Closing the gap between observational research and randomized trials for prevention of Alzheimer's Disease and dementia MPI: Glymour and Power in response to PAR-17-054.html Summary Launching randomized controlled trials (RCTs) for Alzheimer?s disease (AD) prevention is an urgent public health priority. Although cardiovascular risk factor management is among the most promising intervention strategies, there is considerable uncertainty about the optimal eligibility criteria, intervention details, duration, or outcome assessments. Many major trials targeting AD prevention have been disappointing. One possible reason for these disappointments is that observational research has not provided enough information to anticipate whether a proposed RCT would succeed. Observational studies rarely specify populations, exposures, and duration of follow-up with enough detail to guide RCT development. Most observational studies do not have enough information to provide detailed guidance for RCT development. Integration across heterogeneous observational data sources is necessary to achieve the sample size, diversity, and variety of measurements necessary to guide RCT development. In other research areas, simulations have proven useful tools to combine diverse sources of evidence, but in AD prevention, we currently lack tools to systematically combine evidence from heterogeneous data sources in order to guide trial design. This proposal takes advantage of recent advances in causal methods for data integration to overcome the previous barriers and develop a simulation model leveraging all of the information from diverse data sources, including cohorts, clinical administrative data, and registry information.
In AIM 1, we combine information from 8 observational studies, including cohorts, biobanks, and registries, into a unified, flexible, prevention simulation model. This model can simulate effects of hypothetical trials and thereby provide specific guidance for development of effective RCTs for AD prevention. We begin by estimating a structural model using data from the Cardiovascular Health Study (CHS, n=5,888) and the Atherosclerosis Risk in Communities (ARIC, n=15,792) cohorts, which include detailed exposure, outcome, and covariate measures. We will then incorporate data from 6 other sources, with information in total on 1.6 million individuals. We will use a latent variable approach to incorporate alternative measures of exposures, outcomes, and covariates.
In AIM 2, the prevention simulation model will be tested, refined, and validated by comparing simulated and actual findings of the ACCORD-MIND, ACCORDION-MIND, SYST-EUR, HYVET-COG, SCOPE, SHEP, and SPRINT-MIND trials.
AIM 3 will compare a range of hypothetical trials for diabetes and hypertension management to identify interventions most likely to succeed, considering eligibility criteria, intensity and duration of intervention, and outcome measures.
In AIM 4 we develop user-friendly interfaces for the model, allowing incorporation of new evidence from additional data sets, potentially addressing new risk factors, new outcomes, and evaluation of alternative proposed trial designs. The prevention simulation engine will identify which AD prevention RCTs are likely to succeed and thereby accelerate progress towards successful strategies to prevent AD.
Closing the gap between observational research and randomized trials for prevention of Alzheimer's Disease and dementia MPI: Glymour and Power in response to PAR-17-054.html Narrative This proposal will help improve the design of randomized clinical trials to prevent Alzheimer?s disease. We will combine data from 8 different US and international studies to make a better statistical model than would be possible with any single study. This model will help identify whom to enroll, how to intervene, and how long to continue the study to develop a successful trial that prevents Alzheimer?s. The statistical model will be evaluated and fine-tuned by comparing its predictions to outcomes from trials that have already been completed.
