Increased Gradients After Aortic Valve Replacement With The Perceval Valve: The Role of Oversizing

Transcription

1 Accepted Manuscript Increased Gradients After Aortic Valve Replacement With The Perceval Valve: The Role of Oversizing Alfredo Giuseppe Cerillo, MD, Francesca Amoretti, MD, Massimiliano Mariani, MD, Emanuele Cigala, MD, Michele Murzi, MD, Tommaso Gasbarri, MD, Marco Solinas, MD, Dante Chiappino, MD PII: DOI: Reference: ATS S (18) /j.athoracsur To appear in: The Annals of Thoracic Surgery Received Date: 16 August 2017 Revised Date: 8 November 2017 Accepted Date: 18 December 2017 Please cite this article as: Cerillo AG, Amoretti F, Mariani M, Cigala E, Murzi M, Gasbarri T, Solinas M, Chiappino D, Increased Gradients After Aortic Valve Replacement With The Perceval Valve: The Role of Oversizing, The Annals of Thoracic Surgery (2018), doi: /j.athoracsur This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

2

3 Increased Gradients After Aortic Valve Replacement With The Perceval Valve: The Role of Oversizing Short Title: Increased gradients in sutureless AVR Alfredo Giuseppe Cerillo, MD 1, Francesca Amoretti, MD 2, Massimiliano Mariani, MD 3, Emanuele Cigala, MD 3, Michele Murzi, MD 1, Tommaso Gasbarri, MD 1, Marco Solinas, MD 1, Dante Chiappino, MD 2 Operative Units of Cardiac Surgery 1, Radiology 2, and Cardiology 3, G. Pasquinucci Hospital, G. Monasterio Foundation, Via Aurelia Sud, 54100, Massa, Italy Article Word Count: 4492 Abstract Word Count: 247 Address for correspondence: Dr Alfredo Giuseppe Cerillo, G. Pasquinucci Hospital, Via Aurelia Sud 54100, Massa, Italy. acerillo@yahoo.com

4 ABSTRACT Background. Significant underexpansion or distortion of valved stents may be associated with altered leaflet function, leading to increased transprosthetic gradients and, possibly, early structural degeneration. We investigated the relationship between a computed tomography (CT) measure of the degree of oversizing and the early haemodynamic and clinical outcomes in patients undergoing aortic valve replacement with the Perceval sutureless aortic valve. Methods. The degree of oversizing of the implanted prosthesis was calculated as the ratio between the patients' aortic annulus cross-sectional area (CSA) and the ex-vivo CSA of the implanted prosthesis in 151 perceval patients that had a preoperative cardiac CT. This value was then entered in a multivariate analysis to ascertain its role as a predictor of increased postoperative gradient. Results. The operative mortality was 1.3%. Procedural success, defined as having a normofunctioning valve in the proper anatomical location, was achieved in 150 patients (99.3%). The mean transprosthetic gradient was 13.4±5.0 mmhg, and twenty-three patients (15.2%) showed a gradient 20 mmhg at discharge or at the 1 month follow-up. The degree of oversizing of the implanted prosthesis was the single, most important predictor of increased postoperative gradient (OR:1.264; 95% CI: ; p<0.0001). Interestingly, other relevant factors (patients' BSA, prosthesis size) were not associated with increased gradients. Conclusions. Our study demonstrates that in Perceval patients excessive oversizing should be avoided, and suggests that a different sizing algorithm, possibly based on cardiac CT, should be developed. Further studies are needed to optimize the sizing strategy for the Perceval S valve. Key Words: aortic valve replacement, sutureless aortic valve, cardiac ct

