Alzheimer’s Disease-Like Brain Pattern Biomarker: Capturing Risks and Predicting Disease Onset

Alzheimer’s Disease-Like Brain Pattern Biomarker: Capturing Risks and Predicting Disease Onset | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Alzheimer’s Disease-Like Brain Pattern Biomarker: Capturing Risks and Predicting Disease Onset Peter Kochunov, Si Gao, Lauren Salminen, Neda Jahanshad, Talia Nir, and 14 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6770628/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 27 Apr, 2026 Read the published version in Molecular Psychiatry → Version 1 posted 10 You are reading this latest preprint version Abstract Preventing Alzheimer’s disease (AD) requires early-warning biomarkers. We developed a Regional Vulnerability Index (RVI) that quantifies individual brain similarity to AD patients' expected brain deficit patterns. We calculated regional effect sizes to establish brain deficit patterns in amyloid-positive AD cases compared to amyloid-negative healthy controls. The RVI-AD was calculated as a linear index of individual similarity to this established brain pattern in AD. Initially, we demonstrated RVI-AD elevation associated with risk factors in 335 participants (mean age: 49±13 years) in the Amish Connectome Project, followed by an independent sample consisting of 26,010 participants (mean age: 64±7 years) from the UK Biobank. Genetic and cardiovascular risks were evaluated using APOE-e4 genotype and Framingham Cardiovascular Risk Scores (FCVRS), respectively. Additionally, we assessed the risk of converting from MCI to dementia in N=1,932 participants (mean age: ~74) from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Participants with the APOE-e4 allele had significantly elevated RVI-AD indices (p<0.05); FCVRS significantly contributed to higher RVI-AD in an APOE-e4-specific manner (p<0.01), replicable across the samples. In the ADNI cohort, RVI-AD significantly predicted conversion from MCI to dementia in the next decade, particularly within the first 3 years (AUC=74%). In healthy individuals, the RVI-AD index detected the insidious impact of APOE-ε4 and cardiovascular risks in otherwise normally aging cohorts. Elevated RVI-AD also predicted conversion to dementia within 10 years in the older, high-risk cohort. Further development of this brain-pattern similarity-based approach may yield a noninvasive, clinically accessible biomarker to aid early detection of the subtle to more imminent effects of AD risks. Health sciences/Biomarkers Health sciences/Diseases Big data Alzheimer’s disease neuroimaging RVI amyloid cerebrovascular disease Figures Figure 1 Figure 2 Figure 3 Significance Alzheimer’s disease (AD) is the leading cause of cognitive and functional aging-related disability and has limited treatment options. Efforts to prevent or delay its onset are the chief priority for ensuring the health of an aging population. This requires sensitive and specific biomarkers capable of tracking the action of AD risk factors at the stage of clinical normalcy. We developed a Regional Vulnerability Index (RVI) that quantifies individual brain similarity to AD patients' expected brain deficit patterns. We tested two overall hypotheses that RVI-AD is sensitive to the best-known risk factors for AD in young-to-middle-aged individuals (APOE-e4 genotype and cardiovascular risks) and that elevation in RVI-AD can predict conversion to dementia in older individuals with mild cognitive impairment (MCI). Introduction Alzheimer’s disease (AD) is the leading cause of cognitive decline and functional disability among people over the age of 60 years. AD has limited treatment options, and therefore efforts to prevent or delay its onset are the chief priority for ensuring the health of aging population ( 1 , 2 ). The gradual, cumulative brain changes that lead to AD occur over several decades prior to the first signs of cognitive impairment and this provides a window of opportunity for disease-modifying/prevention interventions ( 3 – 6 ). Yet the life-long action of AD risks on an individual is not well-understood and this limits the efforts to evaluate personalized risk mitigation strategies that may delay or prevent AD ( 5 ). Existing approaches are focused on early diagnosis of people at risk for AD and use a combination of amyloid- or tau-sensitive positron emission tomography (PET), blood and cerebrospinal fluid (CSF) analysis, MRI-based brain morphometry, and neurobehavioral clinical/cognitive assessments ( 5 ). These approaches straddle sensitivity versus specificity versus cost, invasiveness, and availability concerns and are generally not suitable for capturing both effects of AD risk factors in typically aging samples and for predicting conversion to AD in high-risk individuals. Structural MRI comes close to this because it is a non-invasive and widely available alternative for assessing preclinical AD risk. However, clinical MRI-based findings in AD, such as hippocampal shrinkage and ventricular enlargement, are non-specific and typically emerge after the onset of prodromal symptoms ( 7 – 11 ). To address these limitations, we propose an alternative strategy that uses structural MRI to capture regional similarities to AD-specific brain changes (i.e., brain deficit patterns)( 12 – 14 ). We first conducted a meta-analysis of regional MRI effect sizes comparing amyloid PET-confirmed AD cases and amyloid-negative controls. From this, we ranked AD’s impact on brain structures and used this regional pattern to develop a Regional Vulnerability Index (RVI) for AD. RVI-AD is measuring the agreement between an individuals’ brain and the characteristic brain patterns in AD, rather than focusing solely on measuring individual brain structures such as the hippocampus or temporal cortex. Other multivariate approaches combine biomarkers, including amyloid pattern similarity scores ( 15 ), into a unifying measure using complex machine and deep learning methods ( 16 – 19 ). In contrast, RVI-AD is a linear measure applied to standard anatomical brain MRI. Here, we used RVI-AD to test the hypotheses that ( 1 ) known risk factors for AD act throughout adulthood, leading to the gradual formation of AD brain deficit patterns, and ( 2 ) the development of these patterns may predict dementia onset. If RVI-AD can capture these processes, it may allow for a mechanistic evaluation of how risk factors influence the brain’s structural progression from healthy aging to AD over time. The first hypothesis was tested by tracking the actions of two established risk factors for AD in healthy adults. The apolipoprotein E (APOE) ε4 allele is the best-validated genetic risk factor for the late-onset form of AD, associated with a 3- to 15-fold elevation of the risk of developing AD ( 20 – 22 ) in people of central European ancestry ( 23 ). APOE is a cholesterol- and triglyceride-transporting lipoprotein with lipid binding sites ( 24 ) and has three major isoforms in humans ( 25 ). Both homozygous and heterozygous carriers of the ε4 isoform have a higher affinity for transporting low-density cholesterol and face a higher risk of developing AD with 30–60% developing dementia by age 85( 26 ). We hypothesized that RVI-AD may capture brain abnormalities linked to AD ( 27 – 31 ) in the brains of healthy ε4 allele carriers. Poor cardiovascular health is another significant risk factor and a key target for individualized efforts for dementia prevention ( 32 – 36 ). Conditions like coronary artery disease and myocardial infarction are associated with a 3- to 10-fold risk elevation for developing AD ( 37 , 38 ). We quantified cardiovascular risks using the Framingham Cardiovascular Risk Score (FCVRS) ( 39 ). FCVRS is widely used to account for the effects of cholesterol, hypertension, and diabetes mellitus, in addition to age and sex, on future cardio- or cerebrovascular disease ( 39 ). FCVRS has been validated for predicting important health outcomes such as myocardial infarction, stroke, and death. The APOE ε4 allele and cardiovascular health are interlinked in AD-related pathogenesis; a higher affinity for low-density cholesterol transport leads to arterial stiffening, hypertension, and risks for neurological sequelae in middle age and beyond ( 40 , 41 ), all of which are known risks for AD ( 36 , 42 ). We tested the first hypothesis by analyzing two cohorts of neurologically healthy, middle-aged to older adults at no immediate risk for AD. We tested if RVI-AD is sensitive to the action of known AD risk factors during neurologically normal state and compared it to effect sizes observed in different brain regions affected by AD, including hippocampus volume and frontal cortical gray matter thickness. Our discovery cohort consisted of Old Order Amish and Mennonite (OOA/M) participants, and the replication cohort consisted of UK Biobank (UKBB) participants. The OOA/M participants, of central European descent, share a rural, farm-based dwelling with low rates of alcohol, tobacco, and illicit substance use ( 43 ). This population has much better genetic, economic, and environmental uniformity compared to general populations. UKBB participants were recruited from urban and suburban populations. The vastly different environmental backgrounds of these two populations should provide a more rigorous replication test for the generalizability of the findings. We tested the second hypothesis using longitudinal data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), which includes older age samples with higher risks for developing AD, to evaluate whether elevation of RVI-AD was associated with an increased risk of future dementia. An ideal early biomarker for AD should capture the risks associated with both APOE ε4 allele and cardiovascular health at a preclinical, healthy stage, as well as predict conversion to dementia when risk is elevated. Methods Risks at Normalcy: Discovery and Replication Samples The discovery sample consisted of 343 OOA/M participants (140M/ 203F, age range 21–72 years, mean ± s.d.: 49.1 ± 12.7 years). Exclusion criteria included a history of epilepsy, cerebrovascular accidents, head injury with cognitive sequelae, intellectual disability, and unstable major medical conditions at the time of the study. Research participants with substance dependence within the past 6 months or current substance use disorder (except nicotine) were also excluded. Other exclusion criteria for this analysis included current or lifetime major psychiatric illness. Study participants gave written informed consent approved by the University of Maryland IRB. The replication sample consisted of participants in the UK Biobank where we analyzed a subset of N = 31,440 participants (mean age = 63.81 ± 7.44 years; 14,813M/16,627F) who were free of neuropsychiatric illnesses and for whom neuroimaging and clinical data were available. This included structural and diffusion white matter imaging phenotypes. Data were collected between 2012 and 2021 ( 36 ). The UK Biobank received ethics approval from the North West Multi-Center Research Ethics Committee (11/NW/03820). Imaging in ACP and UKBB RVI were generated using structural and diffusion imaging data collected using similar imaging protocols and processing using consistent ENIGMA structural and diffusion pipelines, see Supplement. Effect Sizes for Amyloid-Positive AD from ADNI and OASIS NIA-AA has defined biological Alzheimer’s disease based on the presence of abnormal levels of beta-amyloid or tau protein in biofluids or on brain PET scans in subjects who experienced persistent clinical decline in one or more clinical domains, including memory ( 5 ). The regional vulnerability index for AD (RVI-AD) was calculated using the effect sizes derived between amyloid-positive (Aβ+) AD individuals and healthy amyloid-negative (Aβ-) stable controls from two datasets: the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (132 Aβ + dementia cases and 223 Aβ- healthy controls) and the Open Access Series of Imaging Studies (OASIS) (36 Aβ + dementia cases and 180 healthy Aβ- controls). Altogether, imaging data for ADNI and OASIS datasets for 571 subjects–for whom Aβ status was available–served as the basis for the RVI-AD analysis. We calculated Cohen’s d estimates for cortical thickness and subcortical gray matter differences region-by-region between 168 Aβ + dementia cases and 403 Aβ- healthy controls by combining the two datasets using ENIGMA structural imaging pipelines( 43 ) ( Table S1 ). These subjects lacked white matter diffusion imaging phenotypes, so white matter regional effect sizes were derived from a previous report on the ADNI participants. These effect sizes were calculated based on 48 AD cases and 53 controls also using the ENIGMA pipeline, although no specific Aβ status was considered there( 44 ). In total, we extracted 64 regions from the whole brain, including 33 regional cortical thickness measures, 7 subcortical gray matter volumes and 24 regional white matter integrity measurements where data from left and right hemispheres were averaged ( Table S1 ). To calculate Cohen’s d , we performed a linear regression for the diagnosis variable while including age and sex, their interaction, and the total intracranial volume as covariates, and performed t -to- d value conversion. ADNI participants and conversion to dementia ADNI includes an ongoing, longitudinal multisite study of aging( 45 , 46 ). We tested how baseline RVI-AD values were associated with risks of conversion to dementia. Baseline neuroimaging summary data were downloaded from ADNI database ( www.loni.usc.edu/ADNI ) to create RVI-AD measurements for each participant. We also downloaded up to 12 years of follow up clinical data that documented whether the individual status has remained stable or changed to dementia or MCI. The final ADNI sample consisted of (N = 1,932, age = 73.6 ± 7.2, M/F = 1043/889) that included 586 cognitively normal (CN) older healthy controls (age = 73.8 ± 7.2, M/F = 286/318) and 965 participants with mild cognitive impairment (MCI, age = 72.9 ± 7.5, M/F = 569/397) at baseline. 