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

Area of Science:

• Cardiology
• Biomedical Engineering
• Medical Imaging

Background:

• Increased transvalvular gradients in aortic prosthetic valves (APVs) pose diagnostic challenges, often necessitating advanced multimodality imaging (MMI).
• Predictive models can provide crucial insights, particularly in resource-limited environments.
• Differentiating causes of high gradients is vital for effective APV management.

Purpose of the Study:

• To develop and validate a predictive model for differentiating causes of high gradients in mechanical APVs.
• To identify key clinical and echocardiographic parameters predictive of patient-prosthesis mismatch (PPM), thrombus, or pannus formation.
• To assess the model's diagnostic performance using established metrics.

Main Methods:

• Retrospective analysis of 159 patients with high-gradient mechanical APVs.
• Inclusion of clinical data, laboratory findings, time in therapeutic range (TTR), and MMI.
• Development of a multivariate multinomial logistic regression model to predict diagnostic groups (PPM, thrombus, pannus).

Main Results:

• The model identified APV opening angle, acceleration time (AT), APV age, APV size, and effective TTR as significant predictors.
• APV opening angle and AT were the most influential variables, explaining 65% of outcome variation.
• The model achieved a high macro-average multi-class AUC of 0.95, with individual AUCs ranging from 0.94 to 0.98.

Conclusions:

• A novel predictive model effectively distinguishes between PPM, thrombus, and pannus in mechanical APVs.
• Valve opening angle and acceleration time are critical predictors, supporting the use of accessible imaging modalities.
• This model offers a valuable tool for diagnosing high gradients in APVs, especially where advanced imaging is unavailable.