Author + information
- Received July 2, 2015
- Revision received August 18, 2015
- Accepted August 20, 2015
- Published online March 1, 2016.
- Christopher Naoum, MBBSa,
- Jonathon Leipsic, MDa,
- Anson Cheung, MDa,
- Jian Ye, MDa,
- Nicolas Bilbey, MDa,
- George Mak, MBBSa,
- Adam Berger, MBBSa,
- Danny Dvir, MDa,
- Chesnal Arepalli, MDa,
- Jasmine Grewal, MDa,
- David Muller, MBBSb,
- Darra Murphy, MBBSa,
- Cameron Hague, MDa,
- Nicolo Piazza, MDc,
- John Webb, MDa and
- Philipp Blanke, MDa,∗ ()
- aSt. Paul’s Hospital and University of British Columbia, Center for Heart Valve Innovation, Vancouver, British Columbia, Canada
- bSt. Vincent’s Hospital, Sydney, Australia
- cDepartment of Medicine, Division of Cardiology, McGill University Health Centre, Montreal, Quebec, Canada
- ↵∗Reprint requests and correspondence:
Dr. Philipp Blanke, St. Paul’s Hospital, University of British Columbia, 1081 Burrard Street, Vancouver, British Columbia V6Z 1Y6, Canada.
Objectives The aims of this study were to determine D-shaped mitral annulus (MA) dimensions in control subjects without significant cardiac disease and in patients with moderate to severe mitral regurgitation (MR) being considered for transcatheter mitral therapy and to determine predictors of annular size, using cardiac computed tomography.
Background The recently introduced D-shaped method of MA segmentation represents a biomechanically appropriate approach for annular sizing prior to transcatheter mitral valve implantation.
Methods Patients who had retrospectively gated cardiac computed tomography performed at our institution (2012 to 2014) and were free of significant cardiac disease were included as controls (n = 88; 56 ± 11 years of age; 47% female) and were compared with patients with moderate or severe MR due to functional mitral regurgitation (FMR) (n = 27) or mitral valve prolapse (MVP) (n = 32). MA dimensions (projected area, perimeter, intercommissural, and septal-to-lateral distance), maximal left atrial (LA) volumes, and phasic left ventricular volumes were measured.
Results MA dimensions were larger in patients with FMR or MVP compared with controls (area index 4.7 ± 0.6 cm2/m2, 6.0 ± 1.3 cm2/m2, and 7.3 ± 1.7 cm2/m2; perimeter index 59 ± 5 mm/m2, 67 ± 9 mm/m2, and 75 ± 10 mm/m2; intercommissural distance index 20.2 ± 1.9 mm/m2, 21.2 ± 3.1 mm/m2, and 24.7 ± 3.2 mm/m2; septal-to-lateral distance index 14.8 ± 1.6, 18.1 ± 3.3, and 19.5 ± 3.4 mm/m2 in controls and patients with FMR and MVP, respectively; p < 0.05 between controls and MR subgroups). Absolute MA area was 18% larger in patients with MVP than patients with FMR (13.0 ± 2.9 cm2 vs. 11.0 ± 2.3 cm2; p = 0.006). Although LA and left ventricular volumes were both independently associated with MA area index in controls and patients with MVP, only LA volume was associated with annular size in patients with FMR.
Conclusions Moderate to severe MR was associated with increased MA dimensions, especially among patients with MVP compared with control subjects without cardiac disease. Moreover, unlike in controls and patients with MVP, annular enlargement in FMR was more closely associated with LA dilation.
- computed tomography
- mitral annulus
- mitral regurgitation
- transcatheter mitral valve implantation
With the rapid innovation and growing clinical adoption of transcatheter mitral therapies, including transcatheter mitral valve implantation (TMVI), an accurate understanding of mitral annular (MA) dimensions and geometry is becoming increasingly important. Given the saddle-shaped, nonplanar configuration of the MA, 3-dimensional (3D) imaging is required for comprehensive assessment. Although this can be achieved using computed tomography (CT), with its excellent spatial resolution (1–4), limited data exist describing CT values for MA dimensions in patients with significant mitral regurgitation (MR) in whom TMVI may be a potential therapeutic option.
