Faculty Bibliography
Searched for:
in-biosketch:true
person:masura01
Total Results:
83
Multi-ancestry genome-wide meta-analysis of 56,241 individuals identifies known and novel cross-population and ancestry-specific associations as novel risk loci for Alzheimer's disease
BACKGROUND:Limited ancestral diversity has impaired our ability to detect risk variants more prevalent in ancestry groups of predominantly non-European ancestral background in genome-wide association studies (GWAS). We construct and analyze a multi-ancestry GWAS dataset in the Alzheimer's Disease Genetics Consortium (ADGC) to test for novel shared and population-specific late-onset Alzheimer's disease (LOAD) susceptibility loci and evaluate underlying genetic architecture in 37,382 non-Hispanic White (NHW), 6728 African American, 8899 Hispanic (HIS), and 3232 East Asian individuals, performing within ancestry fixed-effects meta-analysis followed by a cross-ancestry random-effects meta-analysis. RESULTS:We identify 13 loci with cross-population associations including known loci at/near CR1, BIN1, TREM2, CD2AP, PTK2B, CLU, SHARPIN, MS4A6A, PICALM, ABCA7, APOE, and two novel loci not previously reported at 11p12 (LRRC4C) and 12q24.13 (LHX5-AS1). We additionally identify three population-specific loci with genome-wide significance at/near PTPRK and GRB14 in HIS and KIAA0825 in NHW. Pathway analysis implicates multiple amyloid regulation pathways and the classical complement pathway. Genes at/near our novel loci have known roles in neuronal development (LRRC4C, LHX5-AS1, and PTPRK) and insulin receptor activity regulation (GRB14). CONCLUSIONS:Using cross-population GWAS meta-analyses, we identify novel LOAD susceptibility loci in/near LRRC4C and LHX5-AS1, both with known roles in neuronal development, as well as several novel population-unique loci. Reflecting the power of diverse ancestry in GWAS, we detect the SHARPIN locus with only 13.7% of the sample size of the NHW GWAS study (n = 409,589) in which this locus was first observed. Continued expansion into larger multi-ancestry studies will provide even more power for further elucidating the genomics of late-onset Alzheimer's disease.
PMCID:12273372
PMID: 40676597
ISSN: 1474-760x
CID: 5897492
Tablet-Based Assessment of Picture Naming in Prodromal Alzheimer's Disease: An Accessible and Effective Tool for Distinguishing Mild Cognitive Impairment from Normal Aging
Effective mild cognitive impairment (MCI) screening requires accessible testing. This study compared two tests for distinguishing MCI patients from controls: Rapid Automatized Naming (RAN) for naming speed and Low Contrast Letter Acuity (LCLA) for sensitivity to low contrast letters. Two RAN tasks were used: the Mobile Universal Lexicon Evaluation System (MULES, picture naming) and Staggered Uneven Number test (SUN, number naming). Both RAN tasks were administered on a tablet and in a paper/pencil format. The tablet format was administered using the Mobile Integrated Cognitive Kit (MICK) application. LCLA was tested at 2.5% and 1.25% contrast. Sixty-four participants (31 MCI, 34 controls; mean age 73.2 ± 6.8 years) were included. MCI patients were slower than controls for paper/pencil (75.0 vs. 53.6 sec, p < 0.001), and tablet MULES (69.0 sec vs. 50.2 sec, p = 0.01). The paper/pencil SUN showed no significant difference (MCI: 59.5 sec vs. controls: 59.9 sec, p = 0.07), nor did tablet SUN (MCI: 59.3 sec vs. controls: 55.7 sec, p = 0.36). MCI patients had worse performance on LCLA testing at 2.5% contrast (33 letters vs. 36, p = 0.04*) and 1.25% (0 letters vs. 14. letters, p < 0.001). Receiver operating characteristic (ROC) analysis showed similar performance of paper/pencil and tablet MULES in distinguishing MCI from controls (AUC = 0.77), outperforming both SUN (AUC = 0.63 paper, 0.59 tablet) and LCLA (2.5% contrast: AUC = 0.65, 1.25% contrast: AUC = 0.72). The MULES, in both formats, may be a valuable screening tool for MCI.
PMID: 40499520
ISSN: 1421-9824
CID: 5868792
The association between measures of sleepiness and subjective cognitive decline symptoms in a diverse population of cognitively normal older adults
Subjective cognitive decline (SCD) is associated with preclinical Alzheimer's disease (AD). Suboptimal sleep is also a risk factor for cognitive decline, but with unclear relationship to SCD. We conducted a retrospective cross-sectional study in a biracial research cohort of 148 cognitively normal older adults who underwent quantification of SCD (Cognitive Change Index; CCI), sleepiness (Epworth Sleepiness Scale; ESS), depression (Geriatric Depression Scale; GDS), and amyloid/tau PET. ESS score was associated with total, amnestic, and non-amnestic CCI scores, after adjustment for GDS, amyloid/tau burden, and race. This supports future longitudinal work on how sleepiness impacts SCD outcomes.
