in the right PFC versus left PFC. For instance, the right PFC exhibits decreased cerebral blood flow with aging (Lu et al., 2011), metabolic decline associated with episodic memory decline in APOE-4 positive elderlies (Small et al., 2000), and reductions in white matter volume related to neuropsychological performance (Brickman et al., 2006). Interestingly, it has been suggested that emotion and cognition are deeply interwoven at molecular, structural, circuit, and behavioral levels, making the distinction between the two increasingly fuzzy and context dependent (Okon-Singer et al., 2015). Additionally, the right hemisphere is proposed to more effectively process negative emotions (e.g., sad, fear), while the left hemisphere is thought to more effectively process positive emotions (e.g., happiness), shown by various test such as facial expression or emotion questionnaire rating in both females and males (Adolphs et al., 2001; Ahern and Schwartz, 1979). Thus, we speculate that our findings of reduced right PFC HbO reflecting reduced right PFC activity may be linked with negative emotions. It is known that older adults show “positive effect,” or tend to be more distracted by positive images than negative images (Kennedy et al., 2020; Kennedy and Mather, 2025). Our results characterized how pTau-217 burden modulates task-evoked hemodynamic activity. The right-lateralized HbO pattern may be relevant because right prefrontal regions are vulnerable to aging- and AD-related changes in cerebral blood flow, metabolism, and white matter integrity (Chen et al., 2026). However, the present study was not designed to test emotional lateralization or positivity-effect hypotheses; thus, these possibilities should be considered speculative and further carefully designed studies are needed for exploration. The task-resting eyes open state difference (e.g., Congruent – eyes open) can remove baseline vascular and metabolic signals, isolating the net HbO response attributable to cognitive processing. This approach enhances physiological interpretability by distinguishing true task-induced activation from resting-state or tonic hemodynamic variability, thereby enabling a more accurate assessment of group differences. We also compared group differences on changes from eyes close to eyes open state (eyes close-eyes open), to reflect neural reactivity, which has been related to reactivity and neural dynamics (Bellato et al., 2020; Petro et al., 2022b). Our results suggested lateralized features of neural dynamic changes. Further studies will be needed to clarify the relationship between “positive effect” and early AD pathology. Association between HbO and heart rate This study reveals distinct association between HR and prefrontal cortex hemodynamics during cognitive processing when stratified by plasma pTau-217 levels, a biomarker linked to AD (Fig. 5). Specifically, individuals with high pTau-217 levels showed a strong positive correlation between post-task HR and HbO in the lateral right PFC, suggesting increased autonomic engagement may be needed to sustain neural activity under pathological conditions. In contrast, those with low pTau-217 exhibited a moderate negative correlation, indicative of more efficient neurovascular and autonomic regulation. These findings suggest that tau-related pathology disrupts the normal coupling between brain and cardiovascular systems, particularly in regions critical for cognitive control (Molloy et al., 2023). The altered HR-HbO relationship in individuals with elevated pTau-217 may reflect early compensatory or maladaptive mechanisms preceding overt cognitive decline (Forte et al., 2019; Guzman et al., 2024; Ku et al., 2024; Ziegler et al., 2009). This highlights the potential of combining brain and heart neurophysiological measures as sensitive, noninvasive markers for detection of early AD pathology-related changes. Functional connectivity The wPLI findings highlight distinct patterns of functional brain connectivity influenced by both regional location and pathological tau burden (Mormino et al., 2011). Phase-based FC measures are supported by prior fNIRS studies showing that phase relationships between hemoglobin signals reflect physiological processes such as cerebral autoregulation and blood flow dynamics and can be used to characterize brain activation and functional networks (Ren et al., 2019). In this context, the observed alterations in wPLI in individuals with elevated pTau-217 may reflect early disruptions in neurovascular coordination. The strongest interhemispheric synchronization was observed between the lateral left and lateral right regions (LL–RL) during resting eyes-open condition (Fig. 6), such that high pTau-217 participants showed slightly lower wPLI values, suggesting early tau-related disruptions in functional integration. This subtle weakening of interhemispheric connectivity could reflect the beginning of network-level changes preceding measurable cognitive decline. In line with our findings, previous studies using fNIRS have reported that PFC-FC was decreased and shifted from the left to the right side in individuals with MCI during verbal fluency tasks, supporting the notion that early AD pathology may weaken the prefrontal FC and cause rightward shift of prefrontal network dynamics (Wang et al., 2024). However, no changes of FC during tasks were observed in our study, possibly from different tasks, or participants at different stages. For instance, cognitively healthy individuals with elevated pTau-217 may possess compensatory mechanisms, such as education, which contribute to early-stage resilience in interhemispheric communication (Aiello Bowles et al., 2019; Verny et al., 2015). Importantly, the broader landscape of resting-state fNIRS research encompasses a diverse set of connectivity and network-level metrics. These include FC approaches such as correlation, coherence, mutual information, cross-correlation, Granger