We recently proposed a D-shaped concept of MA geometry, in which the annulus is truncated along a virtual line connecting both fibrous trigones, as a standardized, reproducible, and more biomechanically appropriate method for MA sizing prior to TMVI (5). An important characteristic of the D-shaped segmentation method is that it yields a more planar annulus that closely resembles the cross-sectional area of current TMVI devices, which is not achieved by conventional (saddle-shaped) analyses. Annular size and geometry and the determinants of MA size in patients with moderate to severe MR have not been studied using the D-shaped method. Moreover, the range of D-shaped MA dimensions in patients without significant cardiac disease is unknown.
Accordingly, we sought to determine annular dimensions, geometry, and drivers of annular size in patients with moderate to severe MR and compare these findings with those of control subjects without significant cardiac disease using retrospectively electrocardiographically (ECG) gated cardiac CT.
The Institutional Review Board approved this retrospective study with a waiver for informed consent. Two study cohorts were identified. Consecutive patients who underwent clinically indicated, retrospectively gated cardiac CT at our institution between August 2012 and February 2014 and were identified as being free of significant cardiac disease on the basis of CT findings and review of available clinical information were included as controls. Only scans performed with retrospective ECG gating were included so that multiphasic data could be analyzed. Exclusion criteria included: 1) known significant mitral valve disease and/or greater than mild MA calcification seen on cardiac CT; 2) clinical history of congestive heart failure and/or reduced measured left ventricular (LV) ejection fraction <50%; 3) obstructive coronary artery disease on cardiac CT (≥70% in any coronary vessel or >50% in the left main) or prior coronary revascularization; 4) history of atrial fibrillation; 5) prior cardiac surgery; 6) complex congenital heart disease; 7) obesity (body mass index [BMI] >35 kg/m2); 8) increased maximal left atrial (LA) volume index (>78 ml/m2, a cutoff representing 2 SDs from the mean value previously reported in healthy subjects ); and/or increased LV mass index (>103 g/m2 for men and >89 g/m2 for women ). Consecutive patients with moderate to severe MR referred for cardiac CT between November 2013 and June 2015 for workup prior to potential TMVI were included. Patients with MR were divided into 2 groups based on MR mechanism (mitral valve prolapse [MVP] or functional mitral regurgitation [FMR]). Patients with a prior aortic and/or mitral valve prosthesis were excluded from the MR group.
Cardiac CT data acquisition
Cardiac CT was performed using a 64-slice helical CT scanner (Discovery high-definition 750 or VCT, GE Healthcare, Milwaukee, Wisconsin). For controls, CT acquisition was undertaken according to the institutional protocol for performing retrospectively gated clinical cardiac CT. For patients with MR, a pre-specified clinical cardiac CT protocol was used. Imaging was performed during a single breath-hold following injection of 80 to 110 ml of intravenous contrast media (Visipaque 320, GE Healthcare) with a triphasic injection (contrast, contrast/saline mix, and saline) for controls and a biphasic injection (contrast and saline) for patients with MR. Tube voltage and current were manually determined (on the basis of BMI) with subsequent ECG modulation of tube current for controls to minimize radiation dose (median [interquartile range] effective dose 9.6 mSv [5.7 to 11.8 mSv] in controls and 14.1 mSv [11.3 to 20.2 mSv] in patients with MR). Scan range extended from the carina to just below the inferior cardiac surface. Axial images were reconstructed at 10% intervals of the cardiac cycle with a slice thickness of 0.625 mm.
CT data analysis
CT measurements were performed offline by batch analysis using dedicated software for MA segmentation (3mensio Structural Heart V7.0, Pie Medical Imaging, Maastricht, the Netherlands) and volumetric analyses (Aquarius iNtuition v4.4, TeraRecon, Foster City, California). Different observers separately assessed MA parameters and cardiac volumes (P.B. and C.N. performed all MA measurements by co-review and consensus agreement; N.B. performed all cardiac volume measurements).
