Amyloid Probability in Alzheimer Disease From Plasma, Cerebrospinal Fluid, and Amyloid Imaging

Amyloid Probability in Alzheimer Disease From Plasma, Cerebrospinal Fluid, and Amyloid Imaging | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Amyloid Probability in Alzheimer Disease From Plasma, Cerebrospinal Fluid, and Amyloid Imaging Khushboo Verma, Satwant Kumar This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8492792/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Blood-based biomarkers are increasingly used to triage patients for amyloid confirmation, yet performance is often reported versus a single comparator despite discordance between amyloid PET and CSF. In a cross-sectional secondary analysis of the Alzheimer’s Disease Neuroimaging Initiative (ADNI), we assembled one PET-anchored PET–CSF–plasma triad per participant (n = 320). Bayesian latent class models integrating PET, CSF and plasma (Aβ42/40 or %p-tau217) estimated pattern-level posterior probabilities of latent amyloid positivity, with prespecified sensitivity analyses for PET–CSF conditional dependence, timing gaps (≤ 7, 8–30, > 30 days) and CSF cutpoints. Concordant PET+/CSF + and PET−/CSF − patterns mapped to probabilities near 1 and 0, whereas discordant patterns yielded intermediate probabilities refined by plasma strata and most sensitive to dependence assumptions. PET–CSF discordance occurred even within ≤ 7 days (12/98; 12%). A CSF Aβ42 coverage analysis showed similar gradients. Pattern-to-probability reporting may aid interpretation without privileging a reference test. Health sciences/Biomarkers Health sciences/Medical research Health sciences/Neurology Biological sciences/Neuroscience Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Determining whether an individual harbors Alzheimer-type brain amyloid has become increasingly important as biomarker confirmation now shapes diagnostic pathways and eligibility for amyloid-targeting therapeutic and research programs 1 – 3 . Blood-based biomarkers (BBMs) are emerging as scalable, low-burden gatekeepers that estimate amyloid likelihood and can triage patients for confirmatory testing, while current appropriate-use recommendations still emphasize confirmation with cerebrospinal fluid (CSF) measures and/or amyloid PET 2 , 3 . In parallel, BBMs are being incorporated into trial screening and enrichment workflows to reduce burden and screen failure, even as PET and CSF remain the predominant confirmatory modalities in many protocols 1 – 4 . However, amyloid “confirmation” is not always unambiguous: amyloid PET and CSF amyloid measures, while generally concordant, can yield discordant classifications in a clinically meaningful minority of individuals (often ~ 10–20%) 5,6 . These discordant profiles are not merely technical noise; multiple longitudinal and biomarker studies suggest they can carry biological and prognostic information, including differential associations with downstream tau pathology and clinical trajectories depending on whether CSF becomes abnormal before PET or vice versa 7 , 8 . As a result, “PET + vs CSF+” is not a trivial labeling choice, discordance creates a real interpretive gap for both clinical decision-making and study design when the two most common confirmatory modalities disagree 5 , 8 . Despite reported discordances diagnostic performance is still commonly framed “versus PET” or “versus CSF,” as if one modality were a definitive ground truth. This convention is increasingly strained: although both Aβ-PET and CSF amyloid assays aim to detect amyloid plaque pathology, they index different (but related) biological processes 9 , 10 , with PET tracking fibrillar plaque burden and CSF reflecting soluble Aβ dynamics 10 . Whereas a clinically meaningful minority of individuals show persistent PET–CSF discordance 7 . When plasma biomarkers are evaluated against a single comparator, discordant and borderline cases are forced into “false-positive/false-negative” bins by design choice rather than biology, and even very high-performing blood tests inherit the uncertainty (and potential ceiling) of an imperfect reference 11 , 12 . In practice, reported “accuracy” can therefore shift with the selected reference test and thresholds, while clinicians and trialists remain without a calibrated probability statement for the common real-world scenario of PET–CSF disagreement 4 , 12 . A natural way to resolve this reference-standard problem is to treat amyloid status as a latent (unobserved) state and jointly model PET, CSF, and plasma as imperfect indicators of that state, rather than forcing one modality to serve as ground truth 13 . Latent class models were developed precisely for diagnostic settings without a single accurate reference standard and can yield directly interpretable outputs such as \(\:\text{P}\left(amyloid+\right)\) given the observed test pattern, with uncertainty 13 , 14 . Importantly, biomarker applications must also confront conditional dependence, for example, PET and CSF may remain correlated even after conditioning on latent amyloid. This assumption violation can materially bias inference if left implicit. To incorporate (and stress-test) such dependence structures Bayesian formulations provide a principled way 14 , 15 . In this context, adding plasma as a third modality is not only clinically pragmatic but also methodologically valuable: an additional indicator can refine posterior probabilities for discordant PET–CSF patterns and improve identifiability under plausible dependence scenarios 15 , 16 . Any integrative approach in this setting must make its assumptions explicit. Latent class inference without a gold standard is identifiable only under modeling constraints, and results can be sensitive to the commonly made (and often violated) assumption that tests are conditionally independent given the latent state, particularly for PET and CSF, which share biology and measurement pipelines 14 , 15 . In addition, amyloid classifications depend on assay-specific cutpoints (especially for CSF), where small threshold shifts can move borderline individuals across the positivity boundary and thereby change apparent discordance rates 17 . Finally, because PET and CSF may not be acquired on the same day in observational cohorts, the time gap between modalities is a plausible contributor to disagreement and should be treated as a prespecified sensitivity axis rather than a post hoc explanation 6 , 8 . Together, these considerations argue for reporting calibrated, pattern-level probabilities with uncertainty and for stress-testing conclusions across dependence, timing, and cutpoint assumptions 14 , 15 . Here, in a cross-sectional secondary analysis of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) 18 , we assembled per-participant PET–CSF–plasma “triads” and used Bayesian latent class models to estimate the posterior probability of a latent amyloid-positive state for each observed PET–CSF–plasma pattern, rather than defining truth by a single modality. Using Elecsys CSF Aβ42/40 as the primary CSF amyloid definition and plasma measurements aligned to the PrecivityAD2 component analytes (plasma Aβ42/40 ratio or plasma %p-tau217), we quantify how plasma strata refine interpretation of discordant PET–CSF profiles 17 , 19 – 21 . We prespecified sensitivity analyses for PET–CSF conditional dependence, modality timing gaps, and CSF cutpoints, and we performed a coverage analysis using Elecsys CSF Aβ42 to address incomplete availability of CSF Aβ42/40 8,17 . Together, these analyses yield a clinically interpretable “pattern-to-probability” output with explicit uncertainty for common real-world discordant and borderline biomarker patterns. Results Cohort and analytic sets The PET-anchored cohort comprised 608 eligible PET–CSF–plasma triads from 320 participants; selecting one prespecified triad per participant yielded 320 triads for analysis ( Fig. 1 ). Participants had a mean (s.d.) age of 71.1 (6.7) years and included 162 women (51%). Elecsys CSF Aβ42/40 was available for 138/320 participants (43%), defining the prespecified primary analysis set for both plasma endpoints ( Fig. 1 ). Included and excluded participants were similar in age, sex, and APOE ε4 frequency, but differed in diagnostic composition: the included set contained a higher proportion of cognitively unimpaired individuals (84/138 [61%] vs 76/182 [42%]) and a lower proportion with mild cognitive impairment (38/138 [28%] vs 93/182 [51%]) ( Table 1 ). Table 1: Representativeness (included vs excluded) Group No. Age, mean (SD) Female, No. (%) APOE ε4+, No. (%) CN, No. (%) MCI, No. (%) AD, No. (%) Other/unknown, No. (%) Included (CSF β-amyloid 42/40 available) 138 71.1 (7.1) 69 (50.0) 47 (34.1) 84 (60.9) 38 (27.5) 15 (10.9) 1 (0.7) Excluded (CSF β-amyloid 42/40 unavailable) 182 71.2 (6.3) 93 (51.1) 60 (33.0) 76 (41.8) 93 (51.1) 8 (4.4) 5 (2.7) Abbreviations: AD, Alzheimer disease dementia; APOE, apolipoprotein E; CN, cognitively unimpaired; CSF, cerebrospinal fluid; MCI, mild cognitive impairment. Pattern-to-probability posteriors In the primary CSF Aβ42/40 analysis set, concordant PET+/CSF+ patterns yielded posterior probabilities of latent amyloid positivity near 1, whereas concordant PET−/CSF− patterns yielded probabilities near 0, across plasma strata for both plasma Aβ42/40 ratio and plasma %p-tau217 ( Fig. 2 ; Supplementary Data 1 ). For plasma Aβ42/40 ratio, the estimated was 0.998 (95% credible interval [CrI], 0.992–1.000) for PET+/CSF+/High (n=22) and 0.001 (95% CrI, <0.01) for PET−/CSF−/Low (n=19). In contrast, discordant PET–CSF patterns mapped to intermediate posterior probabilities with wider uncertainty, and were refined by plasma strata; for example, PET+/CSF−/Intermediate had 0.462 (95% CrI, 0.025–0.961; n=6). Plasma quantile cutpoints used to define low/intermediate/high strata are provided in Supplementary Data 2 . For plasma %p-tau217, was 0.9997 (95% CrI, 0.998–1.000) for PET+/CSF+/High (n=28) and 0.0001 (95% CrI, <0.001) for PET−/CSF−/Low (n=26) ( Supplementary Data 1 ). Discordant patterns again produced intermediate probabilities; for example, PET+/CSF−/Intermediate had 0.325 (95% CrI, 0.010–0.876; n=7) and PET−/CSF+/Intermediate had 0.323 (95% CrI, 0.011–0.865; n=6). Several PET–CSF–plasma combinations were unobserved (n=0) ( Fig. 2 ), reflecting sparse cells for some patterns. Timing-stratified discordance PET–CSF discordance was present across all prespecified PET–CSF timing strata ( Figs. 3–4 ; Supplementary Data 3 ). In the ≤7-day stratum, 12/98 participants were discordant (12%; 95% CI, 7–20%). Discordance was 2/28 (7%; 95% CI, 2–23%) for 8–30 days and 1/12 (8%; 95% CI, 1–35%) for >30 days, with limited precision in longer-gap strata due to smaller denominators. Model-based posterior predictive PET–CSF discordance in the ≤7-day stratum was 0.16 (95% CrI, 0.09–0.23) for plasma Aβ42/40 ratio and 0.15 (95% CrI, 0.09–0.23) for plasma %p-tau217 ( Supplementary Data 4 ). Within the ≤7-day stratum, the intermediate plasma category represented the majority of observations in PET–plasma comparisons (60/96 [63%] for plasma Aβ42/40 ratio; 57/96 [59%] for plasma %p-tau217) ( Supplementary Data 3 ). Restricting to determinate plasma strata (low vs high), mismatch with PET was 6/36 (17%) for plasma Aβ42/40 ratio and 0/39 (0%) for plasma %p-tau217. Similarly, in CSF–plasma comparisons within ≤7 days (n=117), intermediate plasma comprised 75/117 (64%) for plasma Aβ42/40 ratio and 67/117 (57%) for plasma %p-tau217; mismatch among determinate strata was 9/42 (21%) for plasma Aβ42/40 ratio and 0/50 (0%) for plasma %p-tau217. Estimates in longer timing strata were less precise due to smaller denominators ( Supplementary Data 3 ). Sensitivity and coverage analyses Median Centiloids and CSF Aβ42/40 values varied systematically across PET–CSF–plasma patterns, supporting biological coherence of the pattern-to-probability gradients ( Supplementary Fig. 1 ; Supplementary Data 6 ). In prespecified sensitivity analyses, posterior probabilities for concordant PET+/CSF+ and PET−/CSF− patterns remained stably near 1 and near 0, respectively, whereas discordant and borderline patterns showed the greatest sensitivity to model assumptions—including PET–CSF conditional dependence, CSF cutpoint bands, and plasma discretization choices ( Supplementary Figs. 2–5 ; Supplementary Data 5 ). Model convergence diagnostics are reported in Supplementary Data 7 . To address incomplete availability