# paper.pdf ## Metadata - PDFFormatVersion=1.6 - Language=en - IsLinearized=true - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - CreationDate=D:20260623103403+05'30' - Creator=LaTeX with hyperref package - ModDate=D:20260623103442+05'30' - Producer=Acrobat Distiller 26.0 (Windows) - Title=Cortical gray–white matter contrast alterations precede amyloid‐β positivity and macrostructural changes in older adults without dementia - Custom.WPS-ARTICLEDOI=10.1002/alz.71609 - Custom.WPS-JOURNALDOI=10.1002/(ISSN)1552-5279 - Custom.WPS-PROCLEVEL=3 - xmp:createdate=2015-08-18T08:05:36+05:30 - xmp:modifydate=2016-06-21T14:17:29-04:00 - xmp:metadatadate=2016-06-21T14:17:29-04:00 - pdf:producer=Acrobat Distiller 9.0 (Windows) - pdfx:wps-articledoi=10.1002/alz.71609 - pdfx:wps-journaldoi=10.1002/(ISSN)1552-5279 - pdfx:wps-proclevel=3 - jav:journal_article_version=VoR - prism:doi=10.1002/alz.71609 - dc:format=application/pdf - dc:title=Cortical gray–white matter contrast alterations precede amyloid‐β positivity and macrostructural changes in older adults without dementia - xmpmm:documentid=uuid:e6001f36-37ff-3093-bb13-c8b9bb005dda - xmpmm:instanceid=uuid:cfede551-c930-352d-ae67-64db97a58be1 ## Contents ### Page 1 Received: 29 January 2026 Revised: 14 May 2026 Accepted: 18 May 2026 DOI: 10.1002/alz.71609R E S E A R C H A R T I C L E Cortical gray–white matter contrast alterations precede amyloid-β positivity and macrostructural changes in older adults without dementia Leonard Pieperhoff1,2 Luigi Lorenzini1,2,3 Francisco Almeida4,5Mario Tranfa1,2,6,7,8 Ariane Bollack9,10 Prithvi Arunachalam1,2Federico Masserini1,2,11 Robin Wolz12 Sylke Grootoonk12 Craig Ritchie13Mercè Boada14,15 Marta Marquié14,15 Jort Vijverberg2,16 Rik Vandenberghe17 Bernard J. Hanseeuw18,19 Pablo Martinez-Lage20Pierre Payoux21 Pieter Jelle Visser2,16 Michael Schöll22,23,24Giovanni B. Frisoni25 Matteo Pardini3,26 Luca Roccataglia27,28 Andrew W. Stephens29 Chris Buckley9 Gill Farrar9 Frank Jessen30 Juan Domingo Gispert31 Gemma Salvadó31,32 Emma S. Luckett1,2,17,33Henk-Jan MM Mutsaerts1,2 Alle Meije Wink1,2 Lyduine E. Collij1,2Frederik Barkhof1,2 Tiago Gil Oliveira4,5,34 on behalf of the AMYPAD consortium CorrespondenceLeonard Pieperhoff and Frederik Barkhof, de Boelelaan 1117, PK-1X014, Amsterdam Universitair Medische Centra, Amsterdam, 1081HV, the Netherlands. Email: l.pieperhoff@amsterdamumc.nl; f.barkhof@amsterdamumc.nl Tiago Gil Oliveira, Life and Health Sciences Research Institute (ICVS), School of Medicine, University of Minho, Braga, Portugal. Email: tiago@med.uminho.ptFunding informationInnovative Medicines Initiative (IMI) 2 Joint Undertaking, Grant/Award Number: 115952; European Union’s Horizon 2020 Research and Innovation Programme; Alzheimer Nederland, Grant/Award Number: WE.03-2024-27; Alzheimer’s Disease Data Initiative (ADDI); European Federation of Pharmaceutical Industries and AssociationsAbstract INTRODUCTION: Early Alzheimer’s disease (AD) involves subtle cortical changes that may precede atrophy. Magnetic resonance imaging (MRI) microstructural markers may detect earlier pathology than classical morphometry.METHODS: We analyzed cross-sectional MRI and amyloid-β (Aβ) positron emission tomography (PET) data from 1323 non-demented AMYPAD participants. Cortical vol- ume, thickness, gray–white matter contrast (GWC), and mean diffusivity (MD) were related to global Aβ burden and estimated time to Aβ-positivity using regression, correlation, and change-point analyses.RESULTS: Microstructural measures showed stronger age associations than macrostructural measures, whereas all measures were unaffected by apolipopro- tein E (APOE) -ε4 carriership. GWC and MD showed minimal overlap with volume and thickness. Higher Aβ burden was most strongly associated with reduced GWC and cortical thinning. Change-point analyses showed GWC alterations preceded Aβ-positivity by several years. Frederik Barkhof and Tiago Gil Oliveira contributed equally to this work. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2026 The Author(s). Alzheimer’s & Dementia published by Wiley Periodicals LLC on behalf of Alzheimer’s Association.Alzheimer’s Dement. 2026;22:e71609. wileyonlinelibrary.com/journal/alz 1 of 17https://doi.org/10.1002/alz.71609 ### Page 2 2 of 17 PIEPERHOFF ET AL.DISCUSSION: Cortical microstructural MRI, particularly GWC, changes earlier than atrophy and may serve as an early in vivo marker of AD pathology.K E Y W O R D Sdisease progression, MRI, neuroimaging, PET, preclinical Alzheimer’s disease, SILAHighlights∙ Cortical micro- and macro-structural magnetic resonance imaging (MRI) measures capture distinct processes∙ Amyloid burden relates most strongly to gray–white matter contrast loss∙ Microstructural MRI alterations emerge before cortical atrophy∙ Gray–white matter contrast changes precede amyloid-β positivity by several years1 BACKGROUNDAs disease-modifying therapies targeting amyloid-β (Aβ) are targeting earlier disease stages,1 there is an urgent need for imaging markers capable of detecting pathological changes during the earliest dis- ease stages. Traditional macrostructural measures of cortical atrophy, including gray matter (GM) volume and cortical thickness, have been the mainstay of structural MRI assessments in Alzheimer’s disease (AD),2 yet they capture relatively late-stage neurodegeneration and may exhibit early biphasic trajectories, with some studies reporting early transient increases in volume, thickness or density,3–8 potentially reflecting early neuroinflammatory cellular processes.9 In contrast, measures of cortical microstructural integrity may be sensitive to more subtle, early pathological processes such as myelin degradation, synap- tic loss, and neuroinflammatory responses that precede overt tissue loss.10–14 Magnetic resonance imaging (MRI) -derived indices of cortical microstructural integrity, obtainable from both structural and diffu- sion MRI, have shown promise in detecting early neuropathologi- cal/proteinopathic changes. Among these, mean diffusivity (MD) in the cortical GM is a well-established diffusion-based marker that has demonstrated sensitivity to disease onset in presymptomatic familial AD,15 and to cognitive decline and hippocampal volume loss in spo- radic AD,16 possibly reflecting cortical deposition of neurofibrillary tangles.17 Even though less explored than other brain MRI metrics, gray–white matter contrast (GWC) on T1-weighted MRI has emerged as a poten- tially complementary indicator of cortical microstructure. GWC quan- tifies differences in T1-weighted signal intensity between gray and white matter directly adjacent to the gray–white tissue boundary, and has been shown to reduce with aging across most of the cortex inde- pendently of cortical thinning.18,19 This contrast is thought to be driven primarily by intracortical myelin content, however, as an inherently non-specific metric, it may also reflect broader cellular composition of the cortex, such as the distribution of oligodendrocyte precursor cells, endothelial cells and astrocytes in the cortex.19,20 Direct histological validation of GWC against post mortem tissue are still missing, which means the precise biological substrates underlying this signal are not yet fully established. With an increasingly evident role of Aβ-associated oligodendrocyte dysfunction and demyelination in the earliest stages of AD,21–23 GWC could represent a promising early MRI marker of neu- ropathological changes, though thus far it has only been demonstrated in clinical, but not pre-clinical populations.24–26 The extent to which image-derived phenotypes of cortical micro- and macro-structure represent independent biological processes has not been directly studied. While microstructure has been implied to change earlier than macrostructure in previous literature, the differ- ence in the dynamics of their changes have also not been directly compared and only observed across clinical groups.17 Therefore, in a large pan-European cohort mainly encompassing preclinical and early AD cases, we aimed to (i) identify associations between demographic, genetic, clinical participant characteristics, and cortical micro- and macro-structural integrity, as measured by GWC and MD, and volume and thickness, respectively; (ii) quantify the extent of shared/divergent variance between MRI features reflecting cortical micro- and macro- structural integrity, on a global and regional scale; (iii) examine the cross-sectional relation between cortical Aβ burden and MRI fea- tures reflecting cortical micro- and macro-structural integrity; (iv) approximate the time at which MRI features reflecting cortical micro- and macro-structural integrity become affected relative to clinically meaningful amounts of cortical Aβ aggregates.2 METHODS 2.1 ParticipantsCross-sectional data including available T1-weighted MRI, Aβ-positron emission tomography (PET), and available age, sex, Clinical Dementia Rating (CDR) and apolipoprotein E (APOE) -ε4 carriership status for this ### Page 3 PIEPERHOFF ET AL. 3 of 17RESEARCH IN CONTEXT1. Systematic review: The authors reviewed prior litera- ture using PubMed. Previous studies have established cortical thickness and volume as markers of neurode- generation and suggested that microstructural magnetic resonance imaging (MRI) measures, including mean dif- fusivity and gray–white matter contrast (GWC), may be sensitive to early pathology. However, their interrelation- ships, links to amyloid-β burden, and temporal ordering relative to amyloid positivity have not been systemati- cally compared in large preclinical cohorts. 2. Interpretation: Our findings indicate that cortical microstructural MRI measures, particularly GWC, show patterns that are partially independent from classical morphometric measures such as cortical volume and thickness. Alterations in GWC were detectable years before amyloid-β positivity and cortical atrophy. This positions GWC as a sensitive marker of early brain changes in preclinical Alzheimer’s disease. 