5 The Perceval sutureless aortic bioprosthesis (Livanova, Saluggia, Italy) is a bovine pericardium valve mounted in a Nitinol stent, that can be compressed and positioned in a valve delivery system. The stent is deployed under direct vision in the open heart by turning a release screw. After the deployment, a balloon catheter may be used to expand the stent and post-dilate it at a pressure between 2-4 atm (1,2). Similarly to transcatheter aortic valve implantation (TAVI) devices, the anchoring and good sealing of the Perceval bioprosthesis relies on oversizing the stent frame compared to the native aortic annulus. In recent years, cardiac computed tomography (CT) has become the gold standard technique for measuring the aortic annulus in patients undergoing TAVI, and the CT derived axial image of the aortic Virtual basal ring (VBR) is considered as the reference for sizing by most of the transcatheter valves manufacturers (3). Interestingly, the VBR lies exactly on the plane passing through the nadir of the three aortic cusps, that is where, according to the instructions for use (IFU), a correctly positioned Perceval valve should be deployed (2). Since the Perceval valve stent is a self-expandable nitinol stent with relatively low radial force, the area of the aortic annulus VBR could provide a good estimate of the final cross-sectional area of the stent after deployment, and could therefore be used to calculate the degree of oversizing (or underexpansion) of the valve. Recent evidence in TAVI patients suggests that significant underexpansion or distortion of valved stents may be associated with altered leaflet function, leading to increased transprosthetic gradients and possibly predisposing to other negative outcomes, like valve thrombosis, low platelet counts, thrombo-embolic events and early degeneration (4,5). Moreover, there is evidence that excessive oversizing of the Perceval valve is detrimental (6,7). However, a systematic analysis of the relationship between the degree of oversizing and the haemodynamic performance of the Perceval valve is lacking. The aim of the present study is to investigate the relationship between the CT-derived measure of the degree of oversizing and the early haemodynamic and clinical outcomes in patients undergoing sutureless aortic valve replacement with the Perceval sutureless aortic valve.

6 MATERIAL AND METHODS Patients From March 2011 to December 2015, 617 patients underwent aortic valve replacement with the Perceval at our institution. A base-line, preoperative cardiac CT allowing for correct sizing of the aortic annulus was available for 159 patients. Eight patients were excluded from the analysis due to the presence of a bicuspid aortic valve, or because they underwent concomitant mitral procedures. The remaining 151 patients constitute the population of the present study. The demographic and clinical patients' characteristics are reported in Table 1. All the data presented in the study were prospectively collected and entered in our institutional database which includes ten sections that are filled in consecutively by anesthesiologists, surgeons, perfusionist, intensive care unit and ward doctors. The study was approved by the clinical audit committee of the G. Pasquinucci Heart Hospital to meet ethical and legal requirements, and individual consent was waived. Surgical Procedure In the 137 patients undergoing isolated aortic valve replacement (AVR), the operation was performed through a minimally invasive approach (right anterior minithoracotomy in 101 patients and upper partial sternotomy in 36 patients). The remaining 14 patients underwent associated procedures, and were operated through a full sternotomy. The aortic root was accessed through a high aortotomy, at the level of the aortic fat pad or immediately below it. The aortic valve leaflets were resected and the annulus decalcified accurately. Sizing was performed as per the IFU: the size of the prosthesis to be implanted was indicated by the sizer for whom the transparent obturator passed through the annulus but the white obturator did not (3). Although the procedures were performed by six different surgeons, an experienced senior surgeon with more than 100 Perceval implants supervised the sizing process and valve deployment performed by less experienced operators. CT Protocol and Sizing For the purposes of this study, only patients with a suitable preoperative cardiac CT scan performed at our institution were considered. Cardiac CT was performed with a 320 slices multidetector CT scanner (Toshiba Aquilion ONE, Toshiba Medical Systems S.R.L., Rome, Italy). The scan was done using retrospective ECG

