Cerebrovascular reactivity and plasma p‐tau181 in Alzheimer's disease: Insights into APOE‐related vascular phenotypes Authors: Martina Gaia Di Donna, Maria Rosaria Bagnato, Chiara Giuseppina Bonomi, Caterina Motta, Diego Centonze, Alessandro Martorana, Marina Diomedi Published: 2026 ABSTRACT AbstractINTRODUCTIONVascular dysfunction is increasingly recognized as a key contributor to Alzheimer's disease (AD). This study explored cerebrovascular reactivity (CVR), measured by the breath‐holding index (BHI) via Transcranial Doppler (TCD), in relation to plasma AD biomarkers, metabolic dysfunction, cardiovascular burden, and apolipoprotein E (APOE) genotype.METHODSWe enrolled 64 patients with cerebrospinal fluid (CSF)‐confirmed AD (30 APOE ε3, 34 APOE ε4) and 28 age‐matched healthy controls in a cross‐sectional blinded study. All participants underwent TCD to assess middle cerebral artery flow velocity and BHI. AD patients were further evaluated for insulin resistance (TyG index), magnetic resonance imaging markers of small vessel disease (SVD), and plasma biomarkers (phosphorylated tau 181 [p‐tau181], amyloid beta [Aβ]42/Aβ40 ratio – AmyR). Multivariate regressions examined associations between BHI and AD‐related parameters, stratified by APOE genotype.RESULTSBHI was significantly reduced in both AD groups compared to controls (p < 0.001), with no difference between ε3 and ε4 carriers. In ε3 patients, lower BHI was independently associated with higher plasma p‐tau181. In the MRI subgroup analysis, ε3 patients showed a negative correlation between BHI and both TyG and mCSVD while showing a positive correlation with AmyR. None of these associations were observed in ε4 carriers.DISCUSSIONOur findings indicate that CVR impairment is present across AD genotypes but more strongly linked to AD pathology and vascular‐metabolic dysfunction in APOE ε3 carriers. CVR may represent a non‐invasive biomarker of early vascular contributions to cognitive decline in AD. INTRODUCTION The recently revised research diagnostic criteria for Alzheimer's disease (AD) integrate vascular injury into the AD classification system as a non‐AD copathology that significantly contributes to the disease's patophysiology. 1 This update follows decades of evidence supporting the role of vascular dysfunction as a core component of cognitive decline in AD, 2 encompassing age‐related cerebrovascular dysfunction, 3 the effects of vascular risk factors (VRFs), 4 and their interplay with neurodegenerative mechanisms. 5 However, operational implementation remains limited, as current framework mainly depends on structural brain MRI markers, specifically features of small vessel disease (SVD), such as white matter hyperintensities (WMHs) and enlarged perivascular spaces, which capture only the static anatomical footprint of vascular pathology and fail to represent its dynamic functional alterations that may occur earlier in the disease course. Cerebrovascular reactivity (CVR), the capacity of cerebral blood vessels to modulate blood flow in response to vasoactive stimuli, is increasingly recognized as a dynamic and sensitive marker of vascular health. 6 , 7 Specifically, it may serve as an indirect indicator of neurovascular coupling integrity, providing a window into the real‐time performance of cerebral perfusion and ensuring that neuronal activity is matched by adequate perfusion to meet metabolic demands. 7 , 8 Accumulating evidence points to impaired CVR as a key manifestation of vascular dysfunction in AD, possibly contributing to neurovascular uncoupling and subsequent cognitive decline. 9 Importantly, CVR reflects the dynamic adaptability of the cerebrovascular system and may be interpreted as a functional correlate of cerebrovascular resilience, defined as the capacity to buffer physiological and pathological stressors while preserving cognitive performance through structural integrity (brain maintenance) or adaptive compensation (cognitive reserve). Among the available techniques, Transcranial Doppler (TCD) ultrasound with the breath‐holding index (BHI) offers a simple, non‐invasive, cost‐effective, and reproducible method for in vivo CVR assessment. 10 Several studies have reported reduced CVR in AD patients compared to cognitively normal controls. 2 , 11 , 12 Notably, we previously demonstrated an inverse association between BHI and cerebrospinal fluid (CSF) amyloid beta 1‐42 (Aβ1‐42) levels in apolipoprotein E (APOE) ε3 but not in APOE ε4 carriers, 13 suggesting that lower CVR may reflect early amyloid pathology and is also consistent with broader evidence linking APOE ε4 to SVD and generalized cerebrovascular dysfunction. 14 , 15 It is also important to consider that cardiovascular burden may critically impair CVR and overall neurovascular function. 