Assessment of MR severity and mechanism in the MR group
MR severity was graded by review of echocardiographic data according to guidelines for the assessment of MR severity (8). The mechanism of MR was determined by separate review of both echocardiographic data and multiphasic (cine) CT datasets using multiplanar reconstructions to generate 2- and 3-chamber views of the LV. MVP was defined by the presence of systolic excursion of a mitral leaflet more than 2 mm beyond the annular plane in either a 2- or 3-chamber view (9). FMR was defined as LV remodeling (dilation and/or global or regional LV dysfunction) that prevents leaflet coaptation in the absence of a primary mitral valve abnormality (10).
The method for segmentation and assessment of the D-shaped MA has been recently described (5,11). Briefly, mid to late diastolic image reconstructions with the least artifact identified by visual assessment were used for MA segmentation. The MA contour was generated by cubic-spline-interpolation of 16 seeding points manually placed along the insertion of the posterior mitral valve leaflet and along the anterior peak comprising the fibrous aortomitral continuity. The lateral and medial fibrous trigones were then manually identified and the distance between these 2 points defined as the trigone-to-trigone (TT) distance, which separates the anterior compartment of the traditional saddle-shaped annulus from the posterior, D-shaped compartment. MA area and perimeter were computed for the D-shaped component by projection onto the least-squares plane fitted to the 3D annular contour. Total annular perimeter was calculated by adding the TT distance to the posterior 2D perimeter. The septal-to-lateral (SL) distance was defined as the projected distance from the TT line to the posterior peak and the intercommissural (IC) distance as the diameter perpendicular to the SL distance and parallel to the TT distance transecting the centroid of the MA. The IC/SL ratio was also calculated as a measure of overall MA geometry (Figure 1). Intraobserver and interobserver reproducibility of D-shaped MA measurements has been recently documented (5).
LV volumes and mass were measured using a threshold-based, region-growing, 3D segmentation algorithm (Aquarius iNtuition). Endocardial and epicardial contours of the LV were automatically detected with subsequent manual adjustment of the contours and level of the mitral valve plane. LV systolic and diastolic volumes were measured with subsequent calculation of LV stroke volume and ejection fraction. LA size was assessed at end-systole corresponding to maximal LA volume by using a semiautomated attenuation-based algorithm for endocardial border detection with manual correction (Aquarius iNtuition). LA volume excluded the LA appendage and pulmonary veins (12).
Continuous variables are expressed as mean ± SD and categorical variables as number (percentage). Cardiac volumes and MA dimensions are indexed to body surface area (BSA) calculated using the Mosteller formula (13). Indexed values were compared between controls and MR subgroups using an unpaired Student t test or Mann-Whitney U test as appropriate (normality determined using the Kolmogorov-Smirnov method) without adjustment for multiple comparisons.
For exploration of the determinants of the size of the D-shaped annulus beyond body size (a known correlate of MA dimensions ), and in particular to understand the relative contribution of changes in LA and LV sizes, univariate predictors of MA area indexed to BSA were evaluated in the control group using Pearson correlation. Multivariable linear regression was subsequently performed including univariate predictors (p < 0.10) in the model. In cases of significant correlation (R ≥ 0.60) between 2 covariates, the variable with more significant univariate association was included to avoid collinearity. Unstandardized and standardized beta coefficients are reported for individual variables, and the adjusted R2 is reported for the overall model. For assessment of the association between annular size (MA area index) and LA/LV volumes in patients with FMR or MVP, nonparametric (Spearman) correlation was performed, given the smaller sample size in these groups.
Statistical analyses were performed using GraphPad Prism V6.0d (GraphPad Software, La Jolla, California) and SPSS Statistics 22 (IBM Corp., Armonk, New York). A 2-tailed p value <0.05 was considered statistically significant.
Between August 2012 and February 2014, 163 patients underwent retrospectively gated cardiac CT of a total of 2,067 cardiac CT angiograms performed at our institution during that period. The scans were primarily performed for the evaluation of suspected coronary artery disease; 75 patients had coronary artery disease and were excluded from the study, leaving 88 patients in the control cohort. Eighty-five consecutive patients with moderate to severe MR being considered for TMVI were referred for cardiac CT between November 2013 and June 2015, 26 of whom were excluded, resulting in a total of 59 patients in the MR cohort (32 patients with MVP and 27 with FMR) (Figure 2).
