Predicting accumulation and age at onset of amyloid-β from genetic risk and resilience for Alzheimer’s disease

Predicting accumulation and age at onset of amyloid-β from genetic risk and resilience for Alzheimer’s disease | 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 Research Article Predicting accumulation and age at onset of amyloid-β from genetic risk and resilience for Alzheimer’s disease Eleanor K O'Brien, Timothy Cox, Shane Fernandez, Pierrick Bourgeat, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7911284/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Accumulation of brain amyloid beta (Aβ) is a key pathological hallmark of Alzheimer’s disease (AD) and begins many years before cognitive symptoms. Being able to predict the risk of Aβ accumulation, or the age at which this accumulation exceeds a critical threshold, may enable early intervention and treatment to slow or prevent the onset of AD. We utilised published genome-wide association studies (GWAS) to develop polygenic scores (PGS) based on AD risk (PGS risk ) and resilience (PGS resilience ). We tested whether these could predict (i) whether an individual was an accumulator of Aβ (‘Accumulator Status’), and (ii) in accumulators, the age at which brain Aβ is estimated to exceed a threshold of 20 centiloids (CL)(‘Estimated Age at onset of Aβ’; AAO-Aβ) among 2175 participants (1158 with AAO Aβ) from the Alzheimer’s Dementia Onset and Progression in International Cohorts (ADOPIC) study. Additionally, we conducted genome-wide association studies (GWAS) of these traits and developed phenotype-specific PGSs using cross-validation (CV). Higher PGS risk was associated with a greater risk of being an accumulator and a younger AAO-Aβ. When stratified by number of APOE ε4 alleles, PGS risk predicted Accumulator Status in APOE ε4 heterozygotes, and AAO-Aβ in ε4 non-carriers and heterozygotes, with the same directions of effect as were seen in the whole cohort. PGS resilience was not significantly associated with Accumulator Status, but higher PGS resilience was associated with later AAO-Aβ overall and in ε4 heterozygotes. Trait-specific PGSs, developed using CV, were not significantly associated with either trait overall and the direction of association varied across CV folds. Polygenic scores, alongside other risk factors, may be useful for identifying individuals at risk of accumulating Aβ, and predicting the age at which this exceeds a critical threshold. This could provide a window for administering disease-modifying treatment or lifestyle interventions to prevent or delay the onset of AD. Alzheimer's disease amyloid beta accumulation age at onset of amyloid beta polygenic scores risk resilience Figures Figure 1 Figure 2 Figure 3 Figure 4 Background Alzheimer’s disease (AD) is a neurodegenerative disorder, characterised by progressive cognitive decline and memory loss [ 1 ]. It is the most common form of dementia, accounting for around 65% of cases [ 2 ]. The accumulation of amyloid beta (Aβ) in the brain is one of the key pathological hallmarks of AD, beginning years, if not decades, before the onset of clinical symptoms [ 3 – 5 ]. The ‘amyloid cascade hypothesis’ posits that this accumulation precipitates a cascade of events, including tau hyperphosphorylation and formation of neurofibrillary tangles, neuronal loss, cognitive decline and ultimately dementia [ 5 , 6 ]. Predicting whether an individual will accumulate Aβ and the age at which they will reach a critical threshold for brain Aβ that leads to these downstream events could therefore enable timely therapeutic strategies aimed at delaying or preventing the onset and progression of cognitive decline [ 7 ]. Late onset AD (LOAD) occurs sporadically in people aged over 65 years, and makes up 90–95% of AD cases [ 8 ]. LOAD (henceforth referred to as AD) is a complex disease, with risk determined by a combination of genetic, environmental and lifestyle factors [ 9 ]. Among the genetic factors, the Apolipoprotein E ( APOE ) gene on chromosome 19 shows the strongest association with the incidence of AD, with the APOE ε4 allele conferring a dose-dependent increase in risk of AD, and the ε2 allele being protective [ 2 , 10 ]. APOE genotype is also associated with the magnitude and timing of Aβ accumulation, such that ε4 carriage is associated with higher average age-adjusted Aβ burden and younger ages when critical Aβ thresholds are reached [ 11 ]. However, genome-wide association studies (GWAS) have identified more than 80 additional genetic variants that show a significant association with AD risk, albeit with smaller individual effects than APOE [ 12 – 18 ]. GWAS for Aβ accumulation and burden have also identified several significantly associated loci in addition to those within the APOE region [ 19 – 21 ]. These findings suggest that the prediction of AD and related traits can be improved by considering genetic factors beyond APOE . Polygenic scores (PGS; also referred to as ‘genetic scores’ or ‘polygenic risk scores’) aggregate the effects of variants across the genome that are associated with the disease or trait of interest [ 22 – 25 ]. This approach recognises that complex traits are often influenced by many genetic variants, most of small effect, and many of which will not meet the stringent threshold for genome-wide significance [ 26 – 28 ]. The resulting score gives an estimate of genetic susceptibility to a trait or disease, which may then be used, alongside other clinical indicators, to inform decisions around preventative actions or interventions [ 24 ]. PGSs developed from genetic associations with disease incidence may also be useful for predicting endophenotypes linked to that disease (or vice versa ), which can provide further insights into the underlying biological disease mechanisms and aid in earlier diagnosis. In the context of AD, genetic risk of disease (as estimated from GWASs for incidence of AD) may therefore be useful for predicting onset and progression of pathophysiological features such as Aβ that occur earlier in the disease trajectory, prior to diagnosis. In the present study, we develop PGSs based on previously published GWASs for risk of and resilience to AD [ 13 , 29 ]. Here, ‘resilience’ was defined as remaining cognitively unimpaired in the face of high genetic risk for AD [ 29 ]. We test the extent to which these PGSs predict whether an individual accumulates brain Aβ and the age at which Aβ is estimated to reach a critical threshold. We then assess whether we can improve on these predictions using a priori GWASs to develop phenotype-specific PGSs for each of these traits. We seek to develop predictive models that can identify individuals at high risk for early Aβ pathology, to inform personalised approaches to the prevention and treatment of AD. Methods Study population The study included data for 2175 participants from the Alzheimer’s Dementia Onset and Progression in International Cohorts (ADOPIC) study, which combines three longitudinal cohort studies: the Alzheimer’s Disease Neuroimaging Initiative (ADNI)[ 30 – 32 ], the Australian Imaging, Biomarkers and Lifestyle Study of Ageing (AIBL)[ 33 , 34 ], and the Knight ADRC Open Access Series of Imaging Studies (OASIS)[ 35 ]. Full details of these studies, including recruitment criteria and the schedule of assessments have been described previously [ 30 – 35 ]. The studies have all been granted approval by the ethics committees of their respective member institutions, and all participants provided informed written consent. The present study was limited to ADOPIC participants of European ancestry. Genotyping and imputation DNA was extracted from whole blood samples. For ADNI and AIBL participants, this was done using QIAamp DNA blood spin column kits (Qiagen, Valencia, CA, USA) [ 31 , 33 , 34 ], and for OASIS participants using the Autogen FlexSTAR + salt precipitation [ 36 ]. ADNI samples were genotyped at genome-wide single nucleotide polymorphisms (SNPs) on either the Illumina Human610-Quad BeadChip or the Illumina Human OmniExpress BeadChip[ 31 ]. AIBL samples were genotyped on the Axiom Precision Medicine Diversity Array (Applied Biosystems™)[ 33 , 34 ]. OASIS samples were genotyped on one of nine arrays (Illumina Human660W-Quad, Infinium OmniExpressExome-8, Illumina Omni1-Quad, Illumina Human1M-Duo, Infinium Neuro Consortium Array, Infinium CoreExome-24, Infinium Global Screening Array-24, Illumina Human610-Quad, and UK Biobank Axiom array)[ 36 ]. All genotype data were imputed against the TOPMed panel on the TOPMed Imputation Server (University of Michigan, USA) [ 37 , 38 ]. Within cohorts, this was done for each genotyping run, and then the imputed data were combined across the three cohorts to produce a single harmonised genetic data set. Only autosomal SNPs were included in the final genetic data set. We removed SNPs with an imputation quality r 2 ≤ 0.3, samples and SNPs with > 2% missing genotypes after merging, as well as SNPs with minor allele frequency (MAF) < 5%, and those where the p-value from a test for Hardy-Weinberg equilibrium was 0.25. We also used PLINK to conduct a principal components analysis (PCA) on all remaining unrelated individuals to obtain principal components (PCs) to include as covariates to control for population structure in genetic analyses (see below). Genetic data were available for 6157 individuals, of whom 131 were removed due to relatedness, giving a final genetic data set of 6026 individuals (Fig. 1 ). APOE genotype was determined from TaqMan® genotyping assays (Life Technologies, USA) at two SNPs: rs7412 (Assay ID: C__904973_10) and rs429358 (Assay ID: C__3084793_20). Aβ PET Imaging Measures of brain β-amyloid (Aβ) burden were obtained for participants in all three cohorts using positron emission tomography (PET) imaging with one of the following five tracers: 11 C–Pittsburgh compound B (PiB), 18 F-NAV4694 (NAV), 18 F-florbetaben (FBB), 18 F-florbetapir (FBP), or 18 F-flutemetamol (FLUTE). PET images were analysed using CapAIBL [ 41 ] to generate tracer-specific standardised uptake value ratios (SUVRs), which were then transformed to centiloids (CL) as previously described [ 42 , 43 ]. Harmonisation of CL quantification across the three cohorts has been described previously [ 44 ]. Traits Accumulator Status was determined for all individuals in the study with Aβ PET imaging at two or more timepoints (N = 2175; Fig. 1 ). Individuals were classified as “accumulators” if they had Aβ PET ≥ 20 CL at any assessment or showed an increase in Aβ PET across timepoints of ≥ 0.05 CL/year, and as “non-accumulators” otherwise [ 45 ]. Estimated Age at Onset of Aβ (AAO-Aβ) was the age at which participants’ cortical Aβ burden was estimated to have reached 20 CL, obtained by placing them on a natural history curve of Aβ accumulation, as described previously [ 45 ]. AAO-Aβ was estimated for people classified as “accumulators” who also met the following criteria: (i) Their Aβ burden at the final scan was ≥ 10 CL (due to uncertainty around whether people with brain Aβ below this level will reach the 20 CL threshold), and (ii) if brain Aβ did not already exceed the 20 CL threshold, it was predicted to do so within 5 years. In total, AAO-Aβ was estimated for 1158 individuals (Fig. 1 ). Data analysis Development of cross-trait polygenic scores We obtained summary statistics from genome-wide association studies (GWAS) for AD risk [ 13 ] and resilience [ 29 ]. We then used PRSice-2 [ 46 , 47 ] to identify optimal PGSs for each trait (Accumulator Status and AAO-Aβ), based on risk and resilience GWASs. This process finds the p-value threshold for variant selection that results in a set of variants that explains the largest proportion of variance in the target phenotype, and uses these to create a PGS. To determine whether these risk and resilience PGSs could account for significant variation in each trait beyond that explained by APOE (the gene most strongly associated with AD, located on chromosome 19), we excluded chromosome 19 from the genetic data set, and included number of APOE ε4 alleles as a covariate, as recommended by Ware et al (2020) [ 48 ]. Additional covariates were sex and study (which of the three cohort studies the participant came from). To control for population structure, we also included the first three principal components from the PCA. We used an additive genetic model with default clumping and thinning parameters (250kb distance, r 2 threshold 0.1). PGSs were standardised to a mean of 0 and standard deviation of 1. Four PGSs were derived, defined by the GWAS summary stats used (risk or resilience) and the trait being predicted (Accumulator Status or AAO-Aβ): PGS risk−Accumulator , PGS resilience−Accumulator , PGS risk−AAO and PGS resilience−AAO . We then used these optimal PGSs (calculated for the whole population) for the relevant trait in (generalised) linear models to test their association with that trait in three groups that were stratified by number of APOE ε4 alleles: (a) APOE ε4 non-carriers, (b) APOE ε4 heterozygotes, and (c) APOE ε4 homozygotes. These models included the same covariates as described above, excluding APOE ε4 allele count. For Accumulator Status, we fitted logistic regression models using the ‘glm’ function in R with a binomial distribution and logit link function, and for AAO-Aβ we fitted linear regression models using the ‘lm’ function (which assumes a gaussian distribution) in R (version 4.3.3)[ 49 ]. In each case, these were implemented in RStudio (version 2024.12.1 Build 563)[ 50 ]. Beta coefficients for the PGS term in logistic regressions were exponentiated to convert them to odds ratios (OR), which represent the change in odds of being an accumulator with each 1 SD increase in the PGS. For linear regressions on AAO-Aβ, beta coefficients for the PGS term represent the change in estimated AAO-Aβ (in years) with a 1 SD increase in the PGS. To examine the change in odds of being an accumulator and estimated AAO-Aβ between PGS extremes, we also ran models comparing people with scores in the upper and lower quintiles (top and bottom 20%) of the population for each PGS and trait. Threshold PGS scores for the upper and lower quintiles were obtained for all genotyped individuals (N = 6026), and then phenotyped individuals in these ranges were extracted for analysis. In these models, PGS was a categorical predictor with levels “low” (lower quintile) and “high” (upper quintile). Covariates were the same as in models that included the full range of PGS values, and models were again run for all individuals and stratified by number of APOE ε4 alleles. GWAS and development of trait-specific PGS using cross-validation Due to the relatively small size of this study, we used a Monte Carlo cross validation approach [ 51 , 52 ] to derive phenotype-specific PGSs. We created 10 cross validation (CV) runs, where within each run the data were split into discovery (two-thirds of individuals; Accumulator Status N = 1450, AAO-Aβ N = 772) and validation (one-third of individuals; Accumulator Status N = 725, AAO-Aβ N = 386) sets using the ‘sample’ function in R (v 4.3.3) [ 49 ], run in R Studio (v 2024.12.1 Build 563)[ 50 ]. We conducted a genome-wide association study (GWAS) for the trait of interest in the discovery set and used the summary statistics to find the optimal PGS in the corresponding validation set. This process was repeated in each of the 10 CV runs for each trait. We ran GWASs for each trait in each discovery data set in PLINK [ 39 , 40 ], using a logistic regression model for Accumulator Status and linear regression for estimated AAO-Aβ. We assumed an additive genetic model, and included sex, study, and the first three principal components as covariates. To identify the optimal PGS for the trait in the corresponding validation set, we PRSice-2 and followed the same protocol as for the risk and resilience PGSs described above, with the same covariates and clumping parameters, and again with chromosome 19 excluded. Here, the ‘effective allele’ was the allele associated with worse outcomes (higher odds of being an accumulator and earlier estimated AAO-Aβ). Therefore in each case, higher values for these scores are expected to be associated with these worse outcomes. To enable a direct comparison of the performance of the trait-specific PGSs with the risk and resilience PGSs, we tested the association of the optimal risk and resilience PGSs (determined from the whole cohort) with the traits in the validation data set of each CV run, using the same set of covariates as previously. To compare the predictive performance of each PGS for each trait, we summarised performance across all CV runs. For Accumulator Status, we calculated the mean odds ratio and 95% confidence interval, and for estimated AAO-Aβ, we calculate the mean and standard error of the coefficient of the linear relationship of the standardised PGS against the phenotype across the 10 CV runs. For each PGS and trait, the overall R 2 value was estimated as the mean of the R 2 values across CV runs (Nagelkerke’s pseudo R 2 for Accumulator Status), and the overall p-value was estimated from the p-values in each of the 10 CV runs using Stouffer’s method [ 53 ], with p-values weighted by the inverse of the standard error and adjusting for the direction of association in each run. Results Demographics Demographic