16 Indeed, VRFs such as hypertension, diabetes, and aging are well known to negatively affect cerebral perfusion through their contributions to cerebral SVD. 17 However, the exact mechanisms by which these factors and the resulting microvascular damage influence CVR and neurovascular NVU integrity remain incompletely understood, particularly in the context of their potential interactions with AD pathology. In this context, the primary objective of our study was to confirm the link between CVR – assessed via TCD – and amyloidopathy in AD in a biologically defined cohort of AD patients, while accounting for the potential influence of VRFs and MRI markers of SVD. Building upon our previous CSF‐based observations, 13 we aimed to confirm and extend the association between CVR and AD pathology using novel plasma biomarkers, particularly phosphorylated tau181 (p‐tau181), which represent a more accessible and clinically translatable tool for detecting AD‐related processes. Finally, we explored whether these associations differed according to APOE genotype, given its established role in modulating neurovascular integrity. By integrating functional vascular, biochemical, and imaging dimensions, this study sought to clarify the vascular contribution to AD pathophysiology and identify early non‐invasive indicators of cerebrovascular resilience loss with potential utility in clinical and research settings. Accordingly, the aims of the present study were to (i) confirm the relationship of CVR and AD pathology using plasma biomarkers, while accounting for vascular and metabolic factors, and (ii) explore whether these associations differed according to APOE genotype. MATERIALS AND METHODS Detailed methodological procedures for each section are provided in the Supplementary Material. Subjects enrollment Ninety consecutive patients meeting criteria for mild cognitive impairment (MCI) 18 or early dementia due to AD 19 were screened. All participants underwent a standardized diagnostic work‐up (Figure 1). Study flowchart. The diagram outlines the recruitment and selection process for Alzheimer's disease (AD) patients and controls, including inclusion criteria and exclusions due to absent transtemporal window. A total of 64 AD patients and 28 controls completed Transcranial Dopper Breath‐Holding Index assessment. The inclusion criteria were as follows: (1) Mini‐Mental State Examination (MMSE) score of ≥20 to ensure proper execution of the apnea test, which is necessary for BHI evaluation, and (2) a CSF‐based diagnosis of AD, defined as the concomitant increase of p‐tau181 alongside reduced CSF Aβ42/Aβ40. 1 RESEARCH IN CONTEXT Systematic review: We examined prior studies on CVR in AD, with particular focus on TCD–based BHI measurements and their associations with CSF and plasma biomarkers, vascular risk factors, and APOE genotype. Existing literature supports a vascular contribution to AD, but evidence linking CVR to plasma biomarkers and metabolic dysfunction in a genotype‐specific manner is limited. Interpretation: Our results show that CVR impairment is present in AD regardless of APOE status but is preferentially linked to AD plasma biomarkers, metabolic dysfunction, and SVD in APOE ε3 carriers. This identifies CVR as a functional, non‐invasive marker of vascular vulnerability that adds translational value to circulating biomarkers in AD. Future directions: Prospective studies should test whether CVR predicts disease progression and treatment‐related vascular complications. Interventional trials targeting vascular and metabolic health are needed to assess whether modifying CVR can improve clinical outcomes and therapeutic safety in AD. Systematic review: We examined prior studies on CVR in AD, with particular focus on TCD–based BHI measurements and their associations with CSF and plasma biomarkers, vascular risk factors, and APOE genotype. Existing literature supports a vascular contribution to AD, but evidence linking CVR to plasma biomarkers and metabolic dysfunction in a genotype‐specific manner is limited. Systematic review: We examined prior studies on CVR in AD, with particular focus on TCD–based BHI measurements and their associations with CSF and plasma biomarkers, vascular risk factors, and APOE genotype. Existing literature supports a vascular contribution to AD, but evidence linking CVR to plasma biomarkers and metabolic dysfunction in a genotype‐specific manner is limited. Interpretation: Our results show that CVR impairment is present in AD regardless of APOE status but is preferentially linked to AD plasma biomarkers, metabolic dysfunction, and SVD in APOE ε3 carriers. This identifies CVR as a functional, non‐invasive marker of vascular vulnerability that adds translational value to circulating biomarkers in AD. Interpretation: Our results show that CVR impairment is present in AD regardless of APOE status but is preferentially linked to AD plasma biomarkers, metabolic dysfunction, and SVD in APOE ε3 carriers. This identifies CVR as a functional, non‐invasive marker of vascular