causality, transfer entropy, and phase-based measures (e.g., phase locking, phase consistency, phase slope), as well as effective connectivity methods such as dynamic causal modeling and structural equation modeling (Albrecht et al., 2025; Bu et al., 2019; Ghafoor et al., 2019). In addition, graph-theoretical metrics are increasingly used to characterize network topology and integration (Yang et al., 2021). Together, these approaches provide complementary insights into the organization and disruption of brain networks in neurodegenerative conditions. Overall, these findings emphasize the need for longitudinal studies and more fine-grained connectivity analyses to clarify how biomarkers such as pTau-217 influence functional brain networks over time. This study has several limitations. First, the sample size was modest, particularly after extreme-tertile stratification, and participants were predominantly Caucasian women; therefore, statistical power and generalizability are limited. Second, excluding the middle tertile increased contrast between low and high pTau-217 groups but reduced sample size and may not capture the full continuous relationship between pTau-217 and fNIRS measures. Continuous pTau-217 sensitivity analyses were therefore included, but larger cohorts are needed to determine whether the observed effects are linear, threshold-based, or driven by high-biomarker individuals. Third, behavioral, HR-HbO, and wPLI analyses were exploratory and were not corrected for multiple comparisons, so these findings should be interpreted as hypothesis-generating. Fourth, the cross-sectional design prevents conclusions about progression or causality. Fifth, fNIRS signals may be influenced by systemic physiology, scalp blood flow, and extracerebral contamination. Although two short-separation channels were acquired, short-separation regression was not performed. Sixth, HR was measured before and after the task rather than continuously during Stroop performance, so HR-HbO associations do not directly represent beat-to-beat task-synchronized coupling. Seventh, fixed DPF values were used without age- or sex-specific correction; HbO values should therefore be interpreted as relative concentration changes. Seventh, detailed information regarding comorbidities, medication use, physical activity, and other factors potentially affecting cerebral hemodynamics was not systematically available for all participants, limiting our ability to evaluate their potential confounding effects. Future studies should systematically collect and account for these factors. Finally, the analyzed 0.15-1.5 s post-stimulus window captures early or transient hemodynamic activity rather than the full peak hemodynamic response. Replication in larger, longer time window, longitudinal, physiologically monitored cohorts is needed. Conclusion This exploratory fNIRS study provides preliminary evidence that elevated plasma pTau-217 levels, an established blood-based biomarker of early AD pathology, may be associated with early alterations in right prefrontal hemodynamics, functional connectivity, and heart–brain coupling, in CH older adults, consistent with possible compensatory or neurovascular changes. The greater task-rest HbO change in the high pTau-217 group, together with preserved ACC but slower RT, is consistent with possible compensatory recruitment during task engagement. Alternative explanations include altered vascular reactivity, systemic physiological influences, and early neurovascular dysregulation. The exploratory analyses also suggested differences in functional connectivity and HR–HbO coupling associated with pTau-217; however, these findings require confirmation in larger, independent cohorts. Further longitudinal study involving systemic physiological observations will be helpful to clarify specific mechanisms. The observed shift in the coupling between prefrontal oxygenation and HR further highlights early dysregulation of neurocardiac integration linked to tau-related pathology. Collectively, these findings support further investigation of fNIRS as a scalable and cost-effective approach for characterizing pTau-217-associated hemodynamic alterations in CH older adults. CRediT authorship contribution statement Abdulhakim Al-Ezzi: Writing – review & editing, Writing – original draft, Visualization, Software, Methodology, Investigation, Formal analysis, Data curation. Choo Phei Wee: Writing – review & editing, Validation, Methodology, Formal analysis. Xiaomeng Wu: Methodology, Data curation. Natalie Astraea: Writing – review & editing, Supervision, Resources, Data curation. Michael T. Kleinman: Writing – review & editing, Conceptualization. Wendy Mack: Writing – review & editing, Supervision, Methodology. Robert A. Kloner: Writing – review & editing, Supervision, Funding acquisition. Xianghong Arakaki: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Funding acquisition, Conceptualization. Funding This work was supported by the National Institute on Aging [grant number: R01AG063857], HMRI institutional support, and National Center for Advancing Translational Science (NCATS) of the National Institutes of Health [grant number: UL1TR001855]. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments The authors thank the study participants for their altruistic participation in this research. Dr. Anne Nolty and Fuller clinical psychology students including Shant Rising, Caleb Sin, Kaitlyn Bennett, and Rachel Woo conducted neuropsychological tests. David Buennagel helped with recruitment and clinical visits. Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.ynirp.2026.100411. Supplementary data The following is the Supplementary data to this article: Multimedia component 1 Data availability Data will be made available on request. References Associated Data Supplementary Materials Multimedia component 1 Data Availability Statement Data will be made available on request.