Author + information
- Received October 26, 2014
- Revision received January 23, 2015
- Accepted January 26, 2015
- Published online March 1, 2015.
- Marie-Annick Clavel, DVM, PhD∗∗ (, )
- Joseph Malouf, MD∗,
- David Messika-Zeitoun, MD, PhD†,‡,
- Phillip A. Araoz, MD∗,
- Hector I. Michelena, MD∗ and
- Maurice Enriquez-Sarano, MD∗
- ∗Division of Cardiovascular Diseases, Mayo Clinic, Rochester, Minnesota
- †Cardiology Department, AP-HP, Bichat Hospital, Paris, France
- ‡INSERM U698, University Paris 7-Diderot, Paris, France
- ↵∗Reprint requests and correspondence:
Dr. Marie-Annick Clavel, Division of Cardiovascular Diseases, Mayo Clinic, 200 First Street SW, Rochester, Minnesota 55905.
Objectives The aim of this study was to verify the hypothesis that multidetector computed tomography (MDCT) is superior to echocardiography for measuring the left ventricular outflow tract (LVOT) and calculating the aortic valve area (AVA) with regard to hemodynamic correlations and survival outcome prediction after a diagnosis of aortic stenosis (AS).
Background MDCT demonstrated that the LVOT is noncircular, casting doubt on the AVA measurement by 2-dimensional (2D) echocardiography.
Methods A total of 269 patients (76 ± 11 years of age, 61% men) with isolated calcific AS (mean gradient 44 ± 18 mm Hg; ejection fraction 58 ± 15%) underwent Doppler echocardiography and MDCT within the same episode of care. AVA was calculated by echocardiography (AVAEcho) and by MDCT (AVACT) using each technique measurement of LVOT area. In the subset of patients undergoing dynamic 4-dimensional MDCT (n = 135), AVA was calculated with the LVOT measured at 70% and 20% of the R-R interval and measured by planimetry (AVAPlani).
Results Phasic measurements of the LVOT by MDCT yielded slight differences in eccentricity and size (all p < 0.001) but with excellent AVA correlation (r = 0.92, p < 0.0001) and minimal bias (0.05 cm2), whereas the AVAPlani showed poor correlations with all other methods (all r values <0.58). AVACT was larger than AVAEcho (difference 0.12 ± 0.16 cm2; p < 0.0001) but did not improve outcome prediction. Correlation gradient-AVA was slightly better with AVAEcho than AVACT (r = −0.65 with AVAEcho vs. −0.61 with AVACT; p = 0.01), and discordant gradient-AVA was not reduced. For long-term survival, after multivariable adjustment, AVAEcho or AVACT were independently predictive (hazard ratio [HR]: 1.26, 95% confidence interval [CI]: 1.13 to 1.42; p < 0.0001 or HR: 1.18, 95% CI: 1.09 to 1.29 per 0.10 cm2 decrease; p < 0.0001) with a similar prognostic value (p ≥ 0.80). Thresholds for excess mortality differed between methods: AVAEcho ≤1.0 cm2 (HR: 4.67, 95% CI: 2.22 to 10.50; p < 0.0001) versus AVACT ≤1.2 cm2 (HR: 3.16, 95% CI: 1.64 to 6.43; p = 0.005), with simple translation of spline-curve analysis.
Conclusions Head-to-head comparison of MDCT and Doppler echocardiography refutes the hypothesis of MDCT superiority for AVA calculation. AVACT is larger than AVAEcho but does not improve the correlation with transvalvular gradient, the concordance gradient-AVA, or mortality prediction compared with AVAEcho. Larger cut-point values should be used for severe AS if AVACT (<1.2 cm2) is measured versus AVAEcho (<1.0 cm2).
Aortic stenosis (AS) is a frequent valvular disease (1) treatable only by valve replacement, surgical or percutaneous (transcatheter aortic valve replacement [TAVR]) (2,3). Because TAVR does not allow intraprocedural valve sizing, evaluation of the ventriculoaortic transition zone has become an essential goal of aortic valvular imaging to avoid complications of TAVR such as prosthesis embolization, aortic annulus rupture, and paravalvular leak (4–6). This new requirement of thorough pre-procedural aortic apparatus evaluation has led to 3-dimensional (3D) imaging studies using 3D echocardiography (7) or multidetector computed tomography (MDCT) (8), which suggested that the “aortic annulus” was complex, often noncircular (9,10) and concluded that MDCT provided superior annular measurement compared with standard 2D echocardiography (11,12).
