Study population

This retrospective, observational analysis evaluated all consecutive patients undergoing transfemoral TAVI following heart team evaluation at the German Heart Center cardiovascular disease unit in Munich, Germany. Transfemoral TAVI was performed using a minimalistic approach¹⁸ in all cases, while THV size selection was left to the operator's discretion based on size chart, CCTA measurements, anatomical factors including calcium distribution and severity, aortic valve annular size, coronary height, and disease.

All patients with native tricuspid calcified aortic valve disease, and available high-quality CCTA for TAVI were included in this study. Procedural information was obtained from a customized database and screened for all patients treated from January 2014 through December 2021 at the German Heart Center in Munich, Germany. During this period, a total of 2500 transfemoral TAVI cases were performed using commercially available balloon-expandable (1865) and self-expanding (635) THVs. Only those who received the SAPIEN 3 or the SAPIEN 3 Ultra (Edwards Lifesciences, United States) balloon-expandable valves (BEV) were included in this analysis.

The study was performed in full compliance with the principles set forth in the Declaration of Helsinki, and all patients gave their written informed consent to undergo the procedure. Ethical approval was obtained from the Technical University of Munich ethical committee under registry no. OBSERVTAVI (#525/17). CCTA measurements were performed before THV implantation. Angiographic aortic valve annular measurements between the NCC and LCC were performed offline and documented in the database. The baseline clinical and procedural characteristics (including size of the implanted THV and angiographic aortic regurgitation after implantation), and test lab results were obtained from registry data or the clinical records, as appropriate. Regarding the Valve Academic Research Consortium 3 (VARC-3) defined safety and efficacy profile, in-hospital and discharge follow-up was monitored and registered. A 30-day follow-up was established via telephone call, hospital visits or follow-up letter.

Coronary computed tomography angiography measurements

CCTA was analyzed by 1 experienced cardiologist (HA) while a second experienced cardiologist (JM) analyzed a sample of 40 cases to determine inter-observer variability. Using multi-slice computed tomography reconstruction, quantitative measures of the aortic valve annular size (minimum, maximum and mean diameter, perimeter, and area) were obtained based on predefined protocols² using 3-Mensio software (Pie Medical Imaging, The Netherlands). In summary, the 3 hinge points of the aortic cusps were detected and selected. After proper identification of the 3 hinge points, the aortic annulus was seen in automatic multiplanar reconstruction. Annular measurements were obtained 0.5 mm below the hinge points, and the aortic valve annular contour was traced to calculate the perimeter-derived area and diameter (figure 1A). To define the direct one-planar measurement between the NCC and the LCC on the CCTA, a straight line between the red (LCC) and yellow (NCC) hinge points was used to determine length (figure 1B,C). The most appropriate THV size was selected based on size chart recommendations and anatomical considerations (figure 1D). CCTA-based measurements and calculations were used to determine the proper THV size according to commercial size charts supplied by the manufacturer. The mean diameter and area of the aortic annulus were used to select the THV size (23 mm, 26 mm, and 29 mm) and then coded as a binary variable for each size category; when only 1 measurement was within the proposed range for a specific THV size based on the manufacturer's size chart (within the gray zone), the area was used to decide the final THV size.

Figure 1. Central illustration. A: aortic annulus CCTA measurements determining the minimum (19.6 mm), maximum (25.9 mm), and mean (22.8 mm) diameter, perimeter (73 mm), and area (394 mm²). B: CCTA measurement from the lowest point between the left coronary cusp (red dot) and the non-coronary cusp (yellow dot) with a 23.0 mm distance. C: CCTA prediction of angiographic angulation to obtain the standard 3-cusp view (cranial right anterior oblique view of first and seventh nerves) with the left-coronary cusp from 1 side (red dot) and the non-coronary cusp on the other side (yellow dot), and the distance between them (23.0 mm). D: angiographic result after balloon-expandable valve implantation based on the tomographic measurements shown on figure 1 A (SAPIEN 3 Ultra 23 mm). 