of CSF Aβ42/40, we performed a prespecified coverage analysis using Elecsys CSF Aβ42 (n=320). Pattern-level posterior probabilities showed similar gradients across plasma strata to the primary analysis ( Supplementary Fig. 6 ; Supplementary Data 8 ). Observed PET–CSF discordance in the coverage cohort was ~0.19 for ≤7 days and ~0.15 for 8–30 days ( Supplementary Data 9 ), with timing-stratified posterior predictive discordance estimates and timing decompositions shown in Supplementary Data 10 and Supplementary Figs. 7–8 . Coverage plasma quantile cutpoints and key sensitivity posteriors are provided in Supplementary Data 11–12 . Discussion In paired PET–CSF–plasma triads, Bayesian latent class modeling reframes amyloid assessment as estimation of a latent amyloid-positive state and yields directly interpretable, reference-standard–independent posterior probabilities for each observed biomarker pattern 13,14 . Within this framework, concordant PET+/CSF+ and PET−/CSF− profiles mapped to posteriors near 1 and near 0, respectively, whereas PET–CSF discordance, observed in a meaningful minority of individuals in prior cohorts, was mapped to intermediate probabilities that were systematically refined by plasma strata 1,5 . Together, these results support a clinically usable “pattern-to-probability” interpretation of amyloid biomarkers that avoids treating PET or CSF as ground truth in cases where they disagree 14 . Two features of the analysis are particularly important for interpretation. First, allowing residual PET–CSF dependence materially affected posterior probabilities for some discordant and borderline patterns, reinforcing that conditional independence should not be assumed by default in no–gold-standard biomarker settings and is best treated as a prespecified sensitivity axis 14,15 . Second, PET–CSF discordance persisted even when modalities were closely paired in time, suggesting that disagreement is not simply a timing artifact within common acquisition windows; rather, it is consistent with evidence that PET–CSF discordance represents a typical transitional stage in amyloid biomarker evolution (with CSF-first and PET-first pathways) 5,6,10 . Nonetheless, inference about timing effects in longer-gap strata remains limited by sparse denominators, underscoring the need for larger, intentionally time-aligned datasets to isolate temporal contributions to discordance. A central advantage of this framework is that it is explicitly comparator-agnostic. Rather than reporting agreement “vs PET” or “vs CSF”, metrics that are inherently reference-standard–dependent when the comparator itself is imperfect and PET–CSF discordance occurs in routine cohorts, we estimate for the observed PET–CSF–plasma pattern and report uncertainty around that estimate 13,14,22,23 . This is aligned with contemporary biomarker workflows in which blood tests support triage and, when needed, escalate to confirmatory testing, but do not fully eliminate ambiguity in intermediate or discordant cases 1,2,12 . Finally, because CSF Aβ42/40 ratios (which often improve concordance with PET relative to Aβ42 alone) are not universally available across datasets and platforms, we prespecified a CSF Aβ42–based coverage analysis to evaluate whether the pattern-to-probability gradients were preserved when extending to the full cohort while retaining CSF Aβ42/40 as the primary definition 10,17 . Several limitations merit emphasis. First, latent class inference in the absence of a gold standard is identifiable only under explicit modeling constraints (including assumptions about conditional dependence), and posterior probabilities for discordant and borderline patterns can therefore be sensitive to these specifications, even when concordant patterns remain stable 15,24,25 . Second, ADNI is a deeply phenotyped research cohort designed to approximate trial-like sampling, but its case-mix and limited ethnocultural diversity can constrain generalizability and may introduce spectrum effects when transporting pattern-level probabilities to routine clinical populations 25–27 . Third, we discretized continuous plasma markers into low/intermediate/high strata to improve interpretability and align with triage-style reporting; however, categorization can reduce information and precision relative to continuous modeling, and thresholds may not transfer across assays and populations 28,29 . Finally, some pattern cells and longer timing strata were sparse, limiting precision for timing-specific estimates and motivating replication in larger, deliberately time-aligned datasets. These findings have practical implications for both clinical workflows and study reporting. As blood-based biomarkers are increasingly used to triage patients for confirmatory amyloid assessment, a pattern-to-probability output can provide a transparent way to communicate uncertainty 1,3 . This particularly applicable for the common “gray-zone” scenario of PET–CSF disagreement, while remaining compatible with appropriate-use recommendations that still emphasize PET and/or CSF confirmation in many contexts 2,8,9 . More broadly, our results motivate a shift from single-comparator accuracy summaries toward pattern-level probability reporting with prespecified sensitivity analyses, which can improve comparability across studies and reduce implicit overconfidence driven by reference-standard choice 2,12 . Future work should validate these pattern-level posteriors in external, more demographically diverse cohorts; evaluate continuous (rather than discretized) plasma models calibrated to workflow-aligned thresholds; and extend the framework to joint modeling of amyloid with tau and neurodegeneration markers and clinically relevant outcomes, enabling decision support that better matches the multi-modal reality of Alzheimer biomarker implementation 2,12 . Methods Study design and data source This study was a cross-sectional secondary analysis of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) 18 , a multicenter, longitudinal observational cohort. ADNI data were accessed through the ADNI data repository and included acquisitions collected between 2010 and 2022; all analyses were conducted in December 2025. Ethics approvals and consent ADNI study procedures were approved by the institutional review board at each participating site, and all participants (or their legally authorized representatives) provided written informed consent. This secondary analysis used deidentified ADNI participant-level data accessed through the ADNI repository and involved no new participant recruitment or biospecimen collection. The University of Texas at Tyler Institutional Review Board determined this work to be Not Human Subjects Research (IRB #2025-286). Participants and PET-anchored triad construction The analytic unit was a PET-anchored triad consisting of (i) an amyloid PET scan, (ii) a CSF collection, and (iii) a plasma collection from the same participant. Triads were constructed using prespecified symmetric pairing windows of ±90 days for PET–plasma and ±180 days for PET–CSF and plasma–CSF. When multiple plasma and/or CSF observations were eligible for a given PET scan, we selected the PET–CSF–plasma combination that minimized, in order, (1) the maximum absolute pairwise time gap and then (2) the sum of absolute pairwise time gaps across PET–plasma, PET–CSF, and plasma–CSF; deterministic tie-breakers favored smaller gaps and earlier acquisition dates. For the primary analysis, we selected one triad per participant (unique ADNI research identification number) to avoid within-participant correlation and to align the estimand with a cross-sectional probability of latent amyloid positivity at the PET anchor time. Triad selection was deterministic: we prioritized availability of the primary CSF Aβ42/40 measure and then minimized, in order, the maximum and sum of absolute pairwise day gaps. Analyses were performed as complete-case analyses for each plasma endpoint, excluding triads with missing PET status, CSF amyloid status, or the corresponding plasma endpoint. Biomarker measures and definitions Amyloid PET status was defined using the ADNI UC Berkeley amyloid PET processing and summary measures, which provide a binary amyloid classification derived from standardized uptake value ratio (SUVR) summaries after PET–MRI co-registration and intensity normalization to a reference region 30 . For CSF, primary CSF amyloid positivity (CSF-A) was defined using the Roche Elecsys CSF Aβ42/40 ratio and a published ADNI Elecsys CSF-only (“algorithm”) cutpoint (0.0525) 17 . We also evaluated a prespecified multiplicative cutpoint band (0.90×, 0.95×, 1.00×, 1.05×, 1.10×) to quantify sensitivity of inferences to plausible CSF threshold variation 17 . Because Elecsys CSF Aβ42/40 was available for only a subset of participants, we prespecified a coverage/transportability analysis that used Elecsys CSF Aβ42 alone to define CSF amyloid status in the full cohort (n=320), applying a published ADNI Elecsys CSF-only (“algorithm”) cutpoint (963 pg/mL) and the same multiplicative cutpoint band (0.90×–1.10×) 17,19 . Plasma biomarkers were derived from PrecivityAD2 component measurements obtained by high-throughput LC–MS/MS: plasma Aβ42/40 ratio and plasma %p-tau217 20,21 . For interpretability, each plasma endpoint was oriented so that higher values indicated greater likelihood of amyloid positivity and then discretized into low/intermediate/high strata using prespecified within-cohort quantiles (≤20th, 20th–80th, ≥80th percentile); a sensitivity analysis repeated key outputs using workflow-aligned triage thresholds 13,14,16,20 . Outcomes The primary outcome was the pattern-level posterior probability of latent amyloid positivity at the PET anchor time, defined as where denotes the latent amyloid-positive state. Secondary outcomes quantified discordance (observed and posterior predictive) for PET–CSF (primary), and for PET–plasma and CSF–plasma (supporting). Timing strata were prespecified using absolute pairwise day gaps (≤7 days, 8–30 days, and >30 days up to 180 days) and applied separately to each pairwise comparison. For tri-category plasma strata, discordance summaries included the intermediate fraction and, among determinate strata (low vs high), the mismatch rate relative to the corresponding binary comparator. Statistics and reproducibility We used Bayesian latent class models to integrate PET, CSF, and plasma as imperfect indicators of a latent amyloid state ( ) without assuming any single modality as a gold standard 13,14,16 . In the baseline model, PET status, CSF amyloid status, and plasma stratum were assumed conditionally independent given . Because PET and CSF may retain residual correlation even after conditioning on amyloid status, we prespecified a sensitivity model that allowed conditional dependence between PET and CSF by modeling their joint distribution within each latent class; plasma remained modeled as conditionally independent given 14,15 . Priors were prespecified and weakly informative: Beta(1,1) priors for latent prevalence and binary test parameters; Dirichlet(1,1,1) priors for plasma-category probabilities within each latent class; and, in the PET–CSF dependence sensitivity model, Dirichlet(1,1,1,1) priors for the within-class joint PET–CSF probabilities. Posterior sampling used Gibbs sampling with 4 chains (2,000 burn-in iterations; 5,000 retained draws per chain) for the primary and dependence models; timing-stratified models used 2 chains (1,500 burn-in; 3,000 retained draws). Convergence was assessed using split and effective sample size diagnostics, and model fit was evaluated using posterior predictive checks comparing observed versus model-implied pattern frequencies and discordance rates 31–33 . We report posterior means with 95% credible intervals. For observed discordance proportions, we report 95% Wilson confidence intervals. Declarations Data availability The data supporting this study are available from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and can be accessed through the ADNI data portal (ida.loni.usc.edu) subject to ADNI data-use agreements. Because these data were used under license, they are not redistributed with this manuscript. All analytic code required to reproduce the results is publicly available (see Code availability ); qualified researchers can obtain the underlying ADNI datasets directly from ADNI following approval of a data access request. 