vulnerability that adds translational value to circulating biomarkers in AD. Future directions: Prospective studies should test whether CVR predicts disease progression and treatment‐related vascular complications. Interventional trials targeting vascular and metabolic health are needed to assess whether modifying CVR can improve clinical outcomes and therapeutic safety in AD. Future directions: Prospective studies should test whether CVR predicts disease progression and treatment‐related vascular complications. Interventional trials targeting vascular and metabolic health are needed to assess whether modifying CVR can improve clinical outcomes and therapeutic safety in AD. Exclusion criteria included conditions or treatments potentially affecting cerebrovascular reactivity or AD biomarker interpretation, such as drugs interfering with autoregulation, uncontrolled vascular risk factors, prior cerebrovascular events, severe pulmonary disease, significant extra‐ or intracranial stenosis, or MRI features suggestive of severe cerebral amyloid angiopathy (CAA). Patients with clinically significant renal impairment were also excluded to avoid potential alterations in peripheral AD biomarker clearance. Subjects unable to reliably perform the Breath‐Holding Test or lacking an adequate transtemporal acoustic window were also excluded. A total of 70 AD patients and 34 age‐matched healthy controls (HCs) were initially eligible; six per group were excluded for insufficient transtemporal window, resulting in 64 patients and 28 controls. Clinical data, vascular risk factors, and medications were collected. Cognitive impairment was excluded in controls using MMSE (cut‐off ≥ 28). All HCs underwent TCD‐BHI testing; CSF and plasma biomarkers were not collected for ethical reasons. All participants or legal guardians provided written informed consent. The study was conducted according to the Declaration of Helsinki and approved by the local Ethics Committee. CSF, plasma collection, and biomarker analysis CSF and blood samples were collected and stored at −80°C until analysis. CSF Aβ40, Aβ42, p‐tau181, and total tau were measured using a fully automated CLEIA Fujirebio LUMIPULSE® G1200 system. The cut‐offs for amyloid and tau pathology were CSF Aβ42/Aβ40 < 0.0667 (A+) and CSF p‐tau181 > 56 pg/mL (T+). Plasma Aβ42, Aβ40, and p‐tau181 were measured using the same technology. Insulin resistance, triglyceride‐glucose index, and vascular score To assess metabolic dysfunction, we measured insulin resistance using the triglyceride‐glucose (TyG) index, which has been shown to be a reliable clinical surrogate marker of insulin resistance. 20 , 21 For each participant, we directly collected information on the presence (score = 1) or absence (score = 0) of eight cardiovascular risk factors: hypertension, atrial fibrillation, diabetes mellitus (DM), type 2 diabetes mellitus, smoking, stroke or transient ischemic attack, acute myocardial infarction, obstructive sleep apnea, dyslipidemia. 22 Information was obtained through structured interviews and medical record review. Neuroimaging data evaluation A subgroup of 36 patients underwent 1.5T brain MRI within ±3 months from AD diagnosis and were included in the MRI substudy. Further, 1.5T brain MRI sequences (aDWI, T2 FLAIR, and gradient echo) were reviewed to assess SVD markers: lacunae, white matter hyperintensities (WMHs, Fazekas 0 to 3), enlarged perivascular spaces (ePVS), and microbleeds. Two blinded neurologists expert in neurovascular disease (C.G.B. and M.D.) independently assessed MRI markers and calculated an adapted cerebrovascular SVD (CSVD) score, with interrater reliability metrics reported in the original validation studies (1 point each for ≥11 ePVS, ≥1 microbleed, ≥1 lacuna, and Fazekas ≥2). 23 , 24 This score is consistent with previously validated composite SVD scoring systems and aligned with Standards for Reporting Vascular Changes on Neuroimaging‐based frameworks. 25 TCD assessment, apnea test, and BHI evaluation Extracranial and intracranial vessels were evaluated in all patients and healthy controls by expert neurologists blinded to diagnosis (M.G.D.D. and M.R.B.). Carotid plaques and significant intracranial and extracranial stenosis were assessed using validated criteria. 