characteristics of all participants with at least two Aβ PET scans (N = 2175), and of the subset of participants with estimated AAO-Aβ (N = 1158), stratified by cohort study, are shown in Table 1 . Participants from OASIS were on average younger, a higher proportion of them were female, and there were fewer APOE ε4 homozygotes than in ADNI or AIBL. A higher proportion of participants from ADNI were accumulators of Aβ compared to the other cohort studies. In the subset of participants with AAO-Aβ, all of whom were accumulators of Aβ, OASIS participants were still younger and with lower final Aβ burden than those from ADNI and AIBL (Table 1 ). Participant counts and percentages by Accumulator Status, for each APOE ε4 group, are shown in Table 2 . The percentage of participants who were accumulators of Aβ increased with number of APOE ε4 alleles, from 63.1% among non-carriers, to 84% of ε4 heterozygotes and 95.6% of ε4 homozygotes (Table 2 ). Consistent with previous studies, APOE ε4 carriage was strongly associated with both Accumulator Status and estimated AAO-Aβ in a dose-dependent manner, with an increasing number of ε4 alleles associated with higher odds of being an accumulator of Aβ (Fig. 2 , Table S1 ) and earlier estimated AAO-Aβ (Fig. 3 , Table S2). Table 1 Summary of demographic characteristics of (a) all participants included in the study and (b) participants that met the criteria for inclusion in analysis of estimated AAO-Aβ, shown overall and stratified by study cohort. P-values are from comparisons between study cohorts using either a Pearson’s chi square test (sex, accumulator status, no. of APOE ε4 alleles) or a Kruskal-Wallis rank sum test (final age, final PET Aβ, estimated AAO). (a) All participants Characteristic Overall N = 2175 ADNI N = 952 AIBL N = 930 OASIS N = 293 p-value Sex N (%) 0.003 Female 1100 (50.6%) 450 (47.3%) 479 (51.5%) 171 (58.4%) Male 1075 (49.4%) 502 (52.7%) 451 (48.5%) 122 (41.6%) Accumulator status N (%) < 0.001 Non-accumulator 619 (28.5%) 166 (17.4%) 362 (38.9%) 91 (31.1%) Accumulator 1556 (71.5%) 786 (82.6%) 568 (61.1%) 202 (68.9%) Mean final age yrs (SD) 76.0 (7.7) 77.0 (7.7) 76.3 (6.7) 71.5 (8.9) < 0.001 No. of APOE ε4 alleles N (%) 0.021 0 1372 (63.1%) 567 (59.6%) 611 (65.7%) 194 (66.2%) 1 668 (30.7%) 319 (33.5%) 261 (28.1%) 88 (30.0%) 2 135 (6.2%) 66 (6.9%) 58 (6.2%) 11 (3.8%) (b) Participants with estimated Age at Onset of Aβ Characteristic Overall N = 1158 ADNI N = 649 AIBL N = 400 OASIS N = 109 p-value Sex N (%) 0.071 Female 564 (48.7%) 301 (46.4%) 200 (50.0%) 63 (57.8%) Male 594 (51.3%) 348 (53.6%) 200 (50.0%) 46 (42.2%) Mean final age yrs (SD) 77.7 (7.3) 77.9 (7.5) 77.9 (6.8) 75.6 (7.8) 0.018 Mean final PET Aβ (SD) 69.4 (40.0) 71.5 (42.3) 71.3 (37.2) 49.7 (29.0) < 0.001 Mean estimated AAO-Aβ yrs (SD) 66.8 (10.8) 66.5 (11.4) 66.7 (10.2) 69.1 (8.8) 0.100 No. of APOE ε4 alleles N (%) 0.400 0 550 (47.5%) 316 (48.7%) 185 (46.3%) 49 (44.9%) 1 480 (41.5%) 268 (41.3%) 162 (40.5%) 50 (45.9%) 2 128 (11.1%) 65 (10.0%) 53 (13.2%) 10 (9.2%) Table 2 Counts (%) of individuals classified as non-accumulators and accumulators of Aβ, stratified by number of APOE ε4 alleles. Aβ Accumulator Status Number of APOE ε4 alleles TOTAL 0 1 2 Non-accumulator 506 (36.9%) 107 (16.0%) 6 (4.4%) 619 (28.5%) Accumulator 866 (63.1%) 561 (84.0%) 129 (95.6%) 1556 (71.5%) TOTAL 1372 668 135 2175 Polygenic scores derived from risk and resilience GWAS against Accumulator Status and Estimated Age at Onset of Aβ (AAO-Aβ) The optimal risk PGS for Accumulator Status (PGS risk−Accumulator ) was significantly associated with this trait overall, with an odds ratio (OR) of 1.16 (95% CI 1.05–1.29)(Table S1 ), indicating that for an increase of 1 standard deviation (SD) in PGS risk−Accumulator , the odds of being an accumulator of Aβ increase by 16%. When stratified by number of APOE ε4 alleles, the same direction of association was seen in all groups, but this was only significant in APOE ε4 heterozygotes (marginally non-significant in ε4 non-carriers, p = 0.058) (Fig. 2 , Table S1 ). Individuals in the upper quintile of PGS risk−Accumulator were 63% more likely than those in the bottom quintile to be accumulators of Aβ overall (OR = 1.63; 95% CI 1.18–2.26)(Table S1 ), and this difference was also significant within APOE ε4 non-carriers and heterozygotes (Table S1 ). The optimal resilience PGS (PGS resilience−Accumulator ) was not significantly associated with Accumulator Status overall, or for any group when stratified by number of APOE ε4 alleles (Fig. 2 , Table S1 ). We also did not detect a significant difference in the probability of being an accumulator between the upper and lower quintiles of PGS resilience−Accumulator (Table S1 ). It is important to note that only 4.4% (6 out of 135) of APOE ε4 homozygotes in our study were non-accumulators (Table 2 ). This was a much lower percentage than in the other ε4 groups (Table 2 ), and would have limited our power to detect an association with Accumulator Status in this sub-group. The optimal risk and resilience PGSs for estimated AAO-Aβ (PGS risk−AAO and PGS resilience−AAO ) were both significant predictors of this trait overall, with higher PGS risk−AAO and lower PGS resilience−AAO associated with an earlier estimated AAO-Aβ (Fig. 3 , Table S2). A 1 SD increase in PGS risk−AAO and a 1 SD decrease in PGS resilience−AAO were associated with estimated AAO-Aβ 1.3 years and 0.91 years earlier respectively (Fig. 3 , Table S2). When stratified by number of APOE ε4 alleles, higher PGS risk−AAO was associated with an earlier estimated AAO-Aβ in APOE ε4 non-carriers and heterozygotes, but there was no significant association with AAO-Aβ in APOE ε4 homozygotes. Individuals in the upper quintile for PGS risk−AAO had a mean estimated AAO-Aβ 2.9 years earlier than those in the lower quintile after accounting for other covariates, while among APOE ε4 heterozygotes this difference was 4.5 years (Table S2). Higher PGS resilience−AAO was associated with a later estimated AAO-Aβ in APOE ε4 heterozygotes and a marginally non-significant (p = 0.051) increase in estimated AAO-Aβ in ε4 homozygotes, but not associated with AAO-Aβ in APOE ε4 non-carriers (Fig. 3 , Table S2). Individuals in the upper quintile for PGS resilience−AAO had a mean estimated AAO-Aβ 2.2 years later than those in the lower quintile after accounting for other covariates, but there was not a significant difference between these extremes for any of the groups after stratification by number of APOE ε4 alleles (Table S2). Polygenic scores derived from trait-specific GWAS against Accumulator Status and Estimated Age at Onset of Aβ (AAO-Aβ), and comparison with risk and resilience PGSs The mean percentage of people who were accumulators of Aβ across each of the CV runs was 71.4% in the discovery sets (SD 0.50) and 71.8% in the validation sets (SD 1.00)(Table S3), which are both very close to the value of 71.5% for the whole sample (Table 1 ). In the trait-specific GWASs run in each of the 10 CV runs (excluding chromosome 19), we found a total of two SNPs (rs12192157 and rs6900289) associated with Accumulator Status at genome-wide significance ( P < 5 x 10 − 8 ). These closely linked variants were on chromosome 6 and were significant in just one of the CV runs (Table S4). We found one genome-wide significant SNP (rs12022131) for estimated AAO-Aβ in one of the CV runs, located on chromosome 1 (Table S4). The optimal trait-specific PGS for Accumulator Status (PGS Accumulator ) was significantly associated with this trait in four of the 10 CV runs (two of these remained significant after FDR correction), although the direction of the association varied among runs and the overall association was not significant (Fig. 4 , Table S5). The risk and resilience PGSs for Accumulator Status, run in the validation sets of each CV run for comparison, both showed overall significant positive associations with odds of being an accumulator. In contrast to PGS Accumulator , the direction of association was consistent across all 10 CV runs for PGS risk−Accumulator and all but one CV run for PGS resilience−Accumulator , although it was only significant in six CV runs for PGS risk−Accumulator and not in any individual CV run for PGS resilience−Accumulator (Fig. 4 ; Table S5). Similarly, for estimated AAO-Aβ, the optimal trait-specific PGS (PGS AAO ) was not significantly associated with the trait overall, and although the association was significant in four of the CV runs (three after FDR correction), the direction of association was highly variable across runs (Fig. 4 , Table S6). PGS risk−AAO was negatively associated with AAO-Aβ, indicating that individuals with a higher genetic risk of AD had an earlier estimated age at onset of Aβ. PGS resilience−AAO was positively associated with AAO-Aβ, indicating a later estimated age at onset of Aβ in more genetically resilient individuals (Fig. 4 , Table S6). Of the three PGSs for each trait, PGS risk had, on average, the largest effect size and explained the greatest proportion of the variation in both Accumulator Status and estimated AAO-Aβ (Fig. 4 ). Discussion We tested whether polygenic scores based on genetic risk of and resilience to AD could predict Aβ Accumulator Status and estimated AAO-Aβ. We also tested whether PGSs based on phenotype-specific GWASs improved prediction of these traits. Higher genetic risk of AD was associated with higher odds of Aβ accumulation, and earlier estimated AAO-Aβ. Higher genetic resilience was associated with a later estimated AAO-Aβ but was not a significant predictor of Accumulator Status. Phenotype-specific PGSs did not improve on the predictive performance of PGS risk or PGS resilience and were not significant predictors of either trait overall. The associations of PGSs with each trait were seen after accounting for APOE ε4 status, the strongest genetic risk factor for AD [ 2 , 10 ]. Our results contrast with several studies that have found that PGSs did not improve prediction over and above APOE ε4 status for AD and all-cause dementia (ACD) [ 54 ], or for measures of Aβ deposition [ 55 , 56 ]. However, other studies have found that PGSs do result in small but significant improvements over APOE alone in prediction of AD-related traits, including incidence of AD [ 56 , 57 ] and Aβ pathology [ 58 ]. Apparent inconsistencies between these findings may be explained by variation in the methods used to construct PGSs, as well as the specific phenotypes examined. It has been argued that Aβ deposition is largely driven by APOE , and that other genetic contributors to AD become more important at later disease stages [ 54 , 56 ]. However, our results highlight that considering genetic factors beyond APOE can improve prediction of whether and how early individuals will accumulate Aβ. Differences between PGS risk and PGS resilience in their associations with the traits examined, and their interactions with APOE , offer insights into the mechanisms by which the genetic variation captured by these scores confers risk or protection against AD. Increased genetic risk of AD was a stronger predictor of adverse outcomes (higher odds of being an accumulator and earlier estimated AAO-Aβ) in ε4 non-carriers and heterozygotes than in ε4 homozygotes, suggesting it contributes little additional risk in individuals who are already at highest risk due to APOE ε4 homozygosity. By contrast, higher genetic resilience to AD was associated with later estimated AAO-Aβ in ε4 heterozygotes (and a marginally non-significant association in homozygotes), but was not associated with AAO-Aβ in ε4 non-carriers. The original AD resilience GWAS was conducted by limiting the study population to individuals at high genetic risk for AD (defined by a similar risk PGS to that developed here) and contrasting ‘resilient’ (unaffected by AD) individuals with AD cases [ 29 ]. This means that it is effectively a measure of resilience to genetic risk of AD, so it is unsurprising that it is a stronger predictor of AAO-Aβ in APOE ε4 carriers, who are at highest genetic risk [ 2 , 10 ]. PGS resilience was not associated with Accumulator Status, although the overall trend (significant when aggregated across runs in the cross-validation study) was positive. This implies that more genetically resilient individuals are more likely to be accumulators of Aβ, which seems counterintuitive, but is likely another consequence of the fact that for this score, more genetically resilient individuals also have higher genetic risk scores [ 29 ]. A framework previously proposed when considering protective factors for AD, distinguishes between ‘resistance’ and ‘resilience’, where resistance refers to the avoidance of pathological brain changes, while resilience is the ability to cope with accumulating neuropathology and avoid brain atrophy or cognitive decline [ 59 – 61 ]. The lack of association of PGS resilience with Accumulator Status in our study suggests that genetic variation captured by this score does not confer protection against AD by preventing the accumulation of Aβ (‘resistance’), although the association with AAO-Aβ suggests it may slow or delay this accumulation. Further analysis involving a broader range of traits is required to disentangle this. The trait-specific PGSs for both Accumulator Status and estimated AAO-Aβ, derived using a cross-validation approach, were not associated with either trait overall, although several individual CV runs showed significant associations for each trait. However, there was variability in effect direction for these associations. It is likely that unstable signals were owing in part to the sample sizes available for the discovery GWASs in this study (N = 1450 for Accumulator Status and N = 772 for AAO-Aβ), which were much smaller than those in the risk (N = 94 437) and resilience (N = 13 572) GWASs [ 13 , 29 ]. Despite the lack of compelling evidence in the current study, trait-specific PGSs may nevertheless capture unique genetic variation associated with these traits, particularly in larger studies. We identified two closely-linked SNPs on chromosome 6 that were associated with Accumulator Status (rs12192157 and rs6900289), and one SNP on chromosome 1 (rs12022131) that was associated with AAO-Aβ at a genome-wide significant level. While each was significant in only a single CV run, it will be of interest to determine whether a signal is seen in these regions in future studies. To the best of our knowledge, there are currently no known associations of these SNPs with specific traits or diseases. However, the SNPs on chromosome 6 are in close proximity to the LPA gene, which affects plasma concentrations of lipoprotein (a) (Lp(a)) and is strongly associated with cardiovascular disease [ 62 ], a key risk factor for AD [ 63 , 64 ]. Genetic variants, unlike other biomarkers of disease, remain constant across the lifespan, meaning that polygenic scores for diseases and related traits offer the potential to identify high risk individuals at a very early stage, prior to symptom onset [ 25 ]. In AD, a diagnosis is typically made once cognition is impaired, by which time there has been widespread damage to the brain. However, brain Aβ begins accumulating years or decades prior to appearance of cognitive symptoms [ 3 – 5 ]. Therefore, identifying individuals at risk of accumulating Aβ provides an opportunity to administer interventions to prevent or delay onset and progression of disease. While effective treatments for AD have proved elusive, recent years have seen the development of anti-amyloid monoclonal antibodies that remove Aβ from the brain and have been shown to produce modest slowing of cognitive decline in people with mild symptomatic AD [ 65 – 67 ]. Trials are currently evaluating whether administering these treatments at an earlier stage, in asymptomatic individuals, may be more effective and there is hope that it may eventually be possible to prevent AD [ 68 ]. In this instance, PGSs could offer a relatively inexpensive and minimally invasive method to evaluate people’s risk and prioritise them for further screening. A limitation of PGSs is that they typically explain only a small proportion of the total variation in a disease or trait [ 69 ], which was the case in this study (0.2–2.3% for PGS risk ; 0.1–1.6% for PGS resilience , depending on population sub-group). While this limits their utility for making a definitive diagnosis, our results show that PGSs can nevertheless improve the accuracy of prediction of AD-related traits, and may be a useful tool for risk stratification, particularly when considered alongside other risk predictors such as demographics and lifestyle. This study does have several limitations. Within our study population, over 70% of participants were accumulators of Aβ. While data on Accumulator Status in the wider population are scarce, one study found that ~ 20% of healthy adults aged ≥ 60 years had elevated Aβ [ 70 ]. Accumulators of Aβ are therefore almost certainly over-represented in our study population, which is unsurprising given that the component cohorts are enriched for people with cognitive complaints [ 30 – 35 ]. As a result, the likelihood of being an accumulator as predicted by PGS score may be overestimated and should be recalibrated based on a more representative population. Similarly, AAO-Aβ could only be estimated for people who had begun accumulating