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 https://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf. Code availability All code used to construct triads, fit models, and generate figures is available at: https://github.com/SatwantKumar/amyloid-probability-adni. To support reproducibility, the repository includes the full analysis workflow and figure scripts. Acknowledgements Data collection and sharing for this project was funded by the Alzheimer's Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 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. Author contributions S.K. and K.V. conceptualized and designed the study. S.K. acquired the ADNI data, performed data curation, conducted the statistical analyses, and generated the figures and tables. K.V. contributed domain expertise for clinical interpretation of biomarker patterns and results. S.K. and K.V. drafted the manuscript. K.V. and S.K. critically revised the manuscript for important intellectual content. Both authors reviewed and approved the final manuscript. Ethics declarations Competing interests The authors declare no competing interests. References Hansson, O. et al. The Alzheimer’s Association appropriate use recommendations for blood biomarkers in Alzheimer’s disease. Alzheimers Dement 18, 2669–2686 (2022). Palmqvist, S. et al. Alzheimer’s Association Clinical Practice Guideline on the use of blood-based biomarkers in the diagnostic workup of suspected Alzheimer’s disease within specialized care settings. Alzheimer’s and Dementia 21, (2025). Rabinovici, G. D. et al. Updated appropriate use criteria for amyloid and tau PET: A report from the Alzheimer’s Association and Society for Nuclear Medicine and Molecular Imaging Workgroup. Alzheimers Dement 21, e14338 (2025). Angioni, D. et al. Can We Use Blood Biomarkers as Entry Criteria and for Monitoring Drug Treatment Effects in Clinical Trials? A Report from the EU/US CTAD Task Force. Journal of Prevention of Alzheimer’s Disease 10, 418–425 (2023). De Wilde, A. et al. Discordant amyloid-β PET and CSF biomarkers and its clinical consequences. Alzheimer’s Research & Therapy 2019 11:1 11, 78- (2019). Sala, A., Nordberg, A. & Rodriguez-Vieitez, E. Longitudinal pathways of cerebrospinal fluid and positron emission tomography biomarkers of amyloid-β positivity. Molecular Psychiatry 2020 26:10 26, 5864–5874 (2020). Reimand, J., Collij, L., Scheltens, P., Bouwman, F. H. & Ossenkoppele, R. Amyloid-β CSF/PET discordance vs tau load 5 years later: It takes two to tangle. Alzheimer’s & Dementia 16, e037246 (2020). Knopman, D. S. et al. Discrepancies between CSF biomarker and PET determinations of elevated brain amyloid and their prognostic significance. Alzheimers Dement 21, (2025). Leuzy, A. et al. Considerations in the clinical use of amyloid PET and CSF biomarkers for Alzheimer’s disease. Alzheimer’s & Dementia 21, e14528 (2025). Cerman, J. et al. Concordance between amyloid PET and CSF biomarkers in clinical setting: a cross-platform comparison and in-depth analysis of discordant cases. Sci Rep 15, 39963 (2025). Barthélemy, N. R. et al. Highly accurate blood test for Alzheimer’s disease is similar or superior to clinical cerebrospinal fluid tests. Nature Medicine 2024 30:4 30, 1085–1095 (2024). Schindler, S. E. et al. Acceptable performance of blood biomarker tests of amyloid pathology — recommendations from the Global CEO Initiative on Alzheimer’s Disease. Nature Reviews Neurology 2024 20:7 20, 426–439 (2024). Hui, S. L. & Walter, S. D. Estimating the error rates of diagnostic tests. Biometrics 36, 167 (1980). Van Smeden, M., Naaktgeboren, C. A., Reitsma, J. B., Moons, K. G. M. & De Groot, J. A. H. Latent class models in diagnostic studies when there is no reference standard–a systematic review. Am J Epidemiol 179, 423–431 (2014). Dendukuri, N. & Joseph, L. Bayesian approaches to modeling the conditional dependence between multiple diagnostic tests. Biometrics 57, 158–167 (2001). Branscum, A. J., Gardner, I. A. & Johnson, W. O. Estimation of diagnostic-test sensitivity and specificity through Bayesian modeling. Prev Vet Med 68, 145–163 (2005). Dumurgier, J. et al. A Pragmatic, Data-Driven Method to Determine Cutoffs for CSF Biomarkers of Alzheimer Disease Based on Validation Against PET Imaging. Neurology 99, e669 (2022). ADNI | Alzheimer’s Disease Neuroimaging Initiative. https://adni.loni.usc.edu/ . Doecke, J. D. et al. Elecsys CSF biomarker immunoassays demonstrate concordance with amyloid-PET imaging. Alzheimer’s Research & Therapy 2020 12:1 12, 36- (2020). Meyer, M. R. et al. Clinical validation of the PrecivityAD2 blood test: A mass spectrometry-based test with algorithm combining %p-tau217 and Aβ42/40 ratio to identify presence of brain amyloid. Alzheimer’s and Dementia 20, 3179–3192 (2024). Eastwood, S. M. et al. PrecivityAD2TM Blood Test: Analytical Validation of an LC-MS/MS Assay for Quantifying Plasma Phospho-tau217 and Non-Phospho-tau217 Peptide Concentrations That Are Used with Plasma Amyloid-β42/40 in a Multianalyte Assay with Algorithmic Analysis for Detect… Diagnostics (Basel) 14, (2024). Schiller, I. et al. Bias due to composite reference standards in diagnostic accuracy studies. Stat Med 35, 1454–1470 (2016). Kellerhuis, B. E. et al. Drivers of bias in diagnostic test accuracy estimates when using expert panels as a reference standard: a simulation study. BMC Med Res Methodol 25, (2025). Jones, G., Johnson, W. O., Hanson, T. E. & Christensen, R. Identifiability of models for multiple diagnostic testing in the absence of a gold standard. Biometrics 66, 855–863 (2010). Mulherin, S. A. & Miller, W. C. Spectrum bias or spectrum effect? Subgroup variation in diagnostic test evaluation. Ann Intern Med 137, 598–602 (2002). Veitch, D. P. et al. The Alzheimer’s Disease Neuroimaging Initiative in the era of Alzheimer’s disease treatment: A review of ADNI studies from 2021 to 2022. Alzheimer’s & Dementia 20, 652 (2023). Veitch, D. P. et al. Using the Alzheimer’s Disease Neuroimaging Initiative to improve early detection, diagnosis, and treatment of Alzheimer’s disease. Alzheimers Dement 18, 824–857 (2022). Altman, D. G. & Royston, P. The cost of dichotomising continuous variables. BMJ 332, 1080 (2006). Royston, P., Altman, D. G. & Sauerbrei, W. Dichotomizing continuous predictors in multiple regression: a bad idea. Stat Med 25, 127–141 (2006). Weiner, M. W. et al. The Alzheimer’s Disease Neuroimaging Initiative 3: Continued innovation for clinical trial improvement. Alzheimer’s and Dementia 13, 561–571 (2017). Gelman, A., Meng, X. L. & Stern, H. Posterior predictive assessment of model fitness via realized discrepancies. Stat Sin 6, 733–807 (1996). Brooks, S. P. & Gelman, A. General methods for monitoring convergence of iterative simulations)? Journal of Computational and Graphical Statistics 7, 434–455 (1998). Vehtarh, A., Gelman, A., Simpson, D., Carpenter, B. & Burkner, P. C. Rank-Normalization, Folding, and Localization: An Improved Rˆ for Assessing Convergence of MCMC (with Discussion). https://doi.org/10.1214/20 -BA1221 16, 667–718 (2021). Additional Declarations No competing interests reported. Supplementary Files SupplementaryInformationV1SKKV.pdf SupplementaryData4TimingDiscordancePosteriorPredictivePrimaryCSFAbeta4240.csv SupplementaryData7ModelDiagnosticsPrimary.csv SupplementaryData9CoverageTimingDiscordanceObservedCSFAbeta42.csv SupplementaryData1PatternPosteriorTablePrimaryCSFAbeta4240.csv SupplementaryData6PlausibilityByPatternPrimary.csv SupplementaryData10CoverageTimingDiscordancePosteriorPredictiveCSFAbeta42.csv SupplementaryData8CoveragePatternPosteriorTableCSFAbeta42.csv SupplementaryData12CoverageSensitivityKeyPosteriors.csv SupplementaryData11CoveragePlasmaQuantileCutpoints.csv SupplementaryData2PlasmaQuantileCutpointsPrimary.csv SupplementaryData5SensitivityKeyPosteriorsPrimary.csv SupplementaryData3TimingDiscordanceObservedPrimaryCSFAbeta4240.csv Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 21 Apr, 2026 Reviews received at journal 08 Apr, 2026 Reviewers agreed at journal 23 Feb, 2026 Reviews received at journal 20 Feb, 2026 Reviewers agreed at journal 18 Feb, 2026 Reviewers invited by journal 12 Jan, 2026 Editor assigned by journal 06 Jan, 2026 Submission checks completed at journal 03 Jan, 2026 First submitted to journal 31 Dec, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8492792","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":570391078,"identity":"542b6d97-4b24-469f-94e5-ec7c4a0f82b7","order_by":0,"name":"Khushboo Verma","email":"","orcid":"","institution":"University of Texas at Tyler","correspondingAuthor":false,"prefix":"","firstName":"Khushboo","middleName":"","lastName":"Verma","suffix":""},{"id":570391080,"identity":"fa9c81f0-7b9e-4af4-85ef-b4025a217550","order_by":1,"name":"Satwant Kumar","email":"data:image/png;base64,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","orcid":"","institution":"Alzheimer’s Alliance of Smith County","correspondingAuthor":true,"prefix":"","firstName":"Satwant","middleName":"","lastName":"Kumar","suffix":""}],"badges":[],"createdAt":"2026-01-01 02:23:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8492792/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8492792/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100006885,"identity":"66f1f0a1-d528-4541-b511-5742247b0427","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":755182,"visible":true,"origin":"","legend":"","description":"","filename":"ManuscriptFinalSKKV.docx","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/5fd5872e8ef633fb2a365035.docx"},{"id":100361350,"identity":"91d33bdd-8d7c-4f2e-bc46-623a7cf0bbab","added_by":"auto","created_at":"2026-01-16 07:45:00","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5041,"visible":true,"origin":"","legend":"","description":"","filename":"7997329687064997922eea729c3a5de3.json","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/821c677371d34690c630a9d8.json"},{"id":100006886,"identity":"e75fb2ea-c942-4411-b182-4d9d78266f93","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3683,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData10CoverageTimingDiscordancePosteriorPredictiveCSFAbeta42.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/1584ee957a34aa70f2cf918c.csv"},{"id":100006896,"identity":"b110a397-44fe-4108-9a98-36811e95bdd2","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":251,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData11CoveragePlasmaQuantileCutpoints.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/ac5395b89c8629f63d8019c6.csv"},{"id":100006891,"identity":"2e4f558e-1d71-4e2d-89a8-f46cd2e02410","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17249,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData12CoverageSensitivityKeyPosteriors.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/7b523b0fec7c402748c3eabc.csv"},{"id":100006880,"identity":"45533913-c3cf-4c95-8705-b6e00270d90b","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6988,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData1PatternPosteriorTablePrimaryCSFAbeta4240.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/80caa7590856fe2f1140146f.csv"},{"id":100361687,"identity":"17953f1c-b118-467c-a43c-785b4fede24e","added_by":"auto","created_at":"2026-01-16 07:45:30","extension":"csv","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":253,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData2PlasmaQuantileCutpointsPrimary.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/5db53d7626cdd0efd59dec7d.csv"},{"id":100006862,"identity":"10ce7f61-0959-4be4-999b-dc7645af6b11","added_by":"auto","created_at":"2026-01-12 05:41:57","extension":"csv","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4199,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData3TimingDiscordanceObservedPrimaryCSFAbeta4240.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/f047a6fe02b4e37548fc1e4f.csv"},{"id":100006883,"identity":"5585b9fa-f0ce-4b4c-9503-b00e04684527","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4347,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData4TimingDiscordancePosteriorPredictivePrimaryCSFAbeta4240.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/44c251cd4d1b2be5974b10cd.csv"},{"id":100006866,"identity":"6dec9146-32a8-4b07-98ab-6c6bb59becce","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17243,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData5SensitivityKeyPosteriorsPrimary.