Multi-parametric approach to predict prosthetic valve size using CMR and clinical data: insights from SAVR.

The purpose of this investigation was to characterize the CMR and clinical parameters that correlate to prosthetic valve size (PVS) determined at SAVR and develop a multi-parametric model to predict PVS. Sixty-two subjects were included. Linear/area measurements of the aortic annulus were performed on cine CMR images in systole/diastole on long/short axis (SAX) views. Clinical parameters (age, habitus, valve lesion, valve morphology) were recorded. PVS determined intraoperatively was the reference value. Data were analyzed using Spearman correlation. A prediction model combining imaging and clinical parameters was generated. Imaging parameters had moderate to moderately strong correlation to PVS with the highest correlations from systolic SAX mean diameter (r = 0.73, p < 0.0001) and diastolic SAX area (r = 0.73, p < 0.0001). Age was negatively correlated to PVS (r = - 0.47, p = 0.0001). Weight was weakly correlated to PVS (r = 0.27, p = 0.032). AI and bicuspid valve were not predictors of PVS. A model combining clinical and imaging parameters had high accuracy in predicting PVS (R2 = 0.61). Model predicted mean PVS was 23.3 mm (SD 1.1); actual mean PVS was 23.3 mm (SD 1.3). The Spearman r of the model (0.80, 95% CI 0.683-0.874) was significantly higher than systolic SAX area (0.68, 95% CI 0.516-0.795). Clinical parameters like age and habitus impact PVS; valve lesion/morphology do not. A multi-parametric model demonstrated high accuracy in predicting PVS and was superior to a single imaging parameter. A multi-parametric approach to device sizing may have future application in TAVR.