Aortographic measurements

All procedures were performed by experienced TAVI operators using a monoplane digital flat panel detector X-ray system (Allura Xper FD 10 C, Royal Philips, The Netherlands) in a dedicated hybrid cath lab. All fluoroscopic 3-cusp view images were analyzed after completion of the procedure and images with distance measurements saved. Angiographic measurements were obtained offline from the angiographic aortic root injection (native annulus without the implanted THV) using a 5-Fr pigtail catheter placed in the right coronary cusp in the standard 3-cusp view.¹⁴,¹⁵ The distance between the NCC and LCC hinge points was measured by experienced cardiologists (HA and JM) using dedicated Phillips software (figure 2A-H). Angiographic measurements were performed after automatic (based on calibration factor determined by the software) and manual (using the 5-Fr catheter as reference calibration factor) calibration to determine the distance in millimeters.

Figure 2. A: standard 3-cusp cranial right anterior oblique view of first and seventh nerves (red circle) to determine longitudinal measurement (yellow arow) in the best contrasted image (green circle). B: calibration option based on the best image available (yellow arrow). C: measurement option (red arrow) to determine the distance on the aortic annulus (yellow size) on the images saved (green circle). D: saved image (red arrow) to “start analysis” (yellow arrow). E: manual calibration based on the 5-Fr catheter (red arrow and arrowhead) drawing 2 lines over the catheter (red circle) comparing the calibration factor given by the software (0.1451 mm/pixels, yellow arrow) to the one obtained through manual calibration (0.1486 mm/pixels, yellow arrowhead). F: “longitudinal measurement” option should be selected (red arrow) drawing a line at the hinge point of the left and non-coronary cusp (red circle). G: for automatic calibration, select “accept automatic calibration” (yellow arrow); the calibration factor given by the software (0.1451 mm/pixels) will be used for measurement purpose (yellow arrowhead, calibration factor of 0.1451 mm/pixels). H: aortic valve annulus measurement using automatic calibration selecting “longitudinal measurement” (red arrow) and drawing a line between the hinge points of the left and non-coronary cusp (red circle). 

Endpoints

Endpoints were the correlation between angiographic and CCTA measurements of the distance between the NCC and the LCC. The rates of the VARC-3-defined efficacy (technical success, correct position, and intended performance using VARC-3¹⁹ definitions) and safety profile (multiple valves, valve embolization, pacemaker implantation, and more than moderate valvular regurgitation using VARC-3 definitions) in patients with concordant and discordant measurements between angiographic and CT-based measurements were also analyzed.

Rates of in-hospital complications defined as conversion to surgery, perioperative death, life-threating bleeding, major and minor bleeding, major and minor vascular complications, and in-hospital mortality in patients with concordant and discordant measurements were reported. The 30-day mortality rate, chronic heart failure, stroke, valve dysfunction, aortic mean gradient, and aortic regurgitation were reported too.

Statistical analysis

Categorical variables were expressed as frequencies and proportions and compared using the chi-square test or Fisher's exact test, as appropriate. Continuous data were tested for normality with the Shapiro-Wilk test and expressed as mean (± standard deviation) or median (interquartile range [IQR]) as appropriate, and then compared, respectively, using the unpaired t test or the Mann-Whitney U test.