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/ca152aa15671bf43b7fa383f.csv"},{"id":100006864,"identity":"f564c35d-f8cb-46b1-b785-6caff8f431f8","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4870,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData6PlausibilityByPatternPrimary.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/2736970761b9ed92f5cd5977.csv"},{"id":100006890,"identity":"fc9451f3-7f7a-45f1-bf5f-3bd58f423669","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":426,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData7ModelDiagnosticsPrimary.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/bdd2704cb44cb047f20ed288.csv"},{"id":100006863,"identity":"08e56193-5666-4905-884e-9c2154892e89","added_by":"auto","created_at":"2026-01-12 05:41:57","extension":"csv","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6999,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData8CoveragePatternPosteriorTableCSFAbeta42.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/a630a2e74c8caf6696da5d88.csv"},{"id":100006889,"identity":"4e74db6a-035e-4d86-bc2d-1e41d8aaad90","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3543,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData9CoverageTimingDiscordanceObservedCSFAbeta42.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/71c9d2a9b8c06c6cbee7f505.csv"},{"id":100361598,"identity":"e3cac40e-325a-4c6d-a805-c5841c02c0bd","added_by":"auto","created_at":"2026-01-16 07:45:21","extension":"pdf","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":659823,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformationV1SKKV.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/f165b5c089d6a4dd5cc8bd45.pdf"},{"id":100006882,"identity":"f8645d05-62d9-44bb-98eb-f7853e679e0d","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"xml","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":89183,"visible":true,"origin":"","legend":"","description":"","filename":"7997329687064997922eea729c3a5de31enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/682e4dc2e1e5b8fc5351be63.xml"},{"id":100361500,"identity":"e9017173-92c5-4313-80be-9edf5336d255","added_by":"auto","created_at":"2026-01-16 07:45:13","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":42190,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/d682df78d6c28895c6dcdf86.png"},{"id":100006874,"identity":"859deff3-a126-4c09-8bb4-b68ecb5be757","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69529,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/bfa3f9cde30773b0f5d641f8.png"},{"id":100361322,"identity":"f1bdf636-d757-4baf-a629-6669ca747500","added_by":"auto","created_at":"2026-01-16 07:44:55","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40708,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/a6a88997806f460e42ceaece.png"},{"id":100006884,"identity":"8bfb862f-279e-42b3-8a36-b505ce1c0044","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38930,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/1a7aa618fd2c4eab0c99da3b.png"},{"id":100360783,"identity":"3353a808-6736-4c78-bcb0-deb67c7086f4","added_by":"auto","created_at":"2026-01-16 07:43:43","extension":"xml","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":89173,"visible":true,"origin":"","legend":"","description":"","filename":"7997329687064997922eea729c3a5de31structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/c1a5ed42656e428c96191988.xml"},{"id":100360841,"identity":"e8c28aef-2e19-47fa-9f93-d57c6eabab89","added_by":"auto","created_at":"2026-01-16 07:44:00","extension":"html","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":98893,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/6ef0e627795a42a114bec1c8.html"},{"id":100006861,"identity":"50da2f79-7e39-4949-b2ba-0f8f685ed854","added_by":"auto","created_at":"2026-01-12 05:41:57","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":95611,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCohort construction and analytic sets\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFlow diagram summarizing cohort construction from PET-anchored triads and the analytic sets used for the primary analysis (Elecsys CSF Aβ42/40 available) and the prespecified coverage analysis (Elecsys CSF Aβ42 available). Counts are shown as triads and unique participants (research identification numbers).\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/4853e48739bbfe7d56ba225a.png"},{"id":100006870,"identity":"1ce94877-ab59-4ac2-88bc-4109086862e0","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":153445,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePosterior probability of latent amyloid positivity by biomarker pattern\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHeatmaps show the posterior mean probability of latent amyloid positivity (A*) for each observed PET–CSF–plasma pattern in the primary CSF Aβ42/40 analysis. a) Plasma Aβ42/40 ratio strata. b) Plasma %p-tau217 strata. Rows indicate PET and CSF amyloid status; columns indicate plasma strata (low/intermediate/high). Each cell shows the posterior mean and the observed cell count (n). Cells shaded light gray indicate unobserved patterns (n=0).\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/7ba54f5b78687d2dfb6b907e.png"},{"id":100006868,"identity":"42c5174b-b3c6-4bdf-8c6b-f8b0e7db63ff","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":112713,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTiming-stratified discordance across PET, CSF, and plasma Aβ42/40 ratio\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTiming-stratified discordance and mismatch summaries for plasma Aβ42/40 ratio across prespecified absolute time-gap strata (≤7 days, 8–30 days, \u0026gt;30 days). a) PET vs CSF amyloid discordance by |ΔPET–CSF|. b) PET vs plasma mismatch among determinate plasma strata (low vs high) by |ΔPET–plasma|. c) CSF amyloid vs plasma mismatch among determinate strata by |ΔCSF–plasma|. d) Plasma intermediate fraction shown for PET–plasma (filled markers) and CSF amyloid–plasma (open markers). Observed estimates are shown as blue circles with 95% Wilson confidence intervals; model estimates are shown as red squares with 95% posterior predictive credible intervals. Timing strata with fewer than 10 determinate observations are not plotted.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/61d99b7f69dab619af88d5e9.png"},{"id":100006865,"identity":"78b2e280-2952-4450-a097-cc1878f32598","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":108897,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTiming-stratified discordance across PET, CSF, and plasma %p-tau217\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSame format as Fig. 3, but with plasma strata defined by plasma %p-tau217. a) PET vs CSF amyloid discordance. b) PET vs plasma mismatch among determinate plasma strata (low vs high). c) CSF amyloid vs plasma mismatch among determinate strata. d) Plasma intermediate fraction shown for PET–plasma (filled markers) and CSF amyloid–plasma (open markers). Observed estimates are shown as blue circles with 95% Wilson confidence intervals; model estimates are shown as red squares with 95% posterior predictive credible intervals. Timing strata with fewer than 10 determinate observations are not plotted.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/4de54486828db2967613068b.png"},{"id":100380857,"identity":"b305d983-f1df-4b2e-a992-d2ecc15040e3","added_by":"auto","created_at":"2026-01-16 10:35:57","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1287215,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/bd3dbbab-23e4-4d95-9b40-267307e96ede.pdf"},{"id":100360786,"identity":"6630d93d-40f9-4cd0-b14c-8209ca62936d","added_by":"auto","created_at":"2026-01-16 07:43:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":659823,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformationV1SKKV.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/a9a2a24955b770cb7315cba8.pdf"},{"id":100006860,"identity":"907b54e6-7ebc-4f6d-b6ad-6552b522578e","added_by":"auto","created_at":"2026-01-12 05:41:57","extension":"csv","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4347,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData4TimingDiscordancePosteriorPredictivePrimaryCSFAbeta4240.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/be4fdf3da4f40a12ee717984.csv"},{"id":100006887,"identity":"ddc4edd9-2fdf-4213-a3c0-33df6cbd7698","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":426,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData7ModelDiagnosticsPrimary.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/222498265bdf93320db7da97.csv"},{"id":100006895,"identity":"e9b72bd8-83b5-42ed-841f-7ff808cad3ef","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":3543,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData9CoverageTimingDiscordanceObservedCSFAbeta42.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/4497b17f1dc0cad6d00295a1.csv"},{"id":100006877,"identity":"6a76a5ac-add6-414e-8374-aae8028e6a43","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":6988,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData1PatternPosteriorTablePrimaryCSFAbeta4240.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/39897e562ad83433c4793a2e.csv"},{"id":100006898,"identity":"1953c6f2-add7-445b-a845-6506d651bf07","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":4870,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData6PlausibilityByPatternPrimary.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/f2a69d6389ba3960d6422996.csv"},{"id":100006873,"identity":"c63c9111-efbd-4e22-8041-cecd05ad057b","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":3683,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData10CoverageTimingDiscordancePosteriorPredictiveCSFAbeta42.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/ffcb454981793223ecdff528.csv"},{"id":100360861,"identity":"51758b64-8b3e-48a7-a26c-cddc5474a004","added_by":"auto","created_at":"2026-01-16 07:44:04","extension":"csv","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":6999,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData8CoveragePatternPosteriorTableCSFAbeta42.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/72d373470d15bf823d8ef845.csv"},{"id":100006894,"identity":"fcecc34e-f755-4fab-8338-be22f180b4f8","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":17249,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData12CoverageSensitivityKeyPosteriors.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/cdbaf73d849225202143357f.csv"},{"id":100006871,"identity":"2930a400-01e9-4218-9617-6fc056be161f","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":251,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData11CoveragePlasmaQuantileCutpoints.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/a1c3b4bad419937c23f408f1.csv"},{"id":100006879,"identity":"ce1be507-3f81-4d8f-9e25-423d22aef2f1","added_by":"auto","created_at":"2026-01-12 05:41:58","extension":"csv","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":253,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData2PlasmaQuantileCutpointsPrimary.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/599442befe845141c185db1d.csv"},{"id":100006888,"identity":"e0192830-1c6d-4ce3-991d-e0a1dd02fded","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":17243,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData5SensitivityKeyPosteriorsPrimary.