Aβ and were close to, or had exceeded, the 20 CL threshold. The predictive performance of this score in the broader population, including people with low Aβ who may accumulate Aβ, needs to be verified. Furthermore, the lack of an external validation sample for the phenotype-specific PGSs meant that it was necessary to both run the discovery GWAS and develop PGSs within the study population by dividing it into test and validation data sets. Despite utilising the largest existing dataset for these traits, this resulted in small sample sizes relative to those used in the GWASs from which risk and resilience PGSs were derived [ 13 , 29 ]. This was partially addressed by our cross-validation approach. However, the instability of score performance across CV runs is likely due to the small sample. Finally, the study population was limited to people of European ancestry. Differences in LD structure, allele frequencies and genetic architecture can affect the generalisability of genetic predictors across different ancestries [ 71 ], therefore testing in diverse populations would be beneficial. Conclusions Polygenic scores based on genetic risk of AD explained a small but significant proportion of the variation in Accumulator Status and estimated AAO-Aβ, over and above that explained by APOE ε4. The PGS for AD risk may be particularly useful, in combination with other predictors, for identifying individuals at risk of Aβ accumulation and earlier AAO-Aβ, who may benefit from targeted prevention and treatment. Declarations Competing Interests CC is a member of the scientific advisory board of Circular Genomics and owns stocks, and is on the scientific advisory board of ADmit and Alamar, consults for Sanofi, NovoNordisk, and Owkin, and has received research support from GSK, Danaher and EISAI. All other authors report no competing interests relevant to this manuscript. Ethics approval and consent to participate The ADNI, AIBL and OASIS studies have all been granted approval by the ethics committees of their respective member institutions. All participants in this study provided informed written consent. Funding The Alzheimer’s Dementia Onset and Progression in International Cohorts (ADOPIC) study was funded by a National Institute of Health (NIH) grant (R01-AG058676-01A1). Data used in the preparation of this article were obtained from the Australian Imaging, Biomarker and Lifestyle (AIBL) Study database ( https://aibl.org.au/collaboration ). As such, the investigators within AIBL, unless otherwise listed, contributed to the design and implementation of AIBL and/or provided data, but did not participate in the analysis or writing of this report. A complete listing of AIBL investigators can be found at: https://aibl.org.au/about/our-researchers/ . The AIBL Study ( https://aibl.org.au/ ) is a consortium between Austin Health, CSIRO, Edith Cowan University, the Florey Institute (The University of Melbourne), and the National Ageing Research Institute. The study has received partial financial support from the Alzheimer’s Association (US), the Alzheimer’s Drug Discovery Foundation, an Anonymous Foundation, the Science and Industry Endowment Fund, the Dementia Collaborative Research Centres, the Victorian Government’s Operational Infrastructure Support program, the Australian Alzheimer’s Research Foundation (now Alzheimer’s Research Australia), the National Health and Medical Research Council (NHMRC), and The Yulgilbar Foundation. Numerous commercial interactions have supported data collection and analyses. This includes genetic data utilized in this study, which has also been supported by grants awarded to SML by the NHMRC (GNT1161706; GNT2001320). In-kind support has also been provided by Sir Charles Gairdner Hospital, Cogstate Ltd, Hollywood Private Hospital, The University of Melbourne, and St Vincent’s Hospital. Data used in the preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List . pdf. The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer’s disease (AD). Data collection and sharing for the ADNI is funded by the National Institute on Aging (National Institutes of Health Grant U19 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health ( www.fnih.org ). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California. Data were provided in part by OASIS-3: Principal Investigators: T. Benzinger, D. Marcus, J. Morris; NIH P50 AG00561, P30 NS09857781, P01 AG026276, P01 AG003991, R01 AG043434, UL1 TR000448, R01 EB009352. AV-45 doses were provided by Avid Radiopharmaceuticals, a wholly owned subsidiary of Eli Lilly. Author Contribution EKO, SML and TP conceptualised and designed the study. TC generated the estimates of age at onset of Aβ. EKO performed the other statistical analyses, with advice and review by BG, SF, TP and JDD. PB, KN, VLV, VD, CC, AJS, TP and SML contributed to acquisition or curation of data. CLM, CCR, CC, AJS and SML contributed to funding acquisition. EKO, TP, and SML drafted the manuscript. All authors contributed to the revision and editing of the manuscript and approved the submitted version. Acknowledgement We thank all the participants and their families who took part in the AIBL, ADNI and OASIS studies that were included in the ADOPIC consortium cohort. We also thank all the clinicians who may have referred participants. Data Availability The datasets generated and analysed during the current study are not publicly available due to individual cohort restrictions as outlined at: ADNI: https://adni.loni.usc.edu/data-samples/adni-data/; AIBL: https://aibl.org.au/collaboration/#data-access; OASIS: https://sites.wustl.edu/oasisbrains/home/access/. However, data are available from the corresponding author on reasonable request, with the permission of the participating cohort studies. References Knopman DS, et al. Alzheimer disease. Nat reviews Disease primers. 2021;7(1):33. Van Cauwenberghe C, Van Broeckhoven C, Sleegers K. The genetic landscape of Alzheimer disease: clinical implications and perspectives. Genet Sci. 2016;18(5):421–30. Villemagne VL, et al. Amyloid β deposition, neurodegeneration, and cognitive decline in sporadic Alzheimer's disease: a prospective cohort study. Lancet Neurol. 2013;12(4):357–67. Jack CR, et al. Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 2013;12(2):207–16. Selkoe DJ, Hardy J. The amyloid hypothesis of Alzheimer's disease at 25 years. EMBO Mol Med. 2016;8(6):595–608. Hardy JA, Higgins GA. Alzheimer's disease: the amyloid cascade hypothesis. Science. 1992;256(5054):184–5. Sperling RA, Jack CR Jr, Aisen PS. Testing the right target and right drug at the right stage. Science translational medicine, 2011. 3(111): p. 111cm33-111cm33. Harman D. Alzheimer's disease pathogenesis: role of aging. Ann N Y Acad Sci. 2006;1067(1):454–60. Gatz M, et al. Role of genes and environments for explaining Alzheimer disease. Arch Gen Psychiatry. 2006;63(2):168–74. Ward A, et al. Prevalence of apolipoprotein E4 genotype and homozygotes (APOE e4/4) among patients diagnosed with Alzheimer’s disease: a systematic review and meta-analysis. Neuroepidemiology. 2012;38(1):1–17. Lim YY, et al. APOE genotype and early β-amyloid accumulation in older adults without dementia. Neurology. 2017;89(10):1028–34. Lambert J-C, et al. Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer's disease. Nat Genet. 2013;45(12):1452–8. Kunkle BW, et al. Genetic meta-analysis of diagnosed Alzheimer’s disease identifies new risk loci and implicates Aβ, tau, immunity and lipid processing. Nat Genet. 2019;51(3):414–30. Hollingworth P, et al. Common variants at ABCA7, MS4A6A/MS4A4E, EPHA1, CD33 and CD2AP are associated with Alzheimer's disease. Nat Genet. 2011;43(5):429–35. Naj AC, et al. Common variants at MS4A4/MS4A6E, CD2AP, CD33 and EPHA1 are associated with late-onset Alzheimer's disease. Nat Genet. 2011;43(5):436–41. Marioni RE, et al. GWAS on family history of Alzheimer’s disease. Translational psychiatry. 2018;8(1):99. Jansen IE, et al. Genome-wide meta-analysis identifies new loci and functional pathways influencing Alzheimer’s disease risk. Nat Genet. 2019;51(3):404–13. Bellenguez C, et al. New insights into the genetic etiology of Alzheimer’s disease and related dementias. Nat Genet. 2022;54(4):412–36. Ramanan VK, et al. GWAS of longitudinal amyloid accumulation on 18F-florbetapir PET in Alzheimer’s disease implicates microglial activation gene IL1RAP. Brain. 2015;138(10):3076–88. Shade LM, et al. GWAS of multiple neuropathology endophenotypes identifies new risk loci and provides insights into the genetic risk of dementia. Nat Genet. 2024;56(11):2407–21. Yan Q, et al. Genome-wide association study of brain amyloid deposition as measured by Pittsburgh Compound-B (PiB)-PET imaging. Mol Psychiatry. 2021;26(1):309–21. Wray NR, et al. From basic science to clinical application of polygenic risk scores: a primer. JAMA psychiatry. 2021;78(1):101–9. Choi SW, Mak TS-H, O’Reilly PF. Tutorial: a guide to performing polygenic risk score analyses. Nat Protoc. 2020;15(9):2759–72. Maher BS. Polygenic scores in epidemiology: risk prediction, etiology, and clinical utility. Curr Epidemiol Rep. 2015;2:239–44. Torkamani A, Wineinger NE, Topol EJ. The personal and clinical utility of polygenic risk scores. Nat Rev Genet. 2018;19(9):581–90. Visscher PM, et al. 10 years of GWAS discovery: biology, function, and translation. Am J Hum Genet. 2017;101(1):5–22. Zhang Y, et al. Estimation of complex effect-size distributions using summary-level statistics from genome-wide association studies across 32 complex traits. Nat Genet. 2018;50(9):1318–26. Yang J, et al. Common SNPs explain a large proportion of the heritability for human height. Nat Genet. 2010;42(7):565–9. Hou J, et al. Polygenic resilience scores capture protective genetic effects for Alzheimer’s disease. Translational psychiatry. 2022;12(1):296. Weiner MW, et al. The Alzheimer's disease neuroimaging initiative: progress report and future plans. Volume 6. Alzheimer's & Dementia; 2010. pp. 202–11. 3e7. Mueller SG, et al. The Alzheimer's disease neuroimaging initiative. Neuroimaging Clin. 2005;15(4):869–77. Petersen RC, et al. Alzheimer's disease neuroimaging initiative (ADNI): clinical characterization. Neurology. 2010;74(3):201–9. Ellis KA, et al. The Australian Imaging, Biomarkers and Lifestyle (AIBL) study of aging: methodology and baseline characteristics of 1112 individuals recruited for a longitudinal study of Alzheimer's disease. Int Psychogeriatr. 2009;21(4):672–87. Fowler C, et al. Fifteen years of the australian imaging, biomarkers and lifestyle (AIBL) study: progress and observations from 2,359 older adults spanning the spectrum from cognitive normality to Alzheimer’s disease. J Alzheimer's disease Rep. 2021;5(1):443–68. LaMontagne PJ et al. OASIS-3: longitudinal neuroimaging, clinical, and cognitive dataset for normal aging and Alzheimer disease. MedRxiv, 2019: p. 2019.12. 13.19014902. Fernandez MV, et al. Genetic and multi-omic resources for Alzheimer disease and related dementia from the Knight Alzheimer Disease Research Center. Sci data. 2024;11(1):768. Taliun D et al. Sequencing of 53,831 diverse genomes from the NHLBI TOPMed Program. bioRxiv, 2019: p. 563866. Das S, et al. Next-generation genotype imputation service and methods. Nat Genet. 2016;48(10):1284–7. Chang CC et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience, 2015. 4(1): p. s13742-015-0047-8. Purcell S, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81(3):559–75. Bourgeat P, et al. Comparison of MR-less PiB SUVR quantification methods. Neurobiol Aging. 2015;36:S159–66. Bourgeat P, et al. Implementing the centiloid transformation for 11C-PiB and β-amyloid 18F-PET tracers using CapAIBL. NeuroImage. 2018;183:387–93. Su Y, et al. Utilizing the Centiloid scale in cross-sectional and longitudinal PiB PET studies. Volume 19. NeuroImage: Clinical; 2018. pp. 406–16. Bourgeat P et al. β-amyloid PET harmonisation across longitudinal studies: Application to AIBL, ADNI and OASIS3. Neuroimage, 2022. 262: p. 119527. Burnham SC et al. When Does Alzheimer's Disease Start? Robust Estimates Based on Longitudinal Aβ-Amyloid-PET in Three Large International Cohorts. Available at SSRN: http://dx.doi.org/10.2139/ssrn.4757682 , 2024. Choi SW, O'Reilly PF. PRSice-2: Polygenic Risk Score software for biobank-scale data. Gigascience. 2019;8(7):giz082. Euesden J, Lewis CM, O’reilly PF. PRSice: polygenic risk score software. Bioinformatics. 2015;31(9):1466–8. Ware EB, et al. Considering the APOE locus in Alzheimer’s disease polygenic scores in the Health and Retirement Study: a longitudinal panel study. BMC Med Genom. 2020;13:1–13. Core Team R. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing: Vienna, Austria; 2024. Posit team, RStudio: Integrated Development Environment for R. 2025, Posit Software, PBC: Boston, MA, USA. Picard RR, Cook RD. Cross-validation of regression models. J Am Stat Assoc. 1984;79(387):575–83. Xu Q-S, Liang Y-Z. Monte Carlo cross validation. Chemometr Intell Lab Syst. 2001;56(1):1–11. Stouffer S. A study of attitudes. Sci Am. 1949;180(5):11–5. Stocker H, et al. Alzheimer’s polygenic risk scores, APOE, Alzheimer’s disease risk, and dementia-related blood biomarker levels in a population-based cohort study followed over 17 years. Volume 15. Alzheimer's Research & Therapy; 2023. p. 129. 1. Darst BF, et al. Pathway-specific polygenic risk scores as predictors of amyloid-β deposition and cognitive function in a sample at increased risk for Alzheimer’s disease. J Alzheimer’s Disease. 2016;55(2):473–84. Leonenko G, et al. Genetic risk for alzheimer disease is distinct from genetic risk for amyloid deposition. Ann Neurol. 2019;86(3):427–35. Escott-Price V, et al. Common polygenic variation enhances risk prediction for Alzheimer’s disease. Brain. 2015;138(12):3673–84. Ebenau JL, et al. Risk of dementia in APOE ε4 carriers is mitigated by a polygenic risk score. Volume 13. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring,; 2021. p. e12229. 1. Bocancea DI, et al. Measuring resilience and resistance in aging and Alzheimer disease using residual methods: a systematic review and meta-analysis. Neurology. 2021;97(10):474–88. Arenaza-Urquijo EM, Vemuri P. Resistance vs resilience to Alzheimer disease: clarifying terminology for preclinical studies. Neurology. 2018;90(15):695–703. Arenaza-Urquijo EM, Vemuri P. Improving the resistance and resilience framework for aging and dementia studies. Volume 12. Alzheimer's research & therapy; 2020. pp. 1–4. Schmidt K, et al. Structure, function, and genetics of lipoprotein (a). J Lipid Res. 2016;57(8):1339–59. Leszek J, et al. The links between cardiovascular diseases and Alzheimer's disease. Curr Neuropharmacol. 2021;19(2):152–69. de Alves TF. Cardiac disorders as risk factors for Alzheimer's disease. J Alzheimer’s Disease. 2010;20(3):749–63. Delrieu J, Andrieu S, Vellas B. Dementia research in 2023: the year of anti-amyloid immunotherapy. Lancet Neurol. 2024;23(1):13–5. Van Dyck CH, et al. Lecanemab in early Alzheimer’s disease. N Engl J Med. 2023;388(1):9–21. Sims JR, et al. Donanemab in early symptomatic Alzheimer disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA. 2023;330(6):512–27. Reiman EM, et al. A chance to prevent Alzheimer's disease sooner than you think. Lancet Neurol. 2024;23(2):144–5. Kember RL et al. Considerations for the application of polygenic scores to clinical care of individuals with substance use disorders. J Clin Investig, 2024. 134(20). Rodrigue K, et al. β-Amyloid burden in healthy aging: regional distribution and cognitive consequences. Neurology. 2012;78(6):387–95. Kachuri L, et al. Principles and methods for transferring polygenic risk scores across global populations. Nat Rev Genet. 