Joint Longitudinal Hazard Modeling Distinguishes Polygenic Effects on Dementia via Premorbid Cognitive Function and Aging-Related Decline

Alzheimers disease and related dementias (AD/ADRD) are clinical disorders characterized by declines in cognitive function to levels at which everyday living is meaningfully impaired. Genetic variants may confer risk for AD/ADRD by way of steeper rates of aging-related cognitive declines or their association with lower premorbid levels of cognitive function that place individuals closer to a clinically severe threshold of cognitive function. We applied joint longitudinal-hazard models to 26-year data from an ancestrally diverse US sample of N = 11,518 adults to estimate effects of multi-ancestry polygenic indices (PGIs) for a range of established risk factors on dementia hazard and the degree they operate via premorbid cognitive function and longitudinal cognitive decline. APOE e4 and PGIs for ADRD, type 2 diabetes, stroke, LDL cholesterol, and hearing loss operated via rate of change, whereas PGIs for educational attainment, cognitive function, physical activity, BMI, smoking, loneliness, schizophrenia, and externalizing were associated with dementia hazard via premorbid differences in cognitive function. Some results varied by continental ancestry. Our implementation of joint longitudinal-hazard models allowed us to distinguish how different genetic risk factors for AD/ADRD operate via degenerative processes that might be therapeutically targeted during later life and via premorbid levels of cognitive function that would require intervention earlier in the life course.