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/cb53792eb9e704715b7bf73e.csv"},{"id":100006892,"identity":"ec1e910a-98cd-4bba-a851-10e1bd2287a0","added_by":"auto","created_at":"2026-01-12 05:41:59","extension":"csv","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":4199,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryData3TimingDiscordanceObservedPrimaryCSFAbeta4240.csv","url":"https://assets-eu.researchsquare.com/files/rs-8492792/v1/36c73c39e76a3bd0b90fcf17.csv"}],"financialInterests":"No competing interests reported.","formattedTitle":"Amyloid Probability in Alzheimer Disease From Plasma, Cerebrospinal Fluid, and Amyloid Imaging","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDetermining whether an individual harbors Alzheimer-type brain amyloid has become increasingly important as biomarker confirmation now shapes diagnostic pathways and eligibility for amyloid-targeting therapeutic and research programs\u003csup\u003e\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Blood-based biomarkers (BBMs) are emerging as scalable, low-burden gatekeepers that estimate amyloid likelihood and can triage patients for confirmatory testing, while current appropriate-use recommendations still emphasize confirmation with cerebrospinal fluid (CSF) measures and/or amyloid PET\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. In parallel, BBMs are being incorporated into trial screening and enrichment workflows to reduce burden and screen failure, even as PET and CSF remain the predominant confirmatory modalities in many protocols\u003csup\u003e\u003cspan additionalcitationids=\"CR2 CR3\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eHowever, amyloid \u0026ldquo;confirmation\u0026rdquo; is not always unambiguous: amyloid PET and CSF amyloid measures, while generally concordant, can yield discordant classifications in a clinically meaningful minority of individuals (often\u0026thinsp;~\u0026thinsp;10\u0026ndash;20%)\u003csup\u003e5,6\u003c/sup\u003e. These discordant profiles are not merely technical noise; multiple longitudinal and biomarker studies suggest they can carry biological and prognostic information, including differential associations with downstream tau pathology and clinical trajectories depending on whether CSF becomes abnormal before PET or vice versa\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. As a result, \u0026ldquo;PET\u0026thinsp;+\u0026thinsp;vs CSF+\u0026rdquo; is not a trivial labeling choice, discordance creates a real interpretive gap for both clinical decision-making and study design when the two most common confirmatory modalities disagree\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDespite reported discordances diagnostic performance is still commonly framed \u0026ldquo;versus PET\u0026rdquo; or \u0026ldquo;versus CSF,\u0026rdquo; as if one modality were a definitive ground truth. This convention is increasingly strained: although both Aβ-PET and CSF amyloid assays aim to detect amyloid plaque pathology, they index different (but related) biological processes\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e, with PET tracking fibrillar plaque burden and CSF reflecting soluble Aβ dynamics\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Whereas a clinically meaningful minority of individuals show persistent PET\u0026ndash;CSF discordance\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. When plasma biomarkers are evaluated against a single comparator, discordant and borderline cases are forced into \u0026ldquo;false-positive/false-negative\u0026rdquo; bins by design choice rather than biology, and even very high-performing blood tests inherit the uncertainty (and potential ceiling) of an imperfect reference\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. In practice, reported \u0026ldquo;accuracy\u0026rdquo; can therefore shift with the selected reference test and thresholds, while clinicians and trialists remain without a calibrated probability statement for the common real-world scenario of PET\u0026ndash;CSF disagreement\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eA natural way to resolve this reference-standard problem is to treat amyloid status as a latent (unobserved) state and jointly model PET, CSF, and plasma as imperfect indicators of that state, rather than forcing one modality to serve as ground truth\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Latent class models were developed precisely for diagnostic settings without a single accurate reference standard and can yield directly interpretable outputs such as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{P}\\left(amyloid+\\right)\\)\u003c/span\u003e\u003c/span\u003e given the observed test pattern, with uncertainty\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e,\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Importantly, biomarker applications must also confront conditional dependence, for example, PET and CSF may remain correlated even after conditioning on latent amyloid. This assumption violation can materially bias inference if left implicit. To incorporate (and stress-test) such dependence structures Bayesian formulations provide a principled way\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. In this context, adding plasma as a third modality is not only clinically pragmatic but also methodologically valuable: an additional indicator can refine posterior probabilities for discordant PET\u0026ndash;CSF patterns and improve identifiability under plausible dependence scenarios\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAny integrative approach in this setting must make its assumptions explicit. Latent class inference without a gold standard is identifiable only under modeling constraints, and results can be sensitive to the commonly made (and often violated) assumption that tests are conditionally independent given the latent state, particularly for PET and CSF, which share biology and measurement pipelines\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. In addition, amyloid classifications depend on assay-specific cutpoints (especially for CSF), where small threshold shifts can move borderline individuals across the positivity boundary and thereby change apparent discordance rates\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. Finally, because PET and CSF may not be acquired on the same day in observational cohorts, the time gap between modalities is a plausible contributor to disagreement and should be treated as a prespecified sensitivity axis rather than a post hoc explanation\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Together, these considerations argue for reporting calibrated, pattern-level probabilities with uncertainty and for stress-testing conclusions across dependence, timing, and cutpoint assumptions\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eHere, in a cross-sectional secondary analysis of the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI)\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e, we assembled per-participant PET\u0026ndash;CSF\u0026ndash;plasma \u0026ldquo;triads\u0026rdquo; and used Bayesian latent class models to estimate the posterior probability of a latent amyloid-positive state for each observed PET\u0026ndash;CSF\u0026ndash;plasma pattern, rather than defining truth by a single modality. Using Elecsys CSF Aβ42/40 as the primary CSF amyloid definition and plasma measurements aligned to the PrecivityAD2 component analytes (plasma Aβ42/40 ratio or plasma %p-tau217), we quantify how plasma strata refine interpretation of discordant PET\u0026ndash;CSF profiles\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan additionalcitationids=\"CR20\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. We prespecified sensitivity analyses for PET\u0026ndash;CSF conditional dependence, modality timing gaps, and CSF cutpoints, and we performed a coverage analysis using Elecsys CSF Aβ42 to address incomplete availability of CSF Aβ42/40\u003csup\u003e8,17\u003c/sup\u003e. Together, these analyses yield a clinically interpretable \u0026ldquo;pattern-to-probability\u0026rdquo; output with explicit uncertainty for common real-world discordant and borderline biomarker patterns.\u003c/p\u003e"},{"header":"Results","content":"\u003ch3\u003e\u003cem\u003eCohort and analytic sets\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eThe PET-anchored cohort comprised 608 eligible PET\u0026ndash;CSF\u0026ndash;plasma triads from 320 participants; selecting one prespecified triad per participant yielded 320 triads for analysis (\u003cstrong\u003eFig. 1\u003c/strong\u003e). Participants had a mean (s.d.) age of 71.1 (6.7) years and included 162 women (51%). Elecsys CSF A\u0026beta;42/40 was available for 138/320 participants (43%), defining the prespecified primary analysis set for both plasma endpoints (\u003cstrong\u003eFig. 1\u003c/strong\u003e). Included and excluded participants were similar in age, sex, and APOE \u0026epsilon;4 frequency, but differed in diagnostic composition: the included set contained a higher proportion of cognitively unimpaired individuals (84/138 [61%] vs 76/182 [42%]) and a lower proportion with mild cognitive impairment (38/138 [28%] vs 93/182 [51%]) (\u003cstrong\u003eTable 1\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1: Representativeness (included vs excluded)\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGroup\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNo.\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge, mean (SD)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFemale, No. (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAPOE \u0026epsilon;4+, No. (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCN, No. (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMCI, No. (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAD, No. (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOther/unknown, No. (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003eIncluded (CSF \u0026beta;-amyloid 42/40 available)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e138\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e71.1 (7.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e69 (50.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e47 (34.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e84 (60.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e38 (27.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e15 (10.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e1 (0.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003eExcluded (CSF \u0026beta;-amyloid 42/40 unavailable)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e182\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e71.2 (6.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e93 (51.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e60 (33.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e76 (41.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e93 (51.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e8 (4.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 69px;\"\u003e\n \u003cp\u003e5 (2.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAbbreviations: AD, Alzheimer disease dementia; APOE, apolipoprotein E; CN, cognitively unimpaired; CSF, cerebrospinal fluid; MCI, mild cognitive impairment.\u003c/p\u003e\n\u003ch3\u003e\u003cem\u003ePattern-to-probability posteriors\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eIn the primary CSF A\u0026beta;42/40 analysis set, concordant PET+/CSF+ patterns yielded posterior probabilities of latent amyloid positivity near 1, whereas concordant PET\u0026minus;/CSF\u0026minus; patterns yielded probabilities near 0, across plasma strata for both plasma A\u0026beta;42/40 ratio and plasma %p-tau217 (\u003cstrong\u003eFig. 2\u003c/strong\u003e; \u003cstrong\u003eSupplementary Data 1\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eFor plasma A\u0026beta;42/40 ratio, the estimated\u0026nbsp;\u003cimg width=\"62\" height=\"31\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u0026nbsp;was 0.998 (95% credible interval [CrI], 0.992\u0026ndash;1.000) for PET+/CSF+/High (n=22) and 0.001 (95% CrI, \u0026lt;0.01) for PET\u0026minus;/CSF\u0026minus;/Low (n=19). In contrast, discordant PET\u0026ndash;CSF patterns mapped to intermediate posterior probabilities with wider uncertainty, and were refined by plasma strata; for example, PET+/CSF\u0026minus;/Intermediate had\u0026nbsp;\u003cimg width=\"62\" height=\"31\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u0026nbsp;0.462 (95% CrI, 0.025\u0026ndash;0.961; n=6). Plasma quantile cutpoints used to define low/intermediate/high strata are provided in \u003cstrong\u003eSupplementary Data 2\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eFor plasma %p-tau217,\u0026nbsp;\u003cimg width=\"63\" height=\"31\" src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAF8AAAAuCAMAAAB57DWjAAAAAXNSR0IArs4c6QAAAHJQTFRFAAAAAAAAAAA6AABmADo6ADpmADqQAGa2OgAAOgA6OjpmOma2OpDbZgAAZjoAZmY6ZpC2Zrb/kDoAkGY6kLbbkNu2kNv/tmYAtmY6tpBmttv/tv//25A627Zm27aQ2////7Zm/9uQ/9u2/9vb//+2///b0Qd2IwAAAAF0Uk5TAEDm2GYAAAAJcEhZcwAAFiUAABYlAUlSJPAAAAAZdEVYdFNvZnR3YXJlAE1pY3Jvc29mdCBPZmZpY2V/7TVxAAABpklEQVRYR+1V23KDIBAFmzSkTWxq29iLGkz1/3+xwK7AWoOY9KUzMOOMynL27NkLjKWVFEgKJAWSAkmBpEBS4F8o0OS7W3hW69fQ8b7Igvvzrpv8/mStJIf18Am/ur23yVjFN/OAaFFvj/DWF3df9lTN9Ucj+POvLe2NE3cBX+ecZ4hPSEpAQKKSE+3LlYjDP79kT4YorArYwqtBBDckMsba7RuV6yL/98Op2zt8T+USXFUGvxW+3H2x8Q/NJIKYloNWqlyMaqUJA4NBJJkdr8W3QC0ILEE8DAbwu/1OPV4paDN/2XyiuWcqh7ozL98fwtgOwWBmlD39E1SIUEHamrFeq4PpCILWCqXYrfj0PPkqdbh94SptSX4H/lRfH99KbSv5GnzsruGoy29fQGORjC/PLy1Irz4rLA6o25hFpEDcsby2v2wDLsH36hX7qxaUnp0PJdavmlpxA4ixmjsspGdySOKH+abCgv9qfnL+6Kb5RZ1aAY2XQTV4840eGQ24GN0nbEaXiG8R2Ip2FiT5B/ejWLvra4JUk8e21GREMny/R6swb/gDvp8dYgcpk3QAAAAASUVORK5CYII=\" alt=\"image\"\u003e\u0026nbsp;was 0.9997 (95% CrI, 0.998\u0026ndash;1.000) for PET+/CSF+/High (n=28) and 0.0001 (95% CrI, \u0026lt;0.001) for PET\u0026minus;/CSF\u0026minus;/Low (n=26) (\u003cstrong\u003eSupplementary Data 1\u003c/strong\u003e). Discordant patterns again produced intermediate probabilities; for example, PET+/CSF\u0026minus;/Intermediate had\u0026nbsp;\u003cimg width=\"63\" height=\"31\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u0026nbsp;0.325 (95% CrI, 0.010\u0026ndash;0.876; n=7) and PET\u0026minus;/CSF+/Intermediate had\u0026nbsp;\u003cimg width=\"63\" height=\"31\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u0026nbsp;0.323 (95% CrI, 0.011\u0026ndash;0.865; n=6). Several PET\u0026ndash;CSF\u0026ndash;plasma combinations were unobserved (n=0) (\u003cstrong\u003eFig. 2\u003c/strong\u003e), reflecting sparse cells for some patterns.