No abstract

IMPORTANCE Amyloid positron emission tomography (PET) detects amyloid plaques in the brain, a core neuropathological feature of Alzheimer disease. OBJECTIVE To determine if amyloid PET is associated with subsequent changes in the management of patients with mild cognitive impairment (MCI) or dementia of uncertain etiology. DESIGN, SETTING, AND PARTICIPANTS The Imaging Dementia-Evidence for Amyloid Scanning (IDEAS) study was a single-group, multisite longitudinal study that assessed the association between amyloid PET and subsequent changes in clinical management for Medicare beneficiaries with MCI or dementia. Participants were required to meet published appropriate use criteria stating that etiology of cognitive impairment was unknown, Alzheimer disease was a diagnostic consideration, and knowledge of PET results was expected to change diagnosis and management. A total of 946 dementia specialists at 595 US sites enrolled 16 008 patients between February 2016 and September 2017. Patients were followed up through January 2018. Dementia specialists documented their diagnosis and management plan before PET and again 90 (±30) days after PET. EXPOSURES Participants underwent amyloid PET at 343 imaging centers. MAIN OUTCOMES AND MEASURES The primary end point was change in management between the pre-and post-PET visits, as assessed by a composite outcome that included Alzheimer disease drug therapy, other drug therapy, and counseling about safety and future planning. The study was powered to detect a 30% or greater change in the MCI and dementia groups. One of 2 secondary end points is reported: the proportion of changes in diagnosis (from Alzheimer disease to non-Alzheimer disease and vice versa) between pre-and post-PET visits. RESULTS Among 16 008 registered participants, 11 409 (71.3%) completed study procedures and were included in the analysis (median age, 75 years [interquartile range, 71-80]; 50.9% women; 60.5% with MCI). Amyloid PET results were positive in 3817 patients with MCI (55.3%) and 3154 patients with dementia (70.1%). The composite end point changed in 4159 of 6905 patients with MCI (60.2% [95% CI, 59.1%-61.4%]) and 2859 of 4504 patients with dementia (63.5% [95% CI, 62.1%-64.9%]), significantly exceeding the 30% threshold in each group (P < .001, 1-sided). The etiologic diagnosis changed from Alzheimer disease to non-Alzheimer disease in 2860 of 11 409 patients (25.1% [95% CI, 24.3%-25.9%]) and from non-Alzheimer disease to Alzheimer disease in 1201 of 11 409 (10.5% [95% CI, 10.0%-11.1%]). CONCLUSIONS AND RELEVANCE Among Medicare beneficiaries with MCI or dementia of uncertain etiology evaluated by dementia specialists, the use of amyloid PET was associated with changes in clinical management within 90 days. Further research is needed to determine whether amyloid PET is associated with improved clinical outcomes.

IMPORTANCEThe diagnostic evaluation for Alzheimer disease may be improved by a blood-based diagnostic test identifying presence of brain amyloid plaque pathology. OBJECTIVE To determine the clinical performance associated with a diagnostic algorithm incorporating plasma amyloid-β (Aβ) 42:40 ratio, patient age, and apoE proteotype to identify brain amyloid status. DESIGN, SETTING, AND PARTICIPANTS This cohort study includes analysis from 2 independent cross-sectional cohort studies: the discovery cohort of the Plasma Test for Amyloidosis Risk Screening (PARIS) study, a prospective add-on to the Imaging Dementia-Evidence for Amyloid Scanning study, including 249 patients from 2018 to 2019, and MissionAD, a dataset of 437 biobanked patient samples obtained at screenings during 2016 to 2019. Data were analyzed from May to November 2020. EXPOSURES Amyloid detected in blood and by positron emission tomography (PET) imaging. MAIN OUTCOMES AND MEASURESThe main outcome was the diagnostic performance of plasma Aβ42:40 ratio, together with apoE proteotype and age, for identifying amyloid PET status, assessed by accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). RESULTSAll 686 participants (mean [SD] age 73.2 [6.3] years; 368 [53.6%] men; 378 participants [55.1%] with amyloid PET findings) had symptoms of mild cognitive impairment or mild dementia. The AUC of plasma Aβ42:40 ratio for PARIS was 0.79 (95% CI, 0.73-0.85) and 0.86 (95% CI, 0.82-0.89) for MissionAD. Ratio cutoffs for Aβ42:40 based on the Youden index were similar between cohorts (PARIS: 0.089; MissionAD: 0.092). A logistic regression model (LRM) incorporating Aβ42:40 ratio, apoE proteotype, and age improved diagnostic performance within each cohort (PARIS: AUC, 0.86 [95% CI, 0.81-0.91]; MissionAD: AUC, 0.89 [95% CI, 0.86-0.92]), and overall accuracy was 78% (95% CI, 72%-83%) for PARIS and 83% (95% CI, 79%-86%) for MissionAD. The model developed on the prospectively collected samples from PARIS performed well on the MissionAD samples (AUC, 0.88 [95% CI, 0.84-0.91]; accuracy, 78% [95% CI, 74%-82%]). Training the LRM on combined cohorts yielded an AUC of 0.88 (95% CI, 0.85-0.91) and accuracy of 81% (95% CI, 78%-84%). The output of this LRM is the Amyloid Probability Score (APS). For clinical use, 2 APS cutoff values were established yielding 3 categories, with low, intermediate, and high likelihood of brain amyloid plaque pathology. (continued) Key Points Question Is an amyloid probability score based on a mass spectrometrybased blood test measuring plasma amyloid β 42:40 ratio and apoE proteotype plus age, associated with identifying brain amyloidosis among patients with cognitive impairment? Findings In this cohort study of 686 participants from 2 separate studies, the developed Amyloid Probability Score showed high concordance with amyloid PET status, with an area under the curve of 0.88 and overall accuracy of 81%. The test's findings were significantly associated with the presence or absence of brain amyloidosi...