7 gating in spiral technique. Eighty ml of contrast media were injected at 4 ml/sec and bolus tracking was used. The reconstructed slice width was of 1 mm or less, and all the measurements were done in diastole (70% of the cardiac cycle), in order to maximize the image quality, and to avoid variation due to the fact that the systolic images were not available for all patients. The patients CT scan dicom files were downloaded from our institutional images server, anonymized, and independently analyzed. The area and perimeter of the aortic valve virtual basal ring were measured on the 3D multi-planar reconstructions as recommended for TAVI (2, 9). To obtain a measure of the ex vivo cross-sectional area of the Perceval valve, a CT scan of a sample prosthesis for each of the four available measures (S, M, L, XL) was performed. In order to obtain homogeneous measurements and to establish a reference point, the cross-sectional area (CSA) and perimeter of the prosthesis inflow stent were measured at the nadir of the three prosthetic cusps (prosthetic virtual basal ring), on the external surface of the metal stent, with fixed values for the Window Level and the Window Width (Figure 1). All the measurements were blindly repeated three times by four independent investigators, totaling 12 measurements for each size. The average measure was then used for the analysis (table 2). The degree of oversizing of the implanted prosthesis was calculated as the ratio between the patients' native aortic annulus CSA and the ex-vivo CSA of the implanted valve, according to the formula (1-native VBR/prosthesis CSA) x 100. Clinical Management and Haemodynamic Assessment All patients were managed according to our routine institutional protocol. At the end of the surgical procedure, the prosthetic valve function was assessed by an expert cardiologist or cardiac anesthesiologist. Transthoracic echocardiography was always repeated before discharge, and at 1 month postoperatively. Continuous-wave Doppler was used to assess the flow velocity across the prosthetic valve. Peak velocity was averaged from three velocity envelopes, and mean velocity was calculated by averaging the instantaneous velocities measured throughout the velocity complexes. The peak and mean transprosthetic gradients were calculated with the modified Bernoulli equation. Outcome Variables Definitions and Statistical Analysis

8 The clinical and haemodynamic outcomes are defined according to the VARC 2 guidelines (10). Procedure success was defined as having a single, normo-functioning valve in the proper anatomical position. A mean trans-prosthetic gradient of 20 mmhg was defined as increased transprosthetic gradient. Continuous variables are expressed as mean ± standard error, dichotomous variables are expressed as percentages. The association of preoperative, intraoperative and postoperative variables with an increased postoperative gradient was investigated by the Fisher s exact test (dichotomous variables), or by the unpaired Student s t-test (continuous variables). Non-normally distributed continuous variables were analyzed by the Mann Whitney U-test. Factors resulting significantly associated to the end point of the study were then included in a logistic multivariate regression model to ascertain their independent role. Also factors for which the univariate analysis gave a P value 0.1, or of known biologic significance, but failing to meet the critical α level, were included. Odds ratio (OR) and 95% confidence interval (CI) were calculated. Receiver operating characteristic (ROC) curves were calculated to single out the best cutoff value of prosthesis oversizing predicting an increased postoperative gradient. The accuracy of the test was assessed measuring the area under the ROC curve (AUC). The statistical significance of difference of AUC from that of the line of no information was evaluated by Mann Whitney U-statistic. A P value <0.05 was considered significant. Statistical analysis was conducted using SPSS software version 10.1 (SPSS Inc, Chicago, IL). RESULTS Procedural Outcome and Complications The postoperative outcome and complications are reported in table 3. The operative mortality was 1.3%. One patient with multiple comorbidities undergoing combined CABG and AVR developed low cardiac output and multi-organ failure, and died on postoperative day 12. A second patient was discharged to a rehabilitation facility on postoperative day 6 and went home on postoperative day 13. On postoperative day 14 he developed 3 rd degree AV block, and underwent emergency pace-maker implantation at another institution. The post-procedure course was complicated and the patient died on day 45. Procedural success, defined as having a single, normofunctioning valve in the proper anatomical location at the end of the procedure, was achieved in 150 patients (99.3%). One patient that had received an M size prosthesis was reoperated on postoperative day 1 due to the presence of paravalvular leakage with moderate