However, “aortic annulus” measurements, if improper, would have wide-ranging implications regarding AS hemodynamic assessment. Noninvasive aortic valve area (AVA) calculation requires aortic annular (or left ventricular outflow tract [LVOT]) cross-sectional area (LVOTArea) (13), which for Doppler echocardiography uses a disputed single-diameter measurement. AVA measured by Doppler echocardiography (AVAEcho) has long been regarded as validated and as part of routine clinical practice but recently has been criticized as underestimating AVA calculated with the more anatomically sound LVOTArea measured by MDCT (AVACT) (7,11,12). The possibility that a faulty AVAEcho may bear responsibility for “discordant” AS cases, with low gradient despite tight AVA, resonates with other challenges to the authenticity of this syndrome (14). The contrast between AVAEcho being revered as fully validated and as the major independent predictor of outcome in AS and being reviled as anatomically illogical and underestimated versus AVA measured by computed tomography (AVACT) has not been resolved and is crucial to the management of patients with AS.
Thus, the aims of our study were as follows: 1) to verify whether the discordant hemodynamic pattern in AS is associated with aortic “annulus” asymmetry and resolved by use of AVACT versus AVAEcho; 2) to assess whether the association between gradient and AVA is improved by the use of AVACT versus AVAEcho; and 3) to assess whether the association between AVA and survival after a diagnosis of AS is improved by use of AVACT versus AVAEcho. Our main hypothesis is that AVACT will be superior to AVAEcho with regard to discordant AS severity grading, hemodynamic correlations, and clinical outcome impact. However, our secondary hypothesis is that thresholds to define severe AS will be higher by AVACT than by AVAEcho.
We analyzed data for 269 adult AS patients who underwent comprehensive Doppler echocardiography and contrast-enhanced MDCT within the same episode of care (<3 months between evaluations). We excluded children younger than 18 years of age, patients with identified rheumatic disease or endocarditis, and those with moderate or severe mitral valve disease and/or previous valve repair or replacement.
Patients were enrolled in a prospective clinical research study initiated by the Valvular Heart Disease Clinic. Informed consent was obtained according to institutional review board approval. The CT angiography (with contrast) was performed to address clinical questions: 1) uncertainty regarding AS severity; 2) questions regarding LVOT in known AS; 3) in known AS, vascular indications regarding the status of the aorta or peripheral vessels; and 4) questions regarding the presence or severity of coronary artery disease.
Doppler echocardiography measurements
Left ventricular dimensions, peak aortic jet velocity, mean gradient (ΔP), and left ventricular ejection fraction were measured according to recommendations of the American Society of Echocardiography (13). Doppler echocardiographic LVOT diameter was measured at the base of the valve leaflets (Figure 1) and used to calculate LVOTArea using echocardiography:AVAEcho was then calculated by the continuity equation:where VTILVOT and VTIAo are the velocity time integrals of the LVOT and transaortic flow, respectively.
MDCT imaging and measurements
The contrast-enhanced electrocardiography-gated MDCT examinations were performed with a 64-detector scanner (Sensation 64, Siemens Medical Systems, Forchheim, Germany), without the routine use of β-blocker medications (Online Appendix). Dynamic LVOT assessment was performed in patients in whom the entire 4-dimensional MDCT dataset of aortic valve and LVOT was recorded and available for quantitative analysis during all phases of the cardiac cycle (n = 135). Dynamic LVOT measurement involved tracing the LVOT contour at 70% and 20% of the R-R interval, allowing the calculation of AVACT in diastole (AVACT-D) and AVACT in systole (AVACT-S), respectively. Cardiac MDCT was performed using between 80 and 105 ml of isomolar contrast medium.
All LVOT imaging was performed using the same process with dedicated software (Aquarius iNtuition, TeraRecon, Inc., Foster City, California), involving orientation of a cross-sectional plane of LVOT by using 3 orthogonal planes from multiplanar reconstruction (Figures 2A to 2C) and located at or immediately under the lowest implantation base of aortic cusps (Figure 2C). Once this plane was secured, the LVOT contour was traced, 2 orthogonal diameters were directly measured, and the area of LVOT (LVOTAreaCT) was planimetered (Figure 2D). The eccentricity index of the aortic valve annulus was calculated by dividing the smaller diameter by the larger one.