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTiming-stratified discordance\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003ePET\u0026ndash;CSF discordance was present across all prespecified PET\u0026ndash;CSF timing strata (\u003cstrong\u003eFigs. 3\u0026ndash;4\u003c/strong\u003e; \u003cstrong\u003eSupplementary Data\u003c/strong\u003e \u003cstrong\u003e3\u003c/strong\u003e). In the \u0026le;7-day stratum, 12/98 participants were discordant (12%; 95% CI, 7\u0026ndash;20%). Discordance was 2/28 (7%; 95% CI, 2\u0026ndash;23%) for 8\u0026ndash;30 days and 1/12 (8%; 95% CI, 1\u0026ndash;35%) for \u0026gt;30 days, with limited precision in longer-gap strata due to smaller denominators. Model-based posterior predictive PET\u0026ndash;CSF discordance in the \u0026le;7-day stratum was 0.16 (95% CrI, 0.09\u0026ndash;0.23) for plasma A\u0026beta;42/40 ratio and 0.15 (95% CrI, 0.09\u0026ndash;0.23) for plasma %p-tau217 (\u003cstrong\u003eSupplementary Data 4\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eWithin the \u0026le;7-day stratum, the intermediate plasma category represented the majority of observations in PET\u0026ndash;plasma comparisons (60/96 [63%] for plasma A\u0026beta;42/40 ratio; 57/96 [59%] for plasma %p-tau217) (\u003cstrong\u003eSupplementary Data 3\u003c/strong\u003e). Restricting to determinate plasma strata (low vs high), mismatch with PET was 6/36 (17%) for plasma A\u0026beta;42/40 ratio and 0/39 (0%) for plasma %p-tau217. Similarly, in CSF\u0026ndash;plasma comparisons within \u0026le;7 days (n=117), intermediate plasma comprised 75/117 (64%) for plasma A\u0026beta;42/40 ratio and 67/117 (57%) for plasma %p-tau217; mismatch among determinate strata was 9/42 (21%) for plasma A\u0026beta;42/40 ratio and 0/50 (0%) for plasma %p-tau217. Estimates in longer timing strata were less precise due to smaller denominators (\u003cstrong\u003eSupplementary Data 3\u003c/strong\u003e).\u003c/p\u003e\n\u003ch3\u003e\u003cem\u003eSensitivity and coverage analyses\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eMedian Centiloids and CSF A\u0026beta;42/40 values varied systematically across PET\u0026ndash;CSF\u0026ndash;plasma patterns, supporting biological coherence of the pattern-to-probability gradients (\u003cstrong\u003eSupplementary Fig. 1\u003c/strong\u003e; \u003cstrong\u003eSupplementary Data 6\u003c/strong\u003e). In prespecified sensitivity analyses, posterior probabilities for concordant PET+/CSF+ and PET\u0026minus;/CSF\u0026minus; patterns remained stably near 1 and near 0, respectively, whereas discordant and borderline patterns showed the greatest sensitivity to model assumptions\u0026mdash;including PET\u0026ndash;CSF conditional dependence, CSF cutpoint bands, and plasma discretization choices (\u003cstrong\u003eSupplementary Figs. 2\u0026ndash;5\u003c/strong\u003e; \u003cstrong\u003eSupplementary Data 5\u003c/strong\u003e). Model convergence diagnostics are reported in \u003cstrong\u003eSupplementary Data 7\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eTo address incomplete availability of CSF A\u0026beta;42/40, we performed a prespecified coverage analysis using Elecsys CSF A\u0026beta;42 (n=320). Pattern-level posterior probabilities showed similar gradients across plasma strata to the primary analysis (\u003cstrong\u003eSupplementary Fig. 6\u003c/strong\u003e; \u003cstrong\u003eSupplementary Data 8\u003c/strong\u003e). Observed PET\u0026ndash;CSF discordance in the coverage cohort was ~0.19 for \u0026le;7 days and ~0.15 for 8\u0026ndash;30 days (\u003cstrong\u003eSupplementary Data 9\u003c/strong\u003e), with timing-stratified posterior predictive discordance estimates and timing decompositions shown in \u003cstrong\u003eSupplementary Data 10\u003c/strong\u003e and \u003cstrong\u003eSupplementary Figs. 7\u0026ndash;8\u003c/strong\u003e. Coverage plasma quantile cutpoints and key sensitivity posteriors are provided in \u003cstrong\u003eSupplementary Data 11\u0026ndash;12\u003c/strong\u003e.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn paired PET\u0026ndash;CSF\u0026ndash;plasma triads, Bayesian latent class modeling reframes amyloid assessment as estimation of a latent amyloid-positive state and yields directly interpretable, reference-standard\u0026ndash;independent posterior probabilities for each observed biomarker pattern\u003csup\u003e13,14\u003c/sup\u003e. Within this framework, concordant PET+/CSF+ and PET\u0026minus;/CSF\u0026minus; profiles mapped to posteriors near 1 and near 0, respectively, whereas PET\u0026ndash;CSF discordance, observed in a meaningful minority of individuals in prior cohorts, was mapped to intermediate probabilities that were systematically refined by plasma strata\u003csup\u003e1,5\u003c/sup\u003e. Together, these results support a clinically usable \u0026ldquo;pattern-to-probability\u0026rdquo; interpretation of amyloid biomarkers that avoids treating PET or CSF as ground truth in cases where they disagree\u003csup\u003e14\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTwo features of the analysis are particularly important for interpretation. First, allowing residual PET\u0026ndash;CSF dependence materially affected posterior probabilities for some discordant and borderline patterns, reinforcing that conditional independence should not be assumed by default in no\u0026ndash;gold-standard biomarker settings and is best treated as a prespecified sensitivity axis\u003csup\u003e14,15\u003c/sup\u003e. Second, PET\u0026ndash;CSF discordance persisted even when modalities were closely paired in time, suggesting that disagreement is not simply a timing artifact within common acquisition windows; rather, it is consistent with evidence that PET\u0026ndash;CSF discordance represents a typical transitional stage in amyloid biomarker evolution (with CSF-first and PET-first pathways)\u003csup\u003e5,6,10\u003c/sup\u003e. Nonetheless, inference about timing effects in longer-gap strata remains limited by sparse denominators, underscoring the need for larger, intentionally time-aligned datasets to isolate temporal contributions to discordance.\u003c/p\u003e\n\u003cp\u003eA central advantage of this framework is that it is explicitly comparator-agnostic. Rather than reporting agreement \u0026ldquo;vs PET\u0026rdquo; or \u0026ldquo;vs CSF\u0026rdquo;, metrics that are inherently reference-standard\u0026ndash;dependent when the comparator itself is imperfect and PET\u0026ndash;CSF discordance occurs in routine cohorts, we estimate\u0026nbsp;\u003cimg width=\"151\" height=\"31\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u0026nbsp;for the observed PET\u0026ndash;CSF\u0026ndash;plasma pattern and report uncertainty around that estimate\u003csup\u003e13,14,22,23\u003c/sup\u003e. This is aligned with contemporary biomarker workflows in which blood tests support triage and, when needed, escalate to confirmatory testing, but do not fully eliminate ambiguity in intermediate or discordant cases\u003csup\u003e1,2,12\u003c/sup\u003e. Finally, because CSF A\u0026beta;42/40 ratios (which often improve concordance with PET relative to A\u0026beta;42 alone) are not universally available across datasets and platforms, we prespecified a CSF A\u0026beta;42\u0026ndash;based coverage analysis to evaluate whether the pattern-to-probability gradients were preserved when extending to the full cohort while retaining CSF A\u0026beta;42/40 as the primary definition\u0026nbsp;\u003csup\u003e10,17\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eSeveral limitations merit emphasis. First, latent class inference in the absence of a gold standard is identifiable only under explicit modeling constraints (including assumptions about conditional dependence), and posterior probabilities for discordant and borderline patterns can therefore be sensitive to these specifications, even when concordant patterns remain stable\u003csup\u003e15,24,25\u003c/sup\u003e. Second, ADNI is a deeply phenotyped research cohort designed to approximate trial-like sampling, but its case-mix and limited ethnocultural diversity can constrain generalizability and may introduce spectrum effects when transporting pattern-level probabilities to routine clinical populations\u003csup\u003e25\u0026ndash;27\u003c/sup\u003e. Third, we discretized continuous plasma markers into low/intermediate/high strata to improve interpretability and align with triage-style reporting; however, categorization can reduce information and precision relative to continuous modeling, and thresholds may not transfer across assays and populations\u003csup\u003e28,29\u003c/sup\u003e. Finally, some pattern cells and longer timing strata were sparse, limiting precision for timing-specific estimates and motivating replication in larger, deliberately time-aligned datasets.\u003c/p\u003e\n\u003cp\u003eThese findings have practical implications for both clinical workflows and study reporting. As blood-based biomarkers are increasingly used to triage patients for confirmatory amyloid assessment, a pattern-to-probability output can provide a transparent way to communicate uncertainty\u003csup\u003e1,3\u003c/sup\u003e. This particularly applicable for the common \u0026ldquo;gray-zone\u0026rdquo; scenario of PET\u0026ndash;CSF disagreement, while remaining compatible with appropriate-use recommendations that still emphasize PET and/or CSF confirmation in many contexts\u003csup\u003e2,8,9\u003c/sup\u003e. More broadly, our results motivate a shift from single-comparator accuracy summaries toward pattern-level probability reporting with prespecified sensitivity analyses, which can improve comparability across studies and reduce implicit overconfidence driven by reference-standard choice\u003csup\u003e2,12\u003c/sup\u003e. Future work should validate these pattern-level posteriors in external, more demographically diverse cohorts; evaluate continuous (rather than discretized) plasma models calibrated to workflow-aligned thresholds; and extend the framework to joint modeling of amyloid with tau and neurodegeneration markers and clinically relevant outcomes, enabling decision support that better matches the multi-modal reality of Alzheimer biomarker implementation\u003csup\u003e2,12\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch3\u003e\u003cem\u003eStudy design and data source\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eThis study was a cross-sectional secondary analysis of the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI)\u003csup\u003e18\u003c/sup\u003e, a multicenter, longitudinal observational cohort. ADNI data were accessed through the ADNI data repository and included acquisitions collected between 2010 and 2022; all analyses were conducted in December 2025.\u003c/p\u003e\n\u003ch3\u003e\u003cem\u003eEthics approvals and consent\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eADNI study procedures were approved by the institutional review board at each participating site, and all participants (or their legally authorized representatives) provided written informed consent. This secondary analysis used deidentified ADNI participant-level data accessed through the ADNI repository and involved no new participant recruitment or biospecimen collection. The University of Texas at Tyler Institutional Review Board determined this work to be Not Human Subjects Research (IRB #2025-286).