9 to severe aortic regurgitation. The prosthesis was explanted and a bulk annular calcification was removed. An L size prosthesis was then implanted with a good final result. CT Measurements and Prosthesis Oversizing The size of the implanted prosthesis was S in 16 patients (10.5%), M in 46 (30.4%), L in 76 (50.3%) and XL in 13 (8.6%). The mean aortic annulus area at the level of the virtual basal ring was of 3.494±0.002 cm 2, 3.86±0.001 cm 2, 4.66±0.001 cm 2, and 5.57±0.05 cm 2 in S, M, L, and XL patients, respectively. The mean aortic annulus perimeter at the level of the virtual basal ring was of 6.7±0.002 cm, 7.1±0.001, 7.74±0.001 cm, and 8.45±0.04 cm. There was a significant variation in degree of oversizing of the implanted prosthesis, from 6.5% to 49.5% for size S, from 8.9% to 48.9% for size M, from 2.9% to 41.3% for size L, and from 1.1% to 36.5% for size XL. Also, there was a significant overlapping between the four valve sizes (Figure 2, Figure 3). Ecohocardiographic Assessment of Prosthetic Valve Function The mean transprosthetic gradient at discharge was 13.4±5.0 mmhg, and twenty-three patients (15.2%) had a transprosthetic gradient 20 mmhg at discharge or at the 1 month follow-up. Twenty-one of these patient were available for subsequent follow-up. The transprosthetic gradient was reduced in 4, stable in 7 and increased in 10 patients. Factors associated with increased transprosthetic gradient are reported in table 4. At multivariate analysis, the degree of oversizing of the implanted prosthesis was the single, most important predictor of increased postoperative gradient (OR: 1.264; 95% CI: ; p < ). Interestingly, other factors traditionally associated with increased transprosthetic gradients, like the nominal size of the implanted prosthesis or the patients' BSA, showed no impact on the early postoperative gradient at multivariate analysis. According to the ROC curve analysis, the best threshold value of prosthesis oversizing for predicting an increased postoperative gradient was 29.9% (ROC area, ± , P < 0.001). This cutoff value showed a sensitivity of and a specificity of (figure 4): a prosthesis oversizing of more than 30% was associated with a 16-fold increase of the risk of developing an increased postoperative gradient.

10 COMMENT Our study demonstrates that excessive oversizing of the Perceval valve is associated with increased transprosthetic gradients, and should be avoided. Grossly oversized Perceval bioprostheses tend to recoil causing loss of contact between the prosthesis and the annulus, which results in paravalvular leakage and, possibly, in significant aortic regurgitation. Recoiling has been observed in the clinical setting and in the laboratory (6, 7). However, a mean to measure the degree of oversizing in Perceval patients has not been reported to date, and the haemodynamic and clinical effects of different degrees of oversizing when recoil does not occur are not known. In our experience, an oversizing 30% in terms of annular CSA was associated with a 16-fold increase of the risk of developing increased postoperative gradients. In our series, some typical predictors of PPM like the patients' BSA and the size of the implanted prosthesis were not related to the outcome, while a greater degree of oversizing was associated with increased gradients. This apparent paradox is, in our opinion, well explained by the fact that the full expansion of the valved stent inside the patient aortic annulus could be associated with smoother leaflet kinetics, improved valve opening and increased EOA, while the stent compression or deformation that follows the implantation of a prosthesis that is too big with respect to the patient annulus could lead to altered leaflets' kinetics, incomplete valve opening and increased gradients. In fact, the Perceval valve has an in vitro EOA that is much higher than that of conventional sutured bioprostheses (10,11): for the S size, the in vitro EOA ranges from 2.07 to 2.2 cm 2 (10), an in a study of patients with small annuli the measured in vivo EOAI was 1.12 ± 0.2 cm 2 /m 2 versus 0.82 ± 0.1 cm 2 /m 2 for small size conventional, sutured valves (12), indicating that the fluidodynamics of valved stents might be superior to that of conventional surgical valves. The observation that the use of transcatheter aortic valves is associated with higher EOAI and with a reduced incidence of PPM would also support this idea (12). A secondary finding of our study is that intraoperative sizing for the Perceval valve, as recommended by the IFU, does not reflect the true in-vivo measure of the aortic annulus, and could be misleading. Surgical, obturator-based sizing is a complex process that converts several non numerical inputs (visual assessment, tactile feedback, stiffness of the cardiac tissues, the amount and distribution of calcium, the fragility of the aortic wall or the height of the coronary arteries, and others) in a number. Furthermore, both the surgical sizers and the sizing strategy differ significantly between different prostheses, making the sizing process even more complex (13). In our series, there was an impressive degree of overlapping between the four prosthetic sizes