380 participants with dementia at baseline (age = 75.1 ± 7.1, M/F/=207/173) were included as a comparison. These analyses excluded participants with Aβ status data available, as those data had been used to create effect sizes for RVI. Conversion information was not available for all participants because some had dropped out of the study or failed the follow up. For the available data in the CN at baseline group, N = 35 (age = 75.2 ± 3.9. M/F = 17/18) had changed clinical status to dementia since baseline (average conversion period = 4.7 ± 2.5 years) and N = 76 (average age = 75.5 ± 4.6, M/F = 41/49) had clinical status changed to MCI (average conversion period = 2.9 ± 2.6 years). In the MCI at baseline group, N = 348 (age = 74.0 ± 7.0, M/F = 201/147) had clinical status changed to dementia since baseline (average conversion period = 2.3 ± 2.4 years). Genotyping for APOE status. All datasets provided APOE genotypes for all individuals and the genotyping methodology is provided in the Supplement. Framingham cardiovascular risk score The Framingham cardiovascular risk score (FCVRS) is a weighted composite index developed to predict a 10-year risk of a major cardiovascular event such as heart attack, stroke or death( 47 ). FCVRS scores were calculated for all subjects in the discovery and replication datasets using the uniform FCVRS algorithm( 30 ). Only N = 498 subjects in ADNI baseline sample had all of the measures needed to calculate FCVRS. Statistical analysis RVI calculation The Regional Vulnerability Index (RVI) scores were calculated using the ‘RVIpkg’ in [R] software based on our previous publication( 48 ) with some revisions. The original RVI calculated the correlational agreement between an individual’s regional brain measures and the pattern of regional SSD-related brain deficits( 49 ). It was used for patients with a diagnosis and required a sample of healthy controls to perform normalization [70]. It is therefore inappropriate for most of the current samples when participants at baseline had no AD. Accordingly, we computed the modified RVI as a normalized dot product of two vectors: the vector of individual regional deviations (coded as z-score deviations from the mean of the group) and the vector of AD effect sizes (Table S1 ). Specifically, the RVI was calculated for each subject as the dot product between vectors Z and E normalized by the dimensions of the vector using the following equation. $$\:RVI={\sum\:}_{i=1}^{N}({Z}_{i}\times\:{E}_{i})$$ where Z is the vector of deviation from the mean and E is the vector of meta-analytical effect size (Cohen’s d coefficients) for corresponding regional measures for AD. N is the dimension of the vector, i.e., the total number of imaging phenotypes for modalities including the whole-brain, cortical, subcortical, and white matter. Positive RVI values indicate that the regional pattern of an individual coincides with the expected pattern of AD. Therefore, the modified RVI version is similar to the original RVI in terms of relying on the regional effect sizes as the blueprint to define an individual’s similarity to the illness but using a full sample vector alignment rather than relying on healthy control group to achieve normalization. Hypothesis testing We first evaluated if individual RVI-AD derived from Aβ + AD cases and Aβ- healthy controls can significantly identify the impact of the known genetic risk factor for AD - the APOE-ε4 genotype - on the AD-like brain patterns, using a Student’s t -test. This was followed by investigating the impacts of FCVRS and the joint APOE genotype and FCVRS effects on RVI-AD using linear regressions using Model 1: RVI-AD ~ β ε4 ⋅ E4 + β FCVRS ⋅ FCVRS + β ε4*FCVRS ⋅E4*FCVRS (Model 1) Primary analyses were performed on the whole brain RVI-AD with follow up analyses examining tissue-specific RVI-AD. We tested each hypothesis first in the ACP, and then sought replication in the UKBB samples. Next, we used the ADNI dataset to test if higher RVI-AD values in people diagnosed as MCI are associated with risk for conversion to dementia. The primary analysis was in the MCI group, where we compared those who did not convert and remained stable (MCI→MCI) in the following decade to those who converted to dementia (MCI→dementia), using t -tests. We also compared these two groups to those with dementia at baseline and those who maintained the CN status. We then also tested this in the CN group where a small cohort had converted to dementia, although the sample size for converters here was small: we compared those who did not convert and remained CN in the following decade (CN→CN) to those converted to dementia (CN→dementia) and also to those only converted to MCI but not dementia (CN→MCI) using t-tests. To estimate the association between RVI-AD and the time to the onset of dementia, we further conducted logistic regression analysis to investigate the odds ratio (OR) of dementia conversion within MCI cohort (Model 2). Conversion t ~ β RVI RVI-AD + Age (Model 2) Our analysis comprised of two parts. Firstly, we examined the odds ratio (OR) of conversion cases occurring annually. Second, we examined the OR of cumulative dementia conversion cases, over a 12-year follow-up period. For both components of the analysis, logistic regression models were fitted for each annual time point or by averaging of all 12 years of data, adjusting for baseline age. Model 3 expanded the model 2 with two additional dementia risk factors (i.e., APOE-ε4 genotype and FCVRS scores) as predictors to assess if the joint models improved performance in explaining the variation in dementia conversion. Conversion t ~ β RVI RVI-AD + β ε4 ⋅ APOE-ε4 + β FCVRS ⋅ FCVRS + Age (Model 3) Model performance was evaluated using likelihood ratio tests comparing the joint models to the null model. The OR of each predictor reflects the risk associated with a one-unit increase in the predictor. To be able to compare OR between predictors, the predictors were normalized. Note that we cannot directly model the “joint” OR of the predictors, but we can evaluate whether adding APOE4 and/or FCVRS improves the model fit using likelihood ratio tests, where a higher likelihood ratio value of a model indicates a better model fit. As only a portion of the ADNI participants have all the data needed to calculate FCVRS, we used only a subsample with all data available to test the full Model 3. Results APOE Genotype on RVI-AD in healthy samples: discovery and replication Discovery sample The N = 91 APOE-ε4 carriers (38M/ 53F, age: 48.0 ± 12.5) exhibited significantly higher RVI-AD compared to non-carriers (Cohen’s d = 0.29, p = 0.03) (Fig. 1 A). In comparison, the two groups had no significant differences in brain measurements in structures affected by AD, including hippocampal volume (d=-0.01, p = 0.8) and the cortical thickness of the temporal, parietal, and frontal cortical areas (all p > 0.7) (Fig. 1 B-F). The effect sizes of APOE-ε4 on individual brain were significantly correlated with the pattern of effect sizes observed in AD (r = 0.35, p = 0.004, Fig. 1 G). UKBB sample In the UKBB cohort, APOE-ε4 carriers demonstrated significantly higher RVI-AD compared to non-carriers (p = 2·10 − 5 ) (Fig. 1 H). APOE-ε4 carriers had lower hippocampal volume (p = 0.01) but it was not significant after correction for multiple comparisons (Fig. 1 I) and showed no other suggestively significant differences when compared to noncarriers (Fig. 1 J-L). The APOE-ε4 effect sizes for individual brain region measurements were smaller than the RVI-AD effect size (Fig. 1 M). The effect sizes of APOE-ε4 on individual brain regions in UKBB were also significantly associated with the AD effect sizes of these regions as identified by Aβ + AD patients (r = 0.34, p = 0.006, Fig. 1 N). Effects of Cardiovascular Risk on RVI-AD Discovery sample Model 1 was significant in the ACP cohort (F = 4.6, p = 0.0002, Table 1 ), with a significant main effect of FCVRS (p = 0.006) and a significant APOE · FCVRS interaction (p = 0.03, Table 1 ). This significant interaction was due to a stronger correlation between FCVRS and RVI-AD in APOE-ε4 carriers (r = 0.35, p = 4·10 − 5 vs. r = 0.02, p = 0.95) (Fig. 2 A) and the difference in correlation coefficients was significant (z = 2.9, p = 0.004). Notably, the two genotype groups did not significantly differ in general cardiovascular risks as measured by FCVRS (6.5 ± 0.6 vs. 6.5 ± 0.4, p = 0.9), systolic, diastolic, or pulse pressure (117.4 ± 1.5, 72.0 ± 0.9 and 45.4 ± 1.3 vs. 117.2 ± 1.0, 72.6 ± 0.6 and 44.5 ± 0.7; all p > 0.4), suggesting that the APOE-ε4 allele by itself did not significant impact peripheral cardiovascular measures. Table 1 Evaluating Model 1 in participants of Amish Connectome Project (ACP) and UK biobank (UKBB) studies. Values are mean ± SEM of the beta values (is this correct???) Sample Model Statistics APOE4 FCVRS APOE4 x FCVRS ACP (N = 343) Whole-Brain RVI-AD Value F = 4.6, p = 0.002 -0.14 ± 0.9 (t=-0.1,p = 0.9) 0.25 ± 0.09 (t = 2.8,p = 0.006) 0.20 ± 0.09 (t = 2.1,p = 0.03) UKBB (N = 31,440) Whole-Brain RVI-AD Value F = 8.8, p = 7·10 − 6 0.95 ± 0.4 (t = 2.3,p = 0.01) 0.13 ± 0.03 (t = 3.7,p = 2·10 − 4 ) 0.12 ± 0.03 (t = 3.5,p = 5·10 − 4 ) UKBB sample Model 1 was significant in the UKBB cohort (F = 8.8, p = 7·10 − 6 ). We observed significant effects of FCVRS (p = 2·10 − 4 ), APOE-ε4 genotype (p = 0.01), and their interaction (p = 5·10 − 4 ) on RVI-AD (Table 1 ). The correlation between FCVRS and RVI-AD was significant in APOE-ε4 carriers (r = 0.07, p = 4·10 − 7 ) but not in noncarriers (r = 0.00, p = 0.7), and the difference in correlation coefficients was significant (z = 3.7, p = 2·10 − 4 ). (Fig. 2 B). APOE-ε4 carriers and non-carriers did not differ in FVCRS (11.4 ± 0.02 vs. 11.3 ± 0.2, p = 0.07) or systolic, diastolic or pulse pressure (139.9 ± 0.7, 78.0 ± 0.4, and 61.8 ± 0.5 versus 140.9 ± 0.2, 78.7 ± 0.5, and 62.1 ± 0.3, respectively) (all p > 0.2). We also evaluated RVI-AD calculated separately for cortical thickness, subcortical gray matter volumes, and white matter and observed similar trends. Overall, whole-brain RVI-based models demonstrated the highest robustness and consistency in capturing the APOE and FCVRS risks (see Table S2 ). Predicting conversion from MCI to AD in ADNI participants We used logistic regression to evaluate the odds ratio (OR) of conversion to dementia (see Models 2 in Methods) in ADNI participants. Out of N = 965 participants with MCI, N = 335 (age = 74.0 ± 7.0, 201 M/147 F) developed dementia over the next 12 years, with an average conversion period of about 3 years. Participants who converted to dementia had significantly higher baseline RVI compared to those who did not (p = 3·10 − 16 ) (Table 2 , Fig. 3 A). MCI individuals who did not convert to dementia in a decade showed no significant difference in RVI from cognitively normal (CN) elderly adults who remained CN 12 years later (Fig. 3 A). Table 2 Evaluating Models 2 and 3 in the odds ratio (OR) for predicting conversion from MCI to dementia. The values are OR, 95% confidence interval and p-values for MCI subjects whose clinical status changed to dementia over the assessment period. MCI→AD Conversion Models Age RVI-AD APOE4 FCVRS Model 2 (N = 965) Odds Ratio 1.02 (1.01–1.05), p = 0.005 1.74 (1.52–2.03), p = 2·10 − 5 Model 3 (N = 498) Odds Ratio 1.03 (0.99–1.07), p = 0.2 1.35 (1.10–1.64), p = 0.004 1.24 (1.04–1.4), p = 0.01 1.08 (0.90–1.3), p = 0.9 In addition, N = 37 out of the M = 586 cognitively normal ADNI participants (age = 75.2 ± 3.9. M/F = 17/18) also developed dementia (CN→dementia) after an average conversion period of ~ 5 years. This small sample also showed significantly higher baseline RVI-AD compared to CN participants who remained stable (CN→CN) (p = 0.03) (Fig. 3 B). In comparison, the N = 76 CN participants who developed MCI during the same period (CN→MCI) had intermediate RVI values between the CN→dementia and CN→CN (N = 461) groups. We used an annual logistic regression analysis to test the timing of the odds ratio for predicting new conversions to dementia based on RVI-AD. Higher baseline RVI-AD was significantly associated with conversion from MCI to annual new dementia cases in each of the first three years (OR = 2.2 to 1.78, p = 3·10 − 5 to 3·10 − 10 ) (Fig. 3 C). In aggregate, the odds ratio for the 12-year cumulative conversion to dementia was highly significant (OR = 1.8, p = 7·10 − 14 ) as shown by the last data point in Fig. 3 D (which also plotted the OR of annual cumulative conversion). We expanded the logistic regression model to include APOE-ε4 genotype status and FCVRS as additional predictors (see Model 3 in Methods). This analysis was conducted on N = 498 participants with available FCVRS data, out of whom N = 243 MCI participants converted to dementia. Both APOE-ε4 genotype (OR = 1.2, p = 0.01) and RVI-AD (OR = 1.7, p = 5·10 − 5 ) contributed to dementia conversion over the 12-year follow-up period. FCVRS did not contribute significantly to dementia conversion (OR = 0.0). This suggests that in this age range and high-risk stage of MCI, FCVRS effects for conversion may no longer be prominent. We evaluated the area under the curve (AUC) for predicting conversion from MCI to dementia and found that the peak AUC was 74%. Including the APOE genotype only slightly improved the AUC to 75% ( Table S3, see supplement ). In comparison, the model that only had FCVRS + APOE4 status showed AUC of 0.57. The predictive ability of baseline RVI-AD values for conversion to dementia declined with time (Fig. 3 C). To estimate RVI-AD’s long-term predictive value as part of a hypothetical regular screening program, we simulated annual, triennial, and quinquennial structural MRI assessments (Methods in Supplementary Information ). The simulations suggest that once MCI is diagnosed, RVI-AD calculated based on annual MRI had the strongest performance (AUC = 0.80), and even triannual follow-up still had clinically relevant AUC (AUC = 0.75) for a ten-year prediction of conversion to dementia ( Figure S1 ). Discussion We evaluated the regional vulnerability index for Alzheimer’s Disease (RVI-AD) to gauge its ability to track the lifelong contributions of established risk factors that eventually culminate in dementia. RVI-AD quantifies an individual’s similarity to AD brain deficit patterns. We used a meta-analytical sample of AD patients with amyloid positive (Aβ+) status and amyloid negative controls to estimate this pattern. In healthy adults, RVI-AD showed a positive correlation with the subtle actions of known APOE-ε4 genetic effects as well as the cardiovascular risks, as measured by Framingham Cardiovascular Risk Score (FCVRS). We observed a significant additive ε4-by-FCVRS effect on RVI-AD, indicating that comparable levels of cardiovascular risks led to a higher similarity to AD patterns in people with the APOE-ε4 genotype. With a few exceptions, these findings were consistent in both farm-dwelling Amish and in urban and suburban UK participants. Higher RVI-AD in the participants ascertained by Alzheimer’s Disease Neuroimaging Initiative (ADNI) was significantly associated with the elevated risk for conversion to dementia. Simulations suggested that RVI-AD could support a screening program that uses annual-to-triannual structural MRI scans to provide clinically relevant (AUC = 0.75–0.80) assessment of risk for conversion to dementia in people diagnosed with MCI. RVI-AD is a novel and promising biomarker capable of capturing both the preclinical AD risk factors as well as predicting the likelihood of dementia development in individuals with MCI. The presence of illness-like brain patterns reflects the cumulative action of associated risk factors and may be informative of the risk of developing a disorder ( 50 ). We tested this hypothesis using two best-known AD risks: APOE-ε4 genotype and cardiovascular factors. The effects of the APOE-ε4 allele on elevating RVI-AD were stable and replicable in two cohorts where the APOE-ε4’s effects on RVI-AD elevation acted primarily through cardiovascular risks. APOE-ε4 is a significant risk factor for cardio-and-cerebrovascular conditions ( 27 , 57 – 62 ). The burden of cardiovascular factors contributes up to 50% of the risk for developing of late-onset AD( 51 , 63 , 64 ) and while there were no significant differences in FCVRS or blood pressure measures in APOE-ε4 carriers vs. non-carriers, RVI-AD still detected the additional brain effects linked to cardiovascular risks in the carriers. Incremental elevations in FCVRS in APOE-ε4 carriers resulted in a greater resemblance to AD-like brain patterns compared to non-carriers, suggesting that cardiovascular risk factors may act on the brain differently in APOE- ε4 carriers. Thus, RVI may provide an early biomarker to index this genetic + cardiovascular risk effect for developing an AD-like brain pattern at the level of the individual at cognitive and medical normalcy. We tested the second hypothesis in individuals with MCI (mean age ~ 74 years) who were at risk for AD because approximately one third of subjects went on to develop dementia ( 6 , 65 ). MCI subjects who developed dementia had significantly higher baseline RVI-AD indices compared to participants who remained stable. This finding was also replicated in the cognitively normal controls. The predictive accuracy of conversion was strongest within three years after the MRI scan. This suggested that additional neurodegeneration needs to occur for MCI to dementia conversion and that annual-to-triannual MRI assessments in people at risk for dementia could maintain clinically acceptable AUC. A structured RVI-based screening program could supplement or provide a cost-effective alternative to AD-focused PET/CSF assessments. Importantly, MCI participants who did not progress to dementia had RVI-AD values similar to the RVI values for the cognitively normal controls (Fig. 3 A), suggesting that low RVI-AD predicted reduced likelihood for conversion. Approximately half of the MCI individuals never develop dementia ( 6 , 66 , 67 ), leading to a decision