2024;25(1):8–25. Additional Declarations Competing interest reported. CC is a member of the scientific advisory board of Circular Genomics and owns stocks, and is on the scientific advisory board of ADmit and Alamar, consults for Sanofi, NovoNordisk, and Owkin, and has received research support from GSK, Danaher and EISAI. All other authors report no competing interests relevant to this manuscript. Supplementary Files SupptablesgeneticsofAAO.xlsx Cite Share Download PDF Status: Posted Version 1 posted 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-7911284","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":532664414,"identity":"0c2463e8-653b-40cd-a510-54c20499acab","order_by":0,"name":"Eleanor K O'Brien","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0klEQVRIiWNgGAWjYHACNgYGAwYGfgiHmQQtkg2kaQECgwPEauFnYD72uKLAJs/4ePMzCYYK68QG9jMGeLVINrClG54xSCs2O3PMTILhTHpiA08Ofi0GB3jMJBsMDiduu5HDJsHYdjixgYGAFnuIlv+Jm+e/AWr5B9TC/4aALQxgLQcSN0jwALU0ALVIELBF4jDQLw0GyYkzzqQZWyQcSzduk3hWgFcLf3vzsYcNf+wS+9sPP7zxocZatp8/eQNeLagRkcAAi6ZRMApGwSgYBRQBAI8FPytWQIQOAAAAAElFTkSuQmCC","orcid":"","institution":"Edith Cowan University","correspondingAuthor":true,"prefix":"","firstName":"Eleanor","middleName":"K","lastName":"O'Brien","suffix":""},{"id":532664415,"identity":"927ad9ec-7415-4b52-83c0-abb1c6afae96","order_by":1,"name":"Timothy Cox","email":"","orcid":"","institution":"Commonwealth Scientific and Industrial Research Organisation","correspondingAuthor":false,"prefix":"","firstName":"Timothy","middleName":"","lastName":"Cox","suffix":""},{"id":532664416,"identity":"60d3548e-8006-4fe6-89b9-692e1d16a471","order_by":2,"name":"Shane Fernandez","email":"","orcid":"","institution":"Edith Cowan University","correspondingAuthor":false,"prefix":"","firstName":"Shane","middleName":"","lastName":"Fernandez","suffix":""},{"id":532664417,"identity":"16aa92a8-3d1b-4664-a8f8-2f9584986928","order_by":3,"name":"Pierrick Bourgeat","email":"","orcid":"","institution":"Commonwealth Scientific and Industrial Research Organisation","correspondingAuthor":false,"prefix":"","firstName":"Pierrick","middleName":"","lastName":"Bourgeat","suffix":""},{"id":532664418,"identity":"2829b8f5-82f2-4ca3-bf3e-2614ece6663d","order_by":4,"name":"Tenielle Porter","email":"","orcid":"","institution":"Edith Cowan University","correspondingAuthor":false,"prefix":"","firstName":"Tenielle","middleName":"","lastName":"Porter","suffix":""},{"id":532664419,"identity":"8f434b93-d4aa-4292-9ba7-de4b34e9d3e3","order_by":5,"name":"Ben Goudey","email":"","orcid":"","institution":"University of Melbourne","correspondingAuthor":false,"prefix":"","firstName":"Ben","middleName":"","lastName":"Goudey","suffix":""},{"id":532664420,"identity":"7690a123-a19b-4e0a-8978-2c9740dfff74","order_by":6,"name":"James D Doecke","email":"","orcid":"","institution":"Commonwealth Scientific and Industrial Research Organisation","correspondingAuthor":false,"prefix":"","firstName":"James","middleName":"D","lastName":"Doecke","suffix":""},{"id":532664421,"identity":"09b5a92a-5387-4f1d-bf62-b3cc013bb578","order_by":7,"name":"Colin L Masters","email":"","orcid":"","institution":"University of Melbourne","correspondingAuthor":false,"prefix":"","firstName":"Colin","middleName":"L","lastName":"Masters","suffix":""},{"id":532664422,"identity":"54d2e590-9c4f-4c87-a546-6c93eb5666ca","order_by":8,"name":"Jurgen Fripp","email":"","orcid":"","institution":"Commonwealth Scientific and Industrial Research Organisation","correspondingAuthor":false,"prefix":"","firstName":"Jurgen","middleName":"","lastName":"Fripp","suffix":""},{"id":532664423,"identity":"1a878f5d-456c-4b5a-bfb0-ef5858b305f0","order_by":9,"name":"Kwangsik Nho","email":"","orcid":"","institution":"Indiana University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Kwangsik","middleName":"","lastName":"Nho","suffix":""},{"id":532664424,"identity":"b7cca18d-4253-4f22-94d5-d0d0d2f56e1e","order_by":10,"name":"Victor L Villemagne","email":"","orcid":"","institution":"University of Pittsburgh","correspondingAuthor":false,"prefix":"","firstName":"Victor","middleName":"L","lastName":"Villemagne","suffix":""},{"id":532664425,"identity":"a862ed4b-6872-449e-a92e-da07eedb4f8c","order_by":11,"name":"Carlos Cruchaga","email":"","orcid":"","institution":"Washington University","correspondingAuthor":false,"prefix":"","firstName":"Carlos","middleName":"","lastName":"Cruchaga","suffix":""},{"id":532664426,"identity":"99ece69f-2ef9-46c3-92dd-b88e8c8f0ca3","order_by":12,"name":"Christopher C Rowe","email":"","orcid":"","institution":"Austin Health","correspondingAuthor":false,"prefix":"","firstName":"Christopher","middleName":"C","lastName":"Rowe","suffix":""},{"id":532664427,"identity":"2d76b304-0996-4800-bcf0-bf2c09b2c516","order_by":13,"name":"Andrew J Saykin","email":"","orcid":"","institution":"Indiana University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Andrew","middleName":"J","lastName":"Saykin","suffix":""},{"id":532664428,"identity":"e3a910b7-5e40-4bb2-8317-8db7e66dc953","order_by":14,"name":"Vincent Dore","email":"","orcid":"","institution":"Commonwealth Scientific and Industrial Research Organisation","correspondingAuthor":false,"prefix":"","firstName":"Vincent","middleName":"","lastName":"Dore","suffix":""},{"id":532664429,"identity":"99312e5c-7494-45eb-8690-7d0645935cd3","order_by":15,"name":"Simon M Laws","email":"","orcid":"","institution":"Edith Cowan University","correspondingAuthor":false,"prefix":"","firstName":"Simon","middleName":"M","lastName":"Laws","suffix":""}],"badges":[],"createdAt":"2025-10-21 06:53:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7911284/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7911284/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":94136484,"identity":"b72815c7-70c7-465a-a677-9bc4c47b2a9f","added_by":"auto","created_at":"2025-10-22 19:08:44","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1476895,"visible":true,"origin":"","legend":"","description":"","filename":"OBrienetalAAOAB141025.docx","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/c41e44acac2886b348023a9e.docx"},{"id":94136483,"identity":"6b62c6b7-8138-4669-bf16-4444fb00ed6a","added_by":"auto","created_at":"2025-10-22 19:08:44","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16517,"visible":true,"origin":"","legend":"","description":"","filename":"fa0e38782ca34b3087e01e00c12a4f04.json","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/7765074bb32935dd5ff0005f.json"},{"id":94138412,"identity":"7db7183e-2800-4c81-8102-eb72a4fc8852","added_by":"auto","created_at":"2025-10-22 19:24:44","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":123149,"visible":true,"origin":"","legend":"","description":"","filename":"SupptablesgeneticsofAAO.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/35bf05885bb477ae80025564.xlsx"},{"id":94135791,"identity":"8b702ec2-bfee-4939-ba46-bdc2bb256bd7","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159344,"visible":true,"origin":"","legend":"","description":"","filename":"fa0e38782ca34b3087e01e00c12a4f041enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/affd8678a448644bf3ccaca5.xml"},{"id":94135795,"identity":"7b267051-6be2-445d-9c3e-061dfee951ee","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":559920,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/4ac7d9d9179a4aaadd6affa8.jpeg"},{"id":94136488,"identity":"c2f23e79-dced-4e41-8d7b-f834d7ca75b1","added_by":"auto","created_at":"2025-10-22 19:08:44","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":161565,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/72ee0fef8acc1328d3e3e743.png"},{"id":94136489,"identity":"5214cecb-c470-4e3f-96e5-1d432d566a71","added_by":"auto","created_at":"2025-10-22 19:08:44","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":426026,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/353bada3412f7063d1c558e0.jpeg"},{"id":94136487,"identity":"5b1a279b-c4e0-4998-ab3a-b0b212a6d7f5","added_by":"auto","created_at":"2025-10-22 19:08:44","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":188344,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/60317d892d24692031ddb74b.jpeg"},{"id":94135789,"identity":"d26ddfda-5b12-46ea-8969-0acb3787eda9","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52690,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/ccd12f1ebb6daccb1730af79.png"},{"id":94135792,"identity":"11efa687-d3e4-4c60-8138-5af620724642","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31285,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/b755af386e733d0d8ef8edef.png"},{"id":94135796,"identity":"2a81b845-cd34-4a64-88a1-f0e1b291d1fe","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64861,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/5f7b1de0cf3749e1d3a76086.png"},{"id":94135793,"identity":"737ab35c-5057-42f7-9fa1-3948548f859f","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30727,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/b5e01a49d3d4f4563afe3093.png"},{"id":94135800,"identity":"6640d884-6b72-49ac-9886-dacf830703a7","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"xml","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":156831,"visible":true,"origin":"","legend":"","description":"","filename":"fa0e38782ca34b3087e01e00c12a4f041structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/1dd7eb5b2c0f25a9f7410070.xml"},{"id":94135799,"identity":"6944ced6-99c3-4481-88ae-f25f6444cbd2","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"html","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":172628,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/d35f155143ac4fef75ab4c8c.html"},{"id":94135783,"identity":"c436bbf4-6f1c-4bd1-8d25-72d4da306399","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":199834,"visible":true,"origin":"","legend":"\u003cp\u003eSample selection from the ADOPIC study\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/174c4f004696eb8acd77147c.png"},{"id":94135781,"identity":"ec33efea-4fb4-4696-806e-7dbb7c657eba","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":169442,"visible":true,"origin":"","legend":"\u003cp\u003ePlots of standardised polygenic scores (PGS) based on genetic risk (top) and resilience (bottom) to Alzheimer’s disease (AD) against probability of being an accumulator of amyloid β. Lines show predicted values from logistic regression models and shading indicates the 95% confidence interval. Plots on the left show the overall relationship for 2175 participants from the ADOPIC study. Plots on the right show the relationship stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/d44e61c05ae90592106481ca.png"},{"id":94136583,"identity":"02c58182-661b-4084-9279-04c9aff7a2fe","added_by":"auto","created_at":"2025-10-22 19:16:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":232409,"visible":true,"origin":"","legend":"\u003cp\u003ePlots of standardised polygenic scores (PGS) based on genetic risk (top) and resilience (bottom) to Alzheimer’s disease (AD) against estimated age at onset of amyloid β (age when Aβ is estimated to exceed 20 CL). Lines show predicted values from linear regression models and shading indicates the 95% confidence interval. Plots on the left show the overall relationship for 1158 participants from the ADOPIC study. Plots on the right show the relationship stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/21cf46dcac0a9a1f5c24fba8.png"},{"id":94135785,"identity":"503c3dc8-e79a-44e5-882b-3e34371dd95f","added_by":"auto","created_at":"2025-10-22 19:00:44","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":94119,"visible":true,"origin":"","legend":"\u003cp\u003eForest plots of the relationship of standardised polygenic scores (PGS) with (a) Accumulator status (probability of being an accumulator of amyloid β) and (b) Estimated age at onset of Aβ (age when Aβ is estimated to exceed 20 CL). Three PGSs were evaluated for each trait: risk and resilience to AD based on previously published genome-wide association studies (GWAS) and a phenotype-specific PGS based on an \u003cem\u003ea priori\u003c/em\u003e GWAS within the current data set, using a cross-validation (CV) approach. Grey dots and lines indicate the odds ratio and its 95 % confidence interval (accumulator status) or the coefficient and standard error of the linear relationship (age at onset of Aβ) of the standardised PGS against the phenotype within the validation set of each CV run. Coloured diamonds show the mean odds ratio or coefficient for each PGS across the 10 CV runs. Asterisks indicate a significant (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) association of PGS with the trait\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/3ae4f01d922f187c74843b8b.png"},{"id":94473309,"identity":"999ae372-19f6-42ea-b6c3-c67cff9aec19","added_by":"auto","created_at":"2025-10-27 15:43:57","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1719196,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/a2df11f9-f3ba-4524-80ad-400dc05d5da4.pdf"},{"id":94136482,"identity":"9801f955-e3b4-4eb6-965f-98eec943c558","added_by":"auto","created_at":"2025-10-22 19:08:44","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":123149,"visible":true,"origin":"","legend":"","description":"","filename":"SupptablesgeneticsofAAO.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7911284/v1/65febdf42dd4ebb5694628f4.xlsx"}],"financialInterests":"Competing interest reported. CC is a member of the scientific advisory board of Circular Genomics and owns stocks, and is on the scientific advisory board of ADmit and Alamar, consults for Sanofi, NovoNordisk, and Owkin, and has received research support from GSK, Danaher and EISAI. All other authors report no competing interests relevant to this manuscript.","formattedTitle":"Predicting accumulation and age at onset of amyloid-β from genetic risk and resilience for Alzheimer’s disease","fulltext":[{"header":"Background","content":"\u003cp\u003eAlzheimer\u0026rsquo;s disease (AD) is a neurodegenerative disorder, characterised by progressive cognitive decline and memory loss [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. It is the most common form of dementia, accounting for around 65% of cases [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. The accumulation of amyloid beta (Aβ) in the brain is one of the key pathological hallmarks of AD, beginning years, if not decades, before the onset of clinical symptoms [\u003cspan additionalcitationids=\"CR4\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. The \u0026lsquo;amyloid cascade hypothesis\u0026rsquo; posits that this accumulation precipitates a cascade of events, including tau hyperphosphorylation and formation of neurofibrillary tangles, neuronal loss, cognitive decline and ultimately dementia [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Predicting whether an individual will accumulate Aβ and the age at which they will reach a critical threshold for brain Aβ that leads to these downstream events could therefore enable timely therapeutic strategies aimed at delaying or preventing the onset and progression of cognitive decline [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eLate onset AD (LOAD) occurs sporadically in people aged over 65 years, and makes up 90\u0026ndash;95% of AD cases [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. LOAD (henceforth referred to as AD) is a complex disease, with risk determined by a combination of genetic, environmental and lifestyle factors [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Among the genetic factors, the Apolipoprotein E (\u003cem\u003eAPOE\u003c/em\u003e) gene on chromosome 19 shows the strongest association with the incidence of AD, with the \u003cem\u003eAPOE\u003c/em\u003e ε4 allele conferring a dose-dependent increase in risk of AD, and the ε2 allele being protective [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. \u003cem\u003eAPOE\u003c/em\u003e genotype is also associated with the magnitude and timing of Aβ accumulation, such that ε4 carriage is associated with higher average age-adjusted Aβ burden and younger ages when critical Aβ thresholds are reached [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. However, genome-wide association studies (GWAS) have identified more than 80 additional genetic variants that show a significant association with AD risk, albeit with smaller individual effects than \u003cem\u003eAPOE\u003c/em\u003e [\u003cspan additionalcitationids=\"CR13 CR14 CR15 CR16 CR17\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. GWAS for Aβ accumulation and burden have also identified several significantly associated loci in addition to those within the \u003cem\u003eAPOE\u003c/em\u003e region [\u003cspan additionalcitationids=\"CR20\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. These findings suggest that the prediction of AD and related traits can be improved by considering genetic factors beyond \u003cem\u003eAPOE\u003c/em\u003e.\u003c/p\u003e\u003cp\u003ePolygenic scores (PGS; also referred to as \u0026lsquo;genetic scores\u0026rsquo; or \u0026lsquo;polygenic risk scores\u0026rsquo;) aggregate the effects of variants across the genome that are associated with the disease or trait of interest [\u003cspan additionalcitationids=\"CR23 CR24\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. This approach recognises that complex traits are often influenced by many genetic variants, most of small effect, and many of which will not meet the stringent threshold for genome-wide significance [\u003cspan additionalcitationids=\"CR27\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The resulting score gives an estimate of genetic susceptibility to a trait or disease, which may then be used, alongside other clinical indicators, to inform decisions around preventative actions or interventions [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. PGSs developed from genetic associations with disease incidence may also be useful for predicting endophenotypes linked to that disease (or \u003cem\u003evice versa\u003c/em\u003e), which can provide further insights into the underlying biological disease mechanisms and aid in earlier diagnosis. In the context of AD, genetic risk of disease (as estimated from GWASs for incidence of AD) may therefore be useful for predicting onset and progression of pathophysiological features such as Aβ that occur earlier in the disease trajectory, prior to diagnosis.\u003c/p\u003e\u003cp\u003eIn the present study, we develop PGSs based on previously published GWASs for risk of and resilience to AD [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Here, \u0026lsquo;resilience\u0026rsquo; was defined as remaining cognitively unimpaired in the face of high genetic risk for AD [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. We test the extent to which these PGSs predict whether an individual accumulates brain Aβ and the age at which Aβ is estimated to reach a critical threshold. We then assess whether we can improve on these predictions using \u003cem\u003ea priori\u003c/em\u003e GWASs to develop phenotype-specific PGSs for each of these traits. We seek to develop predictive models that can identify individuals at high risk for early Aβ pathology, to inform personalised approaches to the prevention and treatment of AD.