3. Future directions: Future studies should validate these temporal sequences longitudinally, assess whether GWC predicts subsequent amyloid accumulation and cognitive decline, and determine its utility as an outcome marker in preclinical intervention trials. study were drawn from the AMYPAD PNHS v202306 (N = 1323).27 The AMYPAD PNHS (EudraCT: 2018-002277-22) is a European multi-center cohort of non-demented participants to determine the value of Aβ-PET in clinical- and research settings; the study design has been described in detail previously.28,29 Briefly, eligibility crite- ria for inclusion in the AMYPAD PNHS were no history of dementia (CDR < 1), age above 50 years, and being able to undergo an Aβ-PET and MRI scan. The studies were reviewed and approved by the Med- ical Ethical Committee of the University Medical Center Amsterdam, location VUmc and all local sites. The studies were conducted accord- ing to the principles of the Helsinki Declaration of 1975, as revised in 2008, and all human participants gave written informed consent. The median time difference between the MRI and PET acquisition was 50 days (interquartile range [IQR] 99 days). None of the study participants participated in clinical trials of anti-Aβ therapy throughout the duration of the study.2.2 MRI acquisition and quantificationMRI protocols differed across cohorts and sites and are described across site, manufacturer and scanner model for T1-weighted MRI in Table S1 and for diffusion-weighted imaging (DWI) in Table S2. 2.2.1 T1-weighted MRI 3D T1-weighted images were processed using FreeSurfer version 7.1.1.30 In short, the recon-all pipeline involved skull stripping, volumet- ric labeling, intensity normalization, white matter (WM) segmentation, surface-based atlas registration, surface extraction and gyral label- ing. Cortical parcellation was performed using the Desikan–Killiany atlas, and parcellations were visually quality controlled including regional boundary fits, according to a previously published protocol.31 Regional volumetric, thickness and surface area data were computed using mris_anatomical_stats and extracted using aparcstats2table. GWC was computed using FreeSurfer’s pctsurfcon function as described previously,24 sampling the GM and WM adjacent to the gray–white tissue boundary and expressed as GWC = 100×(WM−GM) 0.5×(WM+GM) . Specifically, GM is the average regional GM signal intensity within 35% of GM adjacent to the GM–WM tissue boundary, while WM represents the mean regional WM signal intensity sampled 1 mm below the GM–WM boundary. The inner 35th percentile of GM was chosen for sam- pling signal intensity following prior studies and to ensure sufficient sampling space without risking potential CSF contamination even in thinner regions,24,26,32 with 1 mm of WM similarly being used as the standard parameter to ensure sampled WM remains strictly super- ficial. All metrics were extracted for the 34 symmetric regions of the Desikan–Killiany atlas per hemisphere, resulting in 68 regions of interest (ROIs).33 2.2.2 Diffusion MRI DWI (pre-) processing was performed using QSIprep v0.19.34 Briefly, DWI data underwent denoising and B1 field inhomogeneity correc- tion as implemented in MRtrix3’s “dwidenoise” and “dwibiascorrect”, respectively. Head motion and eddy current correction were per- formed using FSL’s eddy. Using the fieldmap-less symmetric normal- ization approach as implemented in QSIprep, a deformation field was estimated to perform geometric distortion correction and to obtain an unwarped b0 reference for improved DWI to T1w registration. The DWI time-series were then resampled to T1w space. Diffusion tensor imaging scalar maps of MD were computed from the pre-processed DWI scans using FSL DTIFIT. We then computed regional GM MD val- ues by projecting the FreeSurfer-derived Desikan–Killiany parcellation onto the processed MD maps, and averaging scalar values within the 68 ROIs. To minimize potential CSF partial volume contamination, which can bias MD estimates particularly in the context of atrophy,35 only voxels with a gray matter tissue fraction exceeding 80% were included in the regional averaging. 2.2.3 Statistical harmonization To account for the potential confounding effect of scanner, sequence or acquisition parameter differences between different subjects while maintaining inter-subject biological differences, we harmonized ### Page 4 4 of 17 PIEPERHOFF ET AL.all image-derived phenotypes using mean-only Combat harmoniza- tion, implemented in the R package “neuroCombat” version 1.0.13 (Figure S1).36,37 ComBat models batch effects as additive and mul- tiplicative site effects within an empirical Bayes framework; the mean-only variant estimates and removes only the additive compo- nent, leaving variance structure intact. Mean-only rather than full ComBat harmonization was chosen because several batches par- ticularly for MD, comprised few participants, making within-batch variance estimation potentially unreliable and full adjustment poten- tially counterproductive.38 We estimated batches by binning subjects according to echo time and repetition time of the T1-weighted acqui- sition for FreeSurfer-derived measures, then according to echo time and repetition time of the DWI acquisition for MD (Table S3). Age, sex, global CDR, and APOE-ε4 carriership were explicitly modeled as part of the ComBat design matrix to preserve inter-subject biological variance between batches during harmonization. The proportion of variance explained by batch before and after harmonization for each metric is reported in Table S4.2.3 Aβ-PET acquisition and quantificationPET scans were acquired 90–110 minutes p.i. of 185 MBq (± 10%) for [18F]Flutemetamol and 350 MBq (± 20%) for [18F]Florbetaben, consisting of four frames of 5 minutes according to the standard proto- col for each tracer.39,40 Image analysis was performed centrally using IXICO’s automated workflow. Briefly, PET frames were co-registered, averaged, and aligned to the corresponding MRI scan, which was parcellated using a subject-specific multi-atlas approach, that is, the learning embeddings for atlas propagation parcellation procedure.41 Standardized uptake value ratio (SUVR) images were obtained using the whole cerebellum as a reference region in native space. In order to pool Aβ-PET data across sites, SUVR values were transformed to Centiloids (CL) using the standard Global Alzheimer’s Association Interactive Network target region as a measure of global Aβ burden.42 Using sampled iterative local approximation (SILA; https://github. com/Betthauser-Neuro-Lab/SILA-AD-Biomarker),43 estimated time to Aβ-positivity was calculated for all participants. To estimate a par- ticipant’s time to disease onset, the algorithm was applied to the full study dataset of 1387 individuals, of which 763 had longitudinal Aβ- PET data including 98 participants with three scans, with a median maximum follow-up time of 3.2 years (IQR: 2.1–4.4; Figure S2). In short, discrete sampling of CL values versus age data was used to establish the relationship between CL rate of change and CL. Numerical smooth- ing using robust locally estimated scatterplot smoothing (LOESS) and Euler’s method were then applied to numerically integrate these data, generating a non-parametric CL-versus-time curve. To give the inte- grated timeline meaning, the SILA algorithm sets time equal zero to a user-specified threshold, which was set at 24.1 CL, reflecting reliably detectable Aβ.44,45 The estimated years from biomarker positivity is calculated for each person by first solving this curve for time using a person’s observed CL, and subtracting the estimated A+ duration from their age at that scan. A single common accumulation trajectory was assumed for all participants, which is supported by evidence that APOE-ε4 carriership modulates the age of onset of amyloid accumulation, but not its rate of change,43,46 and that women have neither earlier onset nor increased rate of change or burden.43,47 Age-related variation in accumulation rate is captured implicitly through the CL rate-of-change versus CL relationship that SILA derives from the observed data.2.4 Statistical analysisAll analyses were performed in R, version 4.4.1. 2.4.1 Associations between demographic, genetic, clinical participant characteristics, and cortical micro-and macro-structural integrity For the first aim of assessing associations between demographic, genetic, clinical participant characteristics, and cortical micro-and macro-structural integrity, the four metrics of interest (Volume, Thick- ness, GWC, MD) were first z-transformed across the whole cohort, then averaged across the 68 ROIs, while weighting for regional surface area. Two sample independent t-tests for binary, and linear correlations using Pearson’s correlations for continuous variables as well as Spear- man’s rank correlations for Mini-Mental State Examination (MMSE) were computed between each global average metric and the follow- ing measures: age, sex (“Male”/“Female”), APOE-ε4 carriership status (“Yes”/“No”), CDR (“0 – Normal”/“0.5 – Very mild”), MMSE, estimated intracranial volume, lateral ventricular volume, CL. The four global outcome measures were each corrected for age, sex, and APOE-ε4 carriership status by using linear models and retaining the residuals for analyses; global volume was additionally corrected for estimated total intracranial volume. When correlating against the corresponding correction variable, it was left out of the correction process. 