Prosthetic valve thrombosis (PVT) is a serious complication affecting prosthetic heart valves. The transvalvular mean pressure gradient (MPG) derived by Doppler echocardiography is a crucial index to diagnose PVT but may result in false negatives mainly in case of bileaflet mechanical valves (BMVs) in mitral position. This may happen because MPG estimation relies on simplifying assumptions on the transvalvular fluid dynamics or because Doppler examination is manual and operator dependent. A deeper understanding of these issues may allow for improving PVT diagnosis and management. To this aim, we used in vitro and fluid–structure interaction (FSI) modeling to simulate the function of a real mitral BMV in different configurations: normally functioning and stenotic with symmetric and completely asymmetric leaflet opening, respectively. In each condition, the MPG was measured in vitro, computed directly from FSI simulations and derived from the corresponding velocity field through a Doppler-like postprocessing approach. Following verification versus in vitro data, MPG computational data were analyzed to test their dependency on the severity of fluid-dynamic derangements and on the measurement site. Computed MPG clearly discriminated between normally functioning and stenotic configurations. They did not depend markedly on the site of measurement, yet differences below 3 mmHg were found between MPG values at the central and lateral orifices of the BMV. This evidence suggests a mild uncertainty of the Doppler-based evaluation of the MPG due to probe positioning, which yet may lead to false negatives when analyzing subjects with almost normal MPG.