ImportanceRacial and ethnic groups with higher rates of clinical Alzheimer disease (AD) are underrepresented in studies of AD biomarkers, including amyloid positron emission tomography (PET).ObjectiveTo compare amyloid PET positivity among a diverse cohort of individuals with mild cognitive impairment (MCI) or dementia.Design, Setting, and ParticipantsSecondary analysis of the Imaging Dementia–Evidence for Amyloid Scanning (IDEAS), a single-arm multisite cohort study of Medicare beneficiaries who met appropriate-use criteria for amyloid PET imaging between February 2016 and September 2017 with follow-up through January 2018. Data were analyzed between April 2020 and January 2022. This study used 2 approaches: the McNemar test to compare amyloid PET positivity proportions between matched racial and ethnic groups and multivariable logistic regression to assess the odds of having a positive amyloid PET scan. IDEAS enrolled participants at 595 US dementia specialist practices. A total of 21 949 were enrolled and 4842 (22%) were excluded from the present analysis due to protocol violations, not receiving an amyloid PET scan, not having a positive or negative scan, or because of small numbers in some subgroups.ExposuresIn the IDEAS study, participants underwent a single amyloid PET scan.Main Outcomes and MeasuresThe main outcomes were amyloid PET positivity proportions and odds.ResultsData from 17 107 individuals (321 Asian, 635 Black, 829 Hispanic, and 15 322 White) with MCI or dementia and amyloid PET were analyzed between April 2020 and January 2022. The median (range) age of participants was 75 (65-105) years; 8769 participants (51.3%) were female and 8338 (48.7%) were male. In the optimal 1:1 matching analysis (n = 3154), White participants had a greater proportion of positive amyloid PET scans compared with Asian participants (181 of 313; 57.8%; 95% CI, 52.3-63.2 vs 142 of 313; 45.4%; 95% CI, 39.9-50.9, respectively; P = .001) and Hispanic participants (482 of 780; 61.8%; 95% CI, 58.3-65.1 vs 425 of 780; 54.5%; 95% CI, 51.0-58.0, respectively; P = .003) but not Black participants (359 of 615; 58.4%; 95% CI, 54.4-62.2 vs 333 of 615; 54.1%; 95% CI, 50.2-58.0, respectively; P = .13). In the adjusted model, the odds of having a positive amyloid PET scan were lower for Asian participants (odds ratio [OR], 0.47; 95% CI, 0.37-0.59; P &amp;lt; .001), Black participants (OR, 0.71; 95% CI, 0.60-0.84; P &amp;lt; .001), and Hispanic participants (OR, 0.68; 95% CI, 0.59-0.79; P &amp;lt; .001) compared with White participants.Conclusions and RelevanceRacial and ethnic differences found in amyloid PET positivity among individuals with MCI and dementia in this study may indicate differences in underlying etiology of cognitive impairment and guide future treatment and prevention approaches.

Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.