2.4.2 Extent of shared/divergent variance between macro-and micro-structural MRI features For the second aim, which examined the extent of shared/divergent variance between macro-and micro-structural MRI features, we cal- culated pairwise Pearson’s correlations between all possible combina- tions of global and regional average metrics, that is six global combina- tions “Volume-Thickness”, “Volume-GWC”, “Volume-MD”, “Thickness- GWC”, “Thickness-MD”, and “GWC-MD”; and (68×4)2 2 − 68×4 2 = 36856 unique regional combinations between the four metrics across all 68 atlas regions. To eliminate potential confounding effects of pathology, we only included Aβ-negative participants with a global CDR score of 0. Regional and global metrics were also adjusted for age, sex and APOE-ε4 carriership status. All correlations underwent Benjamini–Hochberg false discovery rate (FDR)-correction.48 ### Page 5 PIEPERHOFF ET AL. 5 of 172.4.3 Relation between global aggregations of cortical Aβ and micro- and macro-structural MRI features To understand the cross-sectional relation between global aggrega- tions of cortical Aβ and micro- and macro-structural MRI features, we utilized linear models predicting each regional metric using global CL, correcting for age, sex, and CDR. The models predicting cortical volume were additionally corrected for estimated total intracranial volume. Results were FDR-corrected per metric to account for 68 comparisons. 2.4.4 Approximation of change-points in cross-sectional MRI feature trajectories along time to cortical Aβ positivity Finally, we aimed to estimate the timepoints at which MRI fea- tures reflecting cortical micro- and macro-structural integrity become affected in relation to changes in aggregation of cortical Aβ. We did so in an a priori selection of six ROIs: the precuneus, posterior cingulate, and medial orbitofrontal cortex as three core hubs of Aβ aggregation49; the entorhinal cortex as earliest site of cortical tau spread,50 supe- rior temporal as a region affected by both Aβ and tau pathology,50,51 and lateral-occipital cortex as a typically less or later-affected sensory- association area.52,53 To this end, we used two methods: first, we estimated change-points and their standard error using piecewise terms in a linear model predicting regional phenotypes using estimated time to Aβ-positivity, corrected for sex and CDR, implemented in the R package “segmented”, version 2.1-2.54 Age was not used as a covariate because its variance is already partially modelled within the estimated time to Aβ-positivity; to ensure that identified effects are not predom- inantly due to age effects, we also conducted sensitivity analyses using Centiloid as the outcome while additionally correcting for age. Sec- ond, to approach a more biologically plausible statistical model, we estimated changepoints using a generative additive model with the same original terms, but with a P-splines fit for the estimated time to Aβ-positivity term, implemented in the R package “gamlss” version 5.4- 22.55 The change-point was here defined as the absolute maximum y-value of the second derivative of the smoothed estimated time to Aβ- positivity term, marking its peak curvature, ergo the time-point along the time to Aβ-positivity spectrum at which the rate of change of the metric in question changes the most.56 To estimate a standard error with this second, nonlinear method of change-point estimation, we per- muted the generative additive model 1000 times and used the standard deviation of the change-point estimate, using ordinary bootstrapping implemented in the R package “boot” version 1.3-31.57 Since laterality effects were not specifically tested for, left and right ROIs were aver- aged to reduce the number of statistical tests and increase robustness of metrics. We finally also putatively compared nonlinear change-point estimates across phenotypes using Welch two sample t-tests.2.5 Data availabilityData for this study were drawn from the AMYPAD PNHS v202306, available on the AD Workbench and can be requested via the AD Workbench FAIR portal under https://fair.addi.addatainitiative.org/ #/data/datasets/amypad_pnhs__harmonised_and_derived__v202306. The data access request procedure is described under https://doi.org/10.5281/zenodo.7962924.3 RESULTS 3.1 Demographic and clinical characteristicsBaseline demographics can be found in Table 1. Of the 1323 included participants, the majority were cognitively unimpaired (CDR = 0:N = 1083, 82%), with an overall average MMSE score of 28.87 (± 1.45). Mean age was 68.0 (± 8.7) years, 746 (56%) were female, and 329 (24.9%) were Aβ-positive (CL≥24.1). Individuals with Aβ-positivity more frequently had a CDR = 0.5, lower MMSE scores, and a higher frequency of APOE-ε4 carriership compared with Aβ-negative individ- uals. While volumetric, thickness, and GWC data were available for the whole cohort, diffusion data were only available for a subset of 435 individuals; demographic and clinical characteristics for this subset are reported in Table S5.3.2 Demographic, genetic, and clinical factors affect micro- and macro-structural cortical integrity differentiallyCorrelations or mean differences between demographic, genetic and clinical factors and global cortical averages of volume, thickness, MD, and GWC are reported in Table 2. While cortical volume and thick- ness were modestly negatively correlated with age after correction for sex and APOE-ε4 carriership, MD was strongly positively and GWC strongly negatively correlated to age (Figure 1A). Cortical volume and MD were higher for male compared with female participants, while thickness was higher for female compared with male partici- pants; GWC was not different between male and female participants (Figure 1B). No phenotype was significantly affected by APOE-ε4 carriership (Figure 1C). Participants with a global CDR of 0.5 had lower cortical volume and thickness as well as lower GWC, but no changes in MD (Figure 1D). Lower volume and thickness were associated with lower MMSE score, while MD and GWC were not related to MMSE (Figure 1E). Average lateral ventricular volume decreased with cortical volume and increased with higher MD (Figure 1F). Global Aβ burden measured with CL was related negatively with cortical volume, thickness, and GWC, but not with MD (Figure 1G). ### Page 6 6 of 17 PIEPERHOFF ET AL.TA B L E 1 Demographics and clinical characteristics at baselineVariable OverallN = 1,323* Aβ−N = 994* Aβ+N = 329* p-valueAge, years 68.00 (8.66) 66.47 (8.07) 72.60 (8.76) <0.001†Sex, female 746 (56%) 574 (58%) 172 (52%) 0.083‡Education, years 14.61 (3.96) 14.73 (3.93) 14.26 (4.00) 0.10†CDR <0.001§ 0 - Normal 1083 (82%) 874 (88%) 209 (64%) 0.5 - Very mild 239 (18%) 120 (12%) 119 (36%) MMSE 28.87 (1.45) 29.07 (1.20) 28.20 (1.91) <0.001†(Missing) 125 77 48APOE-ε4 carriership (% carriers) 528 (40%) 320 (32%) 208 (63%) <0.001‡Aβ-PET, Centiloids 18.64 (31.00) 3.56 (9.01) 64.18 (29.31) <0.001†est. total intracranial volume, cm3 1491.8 (175.4) 1490.9 (176.1) 1494.4 (173.6) 0.5†Note: Aβ positivity was defined on a cutoff of CL > 24.1. Abbreviations: Aβ, amyloid β; APOE, apolipoprotein E; CDR, Clinical Dementia Rating; CL, Centiloid; MMSE, Mini-Mental State Examination; PET, positron emission tomography. *Mean (SD); n (%).†Wilcoxon rank sum test.‡Pearson’s chi-squared test. §Fisher’s exact test.TA B L E 2 Correlations or mean differences between demographic, genetic and clinical factors and global cortical MRI averagesParameter Volume (n = 1323) Thickness (n = 1323) MD (n = 435) GWC (n = 1323)Age* −0.44 −0.29 0.54 −0.48Sex (male–female)† 0.10 −0.25 0.28 −0.03APOE-ε4 (noncarrier–carrier)† 0.02 0.07 0.11 −0.01 CDR (0–0.5)† 0.22 0.25 −0.16 0.11MMSE‡ 0.15 0.10 −0.02 −0.01 Lateral ventricle volume* −0.17 −0.04 0.10 −0.04 Aβ-PET, Centiloids* −0.07 −0.11 −0.05 −0.10Note: Significant results are displayed in bold. Abbreviations: Aβ, amyloid-beta; APOE, apolipoprotein E; CDR, Clinical Dementia Rating; GWC, gray–white matter contrast; MD, mean diffusivity; MMSE, Mini-Mental Status Examination. *Pearson’s correlation; Pearson’s R.†Welch Two Sample t test; group mean difference (z).‡Spearman’s correlation; Spearman’s rho.3.3 Micro- and macro-structural MRI features represent mostly independent processesFor global averages in only healthy participants (N = 874; N = 315 with MD data), volume and thickness (R = 0.323, p < 0.00001) as well as GWC and MD (R = -0.522, p < 0.00001) were strongly correlated with each other. Correlations that survived FDR-correction across the 36,856 regional correlations (n = 22,210) are displayed in Figure 2. The strongest regional correlations were generally observed between different regions measured within the same MRI phenotype, that is, Volume–Volume, Thickness–Thickness, GWC–GWC, or MD–MD correlations. Among these, GWC (Rx̄ = 0.782) showed the strongest average intra-phenotype correlations, followed by MD (Rx̄ = 0.374), thickness (Rx̄ = 0.360), and volume (Rx̄ = 0.313). In contrast, correlations between regions across different pheno- types were weaker. Consistent with the global analyses, the strongest inter-phenotype regional correlations were observed between MD and GWC (Rx̄ = −0.206), followed by volume and thickness (Rx̄ = 0.156). These results were highly comparable when also analyzing across the whole cohort (N = 1323) including participants with Aβ pathology and/or CDR above 0 (Figure S3), suggesting that AD pathology overall does not impact how these phenotypes relate to each other. They were ### Page 7 PIEPERHOFF ET AL. 7 of 17F I G U R E 1 Global averages of micro- and macrostructural cortex properties are differentially affected by demographic and clinical characteristics. Relations between global averages of MRI features representing macrostructural integrity (volume, thickness) and microstructural integrity (MD, GWC) are related to (A) age (independent of sex and APOE-ε4 carriership), (B) sex (independent of age and APOE-ε4 carriership), (C)APOE-ε4 (independent of age and sex), and independently of these factors are related to (D) CDR, (E) MMSE, (F) lateral ventricular volume, and (G) cerebral Aβ measured in Centiloid. Pearson correlation R and p-values are displayed for continuous predictors, Spearman rho and p-values for ranked predictors, and significance is indicated for group comparisons using independent sample t-tests. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Aβ, amyloid-beta; APOE, apolipoprotein E; CDR, Clinical Dementia Rating; GWC, gray–white matter contrast; MD, mean diffusivity; MMSE, Mini-Mental Status Examination. additionally highly comparable when analyzing in the subset of healthy participants with available MD (N = 315; Figure S4).3.4 Global Aβ burden is mostly related to boundary-based measures, less to volume and MDWe next examined to what extent regional measures of cortical micro- and macro-structural integrity are related to global Aβ bur- den measured with CL, while controlling for age, sex, CDR, and when applicable estimated intracranial volume. Regional volumes were cross-sectionally mostly unaffected, with minor Aβ-associated reduc- tions mainly in left and right inferior and middle temporal volume, as well as right fusiform, entorhinal, and inferior parietal volume (Figure 3A). Additionally, a mild increase in left pericalcarine volume in relation to Aβ was observed. Cortical thinning, however, was widely observed in relation to Aβ, most pronounced in the bilateral lateral temporal cortices, bilateral precunei, and bilateral entorhinal cortices (Figure 3B). Mild decreases in MD could be found only in the left postcen- tral gyrus (Figure 3C). Widespread Aβ-associated reductions were observed in GWC, with most brain regions being significantly affected. The strongest reductions were observed bilaterally in the parahip- pocampus, middle temporal, inferior temporal, entorhinal cortex, and precuneus (Figure 3D). Adjusting these analyses by MRI-PET time interval did not result in any changes. To assess whether these associations were independent of regional cortical thinning, we additionally corrected GWC for local cortical thickness in the same region. The pattern of Aβ-associated GWC reduc- tions was effectively preserved across all regions (Figure 3E), with the sole exception of left superior parietal GWC. ### Page 8 8 of 17 PIEPERHOFF ET AL.F I G U R E 2 Global and regional inter-phenotypic correlations. FDR-corrected significance-thresholded correlation matrices for regional and global metrics; Correlations that were not significant after FDR-correction are displayed in white. Top right diagonal: Regional correlations across 68 cortical regions of the gyrus-based Desikan–Killiany atlas show similar patterns, with evident left-right coupling (i.e., the same regions across hemispheres tend to correlate stronger than other regions) within each phenotype and for left-volume to right-thickness. Left bottom diagonal: Correlations of global averages (mean of all z-scaled regional phenotypes) of each phenotype and their binned p-values, with evident cross-phenotype correlations mainly between volume and thickness and GWC and mean diffusivity, and a small significant correlation between thickness and mean diffusivity. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. FDR, false discovery rate; GWC, gray–white matter contrast; MD, mean diffusivity. Repeating these analyses in the subset of participants with avail- able MD (N = 435), no significant effects of CL on volume or thickness remained (Figure S5A,B). GWC was predicted by CL in fewer regions (Figure S5D), comprising the bilateral lateral and medial temporal, pericalcarine gyri, as well as left precuneus and left rostral anterior and isthmus cingulate. Additionally adjusting GWC for thickness in this subset did not change any findings (Figure S5E).3.5 Alterations in GWC precede Aβ-positivityFinally, we estimated change-points in phenotype trajectories across estimated time to Aβ-positivity using a linear and non-linear change- point estimation (Table 3), exemplified for thickness of the precuneus in Figure 4A–F. Figure 4A shows the distribution of z-scaled data across time to Aβ-positivity, with its fitted piece-wise linear regression (Figure 4B) and the associated generative additive model (Figure 4C). Nonlinear models of the underlying z-scored data of all phenotypes (Figure 4D) are displayed in Figure 4E, with local absolute maxima indicating their change-points in Figure 4F. Both piece-wise linear and nonlinear methods overall provided highly similar change-points with a mean difference of 0.97 (± 2.32) years to Aβ-positivity across all phenotypes. Results of mean comparisons between nonlinear change- points for all phenotypes and regions are reported in Table 4. For all investigated ROIs, change-points of micro-structural properties preceded those of macro-structural properties. Across investigated regions and methods, GWC on average preceded Aβ-positivity by 7.69 (± 7.24) years, MD preceded Aβ-positivity by 2.83 (± 6.56) years, volume preceded Aβ-positivity by 1.99 (± 3.56) years, and thickness changes followed Aβ-positivity by 0.52 (± 4.78) years. More specifi- cally, GWC was the first changepoint across precuneus (M = −7.68, SD = 6.07), posterior cingulate (M = −6.96, SD = 8.26), entorhi- nal cortex (M = −7.40, SD = 7.22), superior-temporal (M = −7.5, SD = 5.45), and lateral-occipital cortex (M = −5.16, SD = 8.39), while MD was the first affected phenotype for the medial-orbitofrontal cortex (M = −8.024, SD = 7.89). Change-point analyses were also conducted in the subset of par- ticipants with available MD data (N = 435) to facilitate comparability across all metrics (Figure S6). The two methods to estimate change- points remained highly aligned with a change-point difference across the two methods of 0.93 (± 1.83) years to Aβ-positivity across all phe- notypes (Table S6). Across both methods, estimated change-points for GWC and volume remained highly comparable to the full dataset with a mean change-point of 6.62 (± 5.44) years prior to Aβ-positivity for GWC and 2.53 (± 5.22) years prior to Aβ-positivity for volume, whereas thickness was estimated notably earlier at 3.19 (± 5.28) years prior to Aβ-positivity. When assessing change-points in Centiloid while additionally cor- recting for age, we found a highly similar order of phenotype changes (Figure S7), indicating that change-points were not primarily driven by age; whilst GWC remained the earliest affected phenotype with a mean change-point of 2.66 (± 11.79) CL across ROIs, MD was the earliest phenotype to change in the entorhinal cortex. ### Page 9 PIEPERHOFF ET AL. 9 of 17TA B L E 3 Mean change-points of MRI features across time to Aβ-positivityPhenotype Precuneus Posterior cingulate Medial-orbitofrontal Entorhinal Superior-temporal Lateral-occipital PL NL PL NL PL NL PL NL PL NL PL NLVolume −2.47 ± 1.52 −1.67 ± 3.59 −0.80 ± 2.98 −1.54 ± 4.66 −2.61 ± 2.96 −2.33 ± 4.44 −2.47 ± 4.15 −3.19 ± 3.84 −2.23 ± 2.15 −0.09 ± 6.06 −2.47 ± 2.13 −1.99 ± 4.23 Thickness −2.47 ± 1.18 −3.49 ± 5.22 1.85 ± 2.83 4.68 ± 9.02 11.16 ± 3.70 −2.13 ± 7.08 −4.46 ± 4.27 −4.13 ± 5.11 −0.38 ± 1.77 1.03 ± 6.55 3.74 ± 2.24 0.78 ± 8.37 MD −7.24 ± 4.13 −1.81 ± 8.31 −9.01 ± 7.05 −6.37 ± 5.38 −9.03 ± 10.93 −7.02 ± 4.86 7.38 ± 9.00 1.24 ± 7.78 −4.45 ± 6.47 −5.53 ± 6.43 2.60 ± 1.69 5.23 ± 6.72 GWC −8.62 ± 5.12 −7.68 ± 6.07 −8.42 ± 10.31 −6.96 ± 8.26 −9.20 ± 7.29 −5.73 ± 6.86 −8.55 ± 5.89 −7.40 ± 7.22 −8.55 ± 6.37 −7.50 ± 5.45 −8.55 ± 9.62 −5.16 ± 8.39 Abbreviations: Aβ, amyloid β; GWC, gray–white matter contrast; MD, mean diffusivity; MRI, magnetic resonance imaging; NL, non-linear; PL, piecewise-linear.F I G U R E 3 Global Aβ burden is mostly related to boundary-based measures, less to volume and MD. Multivariate linear model t-values of significant effects (after FDR-correction) of global Aβ burden, measured with Centiloid, on local macrostructural [(A) volume and (B) thickness] and microstructural [(C) MD and (D) GWC] properties, corrected for age, sex, clinical dementia rating; and additionally corrected for estimated total intracranial volume when analysing volume. (E) Significant effects of global Aβ burden on GWC after correcting for regional thickness. Aβ, amyloid-beta; FDR, false discovery rate; GWC, gray–white matter contrast; MD, mean diffusivity.4 DISCUSSIONThis study examined the relationship between cortical micro- and macro-structural integrity measures and Aβ pathology in predomi- nantly cognitively unimpaired older adults. Our findings demonstrate that cortical integrity measures can be organized along a struc- tural dimensionality scale, with thickness and volume clustered as macrostructural measures and GWC and MD clustered as micro- structural measures, with unique sensitivities to demographic and clinical factors. Cortical thickness and GWC showed the strongest associations with Aβ burden, with both measures exhibiting signifi- cant Aβ-related alterations across wide-scale cortical regions, partic- ularly in medial temporal and parietal areas. Finally, cross-sectional change-point analyses showed that micro-structural alterations pre- cede macrostructural changes in the AD cascade, with GWC changes emerging up to nine years before Aβ-positivity, followed by MD changes around 3 years prior, while volume reductions occurred closer to Aβ-positivity and thickness changes emerging on average half a year after Aβ. ### Page 10 10 of 17 PIEPERHOFF ET AL.F I G U R E 4 Disease-related phenotype changes occur earlier for GWC than for macrostructural cortex measures. From (A) standardized, averaged phenotype data exemplified by thickness in the precuneus, change points are estimated through (B) piece-wise regression, or using (C) generative additive models: here, (D) underlying standardized phenotype data is first (E) continuously modelled, then inflection points of the first derivative are (F) estimated using local absolute maxima of the second derivative. (G) Forest plots of mean change-points with 95% confidence interval arms. Across selected ROIs, GWC and/or MD changes are consistently estimated to precede volume/thickness changes, with GWC overall being estimated earliest and thickness overall being estimated last. GWC, gray–white matter contrast; MD, mean