11 (Figure 2, Figure 3). Surgical sizing resulted in a mean oversizing of more than 20%, and probably led to the implantation of a too big prosthesis in a significant proportion of patients. Other Authors have reported difficulties with the Perceval S sizing process. Baert and co-workers recently reported valve recoiling due to excessive oversizing in four patients (2.9% of their Perceval population). They suggested to modify the sizing process and to implant the valve size Given by the sizer of which the white obturator pass the annulus with friction (7). This would certainly reduce the number of grossly oversized prostheses, but not the fact that the choice between different valve sizes in patients with similar annular dimensions would remain arbitrary. In our opinion, cardiac CT could offer a simple solution to this problem. In fact, it has become a very standardized process, and it is used with and extremely high rate of success in thousands of TAVI patients every year. Given the surprising forgivingness of the Perceval valve, the development of a CT-based sizing reference chart would not be complex: the valve showed good haemodynamic performance in a wide range of conditions, and still behaved acceptably in the extreme conditions of virtually no to almost 50% oversizing (Table 5). Our study has some limitations than need to be acknowledged. It is a single institution, retrospective analysis of prospectively collected data in a non-consecutive patients population. Only 25% of patients receiving a Perceval valve at our institution had a preoperative cardiac CT, mainly done for coronary assessment, or in the setting of a screening for TAVI. This could have resulted in selection bias. However, the base-line, procedural and postoperative features of this population did not differ significantly from those observed in our general Perceval patients population (14), and we believe that the present findings and conclusions could be easily replicated in other series. Moreover, the implanting surgeons were usually not aware of the CT-derived annular measurements, and the choice of the prosthesis to be implanted was not affected by the availability of CT data. The ex-vivo CSA of the prosthesis as measured at CT, rather than the true measure made by micrometers, was used as a caliper for all our analysis. CT sizing in the presence of metal objects can be tricky, and our sizing protocol could have been affected by metal artifacts. However, precise sizing was not the aim of our research: we tried to establish a relationship between the degree of under-expansion of the valved stent and its haemodynamic performance, and we are convinced that the exactness of the absolute measures was far less important that the derived, relative estimate of oversizing.

12 Our study focused on the early postoperative period. Increased gradients early after aortic valve replacement can be related to many reversible factors, may disappear with time, and are often of limited clinical significance. In our opinion, however, increased gradients in patients with Perceval valves should not be neglected. In fact, as observed above the presence of an increased gradient could indicate altered leaflets kinetics, and may herald other negative outcomes, like subclinical valve thrombosis and early degeneration, at least in some patients (15). The observation that in our series the increased gradients persisted over time would support this concept. By avoiding the need to stitch the annulus and to tie dozens of knots, sutureless aortic valve prostheses have the potential to make aortic valve replacement faster, easier and safer. However, sutureless prostheses are relatively new, and their in-vivo behavior could be different from the expected. Our study demonstrates that in Perceval patients excessive oversizing should be avoided, and suggests that a different sizing algorithm, possibly based on cardiac CT, should be developed. Further studies are needed to optimize the sizing strategy for the Perceval S valve, in order to take advantage of the full potential of this valve.

