Alzheimer’s disease (AD) is the leading cause of dementia in older adults. However, the majority of clinical trials aiming to modify the disease process have failed over the last two decades. This is due, in part, to variation among people with AD in both their clinical features and biological underpinnings. The benefits of treatment may differ with the stage of illness. Some people with AD decline rapidly, while others decline more slowly. Some people have concomitant vascular disease, which may influence cognitive trajectories in the absence of treatment and response to treatment. The ideal participants for AD clinical trials would show cognitive decline in the absence of treatment (i.e., placebo arm) and also would respond to the therapeutic intervention. Identifying such participants for AD trials has proven to be challenging. Our recent studies indicate that by using data collected from patients at the screening visit and machine learning predictive models, we can effectively predict disease progression in the trial population. These models could be used to improve patient selection and enrich AD trials. We are planning to further validate and replicate these predictive models in other trials, and ultimately use them to improve the design of future trials. Future research should be conducted using multimodal data (i.e., clinical tests, MRIs, PET scans, blood-based biomarkers) from new clinical trials, which have collected comprehensive biomarker data, to explore the validity and generalizability of these models. In addition, predictors of treatment response from trials could be used to optimize patient selection in practice.
Despite the high burden of Alzheimer’s disease and increased efforts in research, the success rate of the pharmaceutical randomized clinical trials (RCTs) for dementia drugs has been abysmally low in the last two decades. This is thought to be partly due to clinical heterogeneity and difficulty in predicting which person will have cognitive decline during the one to five years of follow-up in a clinical trial. One strategy to improve design of clinical trials and boost their power is using predictive models that effectively can estimate probability of disease progression and cognitive decline. In recent years, innovative machine learning techniques have been used increasingly in pharmaceutical research and development for prediction of clinical outcomes and response to treatments in various fields of medicine. However, to date, such techniques have not been used in design or conduct of AD trials. This project aims to develop a machine learning platform that can be practically used in design and conduct of future clinical trials. We will use advanced machine learning and deep learning methods and data from previously completed RCTs for treatment of mild-to-moderate AD to develop precise models that can predict disease progression and rate of cognitive decline.