Covariate-specific Breakpoints in Personalized Evaluation of Left Ventricular Enlargement

Abstract

In this thesis, we outline a flexible and interpretable iterative method to fit generalized segmented linear models with covariate-specific breakpoints and evaluate its statistical performance by conducting simulation studies under several data-generating settings. We then apply our methodology to data from the Multi-Ethnic Study of Atherosclerosis (MESA) Study and investigate breakpoints in the association between cardiovascular outcome measures (ASCVD score and 10-year all-cause mortality) and left ventricular mass by gender and BMI. Our results suggest that the estimator is root n consistent. As expected, we also found that there is a positive correlation between BMI and breakpoints of left ventricular mass, after which the risk of mortality considerably increases. In addition, for a given BMI, the breakpoints appear to differ by sex, with males having much higher values. The method allows estimation of an LV enlargement threshold that differs by body size and gender.