diffusivity; ROI, regions of interest. ### Page 11 PIEPERHOFF ET AL. 11 of 17TA B L E 4 Mean differences in estimated change-points of MRI features across time to Aβ-positivityComparison Precuneus Post. cingulate Med. orbitofrontal Entorhinal Sup. temporal Lat. occipital∆ p ∆ p ∆ p ∆ p ∆ p ∆ pGWC > MD −4.83 <0.0001 −0.67 <0.05 1.04 <0.001 −10.56 <0.0001 −2.45 <0.0001 −8.42 <0.0001 GWC > Vol- ume−6.02 <0.0001 −5.41 <0.0001 −3.40 <0.0001 −4.21 <0.0001 −7.41 <0.0001 −3.17 <0.0001 GWC > Thick- ness−4.19 <0.0001 −11.64 <0.0001 −3.60 <0.0001 −3.27 <0.0001 −8.53 <0.0001 −5.95 <0.0001 MD > Volume −1.19 <0.0001 −4.75 <0.0001 −4.44 <0.0001 6.36 <0.0001 −4.97 <0.0001 5.26 <0.0001 MD > Thick- ness 0.64 <0.05 −10.97 <0.0001 −4.65 <0.0001 7.29 <0.0001 −6.08 <0.0001 2.48 <0.0001 Volume >Thickness 1.83 <0.0001 −6.22 <0.0001 −0.20 N.S. 0.94 <0.0001 −1.12 <0.0001 −2.78 <0.0001 Abbreviations: Aβ, amyloid β; GWC, gray–white matter contrast; MD, mean diffusivity; MRI, magnetic resonance imaging. To our knowledge, our change-point analyses provide the first evi- dence validating prior suggestions that micro-structural alterations reflect an earlier stage and possibly separate pathological mechanisms during the pathological cascade than cortical atrophy.15,58 This tem- poral precedence could explain why GWC shows larger effect sizes and broader spatial distribution than thickness in our and prior cross- sectional studies59; it captures an earlier phase of the AD cascade, when microstructural tissue properties are disrupted but gross mor- phometric changes have not yet accumulated. As such, GWC may represent a viable outcome marker in clinical trials targeting preclinical populations.4.1 Distinct micro- and macro-structural mechanisms underlying cortical integrityAmong macrostructural measures of cortical integrity, cortical thick- ness has been previously found to outperform volume in predicting mild cognitive impairment or AD.60 Cortical thickness is highly heri- table but minimally genetically correlated with surface area, despite a substantial amount of overlap of causal variants61,62; hence, cor- tical volume as the combination of these two measures should be related to but still vary considerably from thickness, and has been found to be substantially more polygenic in nature.62 This explains the modest correlation we observed between volume and thickness, which likely reflects their partially overlapping yet distinct genetic and neuroanatomical substrates. While both capture aspects of cor- tical macrostructure, thickness uniquely reflects laminar organization and neuronal packing density including the number of cells within a column.63 Similarly, the microstructural measures GWC and MD, while strongly intercorrelated, are considered to represent distinct biolog- ical processes. MD in the cortical GM is thought to reflect cellular microstructure, with increased diffusivity potentially indicating expan- sion of extracellular space due to neuronal loss, astrocytosis, or inflammatory changes.12,64 A twin study demonstrated that cortical MD across the cortex is significantly heritable, yet shares minimal genetic variance with both cortical thickness and WM MD in most brain regions, in line with our findings of overall minimal regional correlations between thickness and MD.12 This genetic independence reinforces the idea that cortical MD captures unique aspects of tissue integrity beyond what is measured by morphological or GM–WM boundary- based approaches. The mechanisms underlying cortical MD alterations potentially include neuroinflammation, breakdown of cellular mem- branes, shifts in water distribution between intra- and extracellular compartments, and changes in myelinated fibers coursing through the cortical mantle.10–12,65 The relative lack of cross-sectional association between cortical MD and Aβ may be explained by its biphasic trajectory along the Alzheimer’s proteinopathic cascade, resulting in an averag- ing out of linear associations.7 However, a biphasic trajectory is also supported by our change-point models, as both change-point methods estimated robust change-points for MD trajectories along estimated time to Aβ-positivity.4.2 Genetic and pathological correlates of GWC and thicknessGenetic studies have found that both GWC and thickness are signifi- cantly heritable while sharing relatively minimal genetic variance, with only 17 of 66 cortical regions showing mild genetic correlations.66 These findings align with our low regional correlations between GWC and thickness, which is consistent with prior studies.20 Similarly, a genome-wide association study identified 251 SNPs associated with either cortical thickness or GWC, only 42 of which were shared between both phenotypes.67 Genes associated with thickness included those regulating neuronal architecture and connectivity, whereas GWC-associated genes were involved in myelin integrity and tissue repair mechanisms. If GWC is more strongly influenced by genes reg- ulating myelin integrity and cellular homeostasis, processes which may be disrupted early in response to Aβ accumulation,22,23,68,69 this could account for its earlier sensitivity in the AD cascade. Also consistent ### Page 12 12 of 17 PIEPERHOFF ET AL.with this genetic evidence, global amyloid burden remained predic- tive of wide scale decreases in GWC when additionally correcting for local thickness, indicating that the association between AD pathology and GWC is unique from that with thickness. Notably, two large-scale genome-wide association studies have independently identified PLD1as strongly associated with GWC across the cortex, predominantly in temporal, parietal, and prefrontal regions.70,71 PLD1 encodes phospho- lipase D1, a lipid-modulating enzyme that has been shown to impact synaptic plasticity and hippocampal functioning,72 as well as regula- tion of Aβ precursor protein trafficking,73 with increased expression in mitochondrial membranes in AD brains.74 This convergent genetic evidence linking PLD1 to GWC variations suggests that GWC may be particularly sensitive to disruptions in lipid metabolism and membrane integrity which occur early in the AD cascade. In atypical AD patients, GWC has previously shown larger effect sizes and explained additional variance in tau-PET than cortical thick- ness across widespread cortical regions and was abnormal in prefrontal areas where cortical atrophy was minimal, suggesting GWC may detect early pathological changes preceding visible atrophy.59 Consistent with this prior evidence, when re-analyzing the effects of global amy- loid burden in the subset of participants with available MD, effects of amyloid on volume and thickness vanished whereas GWC remained largely predicted by amyloid, hinting that GWC is more affected by and therefore more sensitive to amyloid pathology.4.3 LimitationsSome limitations should be considered when interpreting these find- ings. First, the limited availability of diffusion MRI substantially reduced statistical power for all MD analyses and broadened confi- dence intervals, although the subset remained adequately powered. This limited power is nonetheless an important consideration for com- parability with the other phenotypes, as we found no significant effects of amyloid on thickness when analyzing in this subset, which were oth- erwise abundant. Second, diverse MRI protocols were used across sites in this multi-center study. While we aimed to eliminate site-related bias through statistical harmonization, the harmonization procedure may have differential effects on the various cortical integrity metrics. Indeed, supplementary analyses suggest that volume and thickness may have been harmonized more consistently than MD and GWC, potentially introducing unequal degrees of measurement variance, potentially impacting the relative strength of associations observed between measures and pathology. Third, while our predominantly cog- nitively unimpaired sample provides insights into preclinical changes, the modelled age and disease range capture a specific window of the AD cascade. Relationships between measures and their temporal dynamics may differ in younger or older cohorts, or in cohorts with more advanced pathology. Fourth, we did not account for neurofibril- lary tangle load or vascular (co-)pathology in our analyses. Whereas we analyzed a preclinical cohort where at least overt consequences of tau burden are of low magnitude, high prevalence of neocortical tau has been reported in individuals with 40 to 70 CL,75 and vascular pathol- ogy, in particular WM hyperintensities, are highly frequent in older adults but potentially contribute toward AD progression.76 As such, our analyses of amyloid burden measured by CL are best understood as reflecting generalized AD progression beyond just amyloidosis. Finally, while our change-point estimations suggest a distinct temporal ordering of AD-related phenotypic changes, the reported confidence intervals were also quite large and are of a cross-sectional nature, therefore this temporal ordering should be interpreted with caution and requires longitudinal validation. Change-point analyses especially in preclinical disease stages require large studies, and though our present analyses were comparable or larger to prior cross-sectional studies,77–79 future pooled cohort studies hold promise in providing additional statistical power.4.4 Future directions and clinical implicationsIn summary, the temporal dissociation we identified with GWC pre- ceding thickness by approximately 8 years relative to Aβ-positivity, suggests that GM–WM boundary-based micro- and macro-structural measures reflect distinct pathological changes and may provide unique information when measured at the same time-point. GWC appears to be more sensitive to early microstructural alterations potentially reflecting synaptic loss, gliosis, or myelin changes at the boundary, while thickness reductions reflect later-stage neuronal loss and