The study population was divided into derivation (n = 1256), and validation cohort (n = 40 cases). The study group of interest was obtained from the derivation cohort (n = 393). In the derivation cohort, selection of specific THV sizes (23 mm, 26 mm, and 29mm) based on the gold standard CCTA assessment was categorized as a binary variable and then compared to the THV size selection derived from aortography. Subsequently, logistic regression analysis was performed using the binary variable from the CCTA-based THV selection as a dependent variable while aortographic distance measurements were considered an independent variable. Afterwards, the receiver operating characteristic (ROC) curve was analyzed to identify optimal cut-off criteria (distance in mm, Youden's index) and determine individual diameter ranges based on aortographic distance measurements of each category of THV sizes. The lowest value from the smallest THV and the highest value from the largest THV was determined using the 25^th and 75^th IQR, respectively, taken from the distribution of the derivation population. The suggested THV size was derived using aortography with manual or automatic calibration. Sensitivity, specificity, positive, and negative predictive values, as well as positive and negative likelihood were used to determine diagnostic accuracy index. Bland-Altman plots were used to test correlation between the CT NCC-LCC and the NCC-LCC aortography with manual calibration and NCC-LCC aortography with automatic calibration.

Inter- and intra-observer analysis using intraclass correlation coefficient (with absolute agreement) and Pearson correlation coefficient for dichotomized data were performed in a sample of 40 cases between the 2 independent cardiologists.

To perform internal validation, previously established cut-off values to determine THV size by aortography were applied in a separate cohort of 40 patients (validation cohort) and compared using the gold standard CCTA-based sizing determination.

Pairs of sizing results based on angiography and CCTA were generated as a study group of interest and classified as concordant or discordant after comparison using the chi square test or Fischer's exact test, unpaired t test or the Mann-Whitney U test, as appropriate. Statistical analysis was performed using IBM SPSS Statistics software package (version 27, IBM Corporation, United States). All tests were 2-sided at the 0.05 significance level.

Discussion

This single-center, retrospective, observational study investigated the correlation and diagnostic accuracy between angiographic and tomographic measurements to determine THV size according to 1 single angiographic measurement between the NCC and the LCC in patients treated with BEV.

Regarding this objective, the most salient findings are a) angiographic aortic valve annular size determination based on distance measurements between the NCC and LCC is reproducible; b) diagnostic accuracy between CCTA-based and angiography-based aortic valve annular size determination is of moderate strength (table 5 of the supplementary data); and c) internal validation of previously established diameter ranges for angiography-based aortic valve annular size determination revealed moderate diagnostic accuracy.

We found a moderate overall diagnostic accuracy and correlation between angiographic and CCTA measurements to determine aortic valve annular size for THV selection. The use of angiography only measurements may expand the minimalistic TAVI approach in scenarios where CCTA is not an option or is unavailable.

The gold standard method to size the aortic annulus is direct surgical measurement, which is impossible in the TAVI setting. Hereby, several non-invasive reproducible methods have been used to determine the aortic valve annular size.³,⁵-⁷,²⁰ However, the CCTA has been established as the gold standard diagnostic tool to determine aortic valve annular size dimensions⁴ due to its outstanding reproducibility. Before CCTA was established as the actual gold standard method, angiographic measurements were used demonstrating good correlation with direct perioperative measurements in patients undergoing surgical aortic valve replacement (r = 0.93).¹³ The comparison of transesophageal echocardiography and CCTA and direct perioperative measurements reported by Wang et al.²⁰ showed a moderate correlation. Our study showed moderate diagnostic accuracy and correlation between angiographic and tomographic measurements to determine THV size. Similar results were previously tested in a small sample size of 50 patients where 60% of the valves were properly sized with fair-to-moderate agreement between angiography- and CCTA-guided selections.¹² This provides evidence that angiography measurements could potentially be used in scenarios where CCTA is not available or its application is of increased risk.

Study limitations

This study is limited by its single-center observational nature. A randomized or prospective study may be needed to confirm our results. Furthermore, the use of 1 type of fluoroscopy equipment may limit the applicability of the findings to fluoroscopy equipment from alternative manufacturers. Additionally, due to data storage limitations, angiography was not always available to determine the aortic valve annular size through manual or automatic calibration. Upper and lower values of the suggested size chart were determined based on 75^th and 25^th interquartile range, respectively, due to the lack of upper or lower outliers that would allow us to determine these values.