challenge for disease-modifying treatments that have significant side effects. Present approaches to evaluate individual’s risks, include tau-focused PET scans, volumetric MRI assessments, cognitive batteries, and blood/CSF biomarkers ( 11 , 68 – 70 ). Many of these approaches have high costs and perceived invasiveness ( 73 ). In contrast, RVI-AD uses linear, easily interpretable calculations and measurements from readily available structural MRI. Further validation may enable RVI-AD index to assist in early risk detection and predict more imminent dementia disease onset, especially when combined with other available clinical and biomarker measures. In conclusion, RVI-AD can detect the early, lifetime gradual risk factor effects on the brain before cognitive changes emerge, as well as the more imminent risks for conversion of MCI to AD. AD is a major cause of cognitive dysfunction among older people. With new AD treatments on the horizon, non-invasive brain imaging biomarkers, such as RVI-AD, could identify vulnerable individuals during cognitively normal to prodromal stages, supporting public health efforts to prevent, delay, or reduce the impact of AD. Declarations Funding Support was received from National Institutes of Health grants RF1NS114628, R01NS114628, R01MH112180, R01MH133812, R01MH116948, S10OD023696, R01EB015611, R01MH117601, R01AG095874, U01MH108148, and R01ES033961. These funding sources provided financial support to enable the design and conduct of the study or collection, management, or analysis of the data. None of the funding agencies had a role in interpreting the data or in preparing, reviewing, or approving the manuscript. None had a role in the decision to submit the manuscript for publication. Conflict of interest All authors declare no conflict of interest. References Anonymous (2021) 2021 Alzheimer's disease facts and figures. Alzheimers Dement 17(3):327–406. Majeed J, et al. (2023) Cancer drugs with high repositioning potential for Alzheimer's disease. Expert opinion on emerging drugs . Herholz K, et al. (2002) Discrimination between Alzheimer dementia and controls by automated analysis of multicenter FDG PET. Neuroimage 17(1):302–316. Qiu C (Preventing Alzheimer's disease by targeting vascular risk factors: hope and gap. (1875–8908 (Electronic)). Jack CR, Jr., et al. (2018) NIA-AA Research Framework: Toward a biological definition of Alzheimer's disease. Alzheimers Dement 14(4):535–562. McGrattan AM, et al. (2022) Risk of conversion from mild cognitive impairment to dementia in low- and middle-income countries: A systematic review and meta-analysis. Alzheimer's & dementia (New York, N. Y.) 8(1):e12267. Lo RY, et al. (2011) Longitudinal change of biomarkers in cognitive decline. Arch Neurol 68(10):1257–1266. VAN HOESEN GW & SOLODKIN A (1994) Cellular and Systems Neuroanatomical Changes in Alzheimer's Disease. Annals of the New York Academy of Sciences 747(1):12–35. Solodkin A, et al. (2013) In vivo parahippocampal white matter pathology as a biomarker of disease progression to Alzheimer's disease. The Journal of comparative neurology 521(18):4300–4317. Badea A, et al. (2016) The fornix provides multiple biomarkers to characterize circuit disruption in a mouse model of Alzheimer's disease. Neuroimage 142:498–511. Crystal O, et al. (2023) Detecting conversion from mild cognitive impairment to Alzheimer's disease using FLAIR MRI biomarkers. NeuroImage. Clinical 40:103533. Kress GT, et al. (2023) Preliminary validation of a structural magnetic resonance imaging metric for tracking dementia-related neurodegeneration and future decline. NeuroImage. Clinical 39:103458. Young AL, et al. (2021) Ordinal SuStaIn: Subtype and Stage Inference for Clinical Scores, Visual Ratings, and Other Ordinal Data. Frontiers in artificial intelligence 4:613261. Wang G, et al. (2021) Developing univariate neurodegeneration biomarkers with low-rank and sparse subspace decomposition. Medical image analysis 67:101877. Prosser L, et al. (2020) Amyloid Pattern Similarity Score (AMPSS): A reference region free measure of amyloid PET deposition in Alzheimer’s disease. Alzheimer's & Dementia 16(S4):e042673. Lee JS, et al. (2018) Machine Learning-based Individual Assessment of Cortical Atrophy Pattern in Alzheimer's Disease Spectrum: Development of the Classifier and Longitudinal Evaluation. Sci Rep 8(1):4161. Shi Y, et al. (2022) ASMFS: Adaptive-similarity-based multi-modality feature selection for classification of Alzheimer's disease. Pattern Recognition 126:108566. Feng X, Provenzano FA, & Small SA (2022) A deep learning MRI approach outperforms other biomarkers of prodromal Alzheimer's disease. Alzheimer's research & therapy 14(1):45. Dhinagar N, Thomopoulos S, Laltoo E, & Thompson P (2023) Efficiently Training Vision Transformers on Structural MRI Scans for Alzheimer's Disease Detection . Biffi A, et al. (2010) Genetic variation and neuroimaging measures in Alzheimer disease. Arch Neurol 67(6):677–685. Broer L, et al. (2014) GWAS of longevity in CHARGE consortium confirms APOE and FOXO3 candidacy. The journals of gerontology. Series A, Biological sciences and medical sciences 70(1):110–118. Deelen J, et al. (2011) Genome-wide association study identifies a single major locus contributing to survival into old age; the APOE locus revisited. Aging cell 10(4):686–698. Goate A, et al. (1991) Segregation of a missense mutation in the amyloid precursor protein gene with familial Alzheimer's disease. Nature 349(6311):704–706. Jansen IE, et al. (2022) Genome-wide meta-analysis for Alzheimer's disease cerebrospinal fluid biomarkers. Acta neuropathologica 144(5):821–842. Chakravarty A (2004) Unifying concept for Alzheimer's disease, vascular dementia and normal pressure hydrocephalus - a hypothesis. Medical hypotheses 63(5):827–833. Bateman GA (2004) Pulse wave encephalopathy: a spectrum hypothesis incorporating Alzheimer's disease, vascular dementia and normal pressure hydrocephalus. Medical hypotheses 62(2):182–187. Tai LM, et al. (2016) The role of APOE in cerebrovascular dysfunction. Acta neuropathologica 131(5):709–723. Slooter AJ, et al. (2004) The impact of APOE on myocardial infarction, stroke, and dementia: the Rotterdam Study. Neurology 62(7):1196–1198. Shao A, et al. (2022) Meta-analysis of the association between Apolipoprotein E polymorphism and risks of myocardial infarction. BMC Cardiovasc Disord 22(1):126. D'Agostino RB, Sr., et al. (2008) General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation 117(6):743–753. Lupton MK, et al. (2016) The effect of increased genetic risk for Alzheimer's disease on hippocampal and amygdala volume. Neurobiology of aging 40:68–77. Nir TM, et al. (2022) Effects of ApoE4 and ApoE2 genotypes on subcortical magnetic susceptibility and microstructure in 27,535 participants from the UK Biobank. Pacific Symposium on Biocomputing. Pacific Symposium on Biocomputing 27:121–132. Hua X, et al. (2016) MRI-based brain atrophy rates in ADNI phase 2: acceleration and enrichment considerations for clinical trials. Neurobiology of aging 37:26–37. McKhann G, et al. (1984) Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology 34(7):939–944. Mirra SS, et al. (1991) The Consortium to Establish a Registry for Alzheimer's Disease (CERAD). Part II. Standardization of the neuropathologic assessment of Alzheimer's disease. Neurology 41(4):479–486. Manolio TA, et al. (2012) New Models for Large Prospective Studies: Is There a Better Way? American Journal of Epidemiology 175(9):859–866. Kochunov P, et al. (2006) Retrospective motion correction protocol for high-resolution anatomical MRI. Hum Brain Mapp 27(12):957–962. Glasser MF, et al. (2013) The minimal preprocessing pipelines for the Human Connectome Project. Neuroimage 80:105–124. Sotiropoulos SN, et al. (2013) Advances in diffusion MRI acquisition and processing in the Human Connectome Project. Neuroimage 80:125–143. Ades-Aron B, et al. (2018) Evaluation of the accuracy and precision of the diffusion parameter EStImation with Gibbs and NoisE removal pipeline. Neuroimage 183:532–543. Behrens TE, et al. (2003) Characterization and propagation of uncertainty in diffusion-weighted MR imaging. Magnetic resonance in medicine 50(5):1077–1088. LaMontagne PJ, et al. (2019) OASIS-3: Longitudinal Neuroimaging, Clinical, and Cognitive Dataset for Normal Aging and Alzheimer Disease. medRxiv :2019.2012.2013.19014902. Jahanshad N, et al. (2013) Multi-site genetic analysis of diffusion images and voxelwise heritability analysis: A pilot project of the ENIGMA-DTI working group. Neuroimage 81:455–469. Kochunov P, et al. (2020) ENIGMA-DTI: Translating reproducible white matter deficits into personalized vulnerability metrics in cross-diagnostic psychiatric research. Hum Brain Mapp . Mueller SG, et al. (2005) The Alzheimer's disease neuroimaging initiative. Neuroimaging clinics of North America 15(4):869–877, xi-xii. Petersen RC, et al. (2010) Alzheimer's Disease Neuroimaging Initiative (ADNI): clinical characterization. Neurology 74(3):201–209. Wong ND (2020) Cardiovascular risk assessment: The foundation of preventive cardiology. Am J Prev Cardiol 1:100008. Kochunov P, et al. (2023) Ancestral, Pregnancy, and Negative Early-Life Risks Shape Children's Brain (Dis)similarity to Schizophrenia. Biol Psychiatry 94(4):332–340. Kochunov P, et al. (2019) White Matter in Schizophrenia Treatment Resistance. The American Journal of Psychiatry 176(10):829–838. Kochunov P, et al. (2023) Ancestral, Pregnancy, and Negative Early Life Risks Shape Children's Brain Dis/Similarity to Schizophrenia. Biol Psychiatry 94(4):332–340. Frisoni GB, et al. (2022) The probabilistic model of Alzheimer disease: the amyloid hypothesis revised. Nature reviews. Neuroscience 23(1):53–66. Mahley RW & Huang Y (2012) Apolipoprotein e sets the stage: response to injury triggers neuropathology. Neuron 76(5):871–885. Operto G, et al. (2018) White matter microstructure is altered in cognitively normal middle-aged APOE-ε4 homozygotes. Alzheimer's research & therapy 10(1):48. Operto G, et al. (2019) Interactive effect of age and APOE-ε4 allele load on white matter myelin content in cognitively normal middle-aged subjects. NeuroImage. Clinical 24:101983. Bartzokis G, et al. (2003) White matter structural integrity in healthy aging adults and patients with Alzheimer disease: a magnetic resonance imaging study. Arch Neurol 60(3):393–398. Koutsodendris N, et al. (2023) Neuronal APOE4 removal protects against tau-mediated gliosis, neurodegeneration and myelin deficits. Nature aging 3(3):275–296. Ilveskoski E, et al. (1999) Age-dependent association of apolipoprotein E genotype with coronary and aortic atherosclerosis in middle-aged men: an autopsy study. Circulation 100(6):608–613. Eichner JE, et al. (2002) Apolipoprotein E polymorphism and cardiovascular disease: a HuGE review. Am J Epidemiol 155(6):487–495. Mahley RW (1988) Apolipoprotein E: cholesterol transport protein with expanding role in cell biology. Science 240(4852):622–630. Ngwa JS, et al. (2018) Associations of Pulse and Blood Pressure with Hippocampal Volume by APOE and Cognitive Phenotype: The Alzheimer's Disease Neuroimaging Initiative (ADNI). Dementia and geriatric cognitive disorders 45(1–2):66–78. Bateman GA (2002) Pulse-wave encephalopathy: a comparative study of the hydrodynamics of leukoaraiosis and normal-pressure hydrocephalus. Neuroradiology 44(9):740–748. Scuteri A, Nilsson PM, Tzourio C, Redon J, & Laurent S (2011) Microvascular brain damage with aging and hypertension: pathophysiological consideration and clinical implications. J Hypertens 29(8):1469–1477. George KM, et al. (2023) Association of Early Adulthood Hypertension and Blood Pressure Change With Late-Life Neuroimaging Biomarkers. JAMA network open 6(4):e236431. Gorelick PB, et al. (2011) Vascular contributions to cognitive impairment and dementia: a statement for healthcare professionals from the american heart association/american stroke association. Stroke 42(9):2672–2713. Thaipisuttikul P, Jaikla K, Satthong S, & Wisajun P (2022) Rate of conversion from mild cognitive impairment to dementia in a Thai hospital-based population: A retrospective cohort. Alzheimer's & dementia (New York, N. Y.) 8(1):e12272. Mitchell AJ & Shiri-Feshki M (2009) Rate of progression of mild cognitive impairment to dementia–meta-analysis of 41 robust inception cohort studies. Acta psychiatrica Scandinavica 119(4):252–265. Gauthier S, et al. (2006) Mild cognitive impairment. Lancet 367(9518):1262–1270. Vemuri P, et al. (2009) MRI and CSF biomarkers in normal, MCI, and AD subjects: diagnostic discrimination and cognitive correlations. Neurology 73(4):287–293. Davatzikos C, Bhatt P, Shaw LM, Batmanghelich KN, & Trojanowski JQ (2011) Prediction of MCI to AD conversion, via MRI, CSF biomarkers, and pattern classification. Neurobiology of aging 32(12):2322.e2319-2327. Dayarathna T, et al. (2024) Nanoscale flow cytometry-based quantification of blood-based extracellular vesicle biomarkers distinguishes MCI and Alzheimer's disease. Alzheimers Dement . Barthélemy NR, et al. (2024) Highly accurate blood test for Alzheimer’s disease is similar or superior to clinical cerebrospinal fluid tests. Nature medicine 30(4):1085–1095. Groot C, et al. (2024) Tau Positron Emission Tomography for Predicting Dementia in Individuals With Mild Cognitive Impairment. JAMA neurology 81(8):845–856. Schindler SE & Atri A (2023) The role of cerebrospinal fluid and other biomarker modalities in the Alzheimer’s disease diagnostic revolution. Nature aging 3(5):460–462. Livingston G, et al. (2017) Dementia prevention, intervention, and care. Lancet 390(10113):2673–2734. Nugent KL, et al. (2013) Familial Aggregation of Tobacco Use Behaviors in Amish Men. Nicotine Tob Res . Additional Declarations The authors have declared there is NO conflict of interest to disclose Supplementary Files Supplement1.docx Supplementary Cite Share Download PDF Status: Published Journal Publication published 27 Apr, 2026 Read the published version in Molecular Psychiatry → Version 1 posted Editorial decision: revise 07 Nov, 2025 Review # 2 received at journal 24 Oct, 2025 Reviewer # 2 agreed at journal 08 Oct, 2025 Review # 1 received at journal 23 Sep, 2025 Reviewer # 1 agreed at journal 08 Sep, 2025 Reviewers invited by journal 08 Sep, 2025 Editor assigned by journal 30 May, 2025 Submission checks completed at journal 30 May, 2025 First submitted to journal 29 May, 2025 Unknown event 29 May, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6770628","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":511629526,"identity":"d90755ee-07e3-473d-800b-20ef7c40923f","order_by":0,"name":"Peter Kochunov","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAy0lEQVRIiWNgGAWjYDACHgYGZjCDvQFIGFiQooXnAEiLBClaJBLAJGEd5jyHD34ubLsjZ3Dz+dUNPwokGPjbuxPwarHsbUuWntn2zNjgdk7ZzR6gwyTOnN2AV4vBeR4zZt62w4kzZ+ek3eABajGQyCVWy8wzaTf/EKXlbA9ES78E+7HbRNli2XMsWZrn3GFjfp4cttsyBhI8BP1izpN88DNP2WE5Nvbjz26++WMjx9/eS8BhCCYPmM2DVzmaFvYHBFWPglEwCkbByAQAXdhD2Qe0PnMAAAAASUVORK5CYII=","orcid":"","institution":"The University of Texas Health Science Center at Houston","correspondingAuthor":true,"prefix":"","firstName":"Peter","middleName":"","lastName":"Kochunov","suffix":""},{"id":511629527,"identity":"f88ac016-db2b-44fe-acb7-ae8c770dbd0d","order_by":1,"name":"Si Gao","email":"","orcid":"https://orcid.org/0000-0002-4473-1142","institution":"The University of Texas Health Science Center at Houston","correspondingAuthor":false,"prefix":"","firstName":"Si","middleName":"","lastName":"Gao","suffix":""},{"id":511629528,"identity":"679f8718-3f58-450a-9652-4f67e46252a2","order_by":2,"name":"Lauren Salminen","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Lauren","middleName":"","lastName":"Salminen","suffix":""},{"id":511629529,"identity":"783560bb-09d6-48c9-a430-ba8a1e8f8e8c","order_by":3,"name":"Neda Jahanshad","email":"","orcid":"","institution":"University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Neda","middleName":"","lastName":"Jahanshad","suffix":""},{"id":511629530,"identity":"c58cd569-c176-4f79-9a1f-630921e4719b","order_by":4,"name":"Talia