atro- phy. The convergence of genetic independence, distinct molecular pathways, and differential temporal trajectories collectively supports a model in which GWC and cortical thickness represent comple- mentary but biologically distinct stages of cortical integrity changes during neurodegeneration, each capturing different mechanisms of the disease process. Similar to recent frameworks proposing MRI-based dimensions of cortical thinning as latent vulnerability signatures,80 our results suggest that even in the absence of overt pathology, individu- als may exhibit distinct patterns of microstructural vulnerability that predispose them toward specific disease trajectories. Future research should investigate whether individual GWC tra- jectories predict subsequent Aβ accumulation and cognitive decline, and whether subtype classifications based on regional vulnerabil- ity patterns such as limbic-predominant versus hippocampal-sparing subtypes emerge at the microstructural level before manifesting as atrophy.80 Integrating multimodal imaging, including functional mea- sures at the GM-WM boundary, with genetic risk profiles and fluid biomarkers may enable more precise individual risk prediction and personalized intervention strategies in preclinical stages of AD.AFFILIATIONS1Department of Radiology and Nuclear Medicine, Vrije Universiteit Amsterdam, Amsterdam University Medical Center, Amsterdam, the Netherlands 2Brain imaging, Amsterdam Neuroscience, Amsterdam, the Netherlands 3Department of Neuroscience, Rehabilitation, Ophthalmology, Genetics, Mater- nal and Child Health (DiNOGMI), University of Genoa, Genoa, Italy 4Life and Health Sciences Research Institute (ICVS), School of Medicine, Univer- sity of Minho, Braga, Portugal ### Page 13 PIEPERHOFF ET AL. 13 of 175Department of Neuroradiology, Hospital de Braga, ULS Braga, Braga, Portugal 6Department of Anatomy and Neurosciences, Amsterdam UMC location VUmc, Amsterdam, the Netherlands 7Neurodegeneration, Amsterdam Neuroscience, Amsterdam, the Netherlands 8Department of Advanced Biomedical Sciences, University of Naples “Federico II”, Naples, Italy 9GE HealthCare, Amersham, UK 10Department of Medical Physics and Biomedical Engineering, UCL, London, UK 11“L. Sacco” Department of Biomedical and Clinical Sciences, University of Milan, Milan, Italy 12IXICO plc London, London, UK 13Scottish Brain Sciences, Edinburgh, UK 14Ace Alzheimer Center Barcelona, Universitat Internacional de Catalunya, Barcelona, Spain 15Networking Research Center on Neurodegenerative Diseases (CIBERNED), Instituto de Salud Carlos III, Madrid, Spain 16Alzheimer Center Amsterdam, Neurology, Amsterdam UMC, Amsterdam, the Netherlands 17Laboratory for Cognitive Neurology, KU Leuven, Leuven, Belgium 18Department of Neurology, Cliniques Universitaires Saint-Luc, Brussels, Bel- gium 19Institute of Neuroscience, Université catholique de Louvain, Brussels, Belgium 20Center for Research and Advanced Therapies. CITA Alzheimer Foundation, Donostia-San Sebastián, Spain 21ToNIC Toulouse NeuroImaging Center, Université Toulouse 3, Inserm, Toulouse, France 22Department of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden 23Wallenberg Centre for Molecular and Translational Medicine, University of Gothenburg, Gothenburg, Sweden 24Dementia Research Centre, Institute of Neurology, University College London, London, UK 25Hôpitaux Universitaires de Genève, Geneva, Switzerland 26IRCCS Ospedale Policlinico San Martino, Genoa, Italy 27Department of Health Sciences (DISSAL), University of Genoa, Genoa, Italy 28Department of Neuroradiology, IRCCS Ospedale Policlinico San Martino, Genoa, Italy 29Life Molecular Imaging GmbH, Berlin, Germany 30Department of Psychiatry and Psychotherapy, University Hospital Cologne, Cologne, Germany 31Barcelonaβeta Brain Research Center, Pasqual Maragall Foundation, Barcelona, Spain 32Clinical Memory Research Unit, Department of Clinical Sciences in Malmö, Lund University, Lund, Sweden 33Laboratory for Complex Genetics, KU Leuven, Leuven, Belgium 34ICVS/3B’s - PT Government Associate Laboratory, University of Minho, Braga, PortugalACKNOWLEDGMENTSThis work used data from AMYPAD-PNHS (Amyloid Imaging to Pre- vent Alzheimer’s Disease–prognostic and natural history study). The authors express their most sincere gratitude to the AMYPAD partic- ipants, without whom this research would not have been possible. AMYPAD received funding from the Innovative Medicines Initiative (IMI) 2 Joint Undertaking under grant agreement (grant number 115952). This Joint Undertaking receives support from the European Union’s Horizon 2020 Research and Innovation Programme and EFPIA. This communication reflects the views of the authors and neither IMI nor the European Union and EFPIA are liable for any use that may be made of the information contained herein. This project has received funding support from Alzheimer Nederland (grant number WE.03- 2024-27). This study was supported by the Alzheimer’s Disease Data Initiative (ADDI).CONFLICT OF INTEREST STATEMENTTiago Gil Oliveira has been a consultant for Sonae, Guidepoint, and Lilly, has received fees as a speaker from Eisai, and conference fees covered from Roche. Frederik Barkhof is supported by Engineering and Physical Sciences Research Council (EPSRC), EUJU (IMI), National Institute for Health and Care Research—Biomedical Research Center (NIHR-BRC), General Electric (GE) HealthCare; he is a consultant for Combinostics, IXICO, and Roche; participates on advisory boards of Biogen, Prothena, and Merck; and is a co-founder of Queen Square Analytics. Lyduine E. Collij has received research support and speakers fee from GE Health- Care Ltd. and Springer Healthcare (paid to institution).Gemma Salvadó has received speaker or advisory fees from Springer, GE Healthcare, Biogen, Esteve, Adium and Johnson&Johnson. Michael Schöll receives funding from the Knut and Alice Wallenberg Foundation (Wallenberg Centre for Molecular and Translational Medicine; KAW2014.0363 and KAW2023.0371), the Swedish Research Council (2017-02869, 2021-02678, 2021-06545 and 2023-06188), the European Union’s Horizon Europe research and innovation program under grant agree- ment no 101132933 (AD-RIDDLE) and 101112145 (PROMINENT), the National Institute of Health (R01 AG081394-01), Gates Ventures, the National Research Foundation of Korea (RS-2023-00263612), the Swedish state under the agreement between the Swedish govern- ment and the County Councils, the ALF-agreement (ALFGBG-813971 and ALFGBG-965326), the Swedish Brain Foundation (FO2021-0311), the Swedish Alzheimer Foundation (AF-994900), the Sahlgrenska Academy at the University of Gothenburg, the Västra Götaland Region R&D (VGFOUREG-995510), and Innovation platforms, Sahlgrenska Science Park and the National Institute for Health and Care Research University College London Hospitals Biomedical Research Centre. M.S. has served on advisory boards for Roche and Novo Nordisk, received speaker honoraria from Bioarctic, Eisai, Genentech, Novo Nordisk, and Roche and receives research support (to the institution) from Alz- path, Bioarctic, Novo Nordisk, and Roche (outside scope of submitted work). He is a co-founder of Centile Bioscience and serves as asso- ciate editor with Alzheimer’s Research & Therapy. Pieter Jelle Visser serves as an advisory board member of Eli Lilly and is consultant for Janssen Pharmaceutical. He receives/received research grants from Bristol-Myers Squibb and GE Healthcare, European Commission 6th and 7th Framework programme, the Innovative Medicines Initiative (IMI), European Union Joint Programme–Neurodegenerative Disease Research (JPND), and Zon-Mw. Pierre Payoux received research sup- port and consultancy fees from GE Healthcare.Pablo Martinez-Lage ### Page 14 14 of 17 PIEPERHOFF ET AL.has received honoraria for participating in Advisory Boards and/or edu- cational events, or funding to attend meetings and conferences from Lilly, Nutricia, Roche, Eisai, Grifols, and Esteve. Bernard J. Hanseeuw is consultant to Biogen, Eisai, and Roche. Consulting fees have been paid to his institution. Mercè Boada has consulted for Grifols, Araclon Biotech, Roche, Biogen, Lilly, Merck, and Novo-Nordisk; has served in the Advisory Boards from Grifols, Roche, Lilly, Araclon Biotech, Merck, Biogen, Novo-Nordisk, Bioiberica, Eisai, Servier, and Schwabe Pharma; received fees from lectures from Roche, Biogen, Grifols, Nutricia, Araclon Biotech, Novo-Nordisk, Eisai, Terumo, and Schwabe Pharma; and reports research funding from Life Molecular Imaging, Bioiberica, Grifols, Araclon Biotech, Lilly, Roche, Janssen, Alzehon, Cortyzime, Novo Nordisk, and Schwabe Pharma. M.B. received fund- ing from CIBERNED (Instituto de Salud Carlos III (ISCIII); EU/EFPIA Innovative Medicines Initiative Joint Undertaking, ADAPTED Grant No. 115975; EXIT project, EU Euronanomed3 Program JCT2017 Grant No. AC17/00100; MOPEAD, Innovative Medicine Initiative, Grant. No. 115985; PreDADQoL, ERA-NET (call 2015). Grant no. AC15/00082; TARTAGLIA (Red federada para accelerar la aplicación de la inteligencia artificial en el sistema sanitario español); PREADAPT project, Joint Program for Neurodegenerative Diseases (JPND) Grant No. AC19/00097; GECONEU Grant No. 2023-1-ELO1-KAZZ0-HED- 000032173 co-founded by the European Union; Grants PI13/02434, PI16/01861, BA19/00020, and PI19/01301 from the Acción Estratég- ica en Salud, integrated in the Spanish National RCDCI Plan and financed by Instituto de Salud Carlos III (ISCIII)- Subdirección General de Evaluación and the Fondo Europeo de Desarrollo Regional (FEDER – “Una manera de Hacer Europa”); Fundació “La Caixa” and Grífols (GR@ACE project); and Proyectos de Investigación de Medicina Per- sonalizada (ISCIII), PMP-DEGESCO, Grant N◦ PMP22/00022. Marta Marquié has consulted for F. Hoffmann-La Roche Ltd. and has served in the Spanish Scientific Advisory Board for biomarkers of Araclon Biotech. G.S. has received speaker fees from Springer and