There is no cure for atrial fibrillation (AF), thus prevention of AF and risk communication are key. In risk prediction models, associations between risk factors and AF are commonly expressed as hazard ratios. However, the hazard ratio is challenging to interpret. A novel metric, the difference in restricted mean survival time (RMST), offers a clinically meaningful interpretation and is advantageous for risk communication. The difference in RMSTs between two exposure groups is the mean time without AF lost due to the exposure. In contrast to the hazard ratio, the difference in RMST between risk groups provides an absolute measure of the association between a risk factor and AF. Improved risk communication by reporting the RMST will have a direct impact on cardiovascular public health. The RMST remains underreported in observational studies despite its appealing interpretation. One reason is there are gaps in RMST methods for risk prediction models and complex data scenarios, which are common in cardiovascular research. There is a need to develop new RMST methods with greater flexibility to address statistical challenges in cardiovascular research. We propose to address gaps in RMST methodology for observational studies. Our overall objective is to improve statistical methods for estimating the RMST and improve our understanding of AF epidemiology with these new methods.
Aim 1 is to develop new statistical metrics and data visualizations for the internal and external validation of AF risk prediction models.
Aim 2 is to develop RMST methods that accommodate time- varying risk factors, such as body mass index.
Aim 3 is to develop RMST methods for the competing risk of death. We will assess the performance of our new statistical methods using simulation studies, and illustrate our methods using AF data from the Framingham Heart Study (FHS) and the Atherosclerosis Risk in Communities Study (ARIC). Additionally, we will make our novel methods available to the greater research community by producing R packages. We focus on AF, but our methods can be used for a wide range of diseases. Advancing RMST methods will allow researchers to report the RMST more frequently when communicating AF risk. My mentoring team has outstanding experience in epidemiological research of AF and statistical methods for survival data, and is committed to supporting me in my training and professional development. We have designed a training plan which includes coursework in the prevention strategies, physiology, molecular mechanisms, and epidemiology of cardiovascular disease, and workshops in advanced methods for lifetime data and grant-writing. Through this fellowship, I will develop the skills to achieve my long- term goal of becoming an independent researcher with expertise in cardiovascular disease. After this fellowship, I plan continue advancing risk communication by obtaining a postdoctoral position and applying for a K01 grant to develop RMST methods for individual participant data meta-analysis and combined survival curves.
Developing new statistical methods for the restricted mean survival time will improve risk communication in studies with lifetime data. Our motivating example in cardiovascular research will shed new light on the magnitude of associations between risk factors and atrial fibrillation, and will aid lifestyle change counseling.