Nir","email":"","orcid":"https://orcid.org/0000-0002-7106-7443","institution":"University of Southern California Keck School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Talia","middleName":"","lastName":"Nir","suffix":""},{"id":511629531,"identity":"aed7adff-9701-4f8b-947e-3e1952a12a14","order_by":5,"name":"Paul Thompson","email":"","orcid":"","institution":"Imaging Genetics Center, Mark and Mary Stevens Neuroimaging \u0026 Informatics Institute, Keck School of Medicine of the University of Southern California, Marina del Rey, California, USA","correspondingAuthor":false,"prefix":"","firstName":"Paul","middleName":"","lastName":"Thompson","suffix":""},{"id":511629532,"identity":"908027e4-96a0-4537-8e84-bdef1238f043","order_by":6,"name":"Xiaoming Du","email":"","orcid":"","institution":"University of Maryland Baltimore","correspondingAuthor":false,"prefix":"","firstName":"Xiaoming","middleName":"","lastName":"Du","suffix":""},{"id":511629533,"identity":"d25adb64-c513-41ec-a5f0-bb98bc000e6b","order_by":7,"name":"Bhim Adhikari","email":"","orcid":"","institution":"Maryland Psychiatric Research Center","correspondingAuthor":false,"prefix":"","firstName":"Bhim","middleName":"","lastName":"Adhikari","suffix":""},{"id":511629534,"identity":"72235586-14a0-438a-b74b-0423af254f81","order_by":8,"name":"Alice Kochunov","email":"","orcid":"","institution":"Maryland Psychiatric Research Center","correspondingAuthor":false,"prefix":"","firstName":"Alice","middleName":"","lastName":"Kochunov","suffix":""},{"id":511629535,"identity":"776ad8ec-1274-481f-af7a-91f6a91875df","order_by":9,"name":"Ryan Cassidy","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Ryan","middleName":"","lastName":"Cassidy","suffix":""},{"id":511629536,"identity":"9193aed0-a12c-42de-ae7b-0031ffc033fd","order_by":10,"name":"Yizhou Ma","email":"","orcid":"","institution":"The University of Texas Health Science Center at Houston","correspondingAuthor":false,"prefix":"","firstName":"Yizhou","middleName":"","lastName":"Ma","suffix":""},{"id":511629537,"identity":"67d5732f-1661-478c-ab19-7bc16c46e268","order_by":11,"name":"Joshua Chiappelli","email":"","orcid":"https://orcid.org/0000-0001-6121-7439","institution":"University of Maryland Baltimore","correspondingAuthor":false,"prefix":"","firstName":"Joshua","middleName":"","lastName":"Chiappelli","suffix":""},{"id":511629538,"identity":"e92ea40b-8fbf-4c13-b316-fef6f80f9edc","order_by":12,"name":"Seth Ament","email":"","orcid":"https://orcid.org/0000-0001-6443-7509","institution":"University of Maryland School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Seth","middleName":"","lastName":"Ament","suffix":""},{"id":511629539,"identity":"af013518-d0c1-4dde-96ed-428dff937946","order_by":13,"name":"Yezhi Pan","email":"","orcid":"","institution":"University of Maryland Baltimore","correspondingAuthor":false,"prefix":"","firstName":"Yezhi","middleName":"","lastName":"Pan","suffix":""},{"id":511629540,"identity":"7da00b12-4600-4c6f-8152-ea42c9faa0f4","order_by":14,"name":"Shuo Chen","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Shuo","middleName":"","lastName":"Chen","suffix":""},{"id":511629541,"identity":"3e1034ae-2728-4d19-afbe-9e5c43b227f6","order_by":15,"name":"Alan Shuldiner","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Alan","middleName":"","lastName":"Shuldiner","suffix":""},{"id":511629542,"identity":"7211e452-1182-40dd-955e-85b82064e04c","order_by":16,"name":"Braxton Mitchell","email":"","orcid":"https://orcid.org/0000-0003-4920-4744","institution":"University of Maryland School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Braxton","middleName":"","lastName":"Mitchell","suffix":""},{"id":511629543,"identity":"587cbcd3-f599-4f26-939d-0809f41248e7","order_by":17,"name":"Jair Soares","email":"","orcid":"https://orcid.org/0000-0002-5466-5628","institution":"McGovern Medical School","correspondingAuthor":false,"prefix":"","firstName":"Jair","middleName":"","lastName":"Soares","suffix":""},{"id":511629544,"identity":"78c7e416-7366-490c-87dd-a91c11cea639","order_by":18,"name":"Elliot Hong","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Elliot","middleName":"","lastName":"Hong","suffix":""}],"badges":[],"createdAt":"2025-05-28 19:00:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6770628/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6770628/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41380-026-03617-0","type":"published","date":"2026-04-27T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":91372587,"identity":"d95e65c1-892c-46aa-9a95-5c71b801d734","added_by":"auto","created_at":"2025-09-15 19:02:46","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":245436,"visible":true,"origin":"","legend":"\u003cp\u003eHealthy APOE4 carriers showed replicable Alzheimer’s disease (AD) brain patterns based on the whole-brain regional vulnerability index for AD (RVI-AD). \u003cstrong\u003eA, H\u003c/strong\u003e: RVI-AD in carriers was significantly higher than noncarriers in the Amish Connectome Project (ACP) discovery and the UK Biobank (UKBB) replication samples. Individual regional measures such as hippocampal volumes (\u003cstrong\u003eB, I\u003c/strong\u003e) and frontal, temporal, and parietal cortical thickness (\u003cstrong\u003eC-E, J-L\u003c/strong\u003e) showed weak and insignificant effects in both cohorts. The effect sizes for carriers vs. noncarriers were substantially larger in whole-brain RVI-AD compared to any individual regional measures including the averaging values of the cortical, subcortical, and white matter measures in both cohorts (\u003cstrong\u003eF, M\u003c/strong\u003e). \u003cstrong\u003eG, N\u003c/strong\u003e: Whole-brain RVI-AD was calculated by Alzheimer’s disease effect sizes of different regions (\u003cem\u003ex\u003c/em\u003eaxis), which showed significant correlation with effect sizes of the APOE genotypes in otherwise healthy people in both cohorts. *Statistically significant; # statistically significant difference was replicated across samples\u003cstrong\u003e. \u003c/strong\u003eRegions of interests are labeled in the table.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6770628/v1/c9b2b3e6b440ce10556925f0.png"},{"id":91372064,"identity":"6dc4bdf5-73d7-4e51-af0b-b8a1997a045d","added_by":"auto","created_at":"2025-09-15 18:54:46","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":36631,"visible":true,"origin":"","legend":"\u003cp\u003eThe Framingham cardiovascular risk score (FCVRS) were significantly associated with whole-brain regional vulnerability index for Alzheimer’s disease (RVI-AD) in APOE4 carriers in both the discovery (\u003cstrong\u003eA\u003c/strong\u003e) (left, r=0.35, p=4·10\u003csup\u003e-5\u003c/sup\u003e) and replication (\u003cstrong\u003eB\u003c/strong\u003e) samples (right, r=0.07, p=4·10\u003csup\u003e-7\u003c/sup\u003e). The correlation between RVI-AD and FCVRS in controls were r=0.02 and 0.0 in ACP and UKBB, respectively (p\u0026gt;0.9)\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6770628/v1/f31b8949e076276ef7d59b66.png"},{"id":91370746,"identity":"f2abba03-0f18-4fe5-adb1-e30a5781d0d6","added_by":"auto","created_at":"2025-09-15 18:46:46","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":91950,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA\u003c/strong\u003e. Baseline MCI participants’ whole-brain RVI-AD effects, where conversion (MCI→dementia) cases showed significantly higher baseline RVI-AD than no conversion (MCI→MCI), and importantly, MCI without conversion showed no significant difference to cognitively normal (CN) elderly healthy controls. The CN→CN group from \u003cstrong\u003eB\u003c/strong\u003e and the dementia cases already diagnosed at baseline are included here for easier comparison. \u003cstrong\u003eB\u003c/strong\u003e. Baseline cognitive normal (CN) participants’ whole-brain RVI-AD effects. CN→CN: CN individuals at baseline who remained CN during the 12-year follow-up; CN→MCI: those converted to MCI; CN→dementia: converted to dementia. \u003cstrong\u003eC\u003c/strong\u003e. Mean and s.e. of baseline RVI of annual new AD cases in the MCI participants: high baseline RVI had a significantly higher odds ratio (OR) for predicting conversion from MCI to dementia in the next three years after the baseline MRI. \u003cstrong\u003eD\u003c/strong\u003e. Mean and s.e. of baseline RVI of cumulative cases: high baseline RVI had a significantly higher odds ratio (OR) for predicting conversion from MCI to dementia in each of the 12 year follow-up.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6770628/v1/dfa452c6c47d0654eb931b51.png"},{"id":107972407,"identity":"ccef1155-9642-4112-9dab-bf36a7335f1f","added_by":"auto","created_at":"2026-04-28 07:07:35","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":727740,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6770628/v1/e37ce582-7563-44cd-96fb-f205beea6ff5.pdf"},{"id":91370749,"identity":"0a3342b2-fd04-4eb3-af3d-4ca1f2183e07","added_by":"auto","created_at":"2025-09-15 18:46:46","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":196819,"visible":true,"origin":"","legend":"Supplementary","description":"","filename":"Supplement1.docx","url":"https://assets-eu.researchsquare.com/files/rs-6770628/v1/d846eedd35d550ff75b94fd2.docx"}],"financialInterests":"The authors have declared there is \u003cb\u003eNO\u003c/b\u003e conflict of interest to disclose","formattedTitle":"Alzheimer’s Disease-Like Brain Pattern Biomarker: Capturing Risks and Predicting Disease Onset","fulltext":[{"header":"Significance ","content":"\u003cp\u003eAlzheimer’s disease (AD) is the leading cause of cognitive and functional aging-related disability and has limited treatment options. Efforts to prevent or delay its onset are the chief priority for ensuring the health of an aging population. This requires sensitive and specific biomarkers capable of tracking the action of AD risk factors at the stage of clinical normalcy. We developed a Regional Vulnerability Index (RVI) that quantifies individual brain similarity to AD patients' expected brain deficit patterns. We tested two overall hypotheses that RVI-AD is sensitive to the best-known risk factors for AD in young-to-middle-aged individuals (APOE-e4 genotype and cardiovascular risks) and that elevation in RVI-AD can predict conversion to dementia in older individuals with mild cognitive impairment (MCI).\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eAlzheimer\u0026rsquo;s disease (AD) is the leading cause of cognitive decline and functional disability among people over the age of 60 years. AD has limited treatment options, and therefore efforts to prevent or delay its onset are the chief priority for ensuring the health of aging population (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). The gradual, cumulative brain changes that lead to AD occur over several decades prior to the first signs of cognitive impairment and this provides a window of opportunity for disease-modifying/prevention interventions (\u003cspan additionalcitationids=\"CR4 CR5\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Yet the life-long action of AD risks on an individual is not well-understood and this limits the efforts to evaluate personalized risk mitigation strategies that may delay or prevent AD (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Existing approaches are focused on early diagnosis of people at risk for AD and use a combination of amyloid- or tau-sensitive positron emission tomography (PET), blood and cerebrospinal fluid (CSF) analysis, MRI-based brain morphometry, and neurobehavioral clinical/cognitive assessments (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). These approaches straddle sensitivity versus specificity versus cost, invasiveness, and availability concerns and are generally not suitable for capturing both effects of AD risk factors in typically aging samples and for predicting conversion to AD in high-risk individuals. Structural MRI comes close to this because it is a non-invasive and widely available alternative for assessing preclinical AD risk. However, clinical MRI-based findings in AD, such as hippocampal shrinkage and ventricular enlargement, are non-specific and typically emerge after the onset of prodromal symptoms (\u003cspan additionalcitationids=\"CR8 CR9 CR10\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo address these limitations, we propose an alternative strategy that uses structural MRI to capture regional similarities to AD-specific brain changes (i.e., brain deficit patterns)(\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). We first conducted a meta-analysis of regional MRI effect sizes comparing amyloid PET-confirmed AD cases and amyloid-negative controls. From this, we ranked AD\u0026rsquo;s impact on brain structures and used this regional pattern to develop a Regional Vulnerability Index (RVI) for AD. RVI-AD is measuring the agreement between an individuals\u0026rsquo; brain and the characteristic brain patterns in AD, rather than focusing solely on measuring individual brain structures such as the hippocampus or temporal cortex. Other multivariate approaches combine biomarkers, including amyloid pattern similarity scores (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e), into a unifying measure using complex machine and deep learning methods (\u003cspan additionalcitationids=\"CR17 CR18\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). In contrast, RVI-AD is a linear measure applied to standard anatomical brain MRI. Here, we used RVI-AD to test the hypotheses that (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) known risk factors for AD act throughout adulthood, leading to the gradual formation of AD brain deficit patterns, and (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) the development of these patterns may predict dementia onset. If RVI-AD can capture these processes, it may allow for a mechanistic evaluation of how risk factors influence the brain\u0026rsquo;s structural progression from healthy aging to AD over time.\u003c/p\u003e \u003cp\u003eThe first hypothesis was tested by tracking the actions of two established risk factors for AD in healthy adults. The apolipoprotein E (APOE) ε4 allele is the best-validated genetic risk factor for the late-onset form of AD, associated with a 3- to 15-fold elevation of the risk of developing AD (\u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e) in people of central European ancestry (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). APOE is a cholesterol- and triglyceride-transporting lipoprotein with lipid binding sites (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e) and has three major isoforms in humans (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Both homozygous and heterozygous carriers of the ε4 isoform have a higher affinity for transporting low-density cholesterol and face a higher risk of developing AD with 30\u0026ndash;60% developing dementia by age 85(\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e). We hypothesized that RVI-AD may capture brain abnormalities linked to AD (\u003cspan additionalcitationids=\"CR28 CR29 CR30\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e) in the brains of healthy ε4 allele carriers.