26 , 27 All subjects underwent TCD ultrasound examination to evaluate the bilateral middle cerebral artery mean flow velocity (MCA MFV). A dual 2‐MHz transducer fitted onto a headband and placed on the temporal bone window was used to obtain continuous measurements of the MCA MFV. CVR was assessed at a single time point for all participants, in accordance with the cross‐sectional design of the study. Measurements were calculated before and after the apnea test and monitored with capnography to detect possible abdomino‐thorax respiratory movements, to verify breath‐holding compliance and end‐tidal CO2. CVR to hypercapnia was measured by means of the BHI, 28 calculated by dividing the percentage increase in MFV occurring during breath holding by the length of time (in seconds) subjects held their breath after a normal inspiration. For each patient, BHI was derived as the mean of two consecutive measurements obtained from the same MCA. The side of insonation was chosen according to the best transtemporal acoustic window, ensuring optimal signal quality and stable recording conditions. Subjects were instructed to avoid smoking and coffee intake for 2 h before the test and were trained to perform the procedure. Statistical analysis All statistical analyses were performed using JASP (version 0.18.3.0) and GraphPad Prism (version 9.5.0). Patients were grouped according to APOE genotype. First, demographic, clinical, and CVR data were compared across groups – APOE ε3 carriers, APOE ε4 carriers, and HCs – using the Kruskal‐Wallis test for continuous variables and chi‐squared test for categorical variables. Post hoc comparisons were conducted using Holm correction. To investigate the relationship between plasma AD biomarkers and CVR (as measured by the BHI), multiple linear regression analyses were performed. Analyses were conducted separately for APOE ε3 and APOE ε4 carriers to explore genotype‐specific effects. A backward elimination approach was used, starting with all relevant predictors (age, VS, TyG index, plasma AmyR and p‐tau181) and iteratively removing the least significant variables until the most parsimonious model with only significant predictors was achieved. A separate regression model was run for the subgroup of patients with available MRI data (n = 36) to account for neuroradiological markers of SVD. Here, the modified cerebrovascular SVD (mCSVD) score was included as an additional covariate. The same regression approach was applied as described earlier, and results were again stratified by APOE genotype. This allowed for the evaluation of whether structural cerebrovascular injury contributed to the variability in BHI. Finally, we performed sensitivity analyses using the combined ptau181/AmyR, given its reported relevance in the literature as a robust indicator of AD‐related pathology. 29 RESULTS Comparison between AD and HC: demographic, clinical, and vascular characteristics The study eventually included a total of 64 patients with AD, 30 APOE ε3 and 34 APOE ε4, as well as 28 HCs, as reported in Table 1, which presents demographic information of patients. The groups did not differ in terms of age, VS, and TyG, while the prevalence of females was higher among HCs than in both APOE ε3 and APOE ε4 (χ2 = 7.754, p = 0.021). In the two subgroups of patients with AD, no significant difference was found between APOE ε3 and APOE ε4 in terms of CSF and plasma biomarkers, except for in plasma p‐tau181/AmyR levels, which were higher in APOE ε4 carriers (U = 351.00, p = 0.048). TABLE 1Demographic features, CSF, and plasma dosages and BHI measurements expressed as medians with inter‐quartile ranges (IQRs).APOE ε3 (n = 30)APOE ε4 (n = 34)HC (n = 28)Test StatisticsEffect size p valueAge75.00 (8.00)74.00 (4.75)70.50 (11.00)H = 2.839η² = 0.0090.242Sex (%F)36.66%38.23%67.85%X2 = 7.754V = 0.292 0.021 b , c MMSE23.20 (3.30)24.55 (3.98)29.00 (1.85)H = 34.273ε2 = 0.351 <0.001 b , c VS0.125 (0.250)0.128 (0.250)0.250 (0.125)H = 1.444ε2 = 0.0000.486mCVSD1.07 (0.96)0.95 (0.83)n.a.n.a.n.a.n.a.TyG4.46 (0.27)4.43 (0.36)n.a.n.a.n.a.n.a.Altered TyG (%)33.3%35.3%n.a.n.a.n.a.n.a.CSF AmyR0.047 (0.015)0.048 (0.009)n.a.n.a.n.a.n.a.CSF p‐tau181109.20 (60.40)95.30 (71.25)n.a.n.a.n.a.n.a.CSF lactates1.43 (0.31)1.38 (0.27)n.a.n.a.n.a.n.a.Qalb5.51 (3.54)5.41 (3.93)n.a.n.a.n.a.n.a.Plasma AmyR0.088 (0.026)0.086 (0.022)n.a.n.a.n.a.n.a.Plasma p‐tau1812.130 (1.320)2.645 (1.040)n.a.n.a.n.a.n.a.Plasma p‐tau181/AmyR23.738 (12.651)33.299 (17.103)n.a.n.a.n.a.n.a TCD CVR measurements Rest MCA‐MFV49.00 (11.00)48.50 (16.50)51.00 (13.25)H = 0.662ε2 = 0.0000.718After MCA‐MFV57.00 (12.00)61.50 (21.00)72.50 (25.00)H = 9.336ε2 = 0.080 0.009 b BHI0.631 (0.324)0.685 (0.393)1.272 (0.489)H = 41.277ε2 = 0.427 <0.001 b , c Note: Bold values are statistically significant.Abbreviations: AmyR, Aβ42/Aβ40 ratio; BHI, Breath‐Holding Index; CSF, cerebrospinal fluid; CVR, cerebrovascular reactivity; F, female; H, Kruskal‐Wallis H statistic; HC, healthy control; MCA, middle cerebral artery; mCSVD, modified cerebrovascular small vessel disease; MFV, mean flow velocity; MMSE, Mini‐Mental State Examination; n, number; Qalb, albumin quotient; TyG, triglyceride and glucose index; V, Cramer's V; VS, vascular score; η², eta squared; χ2, chi‐squared. a ε3 versus ε4. b ε3 versus HC. c ε4 versus HC.John Wiley & Sons, Ltd. Demographic features, CSF, and plasma dosages and BHI measurements expressed as medians with inter‐quartile ranges (IQRs). Note: Bold values are statistically significant. Abbreviations: AmyR, Aβ42/Aβ40 ratio; BHI, Breath‐Holding