Adium. M.M. has received funding support from the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skłodowska- Curie grant agreement no. 796706 and the Instituto de Salud Carlos III (ISCIII) Acción Estratégica en Salud, integrated in the Spanish National RCDCI Plan and financed by ISCIII-Subdirección General de Evalu- ación and the Fondo Europeo de Desarrollo Regional (FEDER—Una manera de hacer Europa) grant PI19/00335. Craig Ritchie is the major- ity shareholder, CEO, and Founder of Scottish Brain Sciences. C.R. has received consultancy fees from Biogen, Eisai, MSD, Actinogen, Roche, and Eli Lilly, as well as payment or honoraria from Roche and Eisai. Robin Wolz and Sylke Grootoonk are full-time employees of IXICO plc. Ariane Bollack, Chris Buckley and Gill Farrar are full- time employees of GE HealthCare. Luigi Lorenzini receives funding from the MSCA postdoctoral fellowship (#101204296) and Alzheimer Nederland (WE.03-2025-02). Alle Meije Wink, Henk-Jan M.M. Mut- saerts, Emma S. Luckett, Gemma Salvadó, Juan Domingo Gispert, Frank Jessen, Andrew W. Stephens, Luca Roccataglia, Matteo Par- dini, Giovanni B. Frisoni, Rik Vandenberghe, Jelle Visser, Craig Ritchie, Federico Masserini, Prithvi Arunachalam, Mario Tranfa, and Francisco Almeida report nothing to declare. Leonard Pieperhoff receives fund- ing from Alzheimer Nederland (WE.03-2024-27). Author disclosures are available in the Supporting Information.CONSENT STATEMENTThe studies were conducted according to the principles of the Helsinki Declaration of 1975, as revised in 2008, and all human participants gave written informed consent.ORCIDLeonard Pieperhoff https://orcid.org/0009-0003-9003-4588Luigi Lorenzini https://orcid.org/0000-0002-9756-881XFrancisco Almeida https://orcid.org/0000-0002-8404-4136Mario Tranfa https://orcid.org/0000-0002-4451-4746Ariane Bollack https://orcid.org/0000-0002-9169-7530Prithvi Arunachalam https://orcid.org/0000-0002-0798-6250Federico Masserini https://orcid.org/0000-0002-0746-7426Sylke Grootoonk https://orcid.org/0009-0009-1140-8283Craig Ritchie https://orcid.org/0000-0002-6202-6906Mercè Boada https://orcid.org/0000-0003-2617-3009Marta Marquié https://orcid.org/0000-0002-0660-0950Rik Vandenberghe https://orcid.org/0000-0001-6237-2502Bernard J. Hanseeuw https://orcid.org/0000-0002-3102-6778Pablo Martinez-Lage https://orcid.org/0000-0001-5404-9967Pierre Payoux https://orcid.org/0000-0002-1374-1620Pieter Jelle Visser https://orcid.org/0000-0001-8008-9727Michael Schöll https://orcid.org/0000-0001-7800-1781Giovanni B. Frisoni https://orcid.org/0000-0001-7075-7082Matteo Pardini https://orcid.org/0000-0002-4740-1982Juan Domingo Gispert https://orcid.org/0000-0002-6155-0642Gemma Salvadó https://orcid.org/0000-0002-5210-9230Emma S. Luckett https://orcid.org/0000-0003-1777-4742Henk-Jan MM Mutsaerts https://orcid.org/0000-0003-0894-0307Alle Meije Wink https://orcid.org/0000-0002-8197-0118Lyduine E. Collij https://orcid.org/0000-0001-6263-1762Frederik Barkhof https://orcid.org/0000-0003-3543-3706Tiago Gil Oliveira https://orcid.org/0000-0003-0509-0562REFERENCES1. Therriault J, Schindler SE, Salvadó G, et al. Biomarker-based staging of Alzheimer disease: rationale and clinical applications. Nat Rev Neurol. 2024;20(4):232-244. doi:10.1038/s41582-024-00942-2 2. Schwarz AJ, Sundell KL, Charil A, et al. Magnetic resonance imag- ing measures of brain atrophy from the EXPEDITION3 trial in mild Alzheimer’s disease. Alzheimers Dement (N Y). 2019;5(1):328-337. doi:10.1016/j.trci.2019.05.007 3. Chételat Gaël, Villemagne VL, Pike KE, et al. Larger temporal vol- ume in elderly with high versus low beta-amyloid deposition. Brain. 2010;133(11):3349-3358. doi:10.1093/brain/awq187 4. Ingala S, De Boer C, Masselink LA, et al. Application of the ATN clas- sification scheme in a population without dementia: findings from the EPAD cohort. Alzheimers Dement. 2021;17(7):1189-1204. doi:10. 1002/alz.12292 5. Qiu T, Liu ZQ, Gallego-Rudolf J, et al. Tracking a decade of structural changes in preclinical and prodromal Alzheimer’s disease: insights from amyloid-β pathology. medRxiv. 2025:1-42. doi:10.1101/2025.09. 25.25336640 ### Page 15 PIEPERHOFF ET AL. 15 of 176. Montal V, Vilaplana E, Alcolea D, et al. Cortical microstructural changes along the Alzheimer’s disease continuum. Alzheimers Dement. 2018;14(3):340-351. doi:10.1016/j.jalz.2017.09.013 7. Montal V, Vilaplana E, Pegueroles J, et al. Biphasic cortical macro- and microstructural changes in autosomal dominant Alzheimer’s disease.Alzheimers Dement. 2021;17(4):618-628. doi:10.1002/alz.12224 8. Salvadó G, Shekari M, Falcon C, et al. Brain alterations in the early Alzheimer’s continuum with amyloid-β, tau, glial and neurodegener- ation CSF markers. Brain Commun. 2022;4(3):fcac134. doi:10.1093/ braincomms/fcac134 9. Povala G, Bellaver B, Bastiani MAôDe, et al. Amyloid beta and tau are associated with the dual effect of neuroinflammation on neurode- generation. Alzheimers Dement. 2025;21(10):e70746. doi:10.1002/alz. 70746 10. Vilaplana E, Rodriguez-Vieitez E, Ferreira D, et al. Cortical microstructural correlates of astrocytosis in autosomal-dominant Alzheimer disease. Neurology. 2020;94(19):e2026-e2036. doi:10.1212/WNL.0000000000009405 11. Sampedro F, Kulisevsky J. Intracortical surface-based MR diffusiv- ity to investigate neurologic and psychiatric disorders: a review. J Neuroimaging. 2022;32(1):28-35. doi:10.1111/jon.12930 12. Elman JA, Panizzon MS, Hagler DJ, et al. Genetic and environmental influences on cortical mean diffusivity. Neuroimage. 2017;146:90-99. doi:10.1016/j.neuroimage.2016.11.032 13. Li M, Xu L, Choi S, et al. Functional contrast across the gray-white mat- ter boundary. Nat Commun. 2025;16(1):6077. doi:10.1038/s41467- 025-61251-w 14. Spotorno N, Strandberg O, Stomrud E, et al. Diffusion MRI tracks cor- tical microstructural changes during the early stages of Alzheimer’s disease. Brain. 2024;147(3):961-969. doi:10.1093/brain/awad428 15. Weston PSJ, Poole T, Nicholas JM, et al. Measuring cortical mean diffusivity to assess early microstructural cortical change in presymptomatic familial Alzheimer’s disease. Alzheimers Res Ther. 2020;12(1):112. doi:10.1186/s13195-020-00679-2 16. Torso M, Bozzali M, Zamboni G, Jenkinson M, Chance SA. Alzheimers Disease Neuroimage Initiative. Detection of Alzheimer’s disease using cortical diffusion tensor imaging. Hum Brain Mapp. 2021;42(4):967- 977. doi:10.1002/hbm.25271 17. Sun P, He Z, Li A, et al. Spatial and temporal patterns of cortical mean diffusivity in Alzheimer’s disease and suspected non-Alzheimer’s disease pathophysiology. Alzheimers Dement. 2024;20(10):7048-7061. doi:10.1002/alz.14176 18. Salat DH, Lee SY, van der Kouwe AJ, Greve DN, Fischl B, Rosas HD. Age-associated alterations in cortical gray and white mat- ter signal intensity and gray to white matter contrast. Neuroimage. 2009;48(1):21-28. doi:10.1016/j.neuroimage.2009.06.074 19. Vidal-Piñeiro D, Walhovd KB, Storsve AB, Grydeland Hå, Rohani DA, Fjell AM. Accelerated longitudinal gray/white matter contrast decline in aging in lightly myelinated cortical regions. Hum Brain Mapp. 2016;37(10):3669-3684. doi:10.1002/hbm.23267 20. Parent O, Olafson E, Bussy Aé, et al. High spatial overlap but diverging age-related trajectories of cortical magnetic resonance imaging mark- ers aiming to represent intracortical myelin and microstructure. Hum Brain Mapp. 2023;44(8):3023-3044. doi:10.1002/hbm.26259 21. Cai Z, Xiao M. Oligodendrocytes and Alzheimer’s disease. Int J Neurosci. 2016;126(2):97-104. doi:10.3109/00207454.2015.1025778 22. Huang Z, Jordan JD, Zhang Q. Myelin pathology in Alzheimer’s disease: potential therapeutic opportunities. Aging Dis. 2024;15(2):698-713. doi:10.14336/AD.2023.0628 23. Maitre M, Jeltsch-David H, Okechukwu NG, Klein C, Patte-Mensah C, Mensah-Nyagan AG. Myelin in Alzheimer’s disease: culprit or bystander?. Acta Neuropathol Commun. 2023;11(1):56. doi:10.1186/ s40478-023-01554-5 24. Salat DH, Chen JJ, van der Kouwe AJ, Greve DN, Fischl B, Rosas HD. Hippocampal degeneration is associated with temporal and lim- bic gray matter/white matter tissue contrast in Alzheimer’s disease.Neuroimage. 2011;54(3):1795-1802. doi:10.1016/j.neuroimage.2010. 10.034 25. Flaherty R, Krivensky S, Brickhouse M, et al. The relationship between cortical grey to white matter signal contrast (GWC) and tau pathology in atypical AD. Alzheimers Dement. 2021;17(Suppl 4):e055799. doi:10. 1002/alz.055799 26. Jefferson AL, Gifford KA, Damon S, et al. Gray & white matter tissue contrast differentiates mild cognitive impairment converters from non-converters. Brain Imaging Behav. 2015;9(2):141-148. doi:10. 1007/s11682-014-9291-2 27. García DV, AMYPAD PNHS—Integrated dataset (Raw) v202306. Pub- lished online June 12, 2023. doi:10.5281/ZENODO.8017084 28. Lopes Alves I, Collij LE, Altomare D, et al. Quantitative amyloid PET in Alzheimer’s disease: the AMYPAD prognostic and natural his- tory study. Alzheimers Dement. 2020;16(5):750-758. doi:10.1002/alz. 12069 29. Collij LE, Farrar G, Valléz García D, et al. The amyloid imaging for the prevention of Alzheimer’s disease consortium: a European collabora- tion with global impact. Front Neurol. 2022;13:1063598. doi:10.3389/ fneur.2022.1063598 30. Reuter M, Schmansky NJ, Rosas HD, Fischl B. Within-subject tem- plate estimation for unbiased longitudinal image analysis. Neuroimage. 2012;61(4):1402-1418. doi:10.1016/j.neuroimage.2012.02.084 31. Raamana PR, Theyers A, Selliah T, et al. Visual QC protocol for FreeSurfer cortical parcellations from anatomical MRI. Aper- ture Neuro. 2022;3:1-22. doi:10.52294/1cdce19c-e6db-4684-97cb- ae709da06a3f 32. Kong Li, Herold C, Stieltjes B, et al. Reduced gray to white matter tis- sue intensity contrast in schizophrenia. PLoS One. 2012;7(5):e37016. doi:10.1371/journal.pone.0037016 33. Desikan RS, Ségonne F, Fischl B, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006;31(3):968-980. doi:10. 