13 References 1. Glauber M, Moten SC, Quaini E, Solinas M, Folliguet TA, Meuris B, et al. International Expert Consensus on Sutureless and Rapid Deployment Valves in Aortic Valve Replacement Using Minimally Invasive Approaches. Innovations 2016;11: Pfeiffer S, Sirch J, Vogt F, Fischlein T, Santarpino G. Implantation of the Sorin Perceval sutureless aortic valve: a step by step approach. Minerva Cardioangiol Apr;65(2): Achenbach S, Delgado V, Hausleiter J, Schoenhagen P, Min JK, Leipsic JA. SCCT expert consensus document on computed tomography imaging before transcatheter aortic valve implantation (TA- VI)/transcatheter aortic valve replacement (TAVR). J Cardiovasc Comput Tomogr. 2012;6: Hansson NC, Grove EL, Andersen HR, Leipsic J, Mathiassen ON, Jensen JM, et al. Transcatheter Aortic Valve Thrombosis: Incidence, Predisposing Factors, and Clinical Implications. J Am Coll Cardiol Nov 8;68: Makkar RR, Fontana G, Jilaihawi H, Chakravarty T, Kofoed KF, de Backer O, et al. Possible Subclinical Leaflet Thrombosis in Bioprosthetic Aortic Valves. N Engl J Med ;373: Di Eusanio M, Saia F, Pellicciari G, et al. In the era of the valve-in-valve: is transcatheter aortic valve implantation in sutureless valves feasible? Annals of Cardiothoracic Surgery 2015;4: Baert J, Astarci P, Noirhomme P, de Kerchove L, The Risk of Oversizing with the Perceval S Suturless Bioprosthesis in Small Aortic Annulus. J Thorac Cardiovasc Surg 2017;153: Piazza N, de Jaegere P, Schultz C, Becker AE, Serruys PW, Anderson RH. Anatomy of the aortic valvar complex and its implications for transcatheter implantation of the aortic valve. Circ Cardiovasc Interv Aug;1: Kappetein AP, Head SJ, Généreux P, Piazza N, van Mieghem NM, Blackstone EH, et al. Valve Academic Research Consortium (VARC)-2. Updated standardized endpoint definitions for transcatheter aortic valve implantation: the Valve Academic Research Consortium-2 consensus document (VARC- 2). Eur J Cardiothorac Surg. 2012;42:S45-60.

14 10. Belluschi I, Moriggia S, Giacomini A, Del Forno B, Di Sanzo S, Blasio A, et al. Can Perceval sutureless valve reduce the rate of patient-prosthesis mismatch? Euur J Cardiothorac Surg 2017;51: Tasca G, Vismara R, Mangini A, Romagnoni C, Contino M, Redaelli A, et al. Comparison of the Performance of a Sutureless Bioprosthesis With Two Pericardial Stented Valves on Small Annuli: An In Vitro Study. Ann Thorac Surg. 2017;103: Shalabi A, Spiegelstein D, Sternik L, Feinberg MS, Kogan A, Levin S, et al. Sutureless Versus Stented Valve in Aortic Valve Replacement in Patients With Small Annulus. Ann Thorac Surg. 2016;102: Doenst T, Amorim PA, Al-Alam N, Lehmann S, Mukherjee C, Faerber G. Where is the common sense in aortic valve replacement? A review of hemodynamics and sizing of stented tissue valves. J Thorac Cardiovasc Surg. 2011;142: Concistrè G, Chiaramonti F, Bianchi G, Cerillo A, Murzi M, Margarian R, et al. Aortic valve replacement with Perceval bioprosthesis: single center experience with 617 implants. Ann Thorac Surg 2017, in press. 15. Cerillo AG, Amoretti F, Mariani M, Chiappino D. Sutureless valve thrombosis: the role of stent geometry. Ann Thorac Surg 2017, in press.

15 Table 1. Baseline and operative patients characteristics Age (years) 76±0.4 Sex (f, %) 93 (61.5) BSA (m 2 ) 1.8±0.2 Hypertension, (%) 105 (69.5) Diabetes, (%) 36 (23.8) COPD, (%) 26 (17.2) Chronic renal failure, (%) 4 (2.6) Coronary artery disease, (%) 21 (13.9) Left ventricular dysfunction, (%) 16 (10.5) Peripheral vascular disease, (%) 28 (18.5) EuroSCORE II 4.2±3.9 STS score 3.8±4.1 Aortic VBR area (cm 2 ) 4.37±0.062 Aortic VBR perimeter (cm) 7.48±0.052 Implanted prosthesis Size Surgical approach Size S, (%) 16 (10.5) Size M, (%) 46 (30.4) Size L, (%) 76 (50.3) Size XL, (%) 13 (8.6) Sternotomy, (%) 14 (9.2) Right anterior minithoracotomy, (%) 101 (66.9) Partial sternotomy, (%) 36 (23.8) Associated CABG, (%) 14 (9.2) ACC time (min) 53.7±0.13 CPB time (min) 72.6±0.23 VBR: virtual basal ring; CABG: coronary artery bypass grafting; ACC: aortic cross clamp; CPB: Cardiopulmonary bypass.