Index; CSF, cerebrospinal fluid; CVR, cerebrovascular reactivity; F, female; H, Kruskal‐Wallis H statistic; HC, healthy control; MCA, middle cerebral artery; mCSVD, modified cerebrovascular small vessel disease; MFV, mean flow velocity; MMSE, Mini‐Mental State Examination; n, number; Qalb, albumin quotient; TyG, triglyceride and glucose index; V, Cramer's V; VS, vascular score; η², eta squared; χ2, chi‐squared. ε3 versus ε4. ε3 versus HC. ε4 versus HC. Considering TCD CVR measurements, the Kruskal–Wallis test revealed comparable rest MCA MFV “across groups”. Instead, a significant difference was present in post MCA MFV evaluation (H[2] = 9.336, p = 0.009), with post hoc analysis showing that the APOE ε3 group had significantly lower post MCA‐MFV values compared to HC (p holm = 0.005), while the APOE ε4 group exhibited a similar trend that did not reach significance (p holm = 0.066). Statistically significant differences were also retrieved for BHI (H[2] = 41.277, p < 0.001), with both APOE ε3 and APOE ε4 showing lower values than HC (both p holm < 0.001), but no difference was found between the two AD groups (p holm = 0.730). APOE genotype‐specific associations between BHI, plasma AD biomarkers, and vascular comorbidities A backward multiple linear regression analysis was conducted to investigate the association between plasma biomarkers and BHI, controlling for age, VS, and other relevant covariates (TyG and AmyR). Results showed that plasma p‐tau181 levels were significantly and negatively associated with BHI, even after adjusting for age and other variables (Table 2). This association was specific to APOE ε3 carriers, in whom higher p‐tau181 levels predicted lower BHI values (β = −0.626, p < 0.001. Conversely, in APOE ε4 carriers, p‐tau181 was not significantly associated with BHI, nor were any other included variables. Importantly, VS showed no significant effect on BHI in either APOE subgroup. The sensitivity analysis using p‐tau181/AmyR ratio yielded comparable results, confirming a significant negative association with BHI in APOE ε3 carriers but not in ε4 carriers (Table S1). TABLE 2Multivariate regression analyses in whole cohort.BHI APOE ε3 (n = 30)BHI APOE ε4 (n = 34)95% CI of B 95% CI of B B LowerUpper p β (std) B LowerUpper p β (std) Model 1 Age−0.020−0.0480.0080.151−0.288−0.014−0.0450.0180.379−0.170TyG−0.254−1.1350.6270.556−0.112−0.053−0.7040.5970.868−0.32AmyR−0.155−1.2020.8920.761−0.0510.129−0.7701.0270.7710.058p‐tau181−0.534−0.890−0.177 0.005 −0.534−0.053−0.4770.3710.800−0.048VS0.038−0.9851.0620.9390.0160.335−0.8291.4990.5600.113 Best model Age..........TyG..........AmyR..........p‐tau181−0.625−0.939−0.311 < 0.001 −0.626.....VS.......... R 2 = 0.392 R 2 = 0.000 Note: Bold values are statistically significant.Abbreviations: β (std): standardized beta coefficient; AmyR, Aβ42/Aβ40 ratio; BHI, Breath‐Holding Index; mCSVD, modified cerebrovascular small vessel disease; TyG, triglyceride and glucose index; VS, vascular score.John Wiley & Sons, Ltd. Multivariate regression analyses in whole cohort. Note: Bold values are statistically significant. Abbreviations: β (std): standardized beta coefficient; AmyR, Aβ42/Aβ40 ratio; BHI, Breath‐Holding Index; mCSVD, modified cerebrovascular small vessel disease; TyG, triglyceride and glucose index; VS, vascular score. Multivariate regression analyses of cerebrovascular–imaging relationships in MRI subgroup Evaluating the AD population in terms of neuroradiological findings, we collected data from 36 patients. As shown in the pie chart (Figure 2), the AD population presents only mild features of SVD, despite the poor vascular reactivity. In fact, only 5.7% is affected by a mCSVD score of 3, indicating a severe condition of SVD. The presence of WMHs is mostly predominant as foci punctati, while larger and confluent WMHs are rare. Distribution of modified cerebrovascular small vessel disease (mCSVD) scores among patients with Alzheimer's disease (AD) in the magnetic resonance imaging (MRI) subgroup. Distribution of the MRI subgroup according to mCSVD score categories. Pie chart sectors indicate relative frequency of patients within each score level, expressed as percentages and absolute counts (n). The majority of individuals showed low to moderate SVD burden, with most scoring 1 or 2. A smaller portion of the cohort had no detectable burden (score 0), while only a few reached the highest score of 3. This pattern suggests that severe microvascular pathology is uncommon in this AD population, despite impaired cerebrovascular reactivity. Then we considered the subgroup of patients with available MRI data, stratified again according to APOE genotype (n = 15 APOE ε3 patients and n = 21 APOE ε4 patients) and added the radiological mCSVD score in the model as a covariate (Table 3). Among APOE ε3 carriers, several variables showed significant associations with BHI (Table 3). The TyG index demonstrated