Abstract
Background
Computed tomography (CT) has become the standard imaging modality for pre-procedural aortic annular sizing prior to transcatheter aortic valve replacement (TAVR). We hypothesized that the accuracy of CT derived annular measurements would be greater at sites with higher TAVR procedural volume.
Methods
Within a large integrated health system, TAVR was performed at low (<40 cases), intermediate (40–75 cases), and high-volume sites (>75 cases). 181 patients underwent TAVR with a Sapien XT transcatheter heart valve (THV). Two blinded experienced readers independently remeasured the annulus on CT and compared their measurements to site reported measurements. Hypothetical THV sizes were chosen based on measurements from site CT reports and independent readers’ measurements, and compared to the implanted THV size.
Results
Correlation between site reported measurements and independent readers measurements of mean annulus size varied between low-volume (r = 0.31, p = 0.18), intermediate-volume (r = 0.34, p = 0.01), and high-volume sites (r = 0.96, p < 0.01). On multivariate analysis, interpretation of ≥20 CT scans (OR 0.29; 95% CI 0.03–0.81; p 0.02) and high-volume site (OR 0.16; 95% CI 0.10–0.82; p 0.02) were associated with reduced mismatch between the site predicted THV size and independent readers predicted THV size. Mismatch between site predicted THV size and implanted THV size was associated with a worse 30-day composite of mortality, dialysis-dependent renal failure, cerebrovascular accident, new permanent pacemaker, and hospital readmission (55.3 vs. 38.7%; p = 0.05).
Conclusions
Accuracy of CT aortic annular sizing is improved with higher individual experience and site TAVR volume. These findings should be confirmed in larger, prospective studies.
Highlights
- •
Accuracy of CT aortic annular sizing is improved with higher individual experience and site TAVR volume.
- •
CT readers with experience interpreting ≥20 pre-TAVR CT scans had significantly improved accuracy in identifying the annulus.
- •
Inaccurate aortic annulus sizing on CT was associated with a higher 30-day composite of mortality, dialysis-dependent renal failure, cerebrovascular accident, new permanent pacemaker, and hospital readmission.
1
Introduction
In patients undergoing transcatheter aortic valve replacement (TAVR), computed tomography (CT) has become the standard imaging modality for pre-procedural aortic annular sizing . Identifying the aortic annulus on CT is a complex procedure, involving a multi-step process to generate a double oblique reconstruction at the nadir of the three coronary cusps . Accurate annular sizing is important, as valve undersizing increases the risk of paravalvular regurgitation and excessive oversizing is a risk factor for annular rupture .
In many types of complex procedures, an association exists between volume and outcome, with improved outcomes correlated with a greater frequency at which a procedure is performed . We hypothesized that the accuracy of CT derived aortic annulus measurements would be greater at sites with a higher TAVR procedural volume compared to intermediate and low-volume TAVR sites.
2
Methods
This was an institutional review board approved retrospective study. Within a large integrated health system, TAVR was performed at low (<40 cases), intermediate (40–75 cases), and high-volume sites (>75 cases). 181 consecutive patients underwent TAVR with a Sapien XT (Edwards Lifesciences, Irvine, CA): 21 patients from the low-volume site, 62 patients from the intermediate-volume site, and 98 patients from the high-volume site. Heart teams at each site independently determined TAVR prosthesis sizing for each patient.
Prior to TAVR, all patients underwent contrast material enhanced CT of the heart with retrospective electrocardiography gating on a 64-slice scanner using a standard TAVR protocol , with minor site-specific modifications. Aortic annulus assessment was performed in systole, with imaging post-processing performed on an independent workstation (Vital Images, Minnetonka, MN) using a standard technique . Images were interpreted and reported by a site radiologist.
Clinical and imaging data for each patient were abstracted from the electronic medical record and the picture archiving and communication system. A complete report of CT derived aortic annular measurements was defined as including the annulus minor axis, major axis, mean diameter, area, and area derived diameter . Perimeter measurements were excluded because the post-processing software did not have this capability during part of the study.