PMID: 40170406
ISSN: 1875-8908
CID: 5819022
Sensitivity of unconstrained quantitative magnetization transfer MRI to amyloid burden in preclinical Alzheimer's disease
Magnetization transfer MRI is sensitive to semisolid macromolecules, includingamyloid beta, and has previously been used to discriminate Alzheimer'sdisease (AD) patients from controls. Here, we fit an unconstrained 2-poolquantitative MT (qMT) model, that is, without constraints on the longitudinalrelaxation rate
PMID: 40800521
ISSN: 2837-6056
CID: 5907312
INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE [Meeting Abstract]
ISI:001313316206082
ISSN: 0146-0404
CID: 5765622
Evaluating Large Language Models in extracting cognitive exam dates and scores
Ensuring reliability of Large Language Models (LLMs) in clinical tasks is crucial. Our study assesses two state-of-the-art LLMs (ChatGPT and LlaMA-2) for extracting clinical information, focusing on cognitive tests like MMSE and CDR. Our data consisted of 135,307 clinical notes (Jan 12th, 2010 to May 24th, 2023) mentioning MMSE, CDR, or MoCA. After applying inclusion criteria 34,465 notes remained, of which 765 underwent ChatGPT (GPT-4) and LlaMA-2, and 22 experts reviewed the responses. ChatGPT successfully extracted MMSE and CDR instances with dates from 742 notes. We used 20 notes for fine-tuning and training the reviewers. The remaining 722 were assigned to reviewers, with 309 each assigned to two reviewers simultaneously. Inter-rater-agreement (Fleiss' Kappa), precision, recall, true/false negative rates, and accuracy were calculated. Our study follows TRIPOD reporting guidelines for model validation. For MMSE information extraction, ChatGPT (vs. LlaMA-2) achieved accuracy of 83% (vs. 66.4%), sensitivity of 89.7% (vs. 69.9%), true-negative rates of 96% (vs 60.0%), and precision of 82.7% (vs 62.2%). For CDR the results were lower overall, with accuracy of 87.1% (vs. 74.5%), sensitivity of 84.3% (vs. 39.7%), true-negative rates of 99.8% (98.4%), and precision of 48.3% (vs. 16.1%). We qualitatively evaluated the MMSE errors of ChatGPT and LlaMA-2 on double-reviewed notes. LlaMA-2 errors included 27 cases of total hallucination, 19 cases of reporting other scores instead of MMSE, 25 missed scores, and 23 cases of reporting only the wrong date. In comparison, ChatGPT's errors included only 3 cases of total hallucination, 17 cases of wrong test reported instead of MMSE, and 19 cases of reporting a wrong date. In this diagnostic/prognostic study of ChatGPT and LlaMA-2 for extracting cognitive exam dates and scores from clinical notes, ChatGPT exhibited high accuracy, with better performance compared to LlaMA-2. The use of LLMs could benefit dementia research and clinical care, by identifying eligible patients for treatments initialization or clinical trial enrollments. Rigorous evaluation of LLMs is crucial to understanding their capabilities and limitations.
PMCID:11634005
PMID: 39661652
ISSN: 2767-3170
CID: 5762692
The relationship between anxiety and levels of Alzheimer's disease plasma biomarkers
Anxiety is highly prevalent in Alzheimer's disease (AD), correlating with cerebrospinal fluid/positron emission tomography biomarkers and disease progression. Relationships to plasma biomarkers are unclear. Herein, we compare levels of plasma biomarkers in research participants with and without anxiety at cognitively normal, mild cognitive impairment, and AD dementia stages. We observed significantly higher plasma tau/amyloid-β42 ratio in AD participants with anxiety versus those without, but did not observe differences at other stages or plasma biomarkers. No such relationships were evident with depression. These results support a unique pathophysiological relationship between anxiety and AD that can be reflected in plasma biomarkers, suggestive of heightened neurodegeneration.