\u003c/p\u003e \u003cp\u003ePoor cardiovascular health is another significant risk factor and a key target for individualized efforts for dementia prevention (\u003cspan additionalcitationids=\"CR33 CR34 CR35\" citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e). Conditions like coronary artery disease and myocardial infarction are associated with a 3- to 10-fold risk elevation for developing AD (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e). We quantified cardiovascular risks using the Framingham Cardiovascular Risk Score (FCVRS) (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). FCVRS is widely used to account for the effects of cholesterol, hypertension, and diabetes mellitus, in addition to age and sex, on future cardio- or cerebrovascular disease (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). FCVRS has been validated for predicting important health outcomes such as myocardial infarction, stroke, and death. The APOE ε4 allele and cardiovascular health are interlinked in AD-related pathogenesis; a higher affinity for low-density cholesterol transport leads to arterial stiffening, hypertension, and risks for neurological sequelae in middle age and beyond (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e), all of which are known risks for AD (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWe tested the first hypothesis by analyzing two cohorts of neurologically healthy, middle-aged to older adults at no immediate risk for AD. We tested if RVI-AD is sensitive to the action of known AD risk factors during neurologically normal state and compared it to effect sizes observed in different brain regions affected by AD, including hippocampus volume and frontal cortical gray matter thickness. Our discovery cohort consisted of Old Order Amish and Mennonite (OOA/M) participants, and the replication cohort consisted of UK Biobank (UKBB) participants. The OOA/M participants, of central European descent, share a rural, farm-based dwelling with low rates of alcohol, tobacco, and illicit substance use (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e). This population has much better genetic, economic, and environmental uniformity compared to general populations. UKBB participants were recruited from urban and suburban populations. The vastly different environmental backgrounds of these two populations should provide a more rigorous replication test for the generalizability of the findings.\u003c/p\u003e \u003cp\u003eWe tested the second hypothesis using longitudinal data from the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI), which includes older age samples with higher risks for developing AD, to evaluate whether elevation of RVI-AD was associated with an increased risk of future dementia. An ideal early biomarker for AD should capture the risks associated with both APOE ε4 allele and cardiovascular health at a preclinical, healthy stage, as well as predict conversion to dementia when risk is elevated.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eRisks at Normalcy: Discovery and Replication Samples\u003c/h2\u003e \u003cp\u003eThe discovery sample consisted of 343 OOA/M participants (140M/ 203F, age range 21\u0026ndash;72 years, mean\u0026thinsp;\u0026plusmn;\u0026thinsp;s.d.: 49.1\u0026thinsp;\u0026plusmn;\u0026thinsp;12.7 years). Exclusion criteria included a history of epilepsy, cerebrovascular accidents, head injury with cognitive sequelae, intellectual disability, and unstable major medical conditions at the time of the study. Research participants with substance dependence within the past 6 months or current substance use disorder (except nicotine) were also excluded. Other exclusion criteria for this analysis included current or lifetime major psychiatric illness. Study participants gave written informed consent approved by the University of Maryland IRB.\u003c/p\u003e \u003cp\u003eThe replication sample consisted of participants in the UK Biobank where we analyzed a subset of N\u0026thinsp;=\u0026thinsp;31,440 participants (mean age\u0026thinsp;=\u0026thinsp;63.81\u0026thinsp;\u0026plusmn;\u0026thinsp;7.44 years; 14,813M/16,627F) who were free of neuropsychiatric illnesses and for whom neuroimaging and clinical data were available. This included structural and diffusion white matter imaging phenotypes. Data were collected between 2012 and 2021 (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e). The UK Biobank received ethics approval from the North West Multi-Center Research Ethics Committee (11/NW/03820).\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eImaging in ACP and UKBB\u003c/h3\u003e\n\u003cp\u003eRVI were generated using structural and diffusion imaging data collected using similar imaging protocols and processing using consistent ENIGMA structural and diffusion pipelines, see Supplement.\u003c/p\u003e\n\u003ch3\u003e\u003c/h3\u003e\n\u003cdiv class=\"Heading\"\u003e\u003cem\u003eEffect Sizes for Amyloid-Positive AD from ADNI and OASIS\u003c/em\u003e\u003c/div\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eNIA-AA has defined \u003cem\u003ebiological\u003c/em\u003e Alzheimer\u0026rsquo;s disease based on the presence of abnormal levels of beta-amyloid or tau protein in biofluids or on brain PET scans in subjects who experienced persistent clinical decline in one or more clinical domains, including memory (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). The regional vulnerability index for AD (RVI-AD) was calculated using the effect sizes derived between amyloid-positive (Aβ+) AD individuals and healthy amyloid-negative (Aβ-) stable controls from two datasets: the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI) (132 Aβ\u0026thinsp;+\u0026thinsp;dementia cases and 223 Aβ- healthy controls) and the Open Access Series of Imaging Studies (OASIS) (36 Aβ\u0026thinsp;+\u0026thinsp;dementia cases and 180 healthy Aβ- controls). Altogether, imaging data for ADNI and OASIS datasets for 571 subjects\u0026ndash;for whom Aβ status was available\u0026ndash;served as the basis for the RVI-AD analysis.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003eWe calculated Cohen\u0026rsquo;s \u003cem\u003ed\u003c/em\u003e estimates for cortical thickness and subcortical gray matter differences region-by-region between 168 Aβ\u0026thinsp;+\u0026thinsp;dementia cases and 403 Aβ- healthy controls by combining the two datasets using ENIGMA structural imaging pipelines(\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e) (\u003cb\u003eTable \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e). These subjects lacked white matter diffusion imaging phenotypes, so white matter regional effect sizes were derived from a previous report on the ADNI participants. These effect sizes were calculated based on 48 AD cases and 53 controls also using the ENIGMA pipeline, although no specific Aβ status was considered there(\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e). In total, we extracted 64 regions from the whole brain, including 33 regional cortical thickness measures, 7 subcortical gray matter volumes and 24 regional white matter integrity measurements where data from left and right hemispheres were averaged (\u003cb\u003eTable \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e). To calculate Cohen\u0026rsquo;s \u003cem\u003ed\u003c/em\u003e, we performed a linear regression for the diagnosis variable while including age and sex, their interaction, and the total intracranial volume as covariates, and performed \u003cem\u003et\u003c/em\u003e-to-\u003cem\u003ed\u003c/em\u003e value conversion.\u003c/p\u003e\n\u003ch3\u003eADNI participants and conversion to dementia\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eADNI includes an ongoing, longitudinal multisite study of aging(\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e). We tested how baseline RVI-AD values were associated with risks of conversion to dementia. Baseline neuroimaging summary data were downloaded from ADNI database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003ca href=\"http://www.loni.usc.edu/ADNI\" target=\"_blank\"\u003ewww.loni.usc.edu/ADNI\u003c/a\u003e\u003c/span\u003e\u003cspan address=\"http://www.loni.usc.edu/ADNI\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) to create RVI-AD measurements for each participant. We also downloaded up to 12 years of follow up clinical data that documented whether the individual status has remained stable or changed to dementia or MCI. The final ADNI sample consisted of (N\u0026thinsp;=\u0026thinsp;1,932, age\u0026thinsp;=\u0026thinsp;73.6\u0026thinsp;\u0026plusmn;\u0026thinsp;7.2, M/F\u0026thinsp;=\u0026thinsp;1043/889) that included 586 cognitively normal (CN) older healthy controls (age\u0026thinsp;=\u0026thinsp;73.8\u0026thinsp;\u0026plusmn;\u0026thinsp;7.2, M/F\u0026thinsp;=\u0026thinsp;286/318) and 965 participants with mild cognitive impairment (MCI, age\u0026thinsp;=\u0026thinsp;72.9\u0026thinsp;\u0026plusmn;\u0026thinsp;7.5, M/F\u0026thinsp;=\u0026thinsp;569/397) at baseline. 380 participants with dementia at baseline (age\u0026thinsp;=\u0026thinsp;75.1\u0026thinsp;\u0026plusmn;\u0026thinsp;7.1, M/F/=207/173) were included as a comparison. These analyses excluded participants with Aβ status data available, as those data had been used to create effect sizes for RVI. Conversion information was not available for all participants because some had dropped out of the study or failed the follow up. For the available data in the CN at baseline group, N\u0026thinsp;=\u0026thinsp;35 (age\u0026thinsp;=\u0026thinsp;75.2\u0026thinsp;\u0026plusmn;\u0026thinsp;3.9. M/F\u0026thinsp;=\u0026thinsp;17/18) had changed clinical status to dementia since baseline (average conversion period\u0026thinsp;=\u0026thinsp;4.7\u0026thinsp;\u0026plusmn;\u0026thinsp;2.5 years) and N\u0026thinsp;=\u0026thinsp;76 (average age\u0026thinsp;=\u0026thinsp;75.5\u0026thinsp;\u0026plusmn;\u0026thinsp;4.6, M/F\u0026thinsp;=\u0026thinsp;41/49) had clinical status changed to MCI (average conversion period\u0026thinsp;=\u0026thinsp;2.9\u0026thinsp;\u0026plusmn;\u0026thinsp;2.6 years). In the MCI at baseline group, N\u0026thinsp;=\u0026thinsp;348 (age\u0026thinsp;=\u0026thinsp;74.0\u0026thinsp;\u0026plusmn;\u0026thinsp;7.0, M/F\u0026thinsp;=\u0026thinsp;201/147) had clinical status changed to dementia since baseline (average conversion period\u0026thinsp;=\u0026thinsp;2.3\u0026thinsp;\u0026plusmn;\u0026thinsp;2.4 years).\u003c/p\u003e \u003cp\u003e \u003cb\u003eGenotyping for APOE status.\u003c/b\u003e All datasets provided APOE genotypes for all individuals and the genotyping methodology is provided in the Supplement.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003eFramingham cardiovascular risk score\u003c/h3\u003e\n\u003cp\u003eThe Framingham cardiovascular risk score (FCVRS) is a weighted composite index developed to predict a 10-year risk of a major cardiovascular event such as heart attack, stroke or death(\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e). FCVRS scores were calculated for all subjects in the discovery and replication datasets using the uniform FCVRS algorithm(\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e). Only N\u0026thinsp;=\u0026thinsp;498 subjects in ADNI baseline sample had all of the measures needed to calculate FCVRS.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003eRVI calculation\u003c/h2\u003e \u003cp\u003eThe Regional Vulnerability Index (RVI) scores were calculated using the \u0026lsquo;RVIpkg\u0026rsquo; in [R] software based on our previous publication(\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e) with some revisions. The original RVI calculated the correlational agreement between an individual\u0026rsquo;s regional brain measures and the pattern of regional SSD-related brain deficits(\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e). It was used for patients with a diagnosis and required a sample of healthy controls to perform normalization [70]. It is therefore inappropriate for most of the current samples when participants at baseline had no AD. Accordingly, we computed the modified RVI as a normalized dot product of two vectors: the vector of individual regional deviations (coded as z-score deviations from the mean of the group) and the vector of AD effect sizes (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Specifically, the RVI was calculated for each subject as the dot product between vectors Z and E normalized by the dimensions of the vector using the following equation.\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:RVI={\\sum\\:}_{i=1}^{N}({Z}_{i}\\times\\:{E}_{i})$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere Z is the vector of deviation from the mean and E is the vector of meta-analytical effect size (Cohen\u0026rsquo;s \u003cem\u003ed\u003c/em\u003e coefficients) for corresponding regional measures for AD. N is the dimension of the vector, i.e., the total number of imaging phenotypes for modalities including the whole-brain, cortical, subcortical, and white matter. Positive RVI values indicate that the regional pattern of an individual coincides with the expected pattern of AD. Therefore, the modified RVI version is similar to the original RVI in terms of relying on the regional effect sizes as the blueprint to define an individual\u0026rsquo;s similarity to the illness but using a full sample vector alignment rather than relying on healthy control group to achieve normalization.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e\n\u003ch3\u003eHypothesis testing\u003c/h3\u003e\n\u003cp\u003eWe first evaluated if individual RVI-AD derived from Aβ\u0026thinsp;+\u0026thinsp;AD cases and Aβ- healthy controls can significantly identify the impact of the known genetic risk factor for AD - the APOE-ε4 genotype - on the AD-like brain patterns, using a Student\u0026rsquo;s \u003cem\u003et\u003c/em\u003e-test. This was followed by investigating the impacts of FCVRS and the joint APOE genotype and FCVRS effects on RVI-AD using linear regressions using Model 1:\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eRVI-AD\u0026thinsp;~\u0026thinsp;β\u003csub\u003eε4\u003c/sub\u003e \u0026sdot; E4\u0026thinsp;+\u0026thinsp;β\u003csub\u003eFCVRS\u003c/sub\u003e \u0026sdot; FCVRS\u0026thinsp;+\u0026thinsp;β\u003csub\u003eε4*FCVRS\u003c/sub\u003e\u0026sdot;E4*FCVRS (Model 1)\u003c/h2\u003e \u003cp\u003ePrimary analyses were performed on the whole brain RVI-AD with follow up analyses examining tissue-specific RVI-AD. We tested each hypothesis first in the ACP, and then sought replication in the UKBB samples.\u003c/p\u003e \u003cp\u003eNext, we used the ADNI dataset to test if higher RVI-AD values in people diagnosed as MCI are associated with risk for conversion to dementia. The primary analysis was in the MCI group, where we compared those who did not convert and remained stable (MCI\u0026rarr;MCI) in the following decade to those who converted to dementia (MCI\u0026rarr;dementia), using \u003cem\u003et\u003c/em\u003e-tests. We also compared these two groups to those with dementia at baseline and those who maintained the CN status.\u003c/p\u003e \u003cp\u003eWe then also tested this in the CN group where a small cohort had converted to dementia, although the sample size for converters here was small: we compared those who did not convert and remained CN in the following decade (CN\u0026rarr;CN) to those converted to dementia (CN\u0026rarr;dementia) and also to those only converted to MCI but not dementia (CN\u0026rarr;MCI) using t-tests.\u003c/p\u003e \u003cp\u003eTo estimate the association between RVI-AD and the time to the onset of dementia, we further conducted logistic regression analysis to investigate the odds ratio (OR) of dementia conversion within MCI cohort (Model 2).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e\u003cem\u003eConversion\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;\u003cem\u003e~\u0026thinsp;β\u003c/em\u003e\u003csub\u003e\u003cem\u003eRVI\u003c/em\u003e\u003c/sub\u003e \u003cem\u003eRVI-AD\u0026thinsp;+\u0026thinsp;Age (Model 2)\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eOur analysis comprised of two parts. Firstly, we examined the odds ratio (OR) of conversion cases occurring annually. Second, we examined the OR of cumulative dementia conversion cases, over a 12-year follow-up period. For both components of the analysis, logistic regression models were fitted for each annual time point or by averaging of all 12 years of data, adjusting for baseline age.\u003c/p\u003e \u003cp\u003eModel 3 expanded the model 2 with two additional dementia risk factors (i.e., APOE-ε4 genotype and FCVRS scores) as predictors to assess if the joint models improved performance in explaining the variation in dementia conversion.