Sutureless Aortic Valve Prosthesis Sizing: Estimation and Prediction Using Multidetector-Row Computed Tomography

The sutureless prostheses may facilitate minimally invasive aortic valve replacement because of easy and fast deployment. However, correct device sizing remains a crucial step of this procedure, which may be difficult and time consuming in minimal invasive approaches. We sought to analyze the accuracy of contrast-enhanced preoperative multidetector-row computed tomography (MDCT) in predicting the size of the prosthesis to be implanted in patients undergoing aortic valve replacement through a right anterior minithoracotomy (RAMT).

From January 2011 to September 2013, 235 patients underwent aortic valve surgery as sole procedure with implantation (Sorin Perceval S) in RAMT. Inclusion criterion for this study was presence of preoperative multidetector-row computed tomography (MDCT) with contrast enhancement and Doppler echocardiography. A preoperative MDCT was used to measure the aortic annulus as the diameter derived from either the area (aD) or the circumference (cD) of the virtual basal ring, left ventricular outflow tract (LVOT) diameter derived either from the area (aLVOT) or the circumference (cLVOT). Multidetector-row CT was reviewed by a single operator who was blind to implanted valve size. The operator measured the aortic annulus and LVOT in multiplanar reconstruction modality. Aortic annular diameter and LVOT diameter were retrieved from echocardiographic records. Predictive models were built based on logistic regression; outcome variable was the sutureless valve size, and covariates (annular and LVOT measurements) were used as single and multivariate predictors. A classification tree was built and then pruned with limited nodes to be able to obtain better predictive performance.

We identified 54 patients who had preoperative contrast-enhanced MDCT. Seven patients received a size S, 21 received a size M, and 26 received a size L prosthesis. The mean age of the patients at the time of intervention was 76.3 ± 6.8 years, and the mean logistic EuroSCORE was 10.4% ± 8.7%. Echocardiographic measurements showed lower accuracy compared to MDCT measurements. Echocardiographic LVOT measurement was 61.11% to predict the valve size, whereas annulus measurement was 53.7%. The aLVOT from MDCT had an accuracy of approximately 62.96%, and cLVOT had 64.81% predictive accuracy. Aortic annulus perimeter cD had the highest accuracy to predict the valve size [62.96%, under the curve, 0.61] followed by aortic annular surface aD having an accuracy of approximately 70.37% (under the curve, 0.75). Classification tree models, after pruning with 4 nodes, increased their accuracy (83.33%), and it was easy to interpret and possibly to implement for clinical use.

Multidetector-row CT-derived estimates seem to have higher predictive value for valve size determination in patients undergoing RAMT with the Perceval S prosthesis, thus facilitating this delicate procedure and preventing the selection of wrong candidates. Possibly for precise aortic annulus measurement, contrast-enhanced MDCT is preferable.