1016/j.neuroimage.2006.01.021 34. Cieslak M, Cook PA, He X, et al. QSIPrep: an integrative platform for preprocessing and reconstructing diffusion MRI data. Nat Methods. 2021;18(7):775-778. doi:10.1038/s41592-021-01185-5 35. Henf J, Grothe MJ, Brueggen K, Teipel S, Dyrba M. Mean diffusivity in cortical gray matter in Alzheimer’s disease: the importance of partial volume correction. NeuroImage Clin. 2018;17:579-586. doi:10.1016/j. nicl.2017.10.005 36. Fortin J-P, Cullen N, Sheline YI, et al. Harmonization of corti- cal thickness measurements across scanners and sites. Neuroimage. 2018;167:104-120. doi:10.1016/j.neuroimage.2017.11.024 37. Fortin J-P, Parker D, Tunç B, et al. Harmonization of multi-site diffusion tensor imaging data. Neuroimage. 2017;161:149-170. doi:10.1016/j. neuroimage.2017.08.047 38. Hu F, Chen AA, Horng H, et al. Image harmonization: a review of statistical and deep learning methods for removing batch effects and evaluation metrics for effective harmonization. Neuroimage. 2023;274:120125. doi:10.1016/j.neuroimage.2023.120125 39. Barthel H, Gertz H-J, Dresel S, et al. Cerebral amyloid-β PET with florbetaben (18F) in patients with Alzheimer’s disease and healthy controls: a multicentre phase 2 diagnostic study. Lancet Neurol. 2011;10(5):424-435. doi:10.1016/S1474-4422(11)70077-1 40. Curtis C, Gamez JE, Singh U, et al. Phase 3 trial of flutemetamol labeled with radioactive fluorine 18 imaging and neuritic plaque density. JAMA Neurol. 2015;72(3):287-294. doi:10.1001/jamaneurol.2014.4144 41. Wolz R, Aljabar P, Hajnal JV, Hammers A, Rueckert D, Alzheimer’s Disease Neuroimaging Initiative. LEAP: learning embeddings for ### Page 16 16 of 17 PIEPERHOFF ET AL.atlas propagation. Neuroimage. 2010;49(2):1316-1325. doi:10.1016/j. neuroimage.2009.09.069 42. Klunk WE, Koeppe RA, Price JC, et al. The Centiloid Project: stan- dardizing quantitative amyloid plaque estimation by PET. Alzheimers Dement. 2015;11(1):1-15.e154. doi:10.1016/j.jalz.2014.07.003 43. Betthauser TJ, Bilgel M, Koscik RL, et al. Multi-method investigation of factors influencing amyloid onset and impairment in three cohorts.Brain. 2022;145(11):4065-4079. doi:10.1093/brain/awac213 44. Bateman RJ, Smith J, Donohue MC, et al. Two phase 3 trials of gantenerumab in early Alzheimer’s disease. N Engl J Med. 2023;389(20):1862-1876. doi:10.1056/NEJMoa2304430 45. Collij LE, Bollack A, La Joie R, et al. Centiloid recommendations for clin- ical context-of-use from the AMYPAD consortium. Alzheimers Dement. 2024;20(12):9037-9048. doi:10.1002/alz.14336 46. Vlassenko AG, Mintun MA, Xiong C, et al. Amyloid-beta plaque growth in cognitively normal adults: longitudinal [11C]Pittsburgh compound B data. Ann Neurol. 2011;70(5):857-861. doi:10.1002/ana.22608 47. Buckley RF, Mormino EC, Amariglio RE, et al. Sex, amyloid, and APOEε4 and risk of cognitive decline in preclinical Alzheimer’s disease: findings from three well-characterized cohorts. Alzheimers Dement. 2018;14(9):1193-1203. doi:10.1016/j.jalz.2018.04.010 48. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc. 1995;57(1):289-300. doi:10.1111/j.2517-6161.1995.tb02031.x 49. Palmqvist S, Schöll M, Strandberg O, et al. Earliest accumulation ofβ-amyloid occurs within the default-mode network and concurrently affects brain connectivity. Nat Commun. 2017;8(1):1214. doi:10.1038/ s41467-017-01150-x 50. Braak H, Braak E. Neuropathological stageing of Alzheimer- related changes. Acta Neuropathol. 1991;82(4):239-259. doi:10.1007/bf00308809 51. Thal DR, Rüb U, Orantes M, Braak H. Phases of A beta-deposition in the human brain and its relevance for the development of AD. Neurology. 2002;58(12):1791-1800. doi:10.1212/wnl.58.12.1791 52. Mattsson N, Palmqvist S, Stomrud E, Vogel J, Hansson O. Staging β- amyloid pathology with amyloid positron emission tomography. JAMA Neurol. 2019;76(11):1319-1329. doi:10.1001/jamaneurol.2019.2214 53. Collij LE, Heeman F, Salvadó G, et al. Multitracer model for staging cortical amyloid deposition using PET imaging. Neurology. 2020;95(11):e1538-e1553. doi:10.1212/WNL.0000000000010256 54. Muggeo VMR. Estimating regression models with unknown break- points. Stat Med. 2003;22(19):3055-3071. doi:10.1002/sim.1545 55. Rigby RA, Stasinopoulos DM. Generalized additive models for loca- tion, scale and shape. J R Stat Soc Ser C Appl Stat. 2005;54(3):507-554. doi:10.1111/j.1467-9876.2005.00510.x 56. Wood SN. Generalized Additive Models: An Introduction With R. Chapman and Hall/CRC; 2017. doi:10.1201/9781315370279 57. Davison AC, Hinkley DV. Cambridge Series in Statistical and Probabilistic Mathematics: Bootstrap Methods and Their Application Series Number 1. Cambridge University Press; 2013. doi:10.1017/cbo9780511802843 58. Yoon B, Shim Y-S, Hong Y-J, et al. Comparison of diffusion tensor imag- ing and voxel-based morphometry to detect white matter damage in Alzheimer’s disease. J Neurol Sci. 2011;302(1-2):89-95. doi:10.1016/j. jns.2010.11.012 59. Putcha D, Katsumi Y, Brickhouse M, et al. Gray to white matter sig- nal ratio as a novel biomarker of neurodegeneration in Alzheimer’s disease. NeuroImage Clin. 2023;37:103303. doi:10.1016/j.nicl.2022. 103303 60. Schwarz CG, Gunter JL, Wiste HJ, et al. A large-scale comparison of cortical thickness and volume methods for measuring Alzheimer’s dis- ease severity. NeuroImage Clin. 2016;11:802-812. doi:10.1016/j.nicl. 2016.05.017 61. Panizzon MS, Fennema-Notestine C, Eyler LT, et al. Distinct genetic influences on cortical surface area and cortical thickness. Cereb Cortex. 2009;19(11):2728-2735. doi:10.1093/cercor/bhp026 62. van der Meer D, Frei O, Kaufmann T, et al. Quantifying the poly- genic architecture of the human cerebral cortex: extensive genetic overlap between cortical thickness and surface area. Cereb Cortex. 2020;30(10):5597-5603. doi:10.1093/cercor/bhaa146 63. Rakic P. Specification of cerebral cortical areas. Science. 1988;241(4862):170-176. doi:10.1126/science.3291116 64. Van Camp N, Blockx I, Camón L, et al. A complementary diffusion ten- sor imaging (DTI)-histological study in a model of Huntington’s disease.Neurobiol Aging. 2012;33(5):945-959. doi:10.1016/j.neurobiolaging. 2010.07.001 65. Bartzokis G. Alzheimer’s disease as homeostatic responses to age- related myelin breakdown. Neurobiol Aging. 2011;32(8):1341-1371. doi:10.1016/j.neurobiolaging.2009.08.007 66. Panizzon MS, Fennema-Notestine C, Kubarych TS, et al. Genetic and environmental influences of white and gray matter signal contrast: a new phenotype for imaging genetics?. Neuroimage. 2012;60(3):1686- 1695. doi:10.1016/j.neuroimage.2012.01.122 67. Kim NJ, Chowdhury NF, Buetow KH, Thompson PM, Irimia A. Genetic insights into brain morphology: a genome-wide association study of cortical thickness and T1-weighted MRI gray matter-white mat- ter intensity contrast. Neuroinformatics. 2025;23(2):26. doi:10.1007/ s12021-025-09722-9 68. Depp C, Sun T, Sasmita AO, et al. Myelin dysfunction drives amyloid-β deposition in models of Alzheimer’s disease. Nature. 2023;618(7964):349-357. doi:10.1038/s41586-023-06120-6 69. Papuć E, Rejdak K. The role of myelin damage in Alzheimer’s dis- ease pathology. Arch Med Sci. 2020;16(2):345-351. doi:10.5114/aoms. 2018.76863 70. Smith SM, Douaud Gë, Chen W, et al. An expanded set of genome-wide association studies of brain imaging phenotypes in UK Biobank. Nat Neurosci. 2021;24(5):737-745. doi:10.1038/s41593-021-00826-4 71. Backman JD, Li AH, Marcketta A, et al. Exome sequencing and analysis of 454,787 UK Biobank participants. Nature. 2021;599(7886):628- 634. doi:10.1038/s41586-021-04103-z 72. Santa-Marinha Luí, Castanho I, Silva RR, et al. Phospholipase D1 ablation disrupts mouse longitudinal hippocampal axis organization and functioning. Cell Rep. 2020;30(12):4197-4208.e6. doi:10.1016/j. celrep.2020.02.102 73. Cai D, Netzer WJ, Zhong M, et al. Presenilin-1 uses phospholipase D1 as a negative regulator of beta-amyloid formation. Proc Natl Acad Sci U S A. 2006;103(6):1941-1946. doi:10.1073/pnas.0510708103 74. Jin J-K, Kim N-Ho, Lee Y-J, et al. Phospholipase D1 is up-regulated in the mitochondrial fraction from the brains of Alzheimer’s disease patients. Neurosci Lett. 2006;407(3):263-267. doi:10.1016/j.neulet. 2006.08.062 75. Doré V, Krishnadas N, Bourgeat P, et al. Relationship between amy- loid and tau levels and its impact on tau spreading. Eur J Nucl Med Mol Imaging. 2021;48(7):2225-2232. doi:10.1007/s00259-021-05191-9 76. Prins ND, Scheltens P. White matter hyperintensities, cognitive impairment and dementia: an update. Nat Rev Neurol. 2015;11(3):157- 165. doi:10.1038/nrneurol.2015.10 77. Fortea J, Vilaplana E, Carmona-Iragui M, et al. Clinical and biomarker changes of Alzheimer’s disease in adults with Down syndrome: a cross- sectional study. Lancet. 2020;395(10242):1988-1997. doi:10.1016/ S0140-6736(20)30689-9 78. Zhang T, Nie B, Liu H, et al. Prediction of alzheimer’s disease time to dementia onset using cross-sectional data from spatiotem- poral biomarker progression patterns. Eur J Nucl Med Mol Imaging. 2026;53:4619-4638. doi:10.1007/s00259-026-07797-3 ### Page 17 PIEPERHOFF ET AL. 17 of 1779. Fleisher AS, Chen K, Quiroz YT, et al. Florbetapir PET analysis of amyloid-β deposition in the presenilin 1 E280A autosomal dominant Alzheimer’s disease kindred: a cross-sectional study. Lancet Neurol. 2012;11(12):1057-1065. doi:10.1016/S1474-4422(12)70227-2 80. Lorenzini L, Tranfa M, Pieperhoff L, et al. Cortical thickness sub- types in cognitively unimpaired individuals: differential network and transcriptomic vulnerability to cortical thinning. Alzheimers Dement. 2025;21(10):e70762. doi:10.1002/alz.70762SUPPORTING INFORMATIONAdditional supporting information can be found online in the Support- ing Information section at the end of this article.How to cite this article: Pieperhoff L, Lorenzini L, Almeida F, et al. Cortical gray–white matter contrast alterations precede amyloid-β positivity and macrostructural changes in older adults without dementia. Alzheimer’s Dement. 2026;22:e71609. https://doi.org/10.1002/alz.71609