A predictive model for differentiating causes of elevated mechanical prosthetic aortic valve gradient.

In managing aortic prosthetic valves (APVs), increased transvalvular gradients present significant diagnostic challenges, often requiring advanced multimodality imaging (MMI). Predictive models can offer valuable insights in resource-limited settings. This study aimed to develop a predictive model to differentiate between causes of high mechanical prosthetic aortic valve gradients the causes of high gradients in mechanical prosthetic aortic valves. This retrospective study, conducted at a tertiary cardiology center, included 159 patients with high-gradient mechanical prosthetic aortic valves admitted between February 2020 and April 2024. Clinical evaluations involved included detailed examinations, laboratory findings, time in therapeutic range (TTR), and MMI techniques. Among the patients, 102 had patient-prosthesis mismatch (PPM), 22 had thrombus, and 35 had pannus-related high-gradients. A multivariate multinomial logistic regression model was used to predict the diagnostic groups. The most significant variables, according to partial chi-square values, were APV opening angle, acceleration time (AT), APV age (> 5 years), APV size, and effective TTR. APV opening angle and AT explained 65% of the outcome variation. The model's macro-average multi-class AUC was 0.95, with individual AUC values of 0.98, 0.94, and 0.94 for mismatch PPM, thrombus, and pannus, respectively. The polytomous discrimination index (PDI) was 0.87 overall. This study developed a predictive model to distinguish between PPM, thrombus, and pannus formation in high-gradient mechanical aortic valves. Valve opening angle and acceleration time were the most significant predictors. The model supports the use of simple tools like echocardiography and cinefluoroscopy, especially in settings without access to advanced imaging.