Baseline characteristics for the control group are presented in Table 1. Age was 56 ± 11 years, and 47% of patients were female. LV and LA volumes and LV ejection fraction were consistent with reported values for healthy individuals (6).
In control subjects, mean MA area, MA area index, and IC/SL ratio were 8.9 ± 1.5 cm2, 4.7 ± 0.6 cm/m2, and 1.38 ± 0.14, respectively (Table 2). There was wide intersubject variability noted in MA area (Figure 3). Although annular dimensions were generally larger in men compared with women, the differences largely disappeared after values were indexed to BSA (Table 2).
MA dimensions correlated positively with BSA, as expected (Figure 4). Univariate and multivariate predictors of MA area index in controls are presented in Table 3. Age and sex were not associated with MA area index. Both LV and LA volumes were independently associated with MA area index, with LV systolic volume index (beta = 0.40; p < 0.001) slightly more predictive of MA area index than maximal LA volume index (beta = 0.31; p = 0.001).
MA dimensions and cardiac volumes among MR subgroups and controls are compared in Table 4. Controls were younger compared with both patients with MVP and FMR and had higher mean BMI and BSA values compared with patients with MVP. MA dimensions were generally larger in MR subgroups compared with controls, even after BSA was indexed to account for the differences in BSA. The range of absolute annular areas observed in patients with FMR and MVP is shown in Figure 5.
Annular geometry was also modified in patients with moderate to severe MR compared with controls. Whereas IC and SL distances were both increased, the IC/SL ratio was smaller in both patients with FMR and MVP compared with controls, indicating that annular remodeling in patients with MR involved relatively more SL (anteroposterior) rather than lateral (IC) expansion.
Differences were noted in annular dimensions between MR subgroups (Table 4). Despite LV volumes being significantly larger in patients with FMR compared with patients with MVP, mean MA area was 18% larger in the MVP group. IC/SL ratio was smaller in the FMR group compared with the MVP group. There were no significant differences in annular dimensions between ischemic and nonischemic subgroups of FMR (Online Table).
The associations between annular size and LV and LA volumes were discrepant in patients with MVP and FMR compared with controls (Figure 6). Increasing MA area index in patients with FMR was associated with increasing maximal LA volume index (R = 0.67; p < 0.001) but not with LV volumes. In contrast, MA area index in patients with MVP demonstrated a positive correlation with both LV systolic volume index (R = 0.48; p = 0.005) and maximal LA volume index (R = 0.48; p = 0.005).
In the present study, we reported D-shaped MA dimensions using CT in patients with moderate to severe MR being considered for TMVI and compare these findings with those of subjects without significant cardiac disease. Among controls, we noted wide interindividual variation in MA dimensions, as well as an independent positive association between MA size and LV and LA sizes. Annular dimensions were larger in patients with MR and annular geometry distorted with SL expansion. Although patients with MVP demonstrated a positive correlation between annular size and both LA and LV systolic volumes, in patients with FMR, annular size appeared to only be associated with increasing LA size. Importantly, annular dimensions were larger in MVP compared with FMR, which has implications for annular and therefore device sizing prior to TMVI.
Whereas traditional descriptions of MA geometry regard the annulus to be a saddle-shaped, nonplanar, 3D structure (15), the concept of a D-shaped MA was recently proposed as being more appropriate for TMVI sizing because it better reflects the planar landing zone of TMVI devices. Importantly, D-shaped MA segmentation eliminates the difficulties associated with defining and segmenting the anterior horn. Historically, there has been marked variation observed between surgical and noninvasive imaging definitions of the anterior horn, with surgeons typically estimating a less pronounced horn due to their ability to directly visualize and discriminate between atrial myocardium and fibrous tissue. For these reasons, the D-shape provides a more standardized and reproducible method of annular evaluation and is now being formally used to screen and provide MA sizes in patients being evaluated prior to TMVI (5,16).