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStudy population\u003c/h2\u003e\u003cp\u003eThe study included data for 2175 participants from the Alzheimer\u0026rsquo;s Dementia Onset and Progression in International Cohorts (ADOPIC) study, which combines three longitudinal cohort studies: the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI)[\u003cspan additionalcitationids=\"CR31\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], the Australian Imaging, Biomarkers and Lifestyle Study of Ageing (AIBL)[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], and the Knight ADRC Open Access Series of Imaging Studies (OASIS)[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Full details of these studies, including recruitment criteria and the schedule of assessments have been described previously [\u003cspan additionalcitationids=\"CR31 CR32 CR33 CR34\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. The studies have all been granted approval by the ethics committees of their respective member institutions, and all participants provided informed written consent. The present study was limited to ADOPIC participants of European ancestry.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eGenotyping and imputation\u003c/h3\u003e\n\u003cp\u003eDNA was extracted from whole blood samples. For ADNI and AIBL participants, this was done using QIAamp DNA blood spin column kits (Qiagen, Valencia, CA, USA) [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], and for OASIS participants using the Autogen FlexSTAR\u0026thinsp;+\u0026thinsp;salt precipitation [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eADNI samples were genotyped at genome-wide single nucleotide polymorphisms (SNPs) on either the Illumina Human610-Quad BeadChip or the Illumina Human OmniExpress BeadChip[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. AIBL samples were genotyped on the Axiom Precision Medicine Diversity Array (Applied Biosystems\u0026trade;)[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. OASIS samples were genotyped on one of nine arrays (Illumina Human660W-Quad, Infinium OmniExpressExome-8, Illumina Omni1-Quad, Illumina Human1M-Duo, Infinium Neuro Consortium Array, Infinium CoreExome-24, Infinium Global Screening Array-24, Illumina Human610-Quad, and UK Biobank Axiom array)[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eAll genotype data were imputed against the TOPMed panel on the TOPMed Imputation Server (University of Michigan, USA) [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. Within cohorts, this was done for each genotyping run, and then the imputed data were combined across the three cohorts to produce a single harmonised genetic data set. Only autosomal SNPs were included in the final genetic data set. We removed SNPs with an imputation quality r\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026le;\u0026thinsp;0.3, samples and SNPs with \u0026gt;\u0026thinsp;2% missing genotypes after merging, as well as SNPs with minor allele frequency (MAF)\u0026thinsp;\u0026lt;\u0026thinsp;5%, and those where the p-value from a test for Hardy-Weinberg equilibrium was \u0026lt;\u0026thinsp;1 x 10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e. We used PLINK [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e] to estimate relatedness (pihat) between all pairs of participants and removed one member of each pair of individuals with pihat\u0026thinsp;\u0026gt;\u0026thinsp;0.25. We also used PLINK to conduct a principal components analysis (PCA) on all remaining unrelated individuals to obtain principal components (PCs) to include as covariates to control for population structure in genetic analyses (see below). Genetic data were available for 6157 individuals, of whom 131 were removed due to relatedness, giving a final genetic data set of 6026 individuals (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). \u003cem\u003eAPOE\u003c/em\u003e genotype was determined from TaqMan\u0026reg; genotyping assays (Life Technologies, USA) at two SNPs: rs7412 (Assay ID: C__904973_10) and rs429358 (Assay ID: C__3084793_20).\u003c/p\u003e\n\u003ch3\u003eAβ PET Imaging\u003c/h3\u003e\n\u003cp\u003eMeasures of brain β-amyloid (Aβ) burden were obtained for participants in all three cohorts using positron emission tomography (PET) imaging with one of the following five tracers: \u003csup\u003e11\u003c/sup\u003eC\u0026ndash;Pittsburgh compound B (PiB), \u003csup\u003e18\u003c/sup\u003eF-NAV4694 (NAV), \u003csup\u003e18\u003c/sup\u003eF-florbetaben (FBB), \u003csup\u003e18\u003c/sup\u003eF-florbetapir (FBP), or \u003csup\u003e18\u003c/sup\u003eF-flutemetamol (FLUTE). PET images were analysed using CapAIBL [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e] to generate tracer-specific standardised uptake value ratios (SUVRs), which were then transformed to centiloids (CL) as previously described [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. Harmonisation of CL quantification across the three cohorts has been described previously [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e].\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eTraits\u003c/h3\u003e\n\u003cp\u003e\u003cem\u003eAccumulator Status\u003c/em\u003e was determined for all individuals in the study with Aβ PET imaging at two or more timepoints (N\u0026thinsp;=\u0026thinsp;2175; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Individuals were classified as \u0026ldquo;accumulators\u0026rdquo; if they had Aβ PET\u0026thinsp;\u0026ge;\u0026thinsp;20 CL at any assessment or showed an increase in Aβ PET across timepoints of \u0026ge;\u0026thinsp;0.05 CL/year, and as \u0026ldquo;non-accumulators\u0026rdquo; otherwise [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e].\u003c/p\u003e\u003cp\u003e\u003cem\u003eEstimated Age at Onset of Aβ\u003c/em\u003e (AAO-Aβ) was the age at which participants\u0026rsquo; cortical Aβ burden was estimated to have reached 20 CL, obtained by placing them on a natural history curve of Aβ accumulation, as described previously [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. AAO-Aβ was estimated for people classified as \u0026ldquo;accumulators\u0026rdquo; who also met the following criteria: (i) Their Aβ burden at the final scan was \u0026ge;\u0026thinsp;10 CL (due to uncertainty around whether people with brain Aβ below this level will reach the 20 CL threshold), and (ii) if brain Aβ did not already exceed the 20 CL threshold, it was predicted to do so within 5 years. In total, AAO-Aβ was estimated for 1158 individuals (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003eData analysis\u003c/h2\u003e\u003cdiv id=\"Sec8\" class=\"Section3\"\u003e\u003ch2\u003eDevelopment of cross-trait polygenic scores\u003c/h2\u003e\u003cp\u003eWe obtained summary statistics from genome-wide association studies (GWAS) for AD risk [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] and resilience [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. We then used PRSice-2 [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e] to identify optimal PGSs for each trait (Accumulator Status and AAO-Aβ), based on risk and resilience GWASs. This process finds the p-value threshold for variant selection that results in a set of variants that explains the largest proportion of variance in the target phenotype, and uses these to create a PGS. To determine whether these risk and resilience PGSs could account for significant variation in each trait beyond that explained by \u003cem\u003eAPOE\u003c/em\u003e (the gene most strongly associated with AD, located on chromosome 19), we excluded chromosome 19 from the genetic data set, and included number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles as a covariate, as recommended by Ware et al (2020) [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Additional covariates were sex and study (which of the three cohort studies the participant came from). To control for population structure, we also included the first three principal components from the PCA. We used an additive genetic model with default clumping and thinning parameters (250kb distance, r\u003csup\u003e2\u003c/sup\u003e threshold 0.1). PGSs were standardised to a mean of 0 and standard deviation of 1.\u003c/p\u003e\u003cp\u003eFour PGSs were derived, defined by the GWAS summary stats used (risk or resilience) and the trait being predicted (Accumulator Status or AAO-Aβ): PGS\u003csub\u003erisk\u0026minus;Accumulator\u003c/sub\u003e, PGS\u003csub\u003eresilience\u0026minus;Accumulator\u003c/sub\u003e, PGS\u003csub\u003erisk\u0026minus;AAO\u003c/sub\u003e and PGS\u003csub\u003eresilience\u0026minus;AAO\u003c/sub\u003e. We then used these optimal PGSs (calculated for the whole population) for the relevant trait in (generalised) linear models to test their association with that trait in three groups that were stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles: (a) \u003cem\u003eAPOE\u003c/em\u003e ε4 non-carriers, (b) \u003cem\u003eAPOE\u003c/em\u003e ε4 heterozygotes, and (c) \u003cem\u003eAPOE\u003c/em\u003e ε4 homozygotes. These models included the same covariates as described above, excluding \u003cem\u003eAPOE\u003c/em\u003e ε4 allele count. For Accumulator Status, we fitted logistic regression models using the \u0026lsquo;glm\u0026rsquo; function in R with a binomial distribution and logit link function, and for AAO-Aβ we fitted linear regression models using the \u0026lsquo;lm\u0026rsquo; function (which assumes a gaussian distribution) in R (version 4.3.3)[\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. In each case, these were implemented in RStudio (version 2024.12.1 Build 563)[\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]. Beta coefficients for the PGS term in logistic regressions were exponentiated to convert them to odds ratios (OR), which represent the change in odds of being an accumulator with each 1 SD increase in the PGS. For linear regressions on AAO-Aβ, beta coefficients for the PGS term represent the change in estimated AAO-Aβ (in years) with a 1 SD increase in the PGS.\u003c/p\u003e\u003cp\u003eTo examine the change in odds of being an accumulator and estimated AAO-Aβ between PGS extremes, we also ran models comparing people with scores in the upper and lower quintiles (top and bottom 20%) of the population for each PGS and trait. Threshold PGS scores for the upper and lower quintiles were obtained for all genotyped individuals (N\u0026thinsp;=\u0026thinsp;6026), and then phenotyped individuals in these ranges were extracted for analysis. In these models, PGS was a categorical predictor with levels \u0026ldquo;low\u0026rdquo; (lower quintile) and \u0026ldquo;high\u0026rdquo; (upper quintile). Covariates were the same as in models that included the full range of PGS values, and models were again run for all individuals and stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\n\u003ch3\u003eGWAS and development of trait-specific PGS using cross-validation\u003c/h3\u003e\n\u003cp\u003eDue to the relatively small size of this study, we used a Monte Carlo cross validation approach [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e, \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e] to derive phenotype-specific PGSs. We created 10 cross validation (CV) runs, where within each run the data were split into discovery (two-thirds of individuals; Accumulator Status N\u0026thinsp;=\u0026thinsp;1450, AAO-Aβ N\u0026thinsp;=\u0026thinsp;772) and validation (one-third of individuals; Accumulator Status N\u0026thinsp;=\u0026thinsp;725, AAO-Aβ N\u0026thinsp;=\u0026thinsp;386) sets using the \u0026lsquo;sample\u0026rsquo; function in R (v 4.3.3) [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e], run in R Studio (v 2024.12.1 Build 563)[\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]. We conducted a genome-wide association study (GWAS) for the trait of interest in the discovery set and used the summary statistics to find the optimal PGS in the corresponding validation set. This process was repeated in each of the 10 CV runs for each trait.\u003c/p\u003e\u003cp\u003eWe ran GWASs for each trait in each discovery data set in PLINK [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e], using a logistic regression model for Accumulator Status and linear regression for estimated AAO-Aβ. We assumed an additive genetic model, and included sex, study, and the first three principal components as covariates. To identify the optimal PGS for the trait in the corresponding validation set, we PRSice-2 and followed the same protocol as for the risk and resilience PGSs described above, with the same covariates and clumping parameters, and again with chromosome 19 excluded. Here, the \u0026lsquo;effective allele\u0026rsquo; was the allele associated with worse outcomes (higher odds of being an accumulator and earlier estimated AAO-Aβ). Therefore in each case, higher values for these scores are expected to be associated with these worse outcomes. To enable a direct comparison of the performance of the trait-specific PGSs with the risk and resilience PGSs, we tested the association of the optimal risk and resilience PGSs (determined from the whole cohort) with the traits in the validation data set of each CV run, using the same set of covariates as previously.\u003c/p\u003e\u003cp\u003eTo compare the predictive performance of each PGS for each trait, we summarised performance across all CV runs. For Accumulator Status, we calculated the mean odds ratio and 95% confidence interval, and for estimated AAO-Aβ, we calculate the mean and standard error of the coefficient of the linear relationship of the standardised PGS against the phenotype across the 10 CV runs. For each PGS and trait, the overall R\u003csup\u003e2\u003c/sup\u003e value was estimated as the mean of the R\u003csup\u003e2\u003c/sup\u003e values across CV runs (Nagelkerke\u0026rsquo;s pseudo R\u003csup\u003e2\u003c/sup\u003e for Accumulator Status), and the overall p-value was estimated from the p-values in each of the 10 CV runs using Stouffer\u0026rsquo;s method [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e], with p-values weighted by the inverse of the standard error and adjusting for the direction of association in each run.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eDemographics\u003c/h2\u003e\u003cp\u003eDemographic characteristics of all participants with at least two Aβ PET scans (N\u0026thinsp;=\u0026thinsp;2175), and of the subset of participants with estimated AAO-Aβ (N\u0026thinsp;=\u0026thinsp;1158), stratified by cohort study, are shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Participants from OASIS were on average younger, a higher proportion of them were female, and there were fewer \u003cem\u003eAPOE\u003c/em\u003e ε4 homozygotes than in ADNI or AIBL. A higher proportion of participants from ADNI were accumulators of Aβ compared to the other cohort studies. In the subset of participants with AAO-Aβ, all of whom were accumulators of Aβ, OASIS participants were still younger and with lower final Aβ burden than those from ADNI and AIBL (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Participant counts and percentages by Accumulator Status, for each \u003cem\u003eAPOE\u003c/em\u003e ε4 group, are shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The percentage of participants who were accumulators of Aβ increased with number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles, from 63.1% among non-carriers, to 84% of ε4 heterozygotes and 95.6% of ε4 homozygotes (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Consistent with previous studies, \u003cem\u003eAPOE\u003c/em\u003e ε4 carriage was strongly associated with both Accumulator Status and estimated AAO-Aβ in a dose-dependent manner, with an increasing number of ε4 alleles associated with higher odds of being an accumulator of Aβ (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e) and earlier estimated AAO-Aβ (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, Table S2).\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\u003eSummary of demographic characteristics of (a) all participants included in the study and (b) participants that met the criteria for inclusion in analysis of estimated AAO-Aβ, shown overall and stratified by study cohort. P-values are from comparisons between study cohorts using either a Pearson\u0026rsquo;s chi square test (sex, accumulator status, no. of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles) or a Kruskal-Wallis rank sum test (final age, final PET Aβ, estimated AAO).