a strong negative association with BHI (β = −1.453, 95% CI [−2.083, −0.823], p < 0.001). AmyR showed a positive association (β = 0.970, 95% CI [0.319, 1.620], p = 0.008). Both p‐tau181 (β = −0.309, 95% CI [−0.563, −0.055], p = 0.022) and mCSVD (β = −0.145, 95% CI [−0.259, −0.032], p = 0.018) resulted negatively associated with BHI. Again, the sensitivity analysis using the p‐tau181/AmyR confirmed the genotype‐specific pattern observed in the primary model (Table S2). Given the limited sample size of the MRI subgroup, the following multivariate analyses should be considered exploratory. TABLE 3Multivariate regression analyses in subgroup with available MRI data.BHI APOE ε3 (n = 15)BHI APOE ε4 (n = 21)95% CI of B 95% CI of B B LowerUpper p β (std) B LowerUpper p β (std) Model 1 Age0.006−0.0160.0270.5640.194−0.042−0.1010.0170.147−0.448TyG−1.548−2.548−0.547 0.008 −0.668−0.011−1.1751.1520.984−0.006AmyR1.1520.2752.029 0.017 0.480−0.335−2.2421.5730.711−0.129p‐tau181−0.309−0.590−0.028 0.035 −0.3170.040−0.9571.0370.9320.026VS0.123−0.8771.1240.7790.0590.906−1.1032.9160.3480.267mCSVD−0.145−0.3090.020 0.076 −0.3390.051−0.2830.3850.7470.088 Best model Age..........TyG−1.453−2.083−0.823 <0.001 −0.627.....AmyR0.9700.3191.620 0.008 0.409.....p‐tau181−0.309−0.563−0.055 0.022 −0.317.....VS..........mCSVD−0.145−0.259−0.032 0.018 −0.341 R 2 = 0.905 R 2 = 0.000 Note: Bold values are statistically significant.Abbreviations: β (std): standardized beta coefficient; AmyR, Aβ42/Aβ40 ratio; BHI, Breath‐Holding Index; mCSVD, modified cerebrovascular small vessel disease; TyG, triglyceride and glucose Index; VS, vascular score.John Wiley & Sons, Ltd. Multivariate regression analyses in subgroup with available MRI data. Note: Bold values are statistically significant. Abbreviations: β (std): standardized beta coefficient; AmyR, Aβ42/Aβ40 ratio; BHI, Breath‐Holding Index; mCSVD, modified cerebrovascular small vessel disease; TyG, triglyceride and glucose Index; VS, vascular score. DISCUSSION In this study, we investigated CVR in biologically confirmed AD using TCD and examined its relationship with plasma biomarkers of AD pathology and APOE genotype. Our goal was to clarify the vascular contribution to AD pathophysiology by integrating functional hemodynamic measures with molecular and genetic markers, thereby identifying early, non‐invasive indicators of neurovascular dysfunction that may bridge vascular and amyloid–tau pathways within the disease continuum. In line with previous literature, CVR assessed through the BHI was significantly reduced in AD compared with age‐matched controls, 2 , 12 , 30 confirming the robustness of this index as a sensitive marker of cerebrovascular function. Notably, our study was performed on a strictly biologically defined cohort fulfilling CSF‐based criteria for AD 31 but also considers clinical and radiological characteristics to minimize potential confounding associated with mixed dementia profiles and vascular comorbidities. We believe that this approach advances current knowledge by extending evidence of CVR impairment beyond studies relying solely on clinical AD diagnoses. 2 , 32 , 33 , 34 , 35 Moreover, the absence of differences in vascular scores and metabolic indices across AD subgroups further indicates that CVR impairment is not solely attributable to traditional vascular risk factors or insulin resistance but likely reflects intrinsic neurovascular dysfunction associated with AD pathology. 36 , 37 This observation strengthens the rationale for including CVR measures among early indicators of neurovascular uncoupling in AD. On the other hand, despite clear CVR impairment and differently from previous studies, 12 resting cerebral blood flow (CBF) (MCA MFV) was preserved relative to controls, suggesting that reactivity deficits may precede overt perfusion failure in the early clinical stages of the disease. This observation supports the hypothesis that functional alterations of the neurovascular unit represent an initial, potentially reversible phase of vascular compromise before structural damage and CBF reduction becomes detectable. When considering APOE subgroups, BHI values did not differ significantly between ε3 and ε4 carriers; however, the pattern of associations with plasma biomarkers did reveal genotype‐related differences. Specifically, a significant negative correlation between BHI and both plasma p‐tau181 and p‐tau181/AmyR was observed exclusively in ε3 carriers, suggesting a stronger coupling between vascular function and AD‐related pathology in this group. In contrast, although ε4 carriers exhibited clear CVR impairment, BHI was not significantly related to AD biomarkers or other examined factors. This absence of correlation could possibly reflect a complex vascular phenotype associated with the ε4 genotype, in which widespread endothelial dysfunction, blood–brain barrier disruption, inflammation, and altered CBF regulation may overshadow or mask specific biomarker relationships. 