MA measurements in control subjects without significant cardiac disease
Wide interindividual variation in MA dimensions was seen in control subjects, with a mean value of 8.9 ± 1.5 cm2, but ranging from 5.5 to 13.8 cm2. Prior studies of MA values have reported a broad range of normal values. Early 2D echocardiographic studies found relatively small annular areas, with one study reporting a mean maximal MA area of 7.1 cm2 (indexed 3.8 cm2/m2) (17). More recent studies using 3D echocardiographic techniques have reported normal mean values ranging from 8.4 to 11.8 cm2 (indexed 4.7 to 5.1 cm2/m2) (14,18,19). Cardiac CT studies reporting normative MA values have primarily assessed patients with MR, with only small cohorts of healthy subjects included as controls. Mean values for normal MA area in these studies have ranged from 8.4 to 10.2 cm2 (indexed 4.5 to 5.5 cm2/m2) (1–4,14). In the largest CT series to evaluate healthy subjects (n = 84), Delgado et al. (1) reported mean MA area, anteroposterior (SL), and IC distances of 4.8 ± 0.9 cm2/m2, 12.5 ± 2.1 mm/m2, and 21.6 ± 2.5 mm/m2, respectively. Our mean values for MA area are therefore at the lower end of reported values. This would be largely explained by the deliberate truncation of the anterior horn using our method; however, the broad range of values across studies, allowing for the clinical heterogeneity of study subjects, highlights the difficulties of and lack of standardized annular evaluation.
Sex differences in MA dimensions were also noted in control subjects, with men generally exhibiting larger absolute MA dimensions than women. After correction for BSA, however, these differences largely disappeared, except for SL distance, which was larger in women. Sonne et al. (14) similarly noted no difference in MA dimensions between men and women after correction for BSA, with the exception of medial-lateral MA diameter. In contrast, Mihaila et al. (19) noted larger indexed annular dimensions in men compared with women in a larger cohort of 224 healthy volunteers studied with 3D echocardiography. Combined, these data suggest that although sex-related differences in annular dimensions largely reflect differences in BSA between men and women, subtle differences may still exist beyond body size.
MA dimensions and geometry in patients with MR and the implications for device sizing
MA dimensions were larger in patients with MR, with MA area measuring 43% larger in patients with MVP and 24% larger in patients with FMR, compared with controls. Annular geometry was also distorted in patients with MR with greater SL expansion (reduced IC/SL ratio). Although these results are consistent with those of previous studies using the saddle-shaped annulus (1,3), our findings provide unique sizing and geometric information using the D-shaped annulus, which has implications for the development and implantation of TMVI devices. Of particular importance is the observation of larger MA sizes in patients with MVP compared with patients with FMR because this has implications in relation to the availability of appropriate device sizes for these clearly different conditions. The finding of larger MA dimensions in patients with MVP is consistent with previous 3D echocardiographic studies (20) and is thought to be due to increased outward tension on the MA during systole in the setting of excessive mitral valve tissue (21). Apart from this pathophysiological mechanism, larger annular dimensions in patients with MVP may result from displacement of the posterior segmentation line into the LA due to disjunction of the mitral valve leaflet insertion from the atrioventricular junction (Figure 7), which has been previously reported in association with myxomatous mitral valve disease (22,23). Segmentation of the annulus to reflect the anticipated landing zone as opposed to the anatomic site of MV leaflet insertion is important in the setting of MA disjunction, although further investigation is needed to better understand the impact of disjunction on device sealing and capture.
Determinants of annular size in controls and patients with MR
Age was not associated with changes in MA size in control subjects, consistent with previous studies (14). There was also no association between sex and MA size, likely due to the correction for BSA as discussed above. We did, however, observe an association between MA size and both LA and LV sizes, suggesting independent contribution from both chambers to annular size in controls. These findings were different than the observations in patients with MR. In patients with MVP, a positive correlation between MA area index and both LV and LA sizes was seen; however, in patients with FMR, MA area index was positively associated with LA volume only. The cause of FMR has been attributed to various mechanisms, including systolic leaflet tethering to displaced papillary muscles in a remodeled LV (24), abnormal LV systolic function and shape (25), and annular remodeling (26). Interestingly, in a recent study of patients with MR but structurally normal mitral valves, an independent association between LA enlargement and annular dilation was observed (27), similar to our study, irrespective of the presence of LV dilation and dysfunction.