\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\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\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e\u003cp\u003e(a) All participants\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCharacteristic\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOverall\u003c/p\u003e\u003cp\u003eN\u0026thinsp;=\u0026thinsp;2175\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eADNI\u003c/p\u003e\u003cp\u003eN\u0026thinsp;=\u0026thinsp;952\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAIBL\u003c/p\u003e\u003cp\u003eN\u0026thinsp;=\u0026thinsp;930\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eOASIS\u003c/p\u003e\u003cp\u003eN\u0026thinsp;=\u0026thinsp;293\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ep-value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eSex N (%)\u003c/b\u003e\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\u003ctd align=\"left\" colname=\"c6\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003e0.003\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1100 (50.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e450 (47.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e479 (51.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e171 (58.4%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1075 (49.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e502 (52.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e451 (48.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e122 (41.6%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAccumulator status N (%)\u003c/b\u003e\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\u003ctd align=\"left\" colname=\"c6\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNon-accumulator\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e619 (28.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e166 (17.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e362 (38.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e91 (31.1%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAccumulator\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1556 (71.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e786 (82.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e568 (61.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e202 (68.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMean final age yrs (SD)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e76.0 (7.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e77.0 (7.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e76.3 (6.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e71.5 (8.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eNo. of\u003c/b\u003e \u003cb\u003eAPOE\u003c/b\u003e \u003cb\u003eε4 alleles N (%)\u003c/b\u003e\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\u003ctd align=\"left\" colname=\"c6\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003e0.021\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1372 (63.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e567 (59.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e611 (65.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e194 (66.2%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e668 (30.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e319 (33.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e261 (28.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e88 (30.0%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e135 (6.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66 (6.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e58 (6.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e11 (3.8%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e\u003cp\u003e(b) \u003cb\u003eParticipants with estimated Age at Onset of Aβ\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eCharacteristic\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003eOverall\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eN\u0026thinsp;=\u0026thinsp;1158\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003eADNI\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eN\u0026thinsp;=\u0026thinsp;649\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003eAIBL\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eN\u0026thinsp;=\u0026thinsp;400\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003eOASIS\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eN\u0026thinsp;=\u0026thinsp;109\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003ep-value\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eSex N (%)\u003c/b\u003e\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\u003ctd align=\"left\" colname=\"c6\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003e0.071\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e564 (48.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e301 (46.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e200 (50.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e63 (57.8%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e594 (51.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e348 (53.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e200 (50.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e46 (42.2%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMean final age yrs (SD)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e77.7 (7.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e77.9 (7.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e77.9 (6.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e75.6 (7.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.018\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMean final PET Aβ (SD)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e69.4 (40.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e71.5 (42.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e71.3 (37.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e49.7 (29.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMean estimated AAO-Aβ yrs (SD)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e66.8 (10.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66.5 (11.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e66.7 (10.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e69.1 (8.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.100\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eNo. of\u003c/b\u003e \u003cb\u003eAPOE\u003c/b\u003e \u003cb\u003eε4 alleles N (%)\u003c/b\u003e\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\u003ctd align=\"left\" colname=\"c6\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003e0.400\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e550 (47.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e316 (48.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e185 (46.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e49 (44.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e480 (41.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e268 (41.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e162 (40.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e50 (45.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e128 (11.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e65 (10.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e53 (13.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e10 (9.2%)\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\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\u003eCounts (%) of individuals classified as non-accumulators and accumulators of Aβ, stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles.\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\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eAβ Accumulator Status\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e\u003cp\u003eNumber of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eTOTAL\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eNon-accumulator\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e506 (36.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e107 (16.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e6 (4.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e619 (28.5%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAccumulator\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e866 (63.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e561 (84.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e129 (95.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1556 (71.5%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eTOTAL\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1372\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e668\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e135\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2175\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\u003cem\u003ePolygenic scores derived from risk and resilience GWAS against Accumulator Status and Estimated Age at Onset of Aβ (AAO-Aβ)\u003c/em\u003e\u003c/p\u003e\u003cp\u003eThe optimal risk PGS for Accumulator Status (PGS\u003csub\u003erisk\u0026minus;Accumulator\u003c/sub\u003e) was significantly associated with this trait overall, with an odds ratio (OR) of 1.16 (95% CI 1.05\u0026ndash;1.29)(Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e), indicating that for an increase of 1 standard deviation (SD) in PGS\u003csub\u003erisk\u0026minus;Accumulator\u003c/sub\u003e, the odds of being an accumulator of Aβ increase by 16%. When stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles, the same direction of association was seen in all groups, but this was only significant in \u003cem\u003eAPOE\u003c/em\u003e ε4 heterozygotes (marginally non-significant in ε4 non-carriers, p\u0026thinsp;=\u0026thinsp;0.058) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Individuals in the upper quintile of PGS\u003csub\u003erisk\u0026minus;Accumulator\u003c/sub\u003e were 63% more likely than those in the bottom quintile to be accumulators of Aβ overall (OR\u0026thinsp;=\u0026thinsp;1.63; 95% CI 1.18\u0026ndash;2.26)(Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e), and this difference was also significant within \u003cem\u003eAPOE\u003c/em\u003e ε4 non-carriers and heterozygotes (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). The optimal resilience PGS (PGS\u003csub\u003eresilience\u0026minus;Accumulator\u003c/sub\u003e) was not significantly associated with Accumulator Status overall, or for any group when stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). We also did not detect a significant difference in the probability of being an accumulator between the upper and lower quintiles of PGS\u003csub\u003eresilience\u0026minus;Accumulator\u003c/sub\u003e (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). It is important to note that only 4.4% (6 out of 135) of \u003cem\u003eAPOE\u003c/em\u003e ε4 homozygotes in our study were non-accumulators (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). This was a much lower percentage than in the other ε4 groups (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), and would have limited our power to detect an association with Accumulator Status in this sub-group.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe optimal risk and resilience PGSs for estimated AAO-Aβ (PGS\u003csub\u003erisk\u0026minus;AAO\u003c/sub\u003e and PGS\u003csub\u003eresilience\u0026minus;AAO\u003c/sub\u003e) were both significant predictors of this trait overall, with higher PGS\u003csub\u003erisk\u0026minus;AAO\u003c/sub\u003e and lower PGS\u003csub\u003eresilience\u0026minus;AAO\u003c/sub\u003e associated with an earlier estimated AAO-Aβ (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, Table S2). A 1 SD increase in PGS\u003csub\u003erisk\u0026minus;AAO\u003c/sub\u003e and a 1 SD decrease in PGS\u003csub\u003eresilience\u0026minus;AAO\u003c/sub\u003e were associated with estimated AAO-Aβ 1.3 years and 0.91 years earlier respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, Table S2). When stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles, higher PGS\u003csub\u003erisk\u0026minus;AAO\u003c/sub\u003e was associated with an earlier estimated AAO-Aβ in \u003cem\u003eAPOE\u003c/em\u003e ε4 non-carriers and heterozygotes, but there was no significant association with AAO-Aβ in \u003cem\u003eAPOE\u003c/em\u003e ε4 homozygotes. Individuals in the upper quintile for PGS\u003csub\u003erisk\u0026minus;AAO\u003c/sub\u003e had a mean estimated AAO-Aβ 2.9 years earlier than those in the lower quintile after accounting for other covariates, while among \u003cem\u003eAPOE\u003c/em\u003e ε4 heterozygotes this difference was 4.5 years (Table S2). Higher PGS\u003csub\u003eresilience\u0026minus;AAO\u003c/sub\u003e was associated with a later estimated AAO-Aβ in \u003cem\u003eAPOE\u003c/em\u003e ε4 heterozygotes and a marginally non-significant (p\u0026thinsp;=\u0026thinsp;0.051) increase in estimated AAO-Aβ in ε4 homozygotes, but not associated with AAO-Aβ in \u003cem\u003eAPOE\u003c/em\u003e ε4 non-carriers (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, Table S2). Individuals in the upper quintile for PGS\u003csub\u003eresilience\u0026minus;AAO\u003c/sub\u003e had a mean estimated AAO-Aβ 2.2 years later than those in the lower quintile after accounting for other covariates, but there was not a significant difference between these extremes for any of the groups after stratification by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles (Table S2).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003ePolygenic scores derived from trait-specific GWAS against Accumulator Status and Estimated Age at Onset of Aβ (AAO-Aβ), and comparison with risk and resilience PGSs\u003c/em\u003e\u003c/p\u003e\u003cp\u003eThe mean percentage of people who were accumulators of Aβ across each of the CV runs was 71.4% in the discovery sets (SD 0.50) and 71.8% in the validation sets (SD 1.00)(Table S3), which are both very close to the value of 71.5% for the whole sample (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). In the trait-specific GWASs run in each of the 10 CV runs (excluding chromosome 19), we found a total of two SNPs (rs12192157 and rs6900289) associated with Accumulator Status at genome-wide significance (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 x 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e). These closely linked variants were on chromosome 6 and were significant in just one of the CV runs (Table S4). We found one genome-wide significant SNP (rs12022131) for estimated AAO-Aβ in one of the CV runs, located on chromosome 1 (Table S4).\u003c/p\u003e\u003cp\u003eThe optimal trait-specific PGS for Accumulator Status (PGS\u003csub\u003eAccumulator\u003c/sub\u003e) was significantly associated with this trait in four of the 10 CV runs (two of these remained significant after FDR correction), although the direction of the association varied among runs and the overall association was not significant (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, Table S5). The risk and resilience PGSs for Accumulator Status, run in the validation sets of each CV run for comparison, both showed overall significant positive associations with odds of being an accumulator. In contrast to PGS\u003csub\u003eAccumulator\u003c/sub\u003e, the direction of association was consistent across all 10 CV runs for PGS\u003csub\u003erisk\u0026minus;Accumulator\u003c/sub\u003e and all but one CV run for PGS\u003csub\u003eresilience\u0026minus;Accumulator\u003c/sub\u003e, although it was only significant in six CV runs for PGS\u003csub\u003erisk\u0026minus;Accumulator\u003c/sub\u003e and not in any individual CV run for PGS\u003csub\u003eresilience\u0026minus;Accumulator\u003c/sub\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e; Table S5).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eSimilarly, for estimated AAO-Aβ, the optimal trait-specific PGS (PGS\u003csub\u003eAAO\u003c/sub\u003e) was not significantly associated with the trait overall, and although the association was significant in four of the CV runs (three after FDR correction), the direction of association was highly variable across runs (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, Table S6). PGS\u003csub\u003erisk\u0026minus;AAO\u003c/sub\u003e was negatively associated with AAO-Aβ, indicating that individuals with a higher genetic risk of AD had an earlier estimated age at onset of Aβ. PGS\u003csub\u003eresilience\u0026minus;AAO\u003c/sub\u003e was positively associated with AAO-Aβ, indicating a later estimated age at onset of Aβ in more genetically resilient individuals (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, Table S6). Of the three PGSs for each trait, PGS\u003csub\u003erisk\u003c/sub\u003e had, on average, the largest effect size and explained the greatest proportion of the variation in both Accumulator Status and estimated AAO-Aβ (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eWe tested whether polygenic scores based on genetic risk of and resilience to AD could predict Aβ Accumulator Status and estimated AAO-Aβ. We also tested whether PGSs based on phenotype-specific GWASs improved prediction of these traits. Higher genetic risk of AD was associated with higher odds of Aβ accumulation, and earlier estimated AAO-Aβ. Higher genetic resilience was associated with a later estimated AAO-Aβ but was not a significant predictor of Accumulator Status. Phenotype-specific PGSs did not improve on the predictive performance of PGS\u003csub\u003erisk\u003c/sub\u003e or PGS\u003csub\u003eresilience\u003c/sub\u003e and were not significant predictors of either trait overall.