38 , 39 , 40 , 41 In this context, the ε3 genotype might represent a more “coupled” neurovascular profile, where perfusion responsiveness remains linked to molecular pathology, whereas in ε4 carriers an intrinsic vascular compromise may limit further reactivity modulation, resulting in a ceiling effect (Figure 3). Our results are consistent with prior experimental data showing differential APOE effects on neurovascular regulation, 34 with ε3 carriers displaying higher endothelial nitric oxide synthase expression and less pronounced neuroinflammatory activity compared with ε4 carriers. 42 , 43 , 44 Supporting this interpretation, in the MRI subgroup, CVR correlated with mCSVD burden, insulin resistance, and plasma AD markers only in ε3 carriers. Notably, the regression analysis, controlling for mCSVD burden, revealed an even more pronounced pattern of association between CVR and plasma AD biomarkers in ε3 carriers. Although these analyses were limited by the smaller number of participants per subgroup, they reinforced the notion that the ε3 phenotype integrated metabolic (TyG) and molecular dimensions, with a more “coupled” neurovascular profile in which perfusion responsiveness reflects both metabolic stress and AD pathology, whereas ε4 appears characterized by a more pervasive and less AD‐related vascular impairment. Breath‐Holding Index (BHI), phosphorylated tau 181 (p‐tau181) in Alzheimer's disease (AD) patients according to apolipprotein E (APOE) genotype. The relationship between the BHI and plasma p‐tau181 levels in AD, stratified by APOE genotype – specifically APOE ε4 carriers (a) and APOE ε3 carriers (b) is illustrated. Correlation coefficients and p values are reported to facilitate interpretation. Notably, in APOEε3 carriers (b), a clear inverse association emerges: Higher cerebrovascular reactivity (CVR) correlates with lower p‐tau181 concentrations, suggesting better vascular health in earlier disease stages with a lower pathological burden. Conversely, reduced BHI corresponds to elevated p‐tau181 levels, reflecting more advanced neurodegeneration (p < 0.001; β = −0.626). In sharp contrast, no such relationship is observed in APOE ε4 carriers (a), where CVR appears unrelated to p‐tau181 levels across the full range of biomarker values (p = 0.800; β = −0.048). This striking divergence between genotypes may imply that APOE ε4 carriers experience an earlier or more diffuse impairment of endothelial integrity, obscuring any direct link between CVR and tau pathology. Such findings underscore the possibility of distinct genotype‐driven pathways of vascular involvement in AD. In this scenario, insulin resistance may contribute to impaired cerebrovascular reactivity through several mechanisms, including endothelial dysfunction, reduced nitric oxide bioavailability, and altered neurovascular coupling. 45 , 46 Indeed, the association between TyG and BHI emerged only in APOE ε3 carriers, suggesting that where neurovascular responsiveness remains biologically linked to disease pathology, metabolic–vascular interactions are still detectable, though they may be masked in ε4 carriers by a more pervasive vascular dysfunction. Moreover, as previously mentioned, the relatively mild structural burden observed on MRI suggests that impaired CVR may precede overt SVD, consistent with the concept of cerebrovascular resilience loss–functional dysregulation occurring before irreversible structural injury. Subtle endothelial or microvascular alterations, undetectable by conventional MRI, 47 may nonetheless be captured by dynamic hemodynamic measures such as BHI, underscoring its potential role as early indicators of neurovascular compromise. Taken together, these findings indicate that AD‐related CVR impairment reflects intrinsic neurovascular dysfunction modulated by APOE genotype, contributing to the heterogeneity of vascular resilience and clinical trajectories observed in AD. Future research should combine functional vascular assessment with emerging plasma and imaging markers of endothelial dysfunction 26 , 27 to better delineate genotype‐specific vascular mechanisms and their contribution to disease progression. Our findings should be interpreted considering certain limitations. The relatively small sample size and the cross‐sectional design restrict the ability to infer causality or capture longitudinal dynamics of CVR changes over the disease course. However, the internal biological homogeneity of the cohort and the multimodal integration of vascular, metabolic, and molecular data strengthen the interpretability of our results. Future studies in larger cohorts will be required to formally compare the relative contribution of all the aforementioned factors to plasma tau biomarkers using integrative multivariable models. The absence of MRI data in the control group represents an