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eParticipants and PET-anchored triad construction\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe analytic unit was a PET-anchored triad consisting of (i) an amyloid PET scan, (ii) a CSF collection, and (iii) a plasma collection from the same participant. Triads were constructed using prespecified symmetric pairing windows of \u0026plusmn;90 days for PET\u0026ndash;plasma and \u0026plusmn;180 days for PET\u0026ndash;CSF and plasma\u0026ndash;CSF. When multiple plasma and/or CSF observations were eligible for a given PET scan, we selected the PET\u0026ndash;CSF\u0026ndash;plasma combination that minimized, in order, (1) the maximum absolute pairwise time gap and then (2) the sum of absolute pairwise time gaps across PET\u0026ndash;plasma, PET\u0026ndash;CSF, and plasma\u0026ndash;CSF; deterministic tie-breakers favored smaller gaps and earlier acquisition dates.\u003c/p\u003e\n\u003cp\u003eFor the primary analysis, we selected one triad per participant (unique ADNI research identification number) to avoid within-participant correlation and to align the estimand with a cross-sectional probability of latent amyloid positivity at the PET anchor time. Triad selection was deterministic: we prioritized availability of the primary CSF A\u0026beta;42/40 measure and then minimized, in order, the maximum and sum of absolute pairwise day gaps. Analyses were performed as complete-case analyses for each plasma endpoint, excluding triads with missing PET status, CSF amyloid status, or the corresponding plasma endpoint.\u003c/p\u003e\n\u003ch3\u003e\u003cem\u003eBiomarker measures and definitions\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eAmyloid PET status was defined using the ADNI UC Berkeley amyloid PET processing and summary measures, which provide a binary amyloid classification derived from standardized uptake value ratio (SUVR) summaries after PET\u0026ndash;MRI co-registration and intensity normalization to a reference region\u003csup\u003e30\u003c/sup\u003e. For CSF, primary CSF amyloid positivity (CSF-A) was defined using the Roche Elecsys CSF A\u0026beta;42/40 ratio and a published ADNI Elecsys CSF-only (\u0026ldquo;algorithm\u0026rdquo;) cutpoint (0.0525)\u003csup\u003e17\u003c/sup\u003e. We also evaluated a prespecified multiplicative cutpoint band (0.90\u0026times;, 0.95\u0026times;, 1.00\u0026times;, 1.05\u0026times;, 1.10\u0026times;) to quantify sensitivity of inferences to plausible CSF threshold variation\u003csup\u003e17\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eBecause Elecsys CSF A\u0026beta;42/40 was available for only a subset of participants, we prespecified a coverage/transportability analysis that used Elecsys CSF A\u0026beta;42 alone to define CSF amyloid status in the full cohort (n=320), applying a published ADNI Elecsys CSF-only (\u0026ldquo;algorithm\u0026rdquo;) cutpoint (963 pg/mL) and the same multiplicative cutpoint band (0.90\u0026times;\u0026ndash;1.10\u0026times;)\u003csup\u003e17,19\u003c/sup\u003e. Plasma biomarkers were derived from PrecivityAD2 component measurements obtained by high-throughput LC\u0026ndash;MS/MS: plasma A\u0026beta;42/40 ratio and plasma %p-tau217\u003csup\u003e20,21\u003c/sup\u003e. For interpretability, each plasma endpoint was oriented so that higher values indicated greater likelihood of amyloid positivity and then discretized into low/intermediate/high strata using prespecified within-cohort quantiles (\u0026le;20th, 20th\u0026ndash;80th, \u0026ge;80th percentile); a sensitivity analysis repeated key outputs using workflow-aligned triage thresholds\u003csup\u003e13,14,16,20\u003c/sup\u003e.\u003c/p\u003e\n\u003ch3\u003e\u003cem\u003eOutcomes\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eThe primary outcome was the pattern-level posterior probability of latent amyloid positivity at the PET anchor time, defined as\u003c/p\u003e\n\u003ch3\u003e\u003cimg width=\"417\" height=\"29\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u003c/h3\u003e\n\u003cp\u003ewhere \u003cimg width=\"10\" height=\"29\" src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAA8AAAAsCAMAAACNMBLgAAAAAXNSR0IArs4c6QAAAE5QTFRFAAAAAAAAAAA6AABmADo6ADqQAGa2OgAAOgA6Oma2OpDbZgAAZjoAZrb/kDoAkNu2kNv/tmYAtv//25A62////7Zm/9uQ/9u2//+2///bGyFzpwAAAAF0Uk5TAEDm2GYAAAAJcEhZcwAAFiUAABYlAUlSJPAAAAAZdEVYdFNvZnR3YXJlAE1pY3Jvc29mdCBPZmZpY2V/7TVxAAAAeUlEQVQoU+2QSw6AMAhEwV/VKor1097/oqIlhgOYuJHdK5OZDgD/fHOBgI0Njh7r1Txw5Szv7eQNJ2qiL5dHvxWz5eh7sBxEmqiYVb+7Hizz5ZQIx7zfUCdzohzMug/qwyguIDn6L2XWXsdwFyLxunTSD7GzDV85+wngNwT4PKPyiwAAAABJRU5ErkJggg==\" alt=\"image\"\u003e denotes the latent amyloid-positive state. Secondary outcomes quantified discordance (observed and posterior predictive) for PET\u0026ndash;CSF (primary), and for PET\u0026ndash;plasma and CSF\u0026ndash;plasma (supporting). Timing strata were prespecified using absolute pairwise day gaps (\u0026le;7 days, 8\u0026ndash;30 days, and \u0026gt;30 days up to 180 days) and applied separately to each pairwise comparison. For tri-category plasma strata, discordance summaries included the intermediate fraction and, among determinate strata (low vs high), the mismatch rate relative to the corresponding binary comparator.\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003ch3\u003e\u003cem\u003eStatistics and reproducibility\u003c/em\u003e\u003c/h3\u003e\n\u003cp\u003eWe used Bayesian latent class models to integrate PET, CSF, and plasma as imperfect indicators of a latent amyloid state (\u003cimg width=\"10\" height=\"31\" src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAA8AAAAuCAMAAADA+LPrAAAAAXNSR0IArs4c6QAAAE5QTFRFAAAAAAAAAAA6AABmADo6ADqQAGa2OgAAOgA6Oma2OpDbZgAAZjoAZrb/kDoAkNu2kNv/tmYAtv//25A62////7Zm/9uQ/9u2//+2///bGyFzpwAAAAF0Uk5TAEDm2GYAAAAJcEhZcwAAFiUAABYlAUlSJPAAAAAZdEVYdFNvZnR3YXJlAE1pY3Jvc29mdCBPZmZpY2V/7TVxAAAAeklEQVQoU+2RSRKAMAgEwS1qFMW4JP//qGgoiwd48CC3DlMDQwD++uYFAjZ2seixXs0DV87y3k7ecKIm+nJ59FsxW46+B8tBpImKWfW768EyX06JcMz9DbUyJ8qDWftBfRjFBWSO7qXMmusY7kAkXpdO8iF2NuEr33IC4FcE+HKECnkAAAAASUVORK5CYII=\" alt=\"image\"\u003e) without assuming any single modality as a gold standard\u003csup\u003e13,14,16\u003c/sup\u003e. In the baseline model, PET status, CSF amyloid status, and plasma stratum were assumed conditionally independent given \u003cimg width=\"10\" height=\"31\" src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAA8AAAAuCAMAAADA+LPrAAAAAXNSR0IArs4c6QAAAE5QTFRFAAAAAAAAAAA6AABmADo6ADqQAGa2OgAAOgA6Oma2OpDbZgAAZjoAZrb/kDoAkNu2kNv/tmYAtv//25A62////7Zm/9uQ/9u2//+2///bGyFzpwAAAAF0Uk5TAEDm2GYAAAAJcEhZcwAAFiUAABYlAUlSJPAAAAAZdEVYdFNvZnR3YXJlAE1pY3Jvc29mdCBPZmZpY2V/7TVxAAAAeklEQVQoU+2RSRKAMAgEwS1qFMW4JP//qGgoiwd48CC3DlMDQwD++uYFAjZ2seixXs0DV87y3k7ecKIm+nJ59FsxW46+B8tBpImKWfW768EyX06JcMz9DbUyJ8qDWftBfRjFBWSO7qXMmusY7kAkXpdO8iF2NuEr33IC4FcE+HKECnkAAAAASUVORK5CYII=\" alt=\"image\"\u003e. Because PET and CSF may retain residual correlation even after conditioning on amyloid status, we prespecified a sensitivity model that allowed conditional dependence between PET and CSF by modeling their joint distribution within each latent class; plasma remained modeled as conditionally independent given \u003cimg width=\"10\" height=\"31\" src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAA8AAAAuCAMAAADA+LPrAAAAAXNSR0IArs4c6QAAAE5QTFRFAAAAAAAAAAA6AABmADo6ADqQAGa2OgAAOgA6Oma2OpDbZgAAZjoAZrb/kDoAkNu2kNv/tmYAtv//25A62////7Zm/9uQ/9u2//+2///bGyFzpwAAAAF0Uk5TAEDm2GYAAAAJcEhZcwAAFiUAABYlAUlSJPAAAAAZdEVYdFNvZnR3YXJlAE1pY3Jvc29mdCBPZmZpY2V/7TVxAAAAeklEQVQoU+2RSRKAMAgEwS1qFMW4JP//qGgoiwd48CC3DlMDQwD++uYFAjZ2seixXs0DV87y3k7ecKIm+nJ59FsxW46+B8tBpImKWfW768EyX06JcMz9DbUyJ8qDWftBfRjFBWSO7qXMmusY7kAkXpdO8iF2NuEr33IC4FcE+HKECnkAAAAASUVORK5CYII=\" alt=\"image\"\u003e\u003csup\u003e14,15\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003ePriors were prespecified and weakly informative: Beta(1,1) priors for latent prevalence and binary test parameters; Dirichlet(1,1,1) priors for plasma-category probabilities within each latent class; and, in the PET\u0026ndash;CSF dependence sensitivity model, Dirichlet(1,1,1,1) priors for the within-class joint PET\u0026ndash;CSF probabilities. Posterior sampling used Gibbs sampling with 4 chains (2,000 burn-in iterations; 5,000 retained draws per chain) for the primary and dependence models; timing-stratified models used 2 chains (1,500 burn-in; 3,000 retained draws). Convergence was assessed using split \u003cimg width=\"14\" height=\"32\" src=\"data:image/png;base64,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\" alt=\"image\"\u003eand effective sample size diagnostics, and model fit was evaluated using posterior predictive checks comparing observed versus model-implied pattern frequencies and discordance rates\u003csup\u003e31\u0026ndash;33\u003c/sup\u003e. We report posterior means with 95% credible intervals. For observed discordance proportions, we report 95% Wilson confidence intervals.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cem\u003eData availability\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe data supporting this study are available from the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI) and can be accessed through the ADNI data portal (ida.loni.usc.edu) subject to ADNI data-use agreements. Because these data were used under license, they are not redistributed with this manuscript. All analytic code required to reproduce the results is publicly available (see \u003cem\u003eCode availability\u003c/em\u003e); qualified researchers can obtain the underlying ADNI datasets directly from ADNI following approval of a data access request. 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 https://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCode availability\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAll code used to construct triads, fit models, and generate figures is available at: https://github.com/SatwantKumar/amyloid-probability-adni. To support reproducibility, the repository includes the full analysis workflow and figure scripts.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAcknowledgements\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eData collection and sharing for this project was funded by the Alzheimer\u0026apos;s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 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\u0026apos;s Association; Alzheimer\u0026apos;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 (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer\u0026apos;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\n\u003cp\u003e\u003cem\u003eAuthor contributions\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eS.K. and K.V. conceptualized and designed the study. S.K. acquired the ADNI data, performed data curation, conducted the statistical analyses, and generated the figures and tables. K.V. contributed domain expertise for clinical interpretation of biomarker patterns and results. S.K. and K.V. drafted the manuscript. K.V. and S.K. critically revised the manuscript for important intellectual content. Both authors reviewed and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCompeting interests\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eHansson, O. \u003cem\u003eet al.\u003c/em\u003e The Alzheimer\u0026rsquo;s Association appropriate use recommendations for blood biomarkers in Alzheimer\u0026rsquo;s disease. \u003cem\u003eAlzheimers Dement\u003c/em\u003e 18, 2669\u0026ndash;2686 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePalmqvist, S. \u003cem\u003eet al.\u003c/em\u003e Alzheimer\u0026rsquo;s Association Clinical Practice Guideline on the use of blood-based biomarkers in the diagnostic workup of suspected Alzheimer\u0026rsquo;s disease within specialized care settings. \u003cem\u003eAlzheimer\u0026rsquo;s and Dementia\u003c/em\u003e 21, (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRabinovici, G. D. \u003cem\u003eet al.\u003c/em\u003e Updated appropriate use criteria for amyloid and tau PET: A report from the Alzheimer\u0026rsquo;s Association and Society for Nuclear Medicine and Molecular Imaging Workgroup. \u003cem\u003eAlzheimers Dement\u003c/em\u003e 21, e14338 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAngioni, D. \u003cem\u003eet al.