PMID: 39604275
ISSN: 1875-8908
CID: 5759182
-Penalized Multinomial Regression: Estimation, Inference, and Prediction, With an Application to Risk Factor Identification for Different Dementia Subtypes
High-dimensional multinomial regression models are very useful in practice but have received less research attention than logistic regression models, especially from the perspective of statistical inference. In this work, we analyze the estimation and prediction error of the contrast-based
PMID: 39532663
ISSN: 1097-0258
CID: 5751462
Evaluating Large Language Models in Extracting Cognitive Exam Dates and Scores
IMPORTANCE/UNASSIGNED:Large language models (LLMs) are crucial for medical tasks. Ensuring their reliability is vital to avoid false results. Our study assesses two state-of-the-art LLMs (ChatGPT and LlaMA-2) for extracting clinical information, focusing on cognitive tests like MMSE and CDR. OBJECTIVE/UNASSIGNED:Evaluate ChatGPT and LlaMA-2 performance in extracting MMSE and CDR scores, including their associated dates. METHODS/UNASSIGNED:Our data consisted of 135,307 clinical notes (Jan 12th, 2010 to May 24th, 2023) mentioning MMSE, CDR, or MoCA. After applying inclusion criteria 34,465 notes remained, of which 765 underwent ChatGPT (GPT-4) and LlaMA-2, and 22 experts reviewed the responses. ChatGPT successfully extracted MMSE and CDR instances with dates from 742 notes. We used 20 notes for fine-tuning and training the reviewers. The remaining 722 were assigned to reviewers, with 309 each assigned to two reviewers simultaneously. Inter-rater-agreement (Fleiss' Kappa), precision, recall, true/false negative rates, and accuracy were calculated. Our study follows TRIPOD reporting guidelines for model validation. RESULTS/UNASSIGNED:For MMSE information extraction, ChatGPT (vs. LlaMA-2) achieved accuracy of 83% (vs. 66.4%), sensitivity of 89.7% (vs. 69.9%), true-negative rates of 96% (vs 60.0%), and precision of 82.7% (vs 62.2%). For CDR the results were lower overall, with accuracy of 87.1% (vs. 74.5%), sensitivity of 84.3% (vs. 39.7%), true-negative rates of 99.8% (98.4%), and precision of 48.3% (vs. 16.1%). We qualitatively evaluated the MMSE errors of ChatGPT and LlaMA-2 on double-reviewed notes. LlaMA-2 errors included 27 cases of total hallucination, 19 cases of reporting other scores instead of MMSE, 25 missed scores, and 23 cases of reporting only the wrong date. In comparison, ChatGPT's errors included only 3 cases of total hallucination, 17 cases of wrong test reported instead of MMSE, and 19 cases of reporting a wrong date. CONCLUSIONS/UNASSIGNED:In this diagnostic/prognostic study of ChatGPT and LlaMA-2 for extracting cognitive exam dates and scores from clinical notes, ChatGPT exhibited high accuracy, with better performance compared to LlaMA-2. The use of LLMs could benefit dementia research and clinical care, by identifying eligible patients for treatments initialization or clinical trial enrollments. Rigorous evaluation of LLMs is crucial to understanding their capabilities and limitations.
PMCID:10888985
PMID: 38405784
CID: 5722422
Optimal cutoff scores of the Montreal Cognitive Assessment to detect mild cognitive impairment and dementia in Costa Rican older adults
BACKGROUND/UNASSIGNED:The burden of Alzheimer's disease and related dementias (AD/ADRD) in Costa Rica is expected to become one of the highest in the region. Early detection will help optimize resources and improve primary care interventions. The Montreal Cognitive Assessment (MoCA) has shown good sensitivity for detecting mild cognitive impairment (MCI), but specificity varies depending on the population. This motivated the analysis of different cutoffs to minimize false-positive classifications in a Costa Rican sample for its use in clinical settings. METHODS/UNASSIGNED:Data was analyzed from 516 memory clinic outpatients (148 cognitively normal, 260 MCI, 108 mild AD/ADRD; mean age 66.3 ± 10.8 years) who underwent complete neurological and neuropsychological assessment and were diagnosed by consensus. Optimal MoCA cutoff scores were identified using a multiple cutoff approach. RESULTS/UNASSIGNED:Overall, a cutoff score of ≥ 23 showed better accuracy to distinguish between normal cognition (NC) and MCI (sensitivity 73%, specificity 83%). When analyzed by educational levels, a cutoff score of ≥ 21 showed better accuracy for ≤ 6 years (sensitivity 80%, specificity 76%), ≥23 for 7-12 years (sensitivity 86%, specificity 76%) and ≥ 24 for > 12 years (sensitivity 70%, specificity 85%). For distinguishing MCI from mild AD/ADRD, the optimal overall cutoff score was ≥ 15 (sensitivity 66%, specificity 85%). When stratified by years of education, cutoff scores of ≥ 14 showed better accuracy for ≤ 6 years (sensitivity 70%, specificity 88%), ≥15 for 7-12 years (sensitivity 46%, specificity 95%) and ≥ 17 for > 12 years (sensitivity 67%, specificity 93%). CONCLUSIONS/UNASSIGNED:A MoCA cutoff score of ≥ 23 in the Costa Rican population showed better diagnostic accuracy for detecting MCI and may reduce the false positive rate. Our findings may be helpful for primary care clinical settings and further referral criteria.
PMID: 39387855
ISSN: 1744-411x
CID: 5718242