additional constraint, although these participants were carefully screened for vascular risk factors and cognitive status. APOE genotyping was not available in healthy controls as well, limiting the possibility of genotype‐adjusted comparisons across groups. In addition, TCD‐based assessment of CVR, while highly sensitive, is operator‐dependent and may introduce variability in measurement. Moreover, although capnography was used to monitor end‐tidal CO2 and apnea performance, breath‐holding–based CVR assessment may still be influenced by interindividual variability in chemosensitivity and apnea tolerance, which could contribute to variability. Plasma p‐tau181 is currently one of the most widely validated and clinically implemented blood‐based biomarkers of AD, demonstrating strong correspondence with tau deposition and disease progression in multiple cohorts. 48 Emerging evidence suggests that plasma p‐tau217 may provide higher diagnostic accuracy and closer alignment with core AD pathology. 49 Considering this, we conducted a supplementary sensitivity analysis using the plasma p‐tau181/AmyR, a biomarker that has been shown to be more strongly associated with measures of amyloid‐positron emission tomography burden and longitudinal cognitive decline, revealing associations comparable to those reported for p‐tau217. 29 This analysis yielded results consistent with those obtained using p‐tau181 alone, further supporting the consistency of our findings. However, future investigations incorporating p‐tau217 will be important to determine whether genotype‐specific cerebrovascular–biomarker associations are further strengthened using newer tau markers. Given the cross‐sectional nature of the study, the observed associations do not allow causal inferences, and it remains unclear whether CVR impairment represents a contributing factor to AD pathology or a downstream consequence of neurodegenerative processes; longitudinal data will be required to disentangle the temporal relationship between them. Nonetheless, our findings delineate a biologically grounded vascular signature in AD, in which CVR measured by TCD correlates with plasma biomarkers of AD pathology and metabolic dysfunction in an APOE genotype‐specific manner. These data support the view that functional vascular impairment represents an integral and measurable component of AD biology, reflecting the interplay between molecular pathology and neurovascular resilience. From a translational perspective, BHI may serve a dual role in AD by supporting vascular risk stratification when integrated with plasma biomarkers and by enabling longitudinal monitoring of cerebrovascular dynamics. Its non‐invasive nature and sensitivity to functional vascular changes make it a practical tool for identifying vascular vulnerability and tracking disease progression. This translational dimension is particularly important in the context of emerging anti‐amyloid therapies, where identifying patients with increased vascular susceptibility could help anticipate or mitigate complications such as amyloid‐related imaging abnormalities. Overall, BHI may evolve from a research measure to a bedside index of vascular resilience, complementing plasma biomarkers for patient selection and monitoring. Longitudinal and interventional studies are warranted to determine whether reduced BHI can predict clinical or biomarker disease progression particularly across the transition from prodromal to symptomatic stages and to test whether vascular or metabolic interventions can modify CVR trajectories and their impact on AD progression. AUTHOR CONTRIBUTIONS Conceptualization and study design: Dr. M.G. Di Donna and Dr. M. R. Bagnato (First Authors), Prof. A. Martorana and Prof. M. Diomedi (Senior Authors). TCD data acquisition and analysis: Dr. M.G. Di Donna and Dr. M. R. Bagnato. MRI analysis and radiological scoring: Dr. C.G. Bonomi and Prof. M. Diomedi. Statistical analysis: Dr. C.G. Bonomi and Dr. C. Motta. Interpretation of results: All authors. Writing – original draft: Dr. M.G. Di Donna and Dr. M. R. Bagnato. Writing – review and editing: All authors. Supervision: Prof. A. Martorana and Prof. M. Diomedi (Senior Authors). All authors have read and approved the final version of the manuscript and agree to be accountable for the integrity and accuracy of the work. CONFLICT OF INTEREST STATEMENT The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information. ETHICAL APPROVAL The study design was approved by Area Lazio 2 Institutional Review Board. All participants granted written consent to participate to the study. Study procedures followed were in accordance with the Declaration of Helsinki. GUARANTORS Dr. M.G. Di Donna and Prof. A. Martorana serve as guarantors for the study and accept full responsibility for the work and the decision to submit for publication. Supporting information Supporting Information Supporting Information