AVACT was calculated by using the measured LVOTAreaCT in the continuity equation:In patients with dynamic imaging and full 4-dimensional MDCT volumes, the imaging plane was translated to the tips of the aortic cusps in systole, and planimetry of the aortic valve opening (AVAPlani) was measured.
To assess the respective value of AVA measurements obtained by Doppler echocardiography and MDCT, we examined several endpoints, including the relationship between ΔP and AVAs and the prevalence of discordant AS severity grading. The primary endpoint of this study was the overall survival under medical treatment. Hence, the end of follow-up was marked by aortic valve replacement (AVR) for patients operated on or by death or last known follow-up for patients not operated on. Patients were thus censored (follow-up stopped) at AVR. The secondary endpoint was total mortality during the entire follow-up (i.e., medical and post-AVR) (Online Table 1). Therapeutic decisions in our study were made by the patients’ personal physicians based on all the information available. The decision to not operate immediately after the first evaluation was in this cohort mostly linked to nonsevere AS but also to asymptomatic AS or to symptoms interpreted as unrelated to AS. Only 10% of our patients were considered potentially high operative risk.
Outcome data were obtained from the annual visit of the patient or the patient’s charts, mailed questionnaires or scripted telephone interviews with the patients or physicians, and death certificates. We obtained complete follow-up up to AVR, to death, or to the year preceding freezing of the dataset in 96% of patients (258 of 269), and the 11 patients with incomplete follow-up contributed to outcome information for the duration of known follow-up.
Results are expressed as mean ± SD or percentage. Differences between the 4 methods of AVA measurements were analyzed using 1-way repeated-measures analysis of variance followed by the Tukey test. Correlation and agreement between the different LVOTs and AVAs were determined with the use of the Pearson correlation and Bland-Altman methods, respectively. Relationships between AVAs and ΔP were assessed with multiple regression models, and the equation providing the best fit was retained. Comparison between correlation coefficients used the Wolfe test. The comparison of patient’s classification between AVAEcho and AVACT was done with the McNemar test.
To analyze the effect of AVAEcho and AVACT on mortality, we used spline curve graphs where the x-axis represents AVA and the y-axis the relative risk of mortality. The effect of the clinical, Doppler echocardiographic, and MDCT variables on survival under medical treatment was assessed using Cox proportional hazard models adjusted for age-adjusted Charlson score index (15), sex, symptoms, ΔP, and left ventricular ejection fraction. Analysis of overall mortality during the entire follow-up was done with adjustment for age, sex, symptoms, coronary artery disease, chronic obstructive pulmonary disease, diabetes, hypertension, ΔP, left ventricular ejection fraction, and AVR as a time-dependent variable. All variables in the Cox models verified the proportional hazards assumption on the basis of inspection of trends in the Schoenfeld residuals (all p > 0.32). The results of the Cox proportional hazards were presented with hazard ratio and 95% confidence interval. The accuracy of the Cox model was assessed by the Harrell C index and compared by bootstrapping. A p value <0.05 was considered statistically significant.
Overall, at baseline, patient age was 76 ± 11 years, and 163 patients (61%) were male (Table 1). In terms of comorbidity, prevalence of hypertension (79%), diabetes (29%), coronary artery disease (49%), and chronic pulmonary disease (26%) were as expected in a population of that age. Doppler characteristics were ΔP 44 ± 18 mm Hg and peak velocity 4.2 ± 0.9 m/s.
Evaluation by static MDCT showed that LVOTArea was generally elliptical (93% patients had more than 10% difference between LVOT diameters), with an eccentricity index of 0.79 ± 0.08 in the entire cohort (Table 1). Categorized in the entire cohort using the median LVOT static eccentricity index (i.e., 0.8), patients with the most elliptical annulus (eccentricity index <0.8) had few differences (Table 1) with slightly more severe AS with higher peak velocity (p = 0.03) and ΔP (p = 0.009) and smaller AVA calculated by Doppler echocardiography, and reduced ejection fraction (p = 0.002) compared with patients with more circular annulus.
