The brachial artery flow-mediated dilation (FMD) test is the most widely utilized method to evaluate endothelial function noninvasively in humans by calculating the percent change in diameter (FMD%). However, the underutilized velocity and diameter time course data, coupled with confounding influences in shear exposure, noise, and upward bias, make the FMD test less desirable. In this study, we developed an exposure-response, model-based approach that not only quantifies FMD based on the rich velocity and diameter data, it overcomes previously acknowledged challenges. FMD data were obtained from 15 apparently healthy participants, each exposed to four different cuff occlusion durations. The velocity response following cuff release was described by an exponential model with two parameters defining peak velocity and rate of decay. Shear exposure derived from velocity was used to drive the diameter response model, which consists of additive constriction and dilation terms. Three parameters describing distinct aspects of the vascular response to shear (magnitude of the initial constriction response, and magnitude and time constant of the dilation response) were estimated for both the individuals and population. These parameters are independent of shear exposure. Thus this approach produces identifiable and physiologically meaningful parameters that may provide additional information for comparing differences between experimental groups or over time, and provides a means to completely account for shear exposure. NEW & NOTEWORTHY While flow-mediated dilation (FMD) is a valuable tool for evaluating endothelial function, analytical challenges include confounding influences of shear exposure, upward bias, and underutilization of rich time course data collected during FMD testing. We have developed an exposure-response, model-based approach that quantifies endothelial function based on the velocity and diameter data and fully accounts for shear exposure. It produces physiologically meaningful parameters that may provide useful information for comparing differences between experimental groups or over time.
Bayesian model-based tight clustering for time course data