Leveraging 30 years of Alzheimer's disease clinical trials data to decipher phenotypic response to treatment

PROJECT SUMMARY Looking back to move forward: The overall goal of this proposal is to examine the role of specific phenotypes, or subgroups of people based on a given characteristic, such as sex/gender, APOE4 status, baseline AD biomarker levels, or polygenic hazard scores (PHS), on response to drug or

lifestyle interventions in existing Alzheimer’s disease clinical trials datasets. The precision medicine approach of understanding the role of different phenotypes such as men vs. women, APOE4 carrier vs. non- carrier, and AD polygenic hazard score in understanding Alzheimer’s risk and pathology has gained traction.

There is clearer recognition of sex/gender and APOE genotype differences in tau accumulation, cognitive decline, and most recently in anti-amyloid immunotherapies. Findings from our studies began to hint at critical sex/gender and genetic differences that reside in Alzheimer’s disease risk. Additionally, our team has

developed polygenic hazard scores using factors relevant to AD risk, such as sex/gender, tau PET burden, and multiple risk genotypes. With tremendous strides in advanced technology and methods for genetic analyses, plasma- based biomarkers, and harmonized brain imaging, in this proposal, we will examine the role of key

factors in clinical trials outcomes by leveraging existing Alzheimer’s clinical trials datasets from the last 30-years. These clinical trials largely followed a harmonized set of outcomes focused on cognition and function, with a large subset containing neuroimaging and fluid biomarkers, both in derived and raw form. Data

from 14,602 participants who were randomized to treatment or placebo will be included in this proposal. All datasets are de-identified and were collected within the guidelines of their respective institutional regulations. This sample will be augmented as more data become available from newly completed trials over the project

period. Statistical models will include interaction terms for phenotype-by-arm on ADAS-Cog, PACC (Preclinical Alzheimer’s Cognitive Composite), CDR-SOB (Clinical Dementia Rating Scale Sum-of-Boxes), ADCS-ADL (Activities of Daily Living), in addition to brain MRI, PET, and plasma/CSF levels. Datasets will be analyzed

individually, and collectively in meta-analyses by grouping datasets according to primary mechanism of action. All derived data from biofluids, genetics, and harmonized data will be openly shared with the scientific community as a new resource.