\u003c/p\u003e\u003cp\u003eThe associations of PGSs with each trait were seen after accounting for \u003cem\u003eAPOE\u003c/em\u003e ε4 status, the strongest genetic risk factor for AD [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Our results contrast with several studies that have found that PGSs did not improve prediction over and above \u003cem\u003eAPOE\u003c/em\u003e ε4 status for AD and all-cause dementia (ACD) [\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e], or for measures of Aβ deposition [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e]. However, other studies have found that PGSs do result in small but significant improvements over \u003cem\u003eAPOE\u003c/em\u003e alone in prediction of AD-related traits, including incidence of AD [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e] and Aβ pathology [\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e]. Apparent inconsistencies between these findings may be explained by variation in the methods used to construct PGSs, as well as the specific phenotypes examined. It has been argued that Aβ deposition is largely driven by \u003cem\u003eAPOE\u003c/em\u003e, and that other genetic contributors to AD become more important at later disease stages [\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e]. However, our results highlight that considering genetic factors beyond \u003cem\u003eAPOE\u003c/em\u003e can improve prediction of whether and how early individuals will accumulate Aβ.\u003c/p\u003e\u003cp\u003eDifferences between PGS\u003csub\u003erisk\u003c/sub\u003e and PGS\u003csub\u003eresilience\u003c/sub\u003e in their associations with the traits examined, and their interactions with \u003cem\u003eAPOE\u003c/em\u003e, offer insights into the mechanisms by which the genetic variation captured by these scores confers risk or protection against AD. Increased genetic risk of AD was a stronger predictor of adverse outcomes (higher odds of being an accumulator and earlier estimated AAO-Aβ) in ε4 non-carriers and heterozygotes than in ε4 homozygotes, suggesting it contributes little additional risk in individuals who are already at highest risk due to \u003cem\u003eAPOE\u003c/em\u003e ε4 homozygosity. By contrast, higher genetic resilience to AD was associated with later estimated AAO-Aβ in ε4 heterozygotes (and a marginally non-significant association in homozygotes), but was not associated with AAO-Aβ in ε4 non-carriers. The original AD resilience GWAS was conducted by limiting the study population to individuals at high genetic risk for AD (defined by a similar risk PGS to that developed here) and contrasting \u0026lsquo;resilient\u0026rsquo; (unaffected by AD) individuals with AD cases [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. This means that it is effectively a measure of resilience to genetic risk of AD, so it is unsurprising that it is a stronger predictor of AAO-Aβ in \u003cem\u003eAPOE\u003c/em\u003e ε4 carriers, who are at highest genetic risk [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. PGS\u003csub\u003eresilience\u003c/sub\u003e was not associated with Accumulator Status, although the overall trend (significant when aggregated across runs in the cross-validation study) was positive. This implies that more genetically resilient individuals are \u003cem\u003emore\u003c/em\u003e likely to be accumulators of Aβ, which seems counterintuitive, but is likely another consequence of the fact that for this score, more genetically resilient individuals also have higher genetic risk scores [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. A framework previously proposed when considering protective factors for AD, distinguishes between \u0026lsquo;resistance\u0026rsquo; and \u0026lsquo;resilience\u0026rsquo;, where resistance refers to the avoidance of pathological brain changes, while resilience is the ability to cope with accumulating neuropathology and avoid brain atrophy or cognitive decline [\u003cspan additionalcitationids=\"CR60\" citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e]. The lack of association of PGS\u003csub\u003eresilience\u003c/sub\u003e with Accumulator Status in our study suggests that genetic variation captured by this score does not confer protection against AD by preventing the accumulation of Aβ (\u0026lsquo;resistance\u0026rsquo;), although the association with AAO-Aβ suggests it may slow or delay this accumulation. Further analysis involving a broader range of traits is required to disentangle this.\u003c/p\u003e\u003cp\u003eThe trait-specific PGSs for both Accumulator Status and estimated AAO-Aβ, derived using a cross-validation approach, were not associated with either trait overall, although several individual CV runs showed significant associations for each trait. However, there was variability in effect direction for these associations. It is likely that unstable signals were owing in part to the sample sizes available for the discovery GWASs in this study (N\u0026thinsp;=\u0026thinsp;1450 for Accumulator Status and N\u0026thinsp;=\u0026thinsp;772 for AAO-Aβ), which were much smaller than those in the risk (N\u0026thinsp;=\u0026thinsp;94 437) and resilience (N\u0026thinsp;=\u0026thinsp;13 572) GWASs [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Despite the lack of compelling evidence in the current study, trait-specific PGSs may nevertheless capture unique genetic variation associated with these traits, particularly in larger studies. We identified two closely-linked SNPs on chromosome 6 that were associated with Accumulator Status (rs12192157 and rs6900289), and one SNP on chromosome 1 (rs12022131) that was associated with AAO-Aβ at a genome-wide significant level. While each was significant in only a single CV run, it will be of interest to determine whether a signal is seen in these regions in future studies. To the best of our knowledge, there are currently no known associations of these SNPs with specific traits or diseases. However, the SNPs on chromosome 6 are in close proximity to the \u003cem\u003eLPA\u003c/em\u003e gene, which affects plasma concentrations of lipoprotein (a) (Lp(a)) and is strongly associated with cardiovascular disease [\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e], a key risk factor for AD [\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eGenetic variants, unlike other biomarkers of disease, remain constant across the lifespan, meaning that polygenic scores for diseases and related traits offer the potential to identify high risk individuals at a very early stage, prior to symptom onset [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. In AD, a diagnosis is typically made once cognition is impaired, by which time there has been widespread damage to the brain. However, brain Aβ begins accumulating years or decades prior to appearance of cognitive symptoms [\u003cspan additionalcitationids=\"CR4\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Therefore, identifying individuals at risk of accumulating Aβ provides an opportunity to administer interventions to prevent or delay onset and progression of disease. While effective treatments for AD have proved elusive, recent years have seen the development of anti-amyloid monoclonal antibodies that remove Aβ from the brain and have been shown to produce modest slowing of cognitive decline in people with mild symptomatic AD [\u003cspan additionalcitationids=\"CR66\" citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e]. Trials are currently evaluating whether administering these treatments at an earlier stage, in asymptomatic individuals, may be more effective and there is hope that it may eventually be possible to prevent AD [\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e]. In this instance, PGSs could offer a relatively inexpensive and minimally invasive method to evaluate people\u0026rsquo;s risk and prioritise them for further screening. A limitation of PGSs is that they typically explain only a small proportion of the total variation in a disease or trait [\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e], which was the case in this study (0.2\u0026ndash;2.3% for PGS\u003csub\u003erisk\u003c/sub\u003e; 0.1\u0026ndash;1.6% for PGS\u003csub\u003eresilience\u003c/sub\u003e, depending on population sub-group). While this limits their utility for making a definitive diagnosis, our results show that PGSs can nevertheless improve the accuracy of prediction of AD-related traits, and may be a useful tool for risk stratification, particularly when considered alongside other risk predictors such as demographics and lifestyle.\u003c/p\u003e\u003cp\u003eThis study does have several limitations. Within our study population, over 70% of participants were accumulators of Aβ. While data on Accumulator Status in the wider population are scarce, one study found that ~\u0026thinsp;20% of healthy adults aged\u0026thinsp;\u0026ge;\u0026thinsp;60 years had elevated Aβ [\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e]. Accumulators of Aβ are therefore almost certainly over-represented in our study population, which is unsurprising given that the component cohorts are enriched for people with cognitive complaints [\u003cspan additionalcitationids=\"CR31 CR32 CR33 CR34\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. As a result, the likelihood of being an accumulator as predicted by PGS score may be overestimated and should be recalibrated based on a more representative population. Similarly, AAO-Aβ could only be estimated for people who had begun accumulating Aβ and were close to, or had exceeded, the 20 CL threshold. The predictive performance of this score in the broader population, including people with low Aβ who may accumulate Aβ, needs to be verified. Furthermore, the lack of an external validation sample for the phenotype-specific PGSs meant that it was necessary to both run the discovery GWAS and develop PGSs within the study population by dividing it into test and validation data sets. Despite utilising the largest existing dataset for these traits, this resulted in small sample sizes relative to those used in the GWASs from which risk and resilience PGSs were derived [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. This was partially addressed by our cross-validation approach. However, the instability of score performance across CV runs is likely due to the small sample. Finally, the study population was limited to people of European ancestry. Differences in LD structure, allele frequencies and genetic architecture can affect the generalisability of genetic predictors across different ancestries [\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e], therefore testing in diverse populations would be beneficial.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003ePolygenic scores based on genetic risk of AD explained a small but significant proportion of the variation in Accumulator Status and estimated AAO-Aβ, over and above that explained by \u003cem\u003eAPOE\u003c/em\u003e ε4. The PGS for AD risk may be particularly useful, in combination with other predictors, for identifying individuals at risk of Aβ accumulation and earlier AAO-Aβ, who may benefit from targeted prevention and treatment.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003eCompeting Interests\u003c/h2\u003e\u003cp\u003e CC is a member of the scientific advisory board of Circular Genomics and owns stocks, and is on the scientific advisory board of ADmit and Alamar, consults for Sanofi, NovoNordisk, and Owkin, and has received research support from GSK, Danaher and EISAI. All other authors report no competing interests relevant to this manuscript.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003cp\u003e The ADNI, AIBL and OASIS studies have all been granted approval by the ethics committees of their respective member institutions. All participants in this study provided informed written consent.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e\u003cp\u003eThe Alzheimer\u0026rsquo;s Dementia Onset and Progression in International Cohorts (ADOPIC) study was funded by a National Institute of Health (NIH) grant (R01-AG058676-01A1).\u003c/p\u003e\u003cp\u003eData used in the preparation of this article were obtained from the Australian Imaging, Biomarker and Lifestyle (AIBL) Study database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://aibl.org.au/collaboration\u003c/span\u003e\u003cspan address=\"https://aibl.org.au/collaboration\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). As such, the investigators within AIBL, unless otherwise listed, contributed to the design and implementation of AIBL and/or provided data, but did not participate in the analysis or writing of this report. A complete listing of AIBL investigators can be found at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://aibl.org.au/about/our-researchers/\u003c/span\u003e\u003cspan address=\"https://aibl.org.au/about/our-researchers/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. The AIBL Study (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://aibl.org.au/\u003c/span\u003e\u003cspan address=\"https://aibl.org.au/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) is a consortium between Austin Health, CSIRO, Edith Cowan University, the Florey Institute (The University of Melbourne), and the National Ageing Research Institute. The study has received partial financial support from the Alzheimer\u0026rsquo;s Association (US), the Alzheimer\u0026rsquo;s Drug Discovery Foundation, an Anonymous Foundation, the Science and Industry Endowment Fund, the Dementia Collaborative Research Centres, the Victorian Government\u0026rsquo;s Operational Infrastructure Support program, the Australian Alzheimer\u0026rsquo;s Research Foundation (now Alzheimer\u0026rsquo;s Research Australia), the National Health and Medical Research Council (NHMRC), and The Yulgilbar Foundation. Numerous commercial interactions have supported data collection and analyses. This includes genetic data utilized in this study, which has also been supported by grants awarded to SML by the NHMRC (GNT1161706; GNT2001320). In-kind support has also been provided by Sir Charles Gairdner Hospital, Cogstate Ltd, Hollywood Private Hospital, The University of Melbourne, and St Vincent\u0026rsquo;s Hospital.\u003c/p\u003e\u003cp\u003eData used in the preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report. A complete listing of ADNI investigators can be found at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List\u003c/span\u003e\u003cspan address=\"http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. pdf. The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer\u0026rsquo;s disease (AD). Data collection and sharing for the ADNI is funded by the National Institute on Aging (National Institutes of Health Grant U19 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer\u0026rsquo;s Association; Alzheimer\u0026rsquo;s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research \u0026amp; Development, LLC.; Johnson \u0026amp; Johnson Pharmaceutical Research \u0026amp; Development LLC.; Lumosity; Lundbeck; Merck \u0026amp; Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003ca href=\"https://aibl.org.au/collaboration\" target=\"_blank\"\u003ewww.fnih.org\u003c/a\u003e\u003c/span\u003e\u003cspan address=\"http://www.fnih.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer\u0026rsquo;s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.\u003c/p\u003e\u003cp\u003eData were provided in part by OASIS-3: Principal Investigators: T. Benzinger, D. Marcus, J. Morris; NIH P50 AG00561, P30 NS09857781, P01 AG026276, P01 AG003991, R01 AG043434, UL1 TR000448, R01 EB009352. AV-45 doses were provided by Avid Radiopharmaceuticals, a wholly owned subsidiary of Eli Lilly.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eEKO, SML and TP conceptualised and designed the study. TC generated the estimates of age at onset of Aβ. EKO performed the other statistical analyses, with advice and review by BG, SF, TP and JDD. PB, KN, VLV, VD, CC, AJS, TP and SML contributed to acquisition or curation of data. CLM, CCR, CC, AJS and SML contributed to funding acquisition. EKO, TP, and SML drafted the manuscript. All authors contributed to the revision and editing of the manuscript and approved the submitted version.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003e We thank all the participants and their families who took part in the AIBL, ADNI and OASIS studies that were included in the ADOPIC consortium cohort. We also thank all the clinicians who may have referred participants.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets generated and analysed during the current study are not publicly available due to individual cohort restrictions as outlined at: ADNI: https://adni.loni.usc.edu/data-samples/adni-data/; AIBL: https://aibl.org.au/collaboration/#data-access; OASIS: https://sites.wustl.edu/oasisbrains/home/access/. However, data are available from the corresponding author on reasonable request, with the permission of the participating cohort studies.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eKnopman DS, et al. Alzheimer disease. Nat reviews Disease primers. 2021;7(1):33.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVan Cauwenberghe C, Van Broeckhoven C, Sleegers K. The genetic landscape of Alzheimer disease: clinical implications and perspectives. Genet Sci. 2016;18(5):421\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVillemagne VL, et al. Amyloid β deposition, neurodegeneration, and cognitive decline in sporadic Alzheimer's disease: a prospective cohort study. Lancet Neurol. 