Echocardiographic and dynamic MDCT measurements of AVA
As LVOTArea (Online Appendix), AVA was different in all methods, with echocardiography measuring the smallest values and with considerable dispersion of values for AVAPlani compared with other methods (Figure 3). The correlations between AVAEcho and either AVACT-D or AVACT-S (at 70% and 20% of R-R, respectively) were good and equivalent (both r values = 0.78, p < 0.0001) (Online Figure 1) with modest dispersion. The correlations obtained with AVAPlani and any other method were significant but much poorer (all r values <0.58) (Online Figure 1). Conversely, the correlation between phasic calculations of AVA by MDCT was excellent (r = 0.92, p < 0.0001), with significant but minimal differences (mean 0.05 cm2) (Online Figures 1E and 1F). In view of these excellent dynamic-static LVOT/AVA measurement correlations, static measurements were used in the entire population with regard to grading consistency and survival. Accordingly, in the whole cohort, the correlation between AVAEcho and AVACT was good and with modest dispersion (Online Figure 2).
Relationship between ΔP and AVACT or AVAECHO
First, regression between ΔP and AVA, AVAEcho, or AVACT was analyzed. The correlation between ΔP and AVA was better (Wolfe test, p = 0.01) for AVAEcho (r = −0.65, p < 0.0001) than static AVACT (r = −0.61, p < 0.0001) (Figure 4). The use of indexed AVA provided the same results (see the Online Appendix, Online Figure 3). Thus, AVACT compared with AVAEcho, despite its accounting for LVOT eccentricity, does not improve hemodynamic correlation of AVA-ΔP. When stratified by eccentricity index, the correlation AVA-ΔP tended to be weaker in patients with a more oval LVOT (i.e., eccentricity index <0.8), irrespective of the method used to measure AVA, but regression slopes were not affected, and use of MDCT to assess AVA did not improve AVA-gradient correlations and regressions in any subset (Figure 5). For purposes of completeness, we verified that dynamic MDCT (AVACT-S) did not improve the AVA-gradient relationship (all p values >0.25) over static MDCT, whereas measuring AVA by planimetry worsened the AVA-gradient correlations (all comparisons, p ≤ 0.04).
Second, the usefulness of AVACT to resolve discordant AS severity grading was analyzed. Of our 269 patients, 66 (25%) had a low gradient (<40 mm Hg) despite a tight AVAEcho. This prevalence was not different among patients with a high or low LVOT eccentricity index (p = 0.75) (Table 1). Patients classified as concordant AS assessment (either AVA ≤1.0 cm2 combined with ΔP ≥40 mm Hg or AVA more than 1.0 cm2 with ΔP <40 mm Hg) represented 74% by AVAEcho and 70% by AVACT (p = 0.15), showing that AVACT did not improve the concordance of AS severity grading.
Using specific thresholds for each method did not show improvement of grading by AVACT (Online Appendix).
Impact of AVAECHO and AVACT on survival in AS
During a mean follow-up of 2.0 ± 1.4 years under medical treatment, there were 55 deaths (Online Table 1). On univariable and multivariable analyses, AVA calculated by echocardiography (p < 0.0001) or by MDCT (p < 0.0001) were independent predictors of mortality under medical treatment (Table 2). However, the negative impact on survival under medical treatment occurred for different thresholds for AVAEcho and AVACT, as shown by the spline curves linking relative risk mortality under medical management and the 2 AVA types as continuous variables (Figure 6, Online Appendix). As dichotomized variables, AVAEcho ≤1.0 cm2 (p < 0.0001) and AVACT ≤1.2 cm2 (p = 0.005) were independent predictors of mortality under medical treatment (Table 2, Figure 7). Importantly, as continuous or dichotomized variables, AVAEcho and AVACT showed equivalent power to predict mortality under medical treatment using Harrell C index and net reclassification index (Table 2, Online Appendix).
These results were confirmed in the entire follow-up (3.2 ± 2.5 years) (Online Table 1), with further adjustment for AVR as a time-dependent variable. AVAEcho and AVACT were independent predictors of total mortality as continuous variables (all p values = 0.02) as well as dichotomized variables (all p values ≤0.03).
Our study compares for the first time AVA by using Doppler echocardiography and MDCT obtained during the same episode of care in the same patients with AS with regard to hemodynamic correlations, discordance in AS severity grading, and clinical outcome impact. We first confirm that using static or dynamic MDCT for AVA calculation is equivalent with a minimum bias between phasic measurements. However, the measurement of AVA by planimetry should be avoided given that this measure provides the lowest correlation with the other methods and worsens the association AVA-gradient. Although the larger AVA by planimetry may not be surprising as it measures anatomic (vs. effective) orifice area, the dispersion of values and poor correlations with gradient reflect inconsistent measurements. The most important result of our study addresses the hypothesis that AVACT is superior to AVAEcho, which is not verified by any of the measured endpoints. Indeed, although AVACT is larger than AVAEcho, the AVA-gradient relationship is not improved, and the frequency of discordant AS grading is unaffected, showing that, as a rule, the use of MDCT results in a translation of the correlation curves. Although our study shows for the first time that AVACT independently determines subsequent survival, this method did not improve outcome prediction compared with AVAEcho. However, thresholds affecting mortality were different for AVACT and AVAEcho, with spline curves demonstrating a simple translation of mortality prediction to higher values for AVACT.