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e\u003cem\u003eConversion\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;\u003cem\u003e~\u0026thinsp;β\u003c/em\u003e\u003csub\u003e\u003cem\u003eRVI\u003c/em\u003e\u003c/sub\u003e \u003cem\u003eRVI-AD\u0026thinsp;+\u0026thinsp;β\u003c/em\u003e\u003csub\u003e\u003cem\u003eε4\u003c/em\u003e\u003c/sub\u003e \u003cem\u003e\u0026sdot; APOE-ε4\u0026thinsp;+\u0026thinsp;β\u003c/em\u003e\u003csub\u003e\u003cem\u003eFCVRS\u003c/em\u003e\u003c/sub\u003e \u003cem\u003e\u0026sdot; FCVRS\u0026thinsp;+\u0026thinsp;Age (Model 3)\u003c/em\u003e\u003c/h2\u003e \u003cp\u003eModel performance was evaluated using likelihood ratio tests comparing the joint models to the null model. The OR of each predictor reflects the risk associated with a one-unit increase in the predictor. To be able to compare OR between predictors, the predictors were normalized. Note that we cannot directly model the \u0026ldquo;joint\u0026rdquo; OR of the predictors, but we can evaluate whether adding APOE4 and/or FCVRS improves the model fit using likelihood ratio tests, where a higher likelihood ratio value of a model indicates a better model fit. As only a portion of the ADNI participants have all the data needed to calculate FCVRS, we used only a subsample with all data available to test the full Model 3.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eAPOE Genotype on RVI-AD in healthy samples: discovery and replication\u003c/h2\u003e \u003cdiv id=\"Sec16\" class=\"Section3\"\u003e \u003ch2\u003eDiscovery sample\u003c/h2\u003e \u003cp\u003eThe N\u0026thinsp;=\u0026thinsp;91 APOE-ε4 carriers (38M/ 53F, age: 48.0\u0026thinsp;\u0026plusmn;\u0026thinsp;12.5) exhibited significantly higher RVI-AD compared to non-carriers (Cohen\u0026rsquo;s \u003cem\u003ed\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.29, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.03) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). In comparison, the two groups had no significant differences in brain measurements in structures affected by AD, including hippocampal volume (d=-0.01, p\u0026thinsp;=\u0026thinsp;0.8) and the cortical thickness of the temporal, parietal, and frontal cortical areas (all p\u0026thinsp;\u0026gt;\u0026thinsp;0.7) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB-F). The effect sizes of APOE-ε4 on individual brain were significantly correlated with the pattern of effect sizes observed in AD (r\u0026thinsp;=\u0026thinsp;0.35, p\u0026thinsp;=\u0026thinsp;0.004, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eG).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eUKBB sample\u003c/h2\u003e \u003cp\u003eIn the UKBB cohort, APOE-ε4 carriers demonstrated significantly higher RVI-AD compared to non-carriers (p\u0026thinsp;=\u0026thinsp;2\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eH). APOE-ε4 carriers had lower hippocampal volume (p\u0026thinsp;=\u0026thinsp;0.01) but it was not significant after correction for multiple comparisons (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eI) and showed no other suggestively significant differences when compared to noncarriers (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eJ-L). The APOE-ε4 effect sizes for individual brain region measurements were smaller than the RVI-AD effect size (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eM). The effect sizes of APOE-ε4 on individual brain regions in UKBB were also significantly associated with the AD effect sizes of these regions as identified by Aβ\u0026thinsp;+\u0026thinsp;AD patients (r\u0026thinsp;=\u0026thinsp;0.34, p\u0026thinsp;=\u0026thinsp;0.006, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eN).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eEffects of Cardiovascular Risk on RVI-AD\u003c/h2\u003e \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e \u003ch2\u003eDiscovery sample\u003c/h2\u003e \u003cp\u003eModel 1 was significant in the ACP cohort (F\u0026thinsp;=\u0026thinsp;4.6, p\u0026thinsp;=\u0026thinsp;0.0002, Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), with a significant main effect of FCVRS (p\u0026thinsp;=\u0026thinsp;0.006) and a significant APOE \u0026middot; FCVRS interaction (p\u0026thinsp;=\u0026thinsp;0.03, Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). This significant interaction was due to a stronger correlation between FCVRS and RVI-AD in APOE-ε4 carriers (r\u0026thinsp;=\u0026thinsp;0.35, p\u0026thinsp;=\u0026thinsp;4\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003evs. r\u0026thinsp;=\u0026thinsp;0.02, p\u0026thinsp;=\u0026thinsp;0.95) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA) and the difference in correlation coefficients was significant (z\u0026thinsp;=\u0026thinsp;2.9, p\u0026thinsp;=\u0026thinsp;0.004). Notably, the two genotype groups did not significantly differ in general cardiovascular risks as measured by FCVRS (6.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.6 vs. 6.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.4, p\u0026thinsp;=\u0026thinsp;0.9), systolic, diastolic, or pulse pressure (117.4\u0026thinsp;\u0026plusmn;\u0026thinsp;1.5, 72.0\u0026thinsp;\u0026plusmn;\u0026thinsp;0.9 and 45.4\u0026thinsp;\u0026plusmn;\u0026thinsp;1.3 vs. 117.2\u0026thinsp;\u0026plusmn;\u0026thinsp;1.0, 72.6\u0026thinsp;\u0026plusmn;\u0026thinsp;0.6 and 44.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.7; all p\u0026thinsp;\u0026gt;\u0026thinsp;0.4), suggesting that the APOE-ε4 allele by itself did not significant impact peripheral cardiovascular measures.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEvaluating Model 1 in participants of Amish Connectome Project (ACP) and UK biobank (UKBB) studies. Values are mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SEM of the beta values (is this correct???)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eSample\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eModel Statistics\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eAPOE4\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eFCVRS\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eAPOE4 x FCVRS\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eACP (N\u0026thinsp;=\u0026thinsp;343)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWhole-Brain RVI-AD Value\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eF\u0026thinsp;=\u0026thinsp;4.6, p\u0026thinsp;=\u0026thinsp;0.002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.14\u0026thinsp;\u0026plusmn;\u0026thinsp;0.9\u003c/p\u003e \u003cp\u003e(t=-0.1,p\u0026thinsp;=\u0026thinsp;0.9)\u003c/p\u003e\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.25\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09 (t\u0026thinsp;=\u0026thinsp;2.8,p\u0026thinsp;=\u0026thinsp;0.006)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.20\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09 (t\u0026thinsp;=\u0026thinsp;2.1,p\u0026thinsp;=\u0026thinsp;0.03)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUKBB (N\u0026thinsp;=\u0026thinsp;31,440)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWhole-Brain RVI-AD Value\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eF\u0026thinsp;=\u0026thinsp;8.8, p\u0026thinsp;=\u0026thinsp;7\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.95\u0026thinsp;\u0026plusmn;\u0026thinsp;0.4 (t\u0026thinsp;=\u0026thinsp;2.3,p\u0026thinsp;=\u0026thinsp;0.01)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.13\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03 (t\u0026thinsp;=\u0026thinsp;3.7,p\u0026thinsp;=\u0026thinsp;2\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.12\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03 (t\u0026thinsp;=\u0026thinsp;3.5,p\u0026thinsp;=\u0026thinsp;5\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eUKBB sample\u003c/h2\u003e \u003cp\u003eModel 1 was significant in the UKBB cohort (F\u0026thinsp;=\u0026thinsp;8.8, p\u0026thinsp;=\u0026thinsp;7\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e). We observed significant effects of FCVRS (p\u0026thinsp;=\u0026thinsp;2\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e), APOE-ε4 genotype (p\u0026thinsp;=\u0026thinsp;0.01), and their interaction (p\u0026thinsp;=\u0026thinsp;5\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e) on RVI-AD (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The correlation between FCVRS and RVI-AD was significant in APOE-ε4 carriers (r\u0026thinsp;=\u0026thinsp;0.07, p\u0026thinsp;=\u0026thinsp;4\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;7\u003c/sup\u003e) but not in noncarriers (r\u0026thinsp;=\u0026thinsp;0.00, p\u0026thinsp;=\u0026thinsp;0.7), and the difference in correlation coefficients was significant (z\u0026thinsp;=\u0026thinsp;3.7, p\u0026thinsp;=\u0026thinsp;2\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e). (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). APOE-ε4 carriers and non-carriers did not differ in FVCRS (11.4\u0026thinsp;\u0026plusmn;\u0026thinsp;0.02 vs. 11.3\u0026thinsp;\u0026plusmn;\u0026thinsp;0.2, p\u0026thinsp;=\u0026thinsp;0.07) or systolic, diastolic or pulse pressure (139.9\u0026thinsp;\u0026plusmn;\u0026thinsp;0.7, 78.0\u0026thinsp;\u0026plusmn;\u0026thinsp;0.4, and 61.8\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5 versus 140.9\u0026thinsp;\u0026plusmn;\u0026thinsp;0.2, 78.7\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5, and 62.1\u0026thinsp;\u0026plusmn;\u0026thinsp;0.3, respectively) (all p\u0026thinsp;\u0026gt;\u0026thinsp;0.2). We also evaluated RVI-AD calculated separately for cortical thickness, subcortical gray matter volumes, and white matter and observed similar trends. Overall, whole-brain RVI-based models demonstrated the highest robustness and consistency in capturing the APOE and FCVRS risks (see \u003cb\u003eTable S2\u003c/b\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003ePredicting conversion from MCI to AD in ADNI participants\u003c/h2\u003e \u003cp\u003eWe used logistic regression to evaluate the odds ratio (OR) of conversion to dementia (see Models 2 in Methods) in ADNI participants. Out of N\u0026thinsp;=\u0026thinsp;965 participants with MCI, N\u0026thinsp;=\u0026thinsp;335 (age\u0026thinsp;=\u0026thinsp;74.0\u0026thinsp;\u0026plusmn;\u0026thinsp;7.0, 201 M/147 F) developed dementia over the next 12 years, with an average conversion period of about 3 years. Participants who converted to dementia had significantly higher baseline RVI compared to those who did not (p\u0026thinsp;=\u0026thinsp;3\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;16\u003c/sup\u003e) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). MCI individuals who did not convert to dementia in a decade showed no significant difference in RVI from cognitively normal (CN) elderly adults who remained CN 12 years later (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEvaluating Models 2 and 3 in the odds ratio (OR) for predicting conversion from MCI to dementia. The values are OR, 95% confidence interval and p-values for MCI subjects whose clinical status changed to dementia over the assessment period.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCI\u0026rarr;AD \u003cem\u003eConversion Models\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eAge\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eRVI-AD\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eAPOE4\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eFCVRS\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel 2 (N\u0026thinsp;=\u0026thinsp;965) Odds Ratio\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.02 (1.01\u0026ndash;1.05), p\u0026thinsp;=\u0026thinsp;0.005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.74 (1.52\u0026ndash;2.03), p\u0026thinsp;=\u0026thinsp;2\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel 3 (N\u0026thinsp;=\u0026thinsp;498) Odds Ratio\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.03 (0.99\u0026ndash;1.07), p\u0026thinsp;=\u0026thinsp;0.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.35 (1.10\u0026ndash;1.64), p\u0026thinsp;=\u0026thinsp;0.004\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.24 (1.04\u0026ndash;1.4), p\u0026thinsp;=\u0026thinsp;0.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.08 (0.90\u0026ndash;1.3), p\u0026thinsp;=\u0026thinsp;0.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn addition, N\u0026thinsp;=\u0026thinsp;37 out of the M\u0026thinsp;=\u0026thinsp;586 cognitively normal ADNI participants (age\u0026thinsp;=\u0026thinsp;75.2\u0026thinsp;\u0026plusmn;\u0026thinsp;3.9. M/F\u0026thinsp;=\u0026thinsp;17/18) also developed dementia (CN\u0026rarr;dementia) after an average conversion period of ~\u0026thinsp;5 years. This small sample also showed significantly higher baseline RVI-AD compared to CN participants who remained stable (CN\u0026rarr;CN) (p\u0026thinsp;=\u0026thinsp;0.03) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). In comparison, the N\u0026thinsp;=\u0026thinsp;76 CN participants who developed MCI during the same period (CN\u0026rarr;MCI) had intermediate RVI values between the CN\u0026rarr;dementia and CN\u0026rarr;CN (N\u0026thinsp;=\u0026thinsp;461) groups.\u003c/p\u003e \u003cp\u003eWe used an annual logistic regression analysis to test the timing of the odds ratio for predicting new conversions to dementia based on RVI-AD. Higher baseline RVI-AD was significantly associated with conversion from MCI to annual new dementia cases in each of the first three years (OR\u0026thinsp;=\u0026thinsp;2.2 to 1.78, p\u0026thinsp;=\u0026thinsp;3\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e to 3\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;10\u003c/sup\u003e) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). In aggregate, the odds ratio for the 12-year cumulative conversion to dementia was highly significant (OR\u0026thinsp;=\u0026thinsp;1.8, p\u0026thinsp;=\u0026thinsp;7\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;14\u003c/sup\u003e) as shown by the last data point in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD (which also plotted the OR of annual cumulative conversion).\u003c/p\u003e \u003cp\u003eWe expanded the logistic regression model to include APOE-ε4 genotype status and FCVRS as additional predictors (see Model 3 in Methods). This analysis was conducted on N\u0026thinsp;=\u0026thinsp;498 participants with available FCVRS data, out of whom N\u0026thinsp;=\u0026thinsp;243 MCI participants converted to dementia. Both APOE-ε4 genotype (OR\u0026thinsp;=\u0026thinsp;1.2, p\u0026thinsp;=\u0026thinsp;0.01) and RVI-AD (OR\u0026thinsp;=\u0026thinsp;1.7, p\u0026thinsp;=\u0026thinsp;5\u0026middot;10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e) contributed to dementia conversion over the 12-year follow-up period. FCVRS did not contribute significantly to dementia conversion (OR\u0026thinsp;=\u0026thinsp;0.0). This suggests that in this age range and high-risk stage of MCI, FCVRS effects for conversion may no longer be prominent.\u003c/p\u003e \u003cp\u003eWe evaluated the area under the curve (AUC) for predicting conversion from MCI to dementia and found that the peak AUC was 74%. Including the APOE genotype only slightly improved the AUC to 75% (\u003cb\u003eTable S3, see supplement\u003c/b\u003e). In comparison, the model that only had FCVRS\u0026thinsp;+\u0026thinsp;APOE4 status showed AUC of 0.57.