2013;12(4):357\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJack CR, et al. Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 2013;12(2):207\u0026ndash;16.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSelkoe DJ, Hardy J. The amyloid hypothesis of Alzheimer's disease at 25 years. EMBO Mol Med. 2016;8(6):595\u0026ndash;608.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHardy JA, Higgins GA. Alzheimer's disease: the amyloid cascade hypothesis. Science. 1992;256(5054):184\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSperling RA, Jack CR Jr, Aisen PS. \u003cem\u003eTesting the right target and right drug at the right stage.\u003c/em\u003e Science translational medicine, 2011. 3(111): p. 111cm33-111cm33.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHarman D. Alzheimer's disease pathogenesis: role of aging. Ann N Y Acad Sci. 2006;1067(1):454\u0026ndash;60.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGatz M, et al. Role of genes and environments for explaining Alzheimer disease. Arch Gen Psychiatry. 2006;63(2):168\u0026ndash;74.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWard A, et al. Prevalence of apolipoprotein E4 genotype and homozygotes (APOE e4/4) among patients diagnosed with Alzheimer\u0026rsquo;s disease: a systematic review and meta-analysis. Neuroepidemiology. 2012;38(1):1\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLim YY, et al. APOE genotype and early β-amyloid accumulation in older adults without dementia. Neurology. 2017;89(10):1028\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLambert J-C, et al. Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer's disease. Nat Genet. 2013;45(12):1452\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKunkle BW, et al. Genetic meta-analysis of diagnosed Alzheimer\u0026rsquo;s disease identifies new risk loci and implicates Aβ, tau, immunity and lipid processing. Nat Genet. 2019;51(3):414\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHollingworth P, et al. Common variants at ABCA7, MS4A6A/MS4A4E, EPHA1, CD33 and CD2AP are associated with Alzheimer's disease. Nat Genet. 2011;43(5):429\u0026ndash;35.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNaj AC, et al. Common variants at MS4A4/MS4A6E, CD2AP, CD33 and EPHA1 are associated with late-onset Alzheimer's disease. Nat Genet. 2011;43(5):436\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMarioni RE, et al. GWAS on family history of Alzheimer\u0026rsquo;s disease. Translational psychiatry. 2018;8(1):99.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJansen IE, et al. Genome-wide meta-analysis identifies new loci and functional pathways influencing Alzheimer\u0026rsquo;s disease risk. Nat Genet. 2019;51(3):404\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBellenguez C, et al. New insights into the genetic etiology of Alzheimer\u0026rsquo;s disease and related dementias. Nat Genet. 2022;54(4):412\u0026ndash;36.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRamanan VK, et al. GWAS of longitudinal amyloid accumulation on 18F-florbetapir PET in Alzheimer\u0026rsquo;s disease implicates microglial activation gene IL1RAP. Brain. 2015;138(10):3076\u0026ndash;88.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShade LM, et al. GWAS of multiple neuropathology endophenotypes identifies new risk loci and provides insights into the genetic risk of dementia. Nat Genet. 2024;56(11):2407\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYan Q, et al. Genome-wide association study of brain amyloid deposition as measured by Pittsburgh Compound-B (PiB)-PET imaging. Mol Psychiatry. 2021;26(1):309\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWray NR, et al. From basic science to clinical application of polygenic risk scores: a primer. JAMA psychiatry. 2021;78(1):101\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoi SW, Mak TS-H, O\u0026rsquo;Reilly PF. Tutorial: a guide to performing polygenic risk score analyses. Nat Protoc. 2020;15(9):2759\u0026ndash;72.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaher BS. Polygenic scores in epidemiology: risk prediction, etiology, and clinical utility. Curr Epidemiol Rep. 2015;2:239\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTorkamani A, Wineinger NE, Topol EJ. The personal and clinical utility of polygenic risk scores. Nat Rev Genet. 2018;19(9):581\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVisscher PM, et al. 10 years of GWAS discovery: biology, function, and translation. Am J Hum Genet. 2017;101(1):5\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, et al. Estimation of complex effect-size distributions using summary-level statistics from genome-wide association studies across 32 complex traits. Nat Genet. 2018;50(9):1318\u0026ndash;26.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang J, et al. Common SNPs explain a large proportion of the heritability for human height. Nat Genet. 2010;42(7):565\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHou J, et al. Polygenic resilience scores capture protective genetic effects for Alzheimer\u0026rsquo;s disease. Translational psychiatry. 2022;12(1):296.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWeiner MW, et al. The Alzheimer's disease neuroimaging initiative: progress report and future plans. Volume 6. Alzheimer's \u0026amp; Dementia; 2010. pp. 202\u0026ndash;11. 3e7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMueller SG, et al. The Alzheimer's disease neuroimaging initiative. Neuroimaging Clin. 2005;15(4):869\u0026ndash;77.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePetersen RC, et al. Alzheimer's disease neuroimaging initiative (ADNI): clinical characterization. Neurology. 2010;74(3):201\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEllis KA, et al. The Australian Imaging, Biomarkers and Lifestyle (AIBL) study of aging: methodology and baseline characteristics of 1112 individuals recruited for a longitudinal study of Alzheimer's disease. Int Psychogeriatr. 2009;21(4):672\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFowler C, et al. Fifteen years of the australian imaging, biomarkers and lifestyle (AIBL) study: progress and observations from 2,359 older adults spanning the spectrum from cognitive normality to Alzheimer\u0026rsquo;s disease. J Alzheimer's disease Rep. 2021;5(1):443\u0026ndash;68.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLaMontagne PJ et al. \u003cem\u003eOASIS-3: longitudinal neuroimaging, clinical, and cognitive dataset for normal aging and Alzheimer disease.\u003c/em\u003e MedRxiv, 2019: p. 2019.12. 13.19014902.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFernandez MV, et al. Genetic and multi-omic resources for Alzheimer disease and related dementia from the Knight Alzheimer Disease Research Center. Sci data. 2024;11(1):768.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTaliun D et al. \u003cem\u003eSequencing of 53,831 diverse genomes from the NHLBI TOPMed Program.\u003c/em\u003e bioRxiv, 2019: p. 563866.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDas S, et al. Next-generation genotype imputation service and methods. Nat Genet. 2016;48(10):1284\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChang CC et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience, 2015. 4(1): p. s13742-015-0047-8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePurcell S, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81(3):559\u0026ndash;75.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBourgeat P, et al. Comparison of MR-less PiB SUVR quantification methods. Neurobiol Aging. 2015;36:S159\u0026ndash;66.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBourgeat P, et al. Implementing the centiloid transformation for 11C-PiB and β-amyloid 18F-PET tracers using CapAIBL. NeuroImage. 2018;183:387\u0026ndash;93.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSu Y, et al. Utilizing the Centiloid scale in cross-sectional and longitudinal PiB PET studies. Volume 19. NeuroImage: Clinical; 2018. pp. 406\u0026ndash;16.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBourgeat P et al. \u003cem\u003eβ-amyloid PET harmonisation across longitudinal studies: Application to AIBL, ADNI and OASIS3.\u003c/em\u003e Neuroimage, 2022. 262: p. 119527.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBurnham SC et al. \u003cem\u003eWhen Does Alzheimer's Disease Start? Robust Estimates Based on Longitudinal Aβ-Amyloid-PET in Three Large International Cohorts.\u003c/em\u003e Available at SSRN: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.2139/ssrn.4757682\u003c/span\u003e\u003cspan address=\"10.2139/ssrn.4757682\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, 2024.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoi SW, O'Reilly PF. PRSice-2: Polygenic Risk Score software for biobank-scale data. Gigascience. 2019;8(7):giz082.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEuesden J, Lewis CM, O\u0026rsquo;reilly PF. PRSice: polygenic risk score software. Bioinformatics. 2015;31(9):1466\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWare EB, et al. Considering the APOE locus in Alzheimer\u0026rsquo;s disease polygenic scores in the Health and Retirement Study: a longitudinal panel study. BMC Med Genom. 2020;13:1\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCore Team R. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing: Vienna, Austria; 2024.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePosit team, RStudio: Integrated Development Environment for R. 2025, Posit Software, PBC: Boston, MA, USA.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePicard RR, Cook RD. Cross-validation of regression models. J Am Stat Assoc. 1984;79(387):575\u0026ndash;83.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXu Q-S, Liang Y-Z. Monte Carlo cross validation. Chemometr Intell Lab Syst. 2001;56(1):1\u0026ndash;11.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eStouffer S. A study of attitudes. Sci Am. 1949;180(5):11\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eStocker H, et al. Alzheimer\u0026rsquo;s polygenic risk scores, APOE, Alzheimer\u0026rsquo;s disease risk, and dementia-related blood biomarker levels in a population-based cohort study followed over 17 years. Volume 15. Alzheimer's Research \u0026amp; Therapy; 2023. p. 129. 1.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDarst BF, et al. Pathway-specific polygenic risk scores as predictors of amyloid-β deposition and cognitive function in a sample at increased risk for Alzheimer\u0026rsquo;s disease. J Alzheimer\u0026rsquo;s Disease. 2016;55(2):473\u0026ndash;84.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLeonenko G, et al. Genetic risk for alzheimer disease is distinct from genetic risk for amyloid deposition. Ann Neurol. 2019;86(3):427\u0026ndash;35.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEscott-Price V, et al. Common polygenic variation enhances risk prediction for Alzheimer\u0026rsquo;s disease. Brain. 2015;138(12):3673\u0026ndash;84.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEbenau JL, et al. Risk of dementia in APOE ε4 carriers is mitigated by a polygenic risk score. Volume 13. Alzheimer's \u0026amp; Dementia: Diagnosis, Assessment \u0026amp; Disease Monitoring,; 2021. p. e12229. 1.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBocancea DI, et al. Measuring resilience and resistance in aging and Alzheimer disease using residual methods: a systematic review and meta-analysis. Neurology. 2021;97(10):474\u0026ndash;88.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eArenaza-Urquijo EM, Vemuri P. Resistance vs resilience to Alzheimer disease: clarifying terminology for preclinical studies. Neurology. 2018;90(15):695\u0026ndash;703.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eArenaza-Urquijo EM, Vemuri P. Improving the resistance and resilience framework for aging and dementia studies. Volume 12. Alzheimer's research \u0026amp; therapy; 2020. pp. 1\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchmidt K, et al. Structure, function, and genetics of lipoprotein (a). J Lipid Res. 2016;57(8):1339\u0026ndash;59.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLeszek J, et al. The links between cardiovascular diseases and Alzheimer's disease. Curr Neuropharmacol. 2021;19(2):152\u0026ndash;69.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ede Alves TF. Cardiac disorders as risk factors for Alzheimer's disease. J Alzheimer\u0026rsquo;s Disease. 2010;20(3):749\u0026ndash;63.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDelrieu J, Andrieu S, Vellas B. Dementia research in 2023: the year of anti-amyloid immunotherapy. Lancet Neurol. 2024;23(1):13\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVan Dyck CH, et al. Lecanemab in early Alzheimer\u0026rsquo;s disease. N Engl J Med. 2023;388(1):9\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSims JR, et al. Donanemab in early symptomatic Alzheimer disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA. 2023;330(6):512\u0026ndash;27.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eReiman EM, et al. A chance to prevent Alzheimer's disease sooner than you think. Lancet Neurol. 2024;23(2):144\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKember RL et al. Considerations for the application of polygenic scores to clinical care of individuals with substance use disorders. J Clin Investig, 2024. 134(20).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRodrigue K, et al. β-Amyloid burden in healthy aging: regional distribution and cognitive consequences. Neurology. 2012;78(6):387\u0026ndash;95.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKachuri L, et al. Principles and methods for transferring polygenic risk scores across global populations. Nat Rev Genet. 2024;25(1):8\u0026ndash;25.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Alzheimer's disease, amyloid beta accumulation, age at onset of amyloid beta, polygenic scores, risk, resilience","lastPublishedDoi":"10.21203/rs.3.rs-7911284/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7911284/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAccumulation of brain amyloid beta (Aβ) is a key pathological hallmark of Alzheimer\u0026rsquo;s disease (AD) and begins many years before cognitive symptoms. Being able to predict the risk of Aβ accumulation, or the age at which this accumulation exceeds a critical threshold, may enable early intervention and treatment to slow or prevent the onset of AD. We utilised published genome-wide association studies (GWAS) to develop polygenic scores (PGS) based on AD risk (PGS\u003csub\u003erisk\u003c/sub\u003e) and resilience (PGS\u003csub\u003eresilience\u003c/sub\u003e). We tested whether these could predict (i) whether an individual was an accumulator of Aβ (\u0026lsquo;Accumulator Status\u0026rsquo;), and (ii) in accumulators, the age at which brain Aβ is estimated to exceed a threshold of 20 centiloids (CL)(\u0026lsquo;Estimated Age at onset of Aβ\u0026rsquo;; AAO-Aβ) among 2175 participants (1158 with AAO Aβ) from the Alzheimer\u0026rsquo;s Dementia Onset and Progression in International Cohorts (ADOPIC) study. Additionally, we conducted genome-wide association studies (GWAS) of these traits and developed phenotype-specific PGSs using cross-validation (CV). Higher PGS\u003csub\u003erisk\u003c/sub\u003e was associated with a greater risk of being an accumulator and a younger AAO-Aβ. When stratified by number of \u003cem\u003eAPOE\u003c/em\u003e ε4 alleles, PGS\u003csub\u003erisk\u003c/sub\u003e predicted Accumulator Status in \u003cem\u003eAPOE\u003c/em\u003e ε4 heterozygotes, and AAO-Aβ in ε4 non-carriers and heterozygotes, with the same directions of effect as were seen in the whole cohort. PGS\u003csub\u003eresilience\u003c/sub\u003e was not significantly associated with Accumulator Status, but higher PGS\u003csub\u003eresilience\u003c/sub\u003e was associated with later AAO-Aβ overall and in ε4 heterozygotes. Trait-specific PGSs, developed using CV, were not significantly associated with either trait overall and the direction of association varied across CV folds. Polygenic scores, alongside other risk factors, may be useful for identifying individuals at risk of accumulating Aβ, and predicting the age at which this exceeds a critical threshold. This could provide a window for administering disease-modifying treatment or lifestyle interventions to prevent or delay the onset of AD.\u003c/p\u003e","manuscriptTitle":"Predicting accumulation and age at onset of amyloid-β from genetic risk and resilience for Alzheimer’s disease","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-22 19:00:39","doi":"10.21203/rs.3.rs-7911284/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9aa34cda-cc20-4fac-a64a-7adef2dd4d59","owner":[],"postedDate":"October 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-27T14:26:00+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-22 19:00:39","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7911284","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7911284","identity":"rs-7911284","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}