Two blinded experienced readers (YP, MFM), with one year and five years experience interpreting TAVR CT scans respectively, independently reviewed the CT datasets from each patient and re-measured the aortic annulus minor axis, major axis, and area. Measurement differences were resolved by consensus. Hypothetical transcatheter heart valve (THV) sizes were chosen based on the site measurements and independent readers’ measurements of the mean annular size on CT using the intention for use sizing scheme : 23-mm THV for a 20–23 mm annulus, 26-mm THV for a 23–26 mm annulus and 29-mm THV for a 26–29 mm annulus. For ambiguous mean diameters measuring 23 or 26 mm, both THV sizes were considered appropriate. Mean annular size was chosen for THV sizing, as it was the least commonly missing from site CT reports. Hypothetical THV sizing was used to estimate mismatch between site predicted THV size, independent readers predicted THV size, and the actual implanted THV prosthesis.
Normally-distributed, continuous variables were described using means and standard deviations and non-normally-distributed as median and interquartile range. Normality of continuous variables was evaluated using the Shapiro–Wilk test. Categorical variables were described as proportions. Inter-rater agreement was evaluated using Cohen’s kappa coefficient. Linear correlation was evaluated using Pearson correlation coefficient and Bland–Altman analysis. Comparison between low, intermediate and high-volume centers for continuous variables was made using either one-way ANOVA or Kruskal–Wallis test (based on normal or non-normal distribution respectively), and for categorical variables using chi-squared test. A significant result was then evaluated with a pairwise comparison using Student’s t-test, nonparametric Mann–Whitney test, chi squared or Fisher’s exact test, as appropriate, and adjusted for multiple comparisons using Hommel procedure.
Univariate logistic regression models were used for evaluating associations between baseline characteristics. To adjust for inherent bias and confounding, multivariate logistic regression models were developed within the constraints of overfitting. Baseline characteristics were entered into the multivariate regression model in a forward, stepwise fashion and p-values <0.1 were used to retain a variable in the model. Time-to-event survival analysis was performed using the Kaplan–Meier estimator method, and event-free survival compared using log-rank test to generate p-values. A two-sided alpha of <0.05 was considered significant. Data were analyzed using Stata14.1 (StataCorp, College Station, Texas, USA).
2
Methods
This was an institutional review board approved retrospective study. Within a large integrated health system, TAVR was performed at low (<40 cases), intermediate (40–75 cases), and high-volume sites (>75 cases). 181 consecutive patients underwent TAVR with a Sapien XT (Edwards Lifesciences, Irvine, CA): 21 patients from the low-volume site, 62 patients from the intermediate-volume site, and 98 patients from the high-volume site. Heart teams at each site independently determined TAVR prosthesis sizing for each patient.
Prior to TAVR, all patients underwent contrast material enhanced CT of the heart with retrospective electrocardiography gating on a 64-slice scanner using a standard TAVR protocol , with minor site-specific modifications. Aortic annulus assessment was performed in systole, with imaging post-processing performed on an independent workstation (Vital Images, Minnetonka, MN) using a standard technique . Images were interpreted and reported by a site radiologist.
Clinical and imaging data for each patient were abstracted from the electronic medical record and the picture archiving and communication system. A complete report of CT derived aortic annular measurements was defined as including the annulus minor axis, major axis, mean diameter, area, and area derived diameter . Perimeter measurements were excluded because the post-processing software did not have this capability during part of the study.
Two blinded experienced readers (YP, MFM), with one year and five years experience interpreting TAVR CT scans respectively, independently reviewed the CT datasets from each patient and re-measured the aortic annulus minor axis, major axis, and area. Measurement differences were resolved by consensus. Hypothetical transcatheter heart valve (THV) sizes were chosen based on the site measurements and independent readers’ measurements of the mean annular size on CT using the intention for use sizing scheme : 23-mm THV for a 20–23 mm annulus, 26-mm THV for a 23–26 mm annulus and 29-mm THV for a 26–29 mm annulus. For ambiguous mean diameters measuring 23 or 26 mm, both THV sizes were considered appropriate. Mean annular size was chosen for THV sizing, as it was the least commonly missing from site CT reports. Hypothetical THV sizing was used to estimate mismatch between site predicted THV size, independent readers predicted THV size, and the actual implanted THV prosthesis.
Normally-distributed, continuous variables were described using means and standard deviations and non-normally-distributed as median and interquartile range. Normality of continuous variables was evaluated using the Shapiro–Wilk test. Categorical variables were described as proportions. Inter-rater agreement was evaluated using Cohen’s kappa coefficient. Linear correlation was evaluated using Pearson correlation coefficient and Bland–Altman analysis. Comparison between low, intermediate and high-volume centers for continuous variables was made using either one-way ANOVA or Kruskal–Wallis test (based on normal or non-normal distribution respectively), and for categorical variables using chi-squared test. A significant result was then evaluated with a pairwise comparison using Student’s t-test, nonparametric Mann–Whitney test, chi squared or Fisher’s exact test, as appropriate, and adjusted for multiple comparisons using Hommel procedure.