The complex anatomy of the ventriculoaortic complex
The aortic valve has a complex structure (16). Until recently, little attention has been paid to this anatomic complexity, but recent work showed that the LVOTArea imaged by 3D imaging techniques does not have the previously presumed circular shape (9,10), and our data confirms a generally elliptical shape of the LVOT. The previous rule by 2D Doppler echocardiography, using the parasternal long-axis view, was to measure the diameter, alternatively called aortic annulus and LVOT, at the level of presumed lowest implantation of leaflets (17). 3D imaging techniques directly measuring of the entire area of the LVOT/annulus with orthogonal positioning ascertained by locating the bottom of all aortic cusps suggested that 2D Doppler echocardiography underestimated (8,11,12,18) this measurement. This issue of smaller LVOTArea calculated by 2D than measured by MDCT, which we undeniably confirm, has been labeled as “underestimation of the LVOT area” and has led to calls for the replacement of 2D annular measurement by 3D measurements in the calculation of the AVA (11,12,18). Such appeals are troubling because 2D measurements of AVA have been widely validated (17,19,20). Hence, it is essential to examine from physiological and outcome points of view whether Doppler echocardiography calculation of the AVA is superseded by the “superior” MDCT measurements.
Physiological comparisons of MDCT versus doppler echocardiography measurements
MDCT and Doppler echocardiography measurements of LVOTArea provide different AVAs so that AVAEcho is generally 0.1 to 0.2 cm2 smaller than AVACT (7,11,12). This different AVA calculation obtained by different methods increases the confusion regarding the best threshold of AVA identifying severe AS. Indeed, there is a purported inconsistency between echocardiographic criteria for grading AS, whereby an AVA of 1 cm2 is touted as corresponding to a ΔP of 30 mm Hg, whereas a ΔP of 40 mm Hg would instead correspond to an AVA of 0.8 cm2 (21). Hence, although guidelines indicate 1.0 cm2 as the threshold for severe AS (2,22), it has been proposed (21) that this threshold for AVAEcho should be decreased to 0.8 cm2. This drastic change would conflict with outcome studies showing excess mortality (23) in AS with an AVA ≤1.0 cm2, whereas no study has shown a superiority of AVA ≤0.8 cm2 in predicting mortality. In that regard, our study provides important information. First, it shows that AVACT, despite accounting for the LVOT elliptical shape, provides no superior prediction of survival compared with AVAEcho. Second, our spline analysis shows that the threshold of excess mortality is indeed an AVA ≤1.0 cm2 for AVAEcho, whereas it is an AVA ≤1.2 cm2 for AVACT with spline curves showing a simple translation between these 2 methods. Third, AVAEcho ≤0.8 cm2 and AVACT ≤1.0 cm2 are not independent predictors of mortality. Thus, despite the elliptical shape of the LVOT (9,10), discrepancies between MDCT measurements and direct annular sizing (24) suggest that MDCT and 2D echocardiography do not measure exactly the same structure. An alternate explanation of echocardiography–computed tomography discrepancies may reside with fluid dynamics, as blood velocity is null at the wall level, so that the effective flow area may be more circular than the anatomic elliptical LVOTArea, potentially leading to overestimation of the flow surface and stroke volume by computed tomography. Furthermore, stroke volume and AVA measured by Doppler echocardiography has been validated in studies comparing AVAEcho and AVA calculated by catheterization (19,20,25) or magnetic resonance (26).
Hence, with the hypothesis that MDCT provides superior measurement not verified, measuring LVOT Area by MDCT is not required in clinical routine, unless poor imaging windows preclude aortic annular measurement or TAVR is planned. Indeed, measurement of the mean LVOT diameter and perimeter by MDCT has been linked to better sizing of transcatheter prostheses and to lower post-procedural paravalvular regurgitation (27,28).
