The Diagnostic Efficacy of Left Ventricle Hemodynamic Parameters: Classifying Normal Mitral Valves Against Prosthetic Disease Proxies

Abstract

The dynamics of left ventricle diastolic filling and chamber stiffness are such critical processes in cardiac performance it has been suggested that the parameters of these processes could have important diagnostic power. This study has two objectives: (1) to estimate LV parameters from animal studies and test these parameters for diagnostic efficacy and (2) use Monte Carlo simulation techniques to determine if estimated LV parameters can be used to statistically classify subjects by disease state. The disease state we test are swine fitted with prosthetic mitral valves. Literature data on six swine instrumented for measuring left atrial and ventricle pressures and left ventricle volume before and after mitral valve replacement provide a unique data set for testing the diagnostic efficacy of estimated LV parameters since the prosthetic data can be considered a proxy for a diseased state, for example, mitral valve insufficiency. LV filling models and chamber stiffness models are estimated and their parameters subjected to rigorous statistical tests to elicit their ability to discriminate between normals and animals with prosthetic valve replacement. Additionally, the filling and chamber stiffness models are used to simulate subject-based Monte Carlo data to test empirical distributions for clinical applications of the statistical hypothesis. The E-wave normal and prosthetic filling models each have three parameters, similar structures, and both fit their respective data sets with remarkable accuracy and yet their parameters can be used to readily distinguish between a cohort of normal and a cohort of animals with prosthetic valve replacement. The chamber stiffness models and the E-wave filling models are sufficiently different between normal and animals with prosthetic valve replacement that it is possible to project their parameter vectors onto a scalar axis and develop individual subject z-scores for clinical applications. While it is preferable to compare normal animal data against chronically implanted animals, we argue from a statistical point of view that the filling and stiffness models estimated are sufficiently robust so as to provide evidence that their parameters do indeed have diagnostic power. Further, even in the presence of simulated data with added noise it is possible to make informed decisions as to subject disease classification or disease grade level. While it is recognized a prosthetic valve is an exaggerated disease state it is also true the statistical differences between normal and prosthetic animals were found to be exaggerated.