PO-01-208 A COMPUTATIONAL MODEL TO TRACK CHANGES IN LEFT VENTRICULAR END-DIASTOLIC PRESSURE DURING ATRIAL FIBRILLATION

Abstract

Heart failure and Atrial Fibrillation (AF) commonly occur concomitantly, resulting in significant morbidity and mortality. Thus, monitoring of heart failure status is essential in directing medical therapy in these patients. Left ventricular (LV) filling pressures are one of the most important parameters in assessing worsened or acutely decompensated left heart failure, however current LV pressure monitoring strategies include invasive devices and catheterization. Our study aims to develop a computational model of heart failure in AF and demonstrate that beat-to-beat variations in RR intervals during AF can be exploited to predict changes in LV filling pressure. We used a modified Research CardioVascular SIMulator (RCVSIM) human cardiovascular model to simulate an intact circulatory system in patients with AF over 300 seconds. To simulate beat-by-beat variation of rhythm in AF, the length of each RR-interval was drawn randomly from a Mixed Gaussian Distribution. We modelled worsening HF by changing LV end-systolic compliance in the simulator, using compliances ranging from 1 to 3.5 mL/mmHg, with interval increases in compliance of 0.1 mL/mmHg. Stroke volumes, RR-intervals, and LV end-diastolic pressures are recorded for each simulation. This was repeated for mean heart rates from 60 to 110 beats per minute. Stroke volumes were found to have a strong linear correlation with the previous beat’s RR-interval for all mean heart rates tested (Spearman rho = 0.9-0.97). In Figure 1 below, we show the slope of the RR-interval versus stroke volume (SV/RR-interval) plotted against average LV end-diastolic pressure for all LV systolic compliances noted above and a mean heart rate of 84 beats per minute. A cubic function fit to this plot shows excellent goodness-of-fit with the simulated data with a coefficient of determination (R2) of 0.99. This relationship remained consistent for all mean heart rates tested, and all cubic fits had R2 greater than 0.98. Our study provides a highly accurate model for estimating LV end-diastolic pressures over a broad range of LV systolic heart failure in simulated AF patients using only stroke volume and RR-intervals. Further work using photoplethysmography or continuous non-invasive arterial pressure (CNAP) monitoring could potentially be used to obtain these measures non-invasively, thus leading to a non-invasive method for tracking heart failure in patients with AF.