For image stability, static acquisitions were done in the whole cohort at 70% of the R-R interval, and a large number of patients had only static acquisition for minimizing radiation exposure. In patients with dynamic MDCT, we were able to assess phasic variations of LVOTArea. The impact of phasic variations on AVA calculation is significant but minimal. The fact that these phasic changes did not influence the regression between the AVA and ΔP is reassuring. There is no clinically available method yet to coregister images from 2D echocardiography and MDCT to prove beyond a doubt that measurements provided by these 2 techniques are not of the exact same anatomic structure, but imaging progress will soon allow such overlay and allow verification of this hypothesis. For now, our data showing the equivalent hemodynamic and outcome value of these techniques suggest that the history accumulated with Doppler echocardiography should not be discarded and that guidelines need not be revised for the management and severity assessment of AS. The combined endpoint of AVR or death was not used in the study due to potential bias (AVAEcho guided decision making for AVR) and the lack of increased power with this frequent event (Online Table 1). This study should be confirmed by larger studies with longer follow-up.
Doppler echocardiography is the first-line examination for evaluation of AS severity. Asymmetrical LVOT by 3D imaging raised concerns about 2D echocardiographic AVA calculation accuracy but AVA calculation by MDCT does not improve grading concordance or outcome prediction. There are differences between echocardiography and computed tomography measurements, but echocardiography-measured AVA is not inferior to that calculated using LVOT by MDCT. Moreover, based on survival after diagnosis, thresholds defining severe AS should be different: 1.2 cm2 for AVA by MDCT instead of 1.0 cm2 for AVA by 2D echocardiography. Thus, measurement of LVOT diameter by MDCT is a valuable method to calculate AVA to assess AS severity. However, the use of MDCT is not mandatory in clinical routine for evaluation of AS severity, and echocardiography should remain the first-line of evaluation. Nevertheless, MDCT may be helpful in patients in whom there is a doubt about the aortic annulus diameter measurements by echocardiography for any reasons.
For an expanded Methods section as well as supplemental tables and figures, please see the online version of this article.
Dr. Clavel holds a postdoctoral fellowship grant from Canadian Institute of Health Research. Dr. Malouf worked at Mayo Clinic as a consultant. Dr. Messika-Zeitoun is a consultant for Valtech, Edwards Lifesciences, and Abbott. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- aortic stenosis
- aortic valve area
- aortic valve area measured by Doppler echocardiography
- aortic valve area measured by computed tomography
- aortic valve replacement
- mean gradient
- hazard ratio
- left ventricular outflow tract
- multidetector computed tomography
- transcatheter aortic valve replacement
- Received October 26, 2014.
- Revision received January 23, 2015.
- Accepted January 26, 2015.
- American College of Cardiology Foundation
- Nishimura R.A.,
- Otto C.M.,
- Bonow R.O.,
- et al.
- Vahanian A.,
- Alfieri O.,
- Andreotti F.,
- et al.
- Blanke P.,
- Reinohl J.,
- Schlensak C.,
- et al.
- Santos N.,
- de Agustin J.A.,
- Almeria C.,
- et al.
- Ng A.C.,
- Delgado V.,
- van der Kley F.,
- et al.
- Delgado V.,
- Ng A.C.,
- van de Veire N.R.,
- et al.
- Messika-Zeitoun D.,
- Serfaty J.M.,
- Brochet E.,
- et al.
- Lang R.M.,
- Bierig M.,
- Devereux R.B.,
- et al.
- Jander N.,
- Minners J.,
- Holme I.,
- et al.
- Anderson R.H.
- Skjaerpe T.,
- Hegrenaes L.,
- Hatle L.
- Oh J.K.,
- Taliercio C.P.,
- Holmes D.R. Jr..,
- et al.
- Zoghbi W.A.,
- Farmer K.L.,
- Soto J.G.,
- Nelson J.G.,
- Quinones M.A.
- Minners J.,
- Allgeier M.,
- Gohlke-Baerwolf C.,
- Kienzle R.P.,
- Neumann F.J.,
- Jander N.
- Kempfert J.,
- Van Linden A.,
- Lehmkuhl L.,
- et al.
- Hamm K.,
- Trinkmann F.,
- Heggemann F.,
- et al.
- Jilaihawi H.,
- Kashif M.,
- Fontana G.,
- et al.
- Willson A.B.,
- Webb J.G.,
- Labounty T.M.,
- et al.