Development of a new model based on functional and echocardiographic parameters to predict outcomes after mitral valve surgery for degenerative mitral regurgitation

Background: Surgical correction of severe DMR is indicated only in presence of symptoms and LV function impairment signs as left ventricular ejection fraction (LVEF) and LV end diastolic diameter (LVEDD). However, these parameters may remain normal also in presence of myocardial dysfunction. A multimodal approach, including cardio-pulmonary exercise test (CPET) and exercise stress echocardiography (ESE) may better assess real LV myocardial function. The aim of this study is to develop an artificial intelligence (AI)-based model to predict outcomes after DMR surgery. Methods: We included patients with severe DMR who underwent cardiac surgery at our Heart Surgery unit between 2020 and 2024. All patients underwent a CPET and ESE before surgery. The study outcomes at 6 months were: improvement of NYHA class and recovery of LVEF. A survival random forest model was used to model the outcome probabilities of patients. Multiple imputation was employed to address missing data and maintain dataset integrity. Results: We included 60 patients. Most of them were pauci-symptomatic (NYHA class II in 72%) and had no previous hospital admission for HF. LV function was preserved, but initial LV dilatation occurred. At CPET rest and peak VO2 were 6±1.5 and 17±5 ml/Kg/min respectively and ventilation/CO2 production slope (VE/VCO2) slope corresponded to 33±8.9 while the Oxygen Uptake Efficiency Slope (OUES) was 1.6±0.7 l/min/log(l/(min). At rest, patients presented a mean LVOT-VTI of 18±4 cm, stroke volume (SV) 54±18 ml, and cardiac output (CO) 4.2±1.5 l/min. Those values changed respectively to 24.5±4.9 cm, 72.5±21.6 ml and 9±3.2 l/min at peak. Four patients died before discharge. All patients were in NYHA class I-II at discharge, with LVEF of 55±6.8% and LVEDV of 109±30 ml. According to the random forest analysis NYHA class improvement was mainly influenced by baseline medical therapy, hypertension, EuroSCORE II, Renal Function, baseline vena contracta, VO2 at rest and peak, and VE/VCO2 slope. LVEF fractional improvement was significantly influenced by diabetes, preoperative medical therapy, eGFR, LVOT-VTI and diastolic function. Conclusions: At the time of surgery patients with DMR present reduced functional capacity despite preserved LVEF and mild symptoms. Apart from conventional risk factors, functional capacity and abnormal hemodynamic response to exercise have an impact respectively on NYHA class recovery and LVEF improvement after surgery. More complex models including CPET and ESE parameters are needed to provide a more accurate estimation of surgical correction effects.