\u003c/p\u003e \u003cp\u003eThe predictive ability of baseline RVI-AD values for conversion to dementia declined with time (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). To estimate RVI-AD\u0026rsquo;s long-term predictive value as part of a hypothetical regular screening program, we simulated annual, triennial, and quinquennial structural MRI assessments (Methods in \u003cb\u003eSupplementary Information\u003c/b\u003e). The simulations suggest that once MCI is diagnosed, RVI-AD calculated based on annual MRI had the strongest performance (AUC\u0026thinsp;=\u0026thinsp;0.80), and even triannual follow-up still had clinically relevant AUC (AUC\u0026thinsp;=\u0026thinsp;0.75) for a ten-year prediction of conversion to dementia (\u003cb\u003eFigure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003e We evaluated the regional vulnerability index for Alzheimer\u0026rsquo;s Disease (RVI-AD) to gauge its ability to track the lifelong contributions of established risk factors that eventually culminate in dementia. RVI-AD quantifies an individual\u0026rsquo;s similarity to AD brain deficit patterns. We used a meta-analytical sample of AD patients with amyloid positive (Aβ+) status and amyloid negative controls to estimate this pattern. In healthy adults, RVI-AD showed a positive correlation with the subtle actions of known APOE-ε4 genetic effects as well as the cardiovascular risks, as measured by Framingham Cardiovascular Risk Score (FCVRS). We observed a significant additive ε4-by-FCVRS effect on RVI-AD, indicating that comparable levels of cardiovascular risks led to a higher similarity to AD patterns in people with the APOE-ε4 genotype. With a few exceptions, these findings were consistent in both farm-dwelling Amish and in urban and suburban UK participants. Higher RVI-AD in the participants ascertained by Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI) was significantly associated with the elevated risk for conversion to dementia. Simulations suggested that RVI-AD could support a screening program that uses annual-to-triannual structural MRI scans to provide clinically relevant (AUC\u0026thinsp;=\u0026thinsp;0.75\u0026ndash;0.80) assessment of risk for conversion to dementia in people diagnosed with MCI. RVI-AD is a novel and promising biomarker capable of capturing both the preclinical AD risk factors as well as predicting the likelihood of dementia development in individuals with MCI.\u003c/p\u003e \u003cp\u003eThe presence of illness-like brain patterns reflects the cumulative action of associated risk factors and may be informative of the risk of developing a disorder (\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e). We tested this hypothesis using two best-known AD risks: APOE-ε4 genotype and cardiovascular factors. The effects of the APOE-ε4 allele on elevating RVI-AD were stable and replicable in two cohorts where the APOE-ε4\u0026rsquo;s effects on RVI-AD elevation acted primarily through cardiovascular risks. APOE-ε4 is a significant risk factor for cardio-and-cerebrovascular conditions (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan additionalcitationids=\"CR58 CR59 CR60 CR61\" citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e). The burden of cardiovascular factors contributes up to 50% of the risk for developing of late-onset AD(\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e, \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e) and while there were no significant differences in FCVRS or blood pressure measures in APOE-ε4 carriers vs. non-carriers, RVI-AD still detected the additional brain effects linked to cardiovascular risks in the carriers. Incremental elevations in FCVRS in APOE-ε4 carriers resulted in a greater resemblance to AD-like brain patterns compared to non-carriers, suggesting that cardiovascular risk factors may act on the brain differently in APOE- ε4 carriers. Thus, RVI may provide an early biomarker to index this genetic\u0026thinsp;+\u0026thinsp;cardiovascular risk effect for developing an AD-like brain pattern at the level of the individual at cognitive and medical normalcy.\u003c/p\u003e \u003cp\u003eWe tested the second hypothesis in individuals with MCI (mean age\u0026thinsp;~\u0026thinsp;74 years) who were at risk for AD because approximately one third of subjects went on to develop dementia (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e). MCI subjects who developed dementia had significantly higher baseline RVI-AD indices compared to participants who remained stable. This finding was also replicated in the cognitively normal controls. The predictive accuracy of conversion was strongest within three years after the MRI scan. This suggested that additional neurodegeneration needs to occur for MCI to dementia conversion and that annual-to-triannual MRI assessments in people at risk for dementia could maintain clinically acceptable AUC. A structured RVI-based screening program could supplement or provide a cost-effective alternative to AD-focused PET/CSF assessments. Importantly, MCI participants who did not progress to dementia had RVI-AD values similar to the RVI values for the cognitively normal controls (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA), suggesting that low RVI-AD predicted reduced likelihood for conversion. Approximately half of the MCI individuals never develop dementia (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e), leading to a decision challenge for disease-modifying treatments that have significant side effects. Present approaches to evaluate individual\u0026rsquo;s risks, include tau-focused PET scans, volumetric MRI assessments, cognitive batteries, and blood/CSF biomarkers (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan additionalcitationids=\"CR69\" citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e). Many of these approaches have high costs and perceived invasiveness (\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e). In contrast, RVI-AD uses linear, easily interpretable calculations and measurements from readily available structural MRI. Further validation may enable RVI-AD index to assist in early risk detection and predict more imminent dementia disease onset, especially when combined with other available clinical and biomarker measures.\u003c/p\u003e \u003cp\u003eIn conclusion, RVI-AD can detect the early, lifetime gradual risk factor effects on the brain before cognitive changes emerge, as well as the more imminent risks for conversion of MCI to AD. AD is a major cause of cognitive dysfunction among older people. With new AD treatments on the horizon, non-invasive brain imaging biomarkers, such as RVI-AD, could identify vulnerable individuals during cognitively normal to prodromal stages, supporting public health efforts to prevent, delay, or reduce the impact of AD.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupport was received from National Institutes of Health grants RF1NS114628, R01NS114628, R01MH112180, R01MH133812, R01MH116948, S10OD023696, R01EB015611, R01MH117601, R01AG095874, U01MH108148, and R01ES033961. These funding sources provided financial support to enable the design and conduct of the study or collection, management, or analysis of the data. None of the funding agencies had a role in interpreting the data or in preparing, reviewing, or approving the manuscript. None had a role in the decision to submit the manuscript for publication.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors declare no conflict of interest.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAnonymous (2021) 2021 Alzheimer's disease facts and figures. \u003cem\u003eAlzheimers Dement\u003c/em\u003e 17(3):327\u0026ndash;406.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMajeed J, \u003cem\u003eet al.\u003c/em\u003e (2023) Cancer drugs with high repositioning potential for Alzheimer's disease. \u003cem\u003eExpert opinion on emerging drugs\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHerholz K, \u003cem\u003eet al.\u003c/em\u003e (2002) Discrimination between Alzheimer dementia and controls by automated analysis of multicenter FDG PET. \u003cem\u003eNeuroimage\u003c/em\u003e 17(1):302\u0026ndash;316.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQiu C (Preventing Alzheimer's disease by targeting vascular risk factors: hope and gap. (1875\u0026ndash;8908 (Electronic)).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJack CR, Jr., \u003cem\u003eet al.\u003c/em\u003e (2018) NIA-AA Research Framework: Toward a biological definition of Alzheimer's disease. \u003cem\u003eAlzheimers Dement\u003c/em\u003e 14(4):535\u0026ndash;562.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcGrattan AM, \u003cem\u003eet al.\u003c/em\u003e (2022) Risk of conversion from mild cognitive impairment to dementia in low- and middle-income countries: A systematic review and meta-analysis. \u003cem\u003eAlzheimer's \u0026amp; dementia (New York, N. Y.)\u003c/em\u003e 8(1):e12267.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLo RY, \u003cem\u003eet al.\u003c/em\u003e (2011) Longitudinal change of biomarkers in cognitive decline. \u003cem\u003eArch Neurol\u003c/em\u003e 68(10):1257\u0026ndash;1266.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVAN HOESEN GW \u0026amp; SOLODKIN A (1994) Cellular and Systems Neuroanatomical Changes in Alzheimer's Disease. \u003cem\u003eAnnals of the New York Academy of Sciences\u003c/em\u003e 747(1):12\u0026ndash;35.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSolodkin A, \u003cem\u003eet al.\u003c/em\u003e (2013) In vivo parahippocampal white matter pathology as a biomarker of disease progression to Alzheimer's disease. \u003cem\u003eThe Journal of comparative neurology\u003c/em\u003e 521(18):4300\u0026ndash;4317.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBadea A, \u003cem\u003eet al.\u003c/em\u003e (2016) The fornix provides multiple biomarkers to characterize circuit disruption in a mouse model of Alzheimer's disease. \u003cem\u003eNeuroimage\u003c/em\u003e 142:498\u0026ndash;511.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrystal O, \u003cem\u003eet al.\u003c/em\u003e (2023) Detecting conversion from mild cognitive impairment to Alzheimer's disease using FLAIR MRI biomarkers. \u003cem\u003eNeuroImage. Clinical\u003c/em\u003e 40:103533.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKress GT, \u003cem\u003eet al.\u003c/em\u003e (2023) Preliminary validation of a structural magnetic resonance imaging metric for tracking dementia-related neurodegeneration and future decline. \u003cem\u003eNeuroImage. Clinical\u003c/em\u003e 39:103458.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYoung AL, \u003cem\u003eet al.\u003c/em\u003e (2021) Ordinal SuStaIn: Subtype and Stage Inference for Clinical Scores, Visual Ratings, and Other Ordinal Data. \u003cem\u003eFrontiers in artificial intelligence\u003c/em\u003e 4:613261.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang G, \u003cem\u003eet al.\u003c/em\u003e (2021) Developing univariate neurodegeneration biomarkers with low-rank and sparse subspace decomposition. \u003cem\u003eMedical image analysis\u003c/em\u003e 67:101877.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eProsser L, \u003cem\u003eet al.\u003c/em\u003e (2020) Amyloid Pattern Similarity Score (AMPSS): A reference region free measure of amyloid PET deposition in Alzheimer\u0026rsquo;s disease. \u003cem\u003eAlzheimer's \u0026amp; Dementia\u003c/em\u003e 16(S4):e042673.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee JS, \u003cem\u003eet al.\u003c/em\u003e (2018) Machine Learning-based Individual Assessment of Cortical Atrophy Pattern in Alzheimer's Disease Spectrum: Development of the Classifier and Longitudinal Evaluation. \u003cem\u003eSci Rep\u003c/em\u003e 8(1):4161.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi Y, \u003cem\u003eet al.\u003c/em\u003e (2022) ASMFS: Adaptive-similarity-based multi-modality feature selection for classification of Alzheimer's disease. \u003cem\u003ePattern Recognition\u003c/em\u003e 126:108566.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng X, Provenzano FA, \u0026amp; Small SA (2022) A deep learning MRI approach outperforms other biomarkers of prodromal Alzheimer's disease. \u003cem\u003eAlzheimer's research \u0026amp; therapy\u003c/em\u003e 14(1):45.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDhinagar N, Thomopoulos S, Laltoo E, \u0026amp; Thompson P (2023) \u003cem\u003eEfficiently Training Vision Transformers on Structural MRI Scans for Alzheimer's Disease Detection\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBiffi A, \u003cem\u003eet al.\u003c/em\u003e (2010) Genetic variation and neuroimaging measures in Alzheimer disease. \u003cem\u003eArch Neurol\u003c/em\u003e 67(6):677\u0026ndash;685.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBroer L, \u003cem\u003eet al.\u003c/em\u003e (2014) GWAS of longevity in CHARGE consortium confirms APOE and FOXO3 candidacy. \u003cem\u003eThe journals of gerontology. Series A, Biological sciences and medical sciences\u003c/em\u003e 70(1):110\u0026ndash;118.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeelen J, \u003cem\u003eet al.\u003c/em\u003e (2011) Genome-wide association study identifies a single major locus contributing to survival into old age; the APOE locus revisited. \u003cem\u003eAging cell\u003c/em\u003e 10(4):686\u0026ndash;698.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoate A, \u003cem\u003eet al.\u003c/em\u003e (1991) Segregation of a missense mutation in the amyloid precursor protein gene with familial Alzheimer's disease. \u003cem\u003eNature\u003c/em\u003e 349(6311):704\u0026ndash;706.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJansen IE, \u003cem\u003eet al.\u003c/em\u003e (2022) Genome-wide meta-analysis for Alzheimer's disease cerebrospinal fluid biomarkers. \u003cem\u003eActa neuropathologica\u003c/em\u003e 144(5):821\u0026ndash;842.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChakravarty A (2004) Unifying concept for Alzheimer's disease, vascular dementia and normal pressure hydrocephalus - a hypothesis. \u003cem\u003eMedical hypotheses\u003c/em\u003e 63(5):827\u0026ndash;833.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBateman GA (2004) Pulse wave encephalopathy: a spectrum hypothesis incorporating Alzheimer's disease, vascular dementia and normal pressure hydrocephalus. \u003cem\u003eMedical hypotheses\u003c/em\u003e 62(2):182\u0026ndash;187.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTai LM, \u003cem\u003eet al.\u003c/em\u003e (2016) The role of APOE in cerebrovascular dysfunction. \u003cem\u003eActa neuropathologica\u003c/em\u003e 131(5):709\u0026ndash;723.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSlooter AJ, \u003cem\u003eet al.\u003c/em\u003e (2004) The impact of APOE on myocardial infarction, stroke, and dementia: the Rotterdam Study. \u003cem\u003eNeurology\u003c/em\u003e 62(7):1196\u0026ndash;1198.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShao A, \u003cem\u003eet al.\u003c/em\u003e (2022) Meta-analysis of the association between Apolipoprotein E polymorphism and risks of myocardial infarction. \u003cem\u003eBMC Cardiovasc Disord\u003c/em\u003e 22(1):126.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD'Agostino RB, Sr., \u003cem\u003eet al.\u003c/em\u003e (2008) General cardiovascular risk profile for use in primary care: the Framingham Heart Study. \u003cem\u003eCirculation\u003c/em\u003e 117(6):743\u0026ndash;753.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLupton MK, \u003cem\u003eet al.\u003c/em\u003e (2016) The effect of increased genetic risk for Alzheimer's disease on hippocampal and amygdala volume. \u003cem\u003eNeurobiology of aging\u003c/em\u003e 40:68\u0026ndash;77.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNir TM, \u003cem\u003eet al.\u003c/em\u003e (2022) Effects of ApoE4 and ApoE2 genotypes on subcortical magnetic susceptibility and microstructure in 27,535 participants from the UK Biobank. \u003cem\u003ePacific Symposium on Biocomputing. Pacific Symposium on Biocomputing\u003c/em\u003e 27:121\u0026ndash;132.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHua X, \u003cem\u003eet al.\u003c/em\u003e (2016) MRI-based brain atrophy rates in ADNI phase 2: acceleration and enrichment considerations for clinical trials. \u003cem\u003eNeurobiology of aging\u003c/em\u003e 37:26\u0026ndash;37.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcKhann G, \u003cem\u003eet al.\u003c/em\u003e (1984) Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. \u003cem\u003eNeurology\u003c/em\u003e 34(7):939\u0026ndash;944.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMirra SS, \u003cem\u003eet al.