APPLYING POLYGENIC HAZARD SCORES TO ENRICH CLINICAL TRIAL STUDY POPULATIONS OF THOSE AT RISK OF ALZHEIMER’S DISEASE DEMENTIA (ADD)

Abstract

Clinical trials to prevent dementia in those at increased risk could benefit from improved risk factor stratification with identification of those who can be predicted to experience cognitive and functional decline. The polygenic hazard score (PHS) is known to predict who will develop ADD at a given age, above and beyond APOE genotype alone. Here we model how PHS quartile scores predict decline on the clinical dementia rating scale sum of boxes (CDR-SOB), in those who were cognitively normal (CN) or who had mild cognitive impairment (MCI), and how this would impact required sample sizes for a clinical trial. 509 subjects with a mean age of 75 years were accessed from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Participants (CN (n=189) or MCI (n=320) at baseline) were classified by quartile of PHS based on cumulative incidence rate. Within subject change from baseline on the CDR SOB over 12, 24, and 36 months was assessed using t-statistics comparing the two lower quartiles with the two higher quartiles. MCI subjects in the two highest risk quartiles of PHS show a significantly greater rate of decline starting at 24 months (p<0.05) on the CDR sum-of-the-boxes (CDR-SOB) for MCI subjects. By contrast, there were no significant differences on this by median split in those with normal cognition at baseline. Furthermore, we found using this inclusion criteria could reduce sample sizes required to show a 25% effect size at ≥ 80% power from 485 participants with MCI per arm, to 373 per arm. Enriching clinical trial recruitment by limiting to participants above the median level of risk for incident dementia, according to the PHS, could represent a cost and resource-efficient way to reduce the number of participants required, and could increase signal-to-noise ratio. Replication of these findings with measures more sensitive to preclinical AD may be of further utility.