\u003c/em\u003e Can We Use Blood Biomarkers as Entry Criteria and for Monitoring Drug Treatment Effects in Clinical Trials? A Report from the EU/US CTAD Task Force. \u003cem\u003eJournal of Prevention of Alzheimer\u0026rsquo;s Disease\u003c/em\u003e 10, 418\u0026ndash;425 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Wilde, A. \u003cem\u003eet al.\u003c/em\u003e Discordant amyloid-β PET and CSF biomarkers and its clinical consequences. \u003cem\u003eAlzheimer\u0026rsquo;s Research \u0026amp; Therapy\u003c/em\u003e 2019 \u003cem\u003e11:1\u003c/em\u003e 11, 78- (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSala, A., Nordberg, A. \u0026amp; Rodriguez-Vieitez, E. Longitudinal pathways of cerebrospinal fluid and positron emission tomography biomarkers of amyloid-β positivity. \u003cem\u003eMolecular Psychiatry 2020 26:10\u003c/em\u003e 26, 5864\u0026ndash;5874 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReimand, J., Collij, L., Scheltens, P., Bouwman, F. H. \u0026amp; Ossenkoppele, R. Amyloid-β CSF/PET discordance vs tau load 5 years later: It takes two to tangle. \u003cem\u003eAlzheimer\u0026rsquo;s \u0026amp; Dementia\u003c/em\u003e 16, e037246 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKnopman, D. S. \u003cem\u003eet al.\u003c/em\u003e Discrepancies between CSF biomarker and PET determinations of elevated brain amyloid and their prognostic significance. \u003cem\u003eAlzheimers Dement\u003c/em\u003e 21, (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeuzy, A. \u003cem\u003eet al.\u003c/em\u003e Considerations in the clinical use of amyloid PET and CSF biomarkers for Alzheimer\u0026rsquo;s disease. \u003cem\u003eAlzheimer\u0026rsquo;s \u0026amp; Dementia\u003c/em\u003e 21, e14528 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCerman, J. \u003cem\u003eet al.\u003c/em\u003e Concordance between amyloid PET and CSF biomarkers in clinical setting: a cross-platform comparison and in-depth analysis of discordant cases. \u003cem\u003eSci Rep\u003c/em\u003e 15, 39963 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarth\u0026eacute;lemy, N. R. \u003cem\u003eet al.\u003c/em\u003e Highly accurate blood test for Alzheimer\u0026rsquo;s disease is similar or superior to clinical cerebrospinal fluid tests. \u003cem\u003eNature Medicine 2024 30:4\u003c/em\u003e 30, 1085\u0026ndash;1095 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchindler, S. E. \u003cem\u003eet al.\u003c/em\u003e Acceptable performance of blood biomarker tests of amyloid pathology \u0026mdash; recommendations from the Global CEO Initiative on Alzheimer\u0026rsquo;s Disease. \u003cem\u003eNature Reviews Neurology 2024 20:7\u003c/em\u003e 20, 426\u0026ndash;439 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHui, S. L. \u0026amp; Walter, S. D. Estimating the error rates of diagnostic tests. \u003cem\u003eBiometrics\u003c/em\u003e 36, 167 (1980).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Smeden, M., Naaktgeboren, C. A., Reitsma, J. B., Moons, K. G. M. \u0026amp; De Groot, J. A. H. Latent class models in diagnostic studies when there is no reference standard\u0026ndash;a systematic review. \u003cem\u003eAm J Epidemiol\u003c/em\u003e 179, 423\u0026ndash;431 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDendukuri, N. \u0026amp; Joseph, L. Bayesian approaches to modeling the conditional dependence between multiple diagnostic tests. \u003cem\u003eBiometrics\u003c/em\u003e 57, 158\u0026ndash;167 (2001).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBranscum, A. J., Gardner, I. A. \u0026amp; Johnson, W. O. Estimation of diagnostic-test sensitivity and specificity through Bayesian modeling. \u003cem\u003ePrev Vet Med\u003c/em\u003e 68, 145\u0026ndash;163 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumurgier, J. \u003cem\u003eet al.\u003c/em\u003e A Pragmatic, Data-Driven Method to Determine Cutoffs for CSF Biomarkers of Alzheimer Disease Based on Validation Against PET Imaging. \u003cem\u003eNeurology\u003c/em\u003e 99, e669 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eADNI | Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://adni.loni.usc.edu/\u003c/span\u003e\u003cspan address=\"https://adni.loni.usc.edu/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDoecke, J. D. \u003cem\u003eet al.\u003c/em\u003e Elecsys CSF biomarker immunoassays demonstrate concordance with amyloid-PET imaging. \u003cem\u003eAlzheimer\u0026rsquo;s Research \u0026amp; Therapy\u003c/em\u003e 2020 \u003cem\u003e12:1\u003c/em\u003e 12, 36- (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeyer, M. R. \u003cem\u003eet al.\u003c/em\u003e Clinical validation of the PrecivityAD2 blood test: A mass spectrometry-based test with algorithm combining %p-tau217 and Aβ42/40 ratio to identify presence of brain amyloid. \u003cem\u003eAlzheimer\u0026rsquo;s and Dementia\u003c/em\u003e 20, 3179\u0026ndash;3192 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEastwood, S. M. \u003cem\u003eet al.\u003c/em\u003e PrecivityAD2TM Blood Test: Analytical Validation of an LC-MS/MS Assay for Quantifying Plasma Phospho-tau217 and Non-Phospho-tau217 Peptide Concentrations That Are Used with Plasma Amyloid-β42/40 in a Multianalyte Assay with Algorithmic Analysis for Detect\u0026hellip; \u003cem\u003eDiagnostics (Basel)\u003c/em\u003e 14, (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchiller, I. \u003cem\u003eet al.\u003c/em\u003e Bias due to composite reference standards in diagnostic accuracy studies. \u003cem\u003eStat Med\u003c/em\u003e 35, 1454\u0026ndash;1470 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKellerhuis, B. E. \u003cem\u003eet al.\u003c/em\u003e Drivers of bias in diagnostic test accuracy estimates when using expert panels as a reference standard: a simulation study. \u003cem\u003eBMC Med Res Methodol\u003c/em\u003e 25, (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJones, G., Johnson, W. O., Hanson, T. E. \u0026amp; Christensen, R. Identifiability of models for multiple diagnostic testing in the absence of a gold standard. \u003cem\u003eBiometrics\u003c/em\u003e 66, 855\u0026ndash;863 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMulherin, S. A. \u0026amp; Miller, W. C. Spectrum bias or spectrum effect? Subgroup variation in diagnostic test evaluation. \u003cem\u003eAnn Intern Med\u003c/em\u003e 137, 598\u0026ndash;602 (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVeitch, D. P. \u003cem\u003eet al.\u003c/em\u003e The Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative in the era of Alzheimer\u0026rsquo;s disease treatment: A review of ADNI studies from 2021 to 2022. \u003cem\u003eAlzheimer\u0026rsquo;s \u0026amp; Dementia\u003c/em\u003e 20, 652 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVeitch, D. P. \u003cem\u003eet al.\u003c/em\u003e Using the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative to improve early detection, diagnosis, and treatment of Alzheimer\u0026rsquo;s disease. \u003cem\u003eAlzheimers Dement\u003c/em\u003e 18, 824\u0026ndash;857 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAltman, D. G. \u0026amp; Royston, P. The cost of dichotomising continuous variables. \u003cem\u003eBMJ\u003c/em\u003e 332, 1080 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoyston, P., Altman, D. G. \u0026amp; Sauerbrei, W. Dichotomizing continuous predictors in multiple regression: a bad idea. \u003cem\u003eStat Med\u003c/em\u003e 25, 127\u0026ndash;141 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeiner, M. W. \u003cem\u003eet al.\u003c/em\u003e The Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative 3: Continued innovation for clinical trial improvement. \u003cem\u003eAlzheimer\u0026rsquo;s and Dementia\u003c/em\u003e 13, 561\u0026ndash;571 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGelman, A., Meng, X. L. \u0026amp; Stern, H. Posterior predictive assessment of model fitness via realized discrepancies. \u003cem\u003eStat Sin\u003c/em\u003e 6, 733\u0026ndash;807 (1996).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrooks, S. P. \u0026amp; Gelman, A. General methods for monitoring convergence of iterative simulations)? \u003cem\u003eJournal of Computational and Graphical Statistics\u003c/em\u003e 7, 434\u0026ndash;455 (1998).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVehtarh, A., Gelman, A., Simpson, D., Carpenter, B. \u0026amp; Burkner, P. C. Rank-Normalization, Folding, and Localization: An Improved Rˆ for Assessing Convergence of MCMC (with Discussion). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1214/20\u003c/span\u003e\u003cspan address=\"https://doi.org/10.1214/20\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cem\u003e-BA1221\u003c/em\u003e 16, 667\u0026ndash;718 (2021).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"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":"npj-dementia","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Dementia](https://www.nature.com/npjdementia/)","snPcode":"44400","submissionUrl":"https://submission.springernature.com/new-submission/44400/3","title":"npj Dementia","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8492792/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8492792/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eBlood-based biomarkers are increasingly used to triage patients for amyloid confirmation, yet performance is often reported versus a single comparator despite discordance between amyloid PET and CSF. In a cross-sectional secondary analysis of the Alzheimer\u0026rsquo;s Disease Neuroimaging Initiative (ADNI), we assembled one PET-anchored PET\u0026ndash;CSF\u0026ndash;plasma triad per participant (n\u0026thinsp;=\u0026thinsp;320). Bayesian latent class models integrating PET, CSF and plasma (Aβ42/40 or %p-tau217) estimated pattern-level posterior probabilities of latent amyloid positivity, with prespecified sensitivity analyses for PET\u0026ndash;CSF conditional dependence, timing gaps (\u0026le;\u0026thinsp;7, 8\u0026ndash;30, \u0026gt;\u0026thinsp;30 days) and CSF cutpoints. Concordant PET+/CSF\u0026thinsp;+\u0026thinsp;and PET\u0026minus;/CSF\u0026thinsp;\u0026minus;\u0026thinsp;patterns mapped to probabilities near 1 and 0, whereas discordant patterns yielded intermediate probabilities refined by plasma strata and most sensitive to dependence assumptions. PET\u0026ndash;CSF discordance occurred even within \u0026le;\u0026thinsp;7 days (12/98; 12%). A CSF Aβ42 coverage analysis showed similar gradients. Pattern-to-probability reporting may aid interpretation without privileging a reference test.\u003c/p\u003e","manuscriptTitle":"Amyloid Probability in Alzheimer Disease From Plasma, Cerebrospinal Fluid, and Amyloid Imaging","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-12 05:41:52","doi":"10.21203/rs.3.rs-8492792/v1","editorialEvents":[{"type":"communityComments","content":1},{"type":"decision","content":"Revision requested","date":"2026-04-22T03:44:16+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-08T21:22:20+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"292768916653149606969241271644509402622","date":"2026-02-24T03:56:26+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-20T17:48:06+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"302014783078156357581331559337769897382","date":"2026-02-18T18:26:19+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-12T18:48:07+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-06T18:35:08+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-03T14:43:34+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Dementia","date":"2026-01-01T02:15:25+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-dementia","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Dementia](https://www.nature.com/npjdementia/)","snPcode":"44400","submissionUrl":"https://submission.springernature.com/new-submission/44400/3","title":"npj Dementia","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"264520d7-61f9-4764-ade9-0b58b85d1d61","owner":[],"postedDate":"January 12th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":60695785,"name":"Health sciences/Biomarkers"},{"id":60695786,"name":"Health sciences/Medical research"},{"id":60695787,"name":"Health sciences/Neurology"},{"id":60695788,"name":"Biological sciences/Neuroscience"}],"tags":[],"updatedAt":"2026-04-26T22:53:18+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-12 05:41:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8492792","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8492792","identity":"rs-8492792","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}