\u003c/em\u003e (1991) The Consortium to Establish a Registry for Alzheimer's Disease (CERAD). Part II. Standardization of the neuropathologic assessment of Alzheimer's disease. \u003cem\u003eNeurology\u003c/em\u003e 41(4):479\u0026ndash;486.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManolio TA, \u003cem\u003eet al.\u003c/em\u003e (2012) New Models for Large Prospective Studies: Is There a Better Way? \u003cem\u003eAmerican Journal of Epidemiology\u003c/em\u003e 175(9):859\u0026ndash;866.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKochunov P, \u003cem\u003eet al.\u003c/em\u003e (2006) Retrospective motion correction protocol for high-resolution anatomical MRI. \u003cem\u003eHum Brain Mapp\u003c/em\u003e 27(12):957\u0026ndash;962.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGlasser MF, \u003cem\u003eet al.\u003c/em\u003e (2013) The minimal preprocessing pipelines for the Human Connectome Project. \u003cem\u003eNeuroimage\u003c/em\u003e 80:105\u0026ndash;124.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSotiropoulos SN, \u003cem\u003eet al.\u003c/em\u003e (2013) Advances in diffusion MRI acquisition and processing in the Human Connectome Project. \u003cem\u003eNeuroimage\u003c/em\u003e 80:125\u0026ndash;143.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdes-Aron B, \u003cem\u003eet al.\u003c/em\u003e (2018) Evaluation of the accuracy and precision of the diffusion parameter EStImation with Gibbs and NoisE removal pipeline. \u003cem\u003eNeuroimage\u003c/em\u003e 183:532\u0026ndash;543.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBehrens TE, \u003cem\u003eet al.\u003c/em\u003e (2003) Characterization and propagation of uncertainty in diffusion-weighted MR imaging. \u003cem\u003eMagnetic resonance in medicine\u003c/em\u003e 50(5):1077\u0026ndash;1088.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaMontagne PJ, \u003cem\u003eet al.\u003c/em\u003e (2019) OASIS-3: Longitudinal Neuroimaging, Clinical, and Cognitive Dataset for Normal Aging and Alzheimer Disease. \u003cem\u003emedRxiv\u003c/em\u003e:2019.2012.2013.19014902.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJahanshad N, \u003cem\u003eet al.\u003c/em\u003e (2013) Multi-site genetic analysis of diffusion images and voxelwise heritability analysis: A pilot project of the ENIGMA-DTI working group. \u003cem\u003eNeuroimage\u003c/em\u003e 81:455\u0026ndash;469.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKochunov P, \u003cem\u003eet al.\u003c/em\u003e (2020) ENIGMA-DTI: Translating reproducible white matter deficits into personalized vulnerability metrics in cross-diagnostic psychiatric research. \u003cem\u003eHum Brain Mapp\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMueller SG, \u003cem\u003eet al.\u003c/em\u003e (2005) The Alzheimer's disease neuroimaging initiative. \u003cem\u003eNeuroimaging clinics of North America\u003c/em\u003e 15(4):869\u0026ndash;877, xi-xii.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetersen RC, \u003cem\u003eet al.\u003c/em\u003e (2010) Alzheimer's Disease Neuroimaging Initiative (ADNI): clinical characterization. \u003cem\u003eNeurology\u003c/em\u003e 74(3):201\u0026ndash;209.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWong ND (2020) Cardiovascular risk assessment: The foundation of preventive cardiology. \u003cem\u003eAm J Prev Cardiol\u003c/em\u003e 1:100008.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKochunov P, \u003cem\u003eet al.\u003c/em\u003e (2023) Ancestral, Pregnancy, and Negative Early-Life Risks Shape Children's Brain (Dis)similarity to Schizophrenia. \u003cem\u003eBiol Psychiatry\u003c/em\u003e 94(4):332\u0026ndash;340.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKochunov P, \u003cem\u003eet al.\u003c/em\u003e (2019) White Matter in Schizophrenia Treatment Resistance. \u003cem\u003eThe American Journal of Psychiatry\u003c/em\u003e 176(10):829\u0026ndash;838.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKochunov P, \u003cem\u003eet al.\u003c/em\u003e (2023) Ancestral, Pregnancy, and Negative Early Life Risks Shape Children's Brain Dis/Similarity to Schizophrenia. \u003cem\u003eBiol Psychiatry\u003c/em\u003e 94(4):332\u0026ndash;340.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrisoni GB, \u003cem\u003eet al.\u003c/em\u003e (2022) The probabilistic model of Alzheimer disease: the amyloid hypothesis revised. \u003cem\u003eNature reviews. Neuroscience\u003c/em\u003e 23(1):53\u0026ndash;66.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMahley RW \u0026amp; Huang Y (2012) Apolipoprotein e sets the stage: response to injury triggers neuropathology. \u003cem\u003eNeuron\u003c/em\u003e 76(5):871\u0026ndash;885.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOperto G, \u003cem\u003eet al.\u003c/em\u003e (2018) White matter microstructure is altered in cognitively normal middle-aged APOE-ε4 homozygotes. \u003cem\u003eAlzheimer's research \u0026amp; therapy\u003c/em\u003e 10(1):48.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOperto G, \u003cem\u003eet al.\u003c/em\u003e (2019) Interactive effect of age and APOE-ε4 allele load on white matter myelin content in cognitively normal middle-aged subjects. \u003cem\u003eNeuroImage. Clinical\u003c/em\u003e 24:101983.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBartzokis G, \u003cem\u003eet al.\u003c/em\u003e (2003) White matter structural integrity in healthy aging adults and patients with Alzheimer disease: a magnetic resonance imaging study. \u003cem\u003eArch Neurol\u003c/em\u003e 60(3):393\u0026ndash;398.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoutsodendris N, \u003cem\u003eet al.\u003c/em\u003e (2023) Neuronal APOE4 removal protects against tau-mediated gliosis, neurodegeneration and myelin deficits. \u003cem\u003eNature aging\u003c/em\u003e 3(3):275\u0026ndash;296.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIlveskoski E, \u003cem\u003eet al.\u003c/em\u003e (1999) Age-dependent association of apolipoprotein E genotype with coronary and aortic atherosclerosis in middle-aged men: an autopsy study. \u003cem\u003eCirculation\u003c/em\u003e 100(6):608\u0026ndash;613.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEichner JE, \u003cem\u003eet al.\u003c/em\u003e (2002) Apolipoprotein E polymorphism and cardiovascular disease: a HuGE review. \u003cem\u003eAm J Epidemiol\u003c/em\u003e 155(6):487\u0026ndash;495.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMahley RW (1988) Apolipoprotein E: cholesterol transport protein with expanding role in cell biology. \u003cem\u003eScience\u003c/em\u003e 240(4852):622\u0026ndash;630.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNgwa JS, \u003cem\u003eet al.\u003c/em\u003e (2018) Associations of Pulse and Blood Pressure with Hippocampal Volume by APOE and Cognitive Phenotype: The Alzheimer's Disease Neuroimaging Initiative (ADNI). \u003cem\u003eDementia and geriatric cognitive disorders\u003c/em\u003e 45(1\u0026ndash;2):66\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBateman GA (2002) Pulse-wave encephalopathy: a comparative study of the hydrodynamics of leukoaraiosis and normal-pressure hydrocephalus. \u003cem\u003eNeuroradiology\u003c/em\u003e 44(9):740\u0026ndash;748.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScuteri A, Nilsson PM, Tzourio C, Redon J, \u0026amp; Laurent S (2011) Microvascular brain damage with aging and hypertension: pathophysiological consideration and clinical implications. \u003cem\u003eJ Hypertens\u003c/em\u003e 29(8):1469\u0026ndash;1477.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeorge KM, \u003cem\u003eet al.\u003c/em\u003e (2023) Association of Early Adulthood Hypertension and Blood Pressure Change With Late-Life Neuroimaging Biomarkers. \u003cem\u003eJAMA network open\u003c/em\u003e 6(4):e236431.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGorelick PB, \u003cem\u003eet al.\u003c/em\u003e (2011) Vascular contributions to cognitive impairment and dementia: a statement for healthcare professionals from the american heart association/american stroke association. \u003cem\u003eStroke\u003c/em\u003e 42(9):2672\u0026ndash;2713.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThaipisuttikul P, Jaikla K, Satthong S, \u0026amp; Wisajun P (2022) Rate of conversion from mild cognitive impairment to dementia in a Thai hospital-based population: A retrospective cohort. \u003cem\u003eAlzheimer's \u0026amp; dementia (New York, N. Y.)\u003c/em\u003e 8(1):e12272.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMitchell AJ \u0026amp; Shiri-Feshki M (2009) Rate of progression of mild cognitive impairment to dementia\u0026ndash;meta-analysis of 41 robust inception cohort studies. \u003cem\u003eActa psychiatrica Scandinavica\u003c/em\u003e 119(4):252\u0026ndash;265.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGauthier S, \u003cem\u003eet al.\u003c/em\u003e (2006) Mild cognitive impairment. \u003cem\u003eLancet\u003c/em\u003e 367(9518):1262\u0026ndash;1270.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVemuri P, \u003cem\u003eet al.\u003c/em\u003e (2009) MRI and CSF biomarkers in normal, MCI, and AD subjects: diagnostic discrimination and cognitive correlations. \u003cem\u003eNeurology\u003c/em\u003e 73(4):287\u0026ndash;293.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDavatzikos C, Bhatt P, Shaw LM, Batmanghelich KN, \u0026amp; Trojanowski JQ (2011) Prediction of MCI to AD conversion, via MRI, CSF biomarkers, and pattern classification. \u003cem\u003eNeurobiology of aging\u003c/em\u003e 32(12):2322.e2319-2327.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDayarathna T, \u003cem\u003eet al.\u003c/em\u003e (2024) Nanoscale flow cytometry-based quantification of blood-based extracellular vesicle biomarkers distinguishes MCI and Alzheimer's disease. \u003cem\u003eAlzheimers Dement\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarth\u0026eacute;lemy NR, \u003cem\u003eet al.\u003c/em\u003e (2024) Highly accurate blood test for Alzheimer\u0026rsquo;s disease is similar or superior to clinical cerebrospinal fluid tests. \u003cem\u003eNature medicine\u003c/em\u003e 30(4):1085\u0026ndash;1095.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGroot C, \u003cem\u003eet al.\u003c/em\u003e (2024) Tau Positron Emission Tomography for Predicting Dementia in Individuals With Mild Cognitive Impairment. \u003cem\u003eJAMA neurology\u003c/em\u003e 81(8):845\u0026ndash;856.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchindler SE \u0026amp; Atri A (2023) The role of cerebrospinal fluid and other biomarker modalities in the Alzheimer\u0026rsquo;s disease diagnostic revolution. \u003cem\u003eNature aging\u003c/em\u003e 3(5):460\u0026ndash;462.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLivingston G, \u003cem\u003eet al.\u003c/em\u003e (2017) Dementia prevention, intervention, and care. \u003cem\u003eLancet\u003c/em\u003e 390(10113):2673\u0026ndash;2734.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNugent KL, \u003cem\u003eet al.\u003c/em\u003e (2013) Familial Aggregation of Tobacco Use Behaviors in Amish Men. \u003cem\u003eNicotine Tob Res\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"molecular-psychiatry","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"mp","sideBox":"Learn more about [Molecular Psychiatry](http://www.nature.com/mp/)","snPcode":"41380","submissionUrl":"https://mts-mp.nature.com/cgi-bin/main.plex","title":"Molecular Psychiatry","twitterHandle":"@molpsychiatry","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Big data, Alzheimer’s disease, neuroimaging, RVI, amyloid, cerebrovascular disease ","lastPublishedDoi":"10.21203/rs.3.rs-6770628/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6770628/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePreventing Alzheimer’s disease (AD) requires early-warning biomarkers. We developed a Regional Vulnerability Index (RVI) that quantifies individual brain similarity to AD patients' expected brain deficit patterns. We calculated regional effect sizes to establish brain deficit patterns in amyloid-positive AD cases compared to amyloid-negative healthy controls. The RVI-AD was calculated as a linear index of individual similarity to this established brain pattern in AD. Initially, we demonstrated RVI-AD elevation associated with risk factors in 335 participants (mean age: 49±13 years) in the Amish Connectome Project, followed by an independent sample consisting of 26,010 participants (mean age: 64±7 years) from the UK Biobank. Genetic and cardiovascular risks were evaluated using APOE-e4 genotype and Framingham Cardiovascular Risk Scores (FCVRS), respectively. Additionally, we assessed the risk of converting from MCI to dementia in N=1,932 participants (mean age: ~74) from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Participants with the APOE-e4 allele had significantly elevated RVI-AD indices (p\u0026lt;0.05); FCVRS significantly contributed to higher RVI-AD in an APOE-e4-specific manner (p\u0026lt;0.01), replicable across the samples. In the ADNI cohort, RVI-AD significantly predicted conversion from MCI to dementia in the next decade, particularly within the first 3 years (AUC=74%). In healthy individuals, the RVI-AD index detected the insidious impact of APOE-ε4 and cardiovascular risks in otherwise normally aging cohorts. Elevated RVI-AD also predicted conversion to dementia within 10 years in the older, high-risk cohort. Further development of this brain-pattern similarity-based approach may yield a noninvasive, clinically accessible biomarker to aid early detection of the subtle to more imminent effects of AD risks.\u003c/p\u003e","manuscriptTitle":"Alzheimer’s Disease-Like Brain Pattern Biomarker: Capturing Risks and Predicting Disease Onset","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-15 18:46:42","doi":"10.21203/rs.3.rs-6770628/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"revise","date":"2025-11-07T18:21:57+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"This content is not available.","date":"2025-10-24T15:59:45+00:00","index":2,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2025-10-08T07:44:34+00:00","index":2,"fulltext":"This content is not available."},{"type":"editorInvitedReview","content":"This content is not available.","date":"2025-09-23T14:15:28+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2025-09-08T07:17:15+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewersInvited","content":"","date":"2025-09-08T05:33:13+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-05-30T15:18:13+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-05-30T15:13:03+00:00","index":"","fulltext":""},{"type":"submitted","content":"Molecular Psychiatry","date":"2025-05-29T14:37:48+00:00","index":"","fulltext":""},{"type":"checksFailed","content":"","date":"2025-05-29T09:17:45+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"molecular-psychiatry","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"mp","sideBox":"Learn more about [Molecular Psychiatry](http://www.nature.com/mp/)","snPcode":"41380","submissionUrl":"https://mts-mp.nature.com/cgi-bin/main.plex","title":"Molecular Psychiatry","twitterHandle":"@molpsychiatry","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"f7c52e8a-6149-499a-b5dd-fcd39799307a","owner":[],"postedDate":"September 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":54333633,"name":"Health sciences/Biomarkers"},{"id":54333634,"name":"Health sciences/Diseases"}],"tags":[],"updatedAt":"2026-04-28T07:07:30+00:00","versionOfRecord":{"articleIdentity":"rs-6770628","link":"https://doi.org/10.1038/s41380-026-03617-0","journal":{"identity":"molecular-psychiatry","isVorOnly":false,"title":"Molecular Psychiatry"},"publishedOn":"2026-04-27 04:00:00","publishedOnDateReadable":"April 27th, 2026"},"versionCreatedAt":"2025-09-15 18:46:42","video":"","vorDoi":"10.1038/s41380-026-03617-0","vorDoiUrl":"https://doi.org/10.1038/s41380-026-03617-0","workflowStages":[]},"version":"v1","identity":"rs-6770628","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6770628","identity":"rs-6770628","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}