16 Table 2. Average CT derived area and perimeter of the Perceval sutureless prosthesis CSA area (cm 2 ), mean±se CSA perimeter (cm), mean±se Size S 4.429± ±0.004 Size M 5.235± ±0.005 Size L 6.21± ±0.003 Size XL 7.05± ±0.006

17 Table 3. Early outcome and complications Operative mortality, (%) 2 (1.3) Procedural Success, (%) 150 (99.3) Stroke, (%) 4 (2.6) Acute renal failure, (%) 1 (0.6) Acute respiratory failure, (%) 6 (3.9) Pace-maker implantation, (%) 7 (4.6) Reoperation for bleeding, (%) 4 (2.6) Time on ventilator (hours), mean±se 8±0.2 ICU stay (days), mean±se 1±0.06 Hospital stay (days), mean±se 5.2±0.2 Echocardiographic data Left ventricular ejection fraction, mean±se 55.4±0.8 Mean pressure gradient, mean±se 13.4±5.0 Transprosthetic gradient > 20 mmhg, (%) 23 (15.2) Residual aortic regurgitation Mild 12 (7.9) Moderate* 2 (1.3) Severe - ICU: intensive care unit. * Moderate residual aortic regurgitation was intraprosthetic in both cases

18 Table 4. Predictors of increased postoperative gradient Univariate Multivariate OR 95% CI p OR 95% CI p Age ns Sex (f) BSA Prosthesis size VBR area (cm 2 ) ns VBR perimeter (cm) ns Oversizing (%) < < Oversizing > 30% <0.0001

19 Table 5. A sizing algorithm for the Perceval S valve. The model is based on the CT-derived virtual basal ring area (mm 2 ), and is designed to obtain an oversizing ranging from 5% to 25%. Min Max S M L XL

20 FIGURE LEGENDS Figure 1. An ex-vivo Perceval S valve sized at CT. (a) To obtain a good visualization of the prosthetic cusps, a Brain preset (Window Level: 50; Window Width: 100) was used. (b) The nadir of the first cusp is identified and marked. (c) The operation is repeated on the nadir of the remaining cusps, and the plane of the prosthetic Virtual basal ring is identified. (d) To reduce the metal artifacts, the measurements of the stent crosssectional area were performed using a Bone preset (Window Level: 300; Window Width: 1500). Figure 2. Scatterplot demonstrating the distribution of the patients' aortic VBR area (top) and perimeter (bottom) according to the prosthesis size implanted. There is a considerable overlapping between the four valve sizes. Figure 3. Double oblique axial CT images of the virtual basal ring from six different patients. Upper line: three patients with virtually identical aortic annular sizes (3.7 cm 2, red boxes) that received three different prosthesis sizes (S, M and L respectively). The ex-vivo area of the implanted prostheses (grey boxes) has been represented with three closed circles for comparison. Middle line: three patients with virtually identical aortic annular sizes (4.1 cm 2, red boxes) that received three different prosthesis sizes (S, M and L respectively). Bottom line: the VBR area of this patient (4.14 cm 2, center) is depicted against the ex-vivo area of the S, M and L Perceval valve (left). The patient received a M valve. The post-implant area of the VBR was 4.18 cm 2, indicating minimal expansion of the aortic annulus after the implantation. The patient underwent valvein-valve implantation with a 23 mm Sapien 3 valve (Edwards Lifesciences, Irvine, Ca) for valve failure three years after the original operation. Figure 4. Receiver operating characteristic curve for the prediction of increased postoperative gradients. The best predictive value of oversizing was 29.9%. AUC=0.885; 95% CI= P<

21

22

23

24