Early, accurate prediction of Alzheimer’s disease (AD) risk at the individual level remains a major barrier to both clinical monitoring and the design of AD therapeutic trials. While recent advances in blood-based biomarkers have enabled accurate prediction of amyloid and tau pathology in a patient’s brain, it is unclear whether even the most cutting-edge biomarkers can predict which patients will go on to develop cognitive symptoms and impairment. Additionally, many studies aimed at addressing this gap rely on individual cohorts, limiting our ability to generalize any findings to the broader clinical populations they will actually be used in. Dr. Buckley aims to address these gaps through this project by leveraging a wide range of cohorts to develop a clinically viable model for interpreting blood-based biomarkers in real-world contexts.
Overall, Dr. Buckley’s goal is to develop a clinically actionable model of absolute risk for cognitive impairment. Her approach aims to address a pair of issues that studies of this kind have previously encountered. Firstly, she will use clinical data from multiple cohorts to overcome the difficulties of generalizing results from a single cohort. While we typically think of a cohort’s heterogeneity based on patient demographics, there are other sources of variance as well. This can include a wide range of factors, from the test brand for the biomarker, to the time of day the sample is collected, to the accuracy of the patient’s other health information. This reality of our healthcare system needs to be accounted for in any models seeking to accurately predict risk on a broad scale. Secondly, she will seek to design models to predict risk at 2-, 5-, and 10-year intervals. Many studies focus on a single time point to predict risk, but this means the efficacy of their predictions is unknown outside that time frame. By looking at three different time frames, Dr. Buckley will be able to provide a clearer picture of the reliability of her model over time. As a result, this improves the quantity and quality of information a clinician can derive from the tests, relay to the patient, and use to develop a prevention and treatment strategy.
The project consists of two overarching aims. The first aim involves developing models based on the available information. Specifically, Dr. Buckley outlines four tiers of clinical information, with each also including the information from the tiers below it. Tier 1 will factor in the patient’s demographics, APOE genotype, and baseline cognition. The second tier adds blood-based biomarker information to Tier 1, and Tier 3 adds neuroimaging biomarkers, including amyloid PET and MRI. Finally, Tier 4 will account for proteomic and genomic features that, while incredibly informative, are not routinely collected for all patients. Each tiered model will then be used to estimate the risk of cognitive progression over 2, 5, and 10 years. Tiering the models enables Dr. Buckley to compare the incremental value of each additional data type, allowing her to identify the most impactful datasets for accurate risk prediction. The second aim of the project will then focus on translating the tiered models into a clinically deployable, web-based platform. This is an important step toward using these models in everyday clinical settings. She intends to test the platform using a series of simulated cases relevant to different scenarios, including patient stratification for an AD prevention trial. This validation will be crucial in demonstrating that the platform is effective both on a personal patient level and a broader trial design level.
The project aims to develop a comprehensive set of models to overcome previous challenges in the field and produce a clinically deployable model for accurately predicting worsening cognitive symptoms in patients, based on real-world clinical data and biomarker information.