Univariate logistic regression models were used for evaluating associations between baseline characteristics. To adjust for inherent bias and confounding, multivariate logistic regression models were developed within the constraints of overfitting. Baseline characteristics were entered into the multivariate regression model in a forward, stepwise fashion and p-values <0.1 were used to retain a variable in the model. Time-to-event survival analysis was performed using the Kaplan–Meier estimator method, and event-free survival compared using log-rank test to generate p-values. A two-sided alpha of <0.05 was considered significant. Data were analyzed using Stata14.1 (StataCorp, College Station, Texas, USA).
3
Results
Table 1 shows baseline patient characteristics and CT parameters at the three sites. Patients at the high-volume site had higher average Society of Thoracic Surgeons Predicted Risk of Mortality score (STS-PROM; p < 0.01) and New York Heart Association class (NYHA; p < 0.01), whereas more patients had a history of smoking and atrial fibrillation at the low-volume site (p < 0.01). On imaging, the high volume site used less contrast (p < 0.01) and had higher contrast opacification of the annulus (p < 0.01) relative to the other sites. Incomplete reporting of the aortic annular measurements occurred at all sites, primarily due to not reporting annular area, which was missing in 43% of patients from the low-volume site, 82% of patients from the intermediate-volume site, and 10% of patients from the high-volume site (p < 0.01). Aortic annulus mean diameter was reported in all patients except for two cases at the low volume site (p < 0.01).
| Variable | Low-Volume Site | Intermediate-Volume Site | High-Volume Site | p-value |
|---|---|---|---|---|
| CLINICAL | ||||
| Age, Median (IQR) | 82 (80–85) | 86.5 (81–89) | 83 (75–87) | 0.01 |
| Male gender (%) | 57 | 40 | 53 | 0.22 |
| Caucasian (%) | 95 | 90 | 91 | 0.61 |
| BMI, Median (IQR) | 27.9 (26–29.8) | 27.3 (23.8–31.2) | 26.7 (24.3–31.3) | 0.63 |
| Diabetes Mellitus (%) | 33.3 | 45 | 37 | 0.48 |
| CAD (%) | 86 | 71 | 62 | 0.09 |
| CABG (%) | 33.3 | 33.9 | 24.9 | 0.04 |
| PVD (%) | 66.7 | 53.2 | 59.2 | 0.53 |
| Smoker (%) | 71.4 | 37.1 | 48 | 0.02 |
| CKD, n (%) | 7 (33) | 18 (29) | 37 (37) | 0.33 |
| Afib, n (%) | 12 (57) | 23 (37) | 14 (14) | <0.01 |
| LVEF, Median (IQR) | 30 (20–40) | 38.5 (30–40) | 25 (25–40) | 0.26 |
| STS PROM, Median (IQR) | 5.3 (3.9–8.9) | 5 (3.6–7.1) | 8 (5.1–11.1) | <0.01 |
| NYHA Class (%) | <0.01 | |||
| Class I | 4.8 | 0 | 2 | |
| Class II | 4.8 | 4.8 | 4.1 | |
| Class III | 66.8 | 80.7 | 45.9 | |
| Class IV | 23.8 | 14.5 | 47.9 | |
| IMAGING | ||||
| Contrast Volume | 120 (100–130) | 100 (75–125) | 85 (80–85) | <0.01 |
| Annulus HU, Median (IQR) | 361 (341–386) | 379 (338–425) | 451 (358–523) | <0.01 |
| Heart Rate, Mean | 69 | 73 | 69 | 0.57 |
| CT Readers, n | 3 | 10 | 1 | <0.01 |
| Missing Annulus Minor, n (%) | 3 (14.3) | 1 (1.6) | 0 | <0.01 |
| Missing Annulus Major, n (%) | 3 (14.3) | 0 | 0 | <0.01 |
| Missing Annulus Mean, n (%) | 2 (9.5) | 0 | 0 | <0.01 |
| Missing Annulus Area, n (%) | 9 (42.9) | 51 (82.3) | 10 (10.2) | <0.01 |
| Missing Area Derived Diameter, n (%) | 7 (33.3) | 36 (58.1) | 10 (10.2) | <0.01 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