MA dimensions were measured at mid to late diastole; however, prior studies have reported intracycle variation in MA area, with larger values previously observed in diastole (2,18). Patients with MR were being considered for TMVI because they were deemed unfit for surgery. Similarly, controls were only included if they underwent CT with retrospective ECG gating, which represented a small proportion of the overall sample. These referral and selection biases limit the applicability of our data to similar patients. Hypertensive patients were not excluded from the controls cohort (history of hypertension present in 44%), and this condition may be associated with diastolic dysfunction and therefore atrioventricular remodeling. However, patients with increased LA volume index, which is sensitive for the detection of severe diastolic dysfunction (28), and increased LV mass index were excluded. Moreover, we reviewed the electronic charts of controls to ensure that subjects with an echocardiographic report did not have evidence of severe diastolic dysfunction, minimizing the impact of such changes on our results. Although CT provides 3D data with high spatial resolution, it is somewhat limited by its ionizing radiation. Although the present study provides important insights into potential interactions between atrioventricular and annular remodeling, our results should be interpreted cautiously due to the small sample sizes, particularly in the FMR cohort. Moreover, the ability of D-shaped annular segmentation to appropriately size the annulus for the purposes of device selection and the subsequent impact on clinical outcomes needs to be addressed in future studies.
Significant interindividual variability in D-shaped MA dimensions was seen in patients without significant cardiac disease. Among patients with moderate to severe MR, significant MA enlargement was observed and was associated with SL (anteroposterior) MA expansion. Importantly, patients with MVP exhibited larger MA dimensions than patients with FMR, and the drivers of annular enlargement appeared to be different in these cohorts, with LA dilation contributing more significantly in patients with FMR. Our findings provide insights into the size and geometry of the D-shaped annulus and the drivers of MA dilation in patients being considered for transcatheter mitral therapy.
COMPETENCY IN MEDICAL KNOWLEDGE: The D-shaped approach to MA segmentation is considered a more biomechanically appropriate method currently used to screen and size patients being considered for transcatheter mitral valve therapy. Patients with moderate to severe MR exhibit larger D-shaped MA dimensions, as well as annular remodeling characterized by SL expansion, than control subjects, especially patients with MVP. Unlike in controls and patients with MVP, D-shaped annular enlargement in FMR is more closely associated with LA dilation.
TRANSLATIONAL OUTLOOK: Further studies are needed to elucidate the impact of this knowledge of D-shaped MA dimensions, geometry, and drivers of annular size on device performance and clinical outcomes following TMVI.
Dr. Leipsic has served as a consultant to Edwards Lifesciences and Neovasc Inc.; and has provided CT core laboratory services to Edwards Lifesciences, Neovasc Inc., and Tendyne Holdings Inc. Dr. Cheung has served as a consultant to Edwards Lifesciences and Neovasc Inc. Drs. Ye and Webb have served as consultants to Edwards Lifesciences. Dr. Piazza has served on scientific advisory boards for Medtronic; has served as a consultant for HighLife SAS; and owns equity shares in HighLife SAS. Dr. Blanke has served as a consultant to Edwards Lifesciences, Neovasc Inc., Tendyne Holdings Inc., and Circle Imaging; and has provided CT core laboratory services to Edwards Lifesciences, Neovasc Inc., and Tendyne Holdings Inc. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- body mass index
- body surface area
- computed tomography
- functional mitral regurgitation
- left atrial/atrium
- left ventricle/ventricular
- mitral annular/annulus
- mitral regurgitation
- mitral valve prolapse
- transcatheter mitral valve implantation
- Received July 2, 2015.
- Revision received August 18, 2015.
- Accepted August 20, 2015.
- American College of Cardiology Foundation
- Delgado V.,
- Tops L.F.,
- Schuijf J.D.,
- et al.
- Beaudoin J.,
- Thai W.E.,
- Wai B.,
- Handschumacher M.D.,
- Levine R.A.,
- Truong Q.A.
- Lin F.Y.,
- Devereux R.B.,
- Roman M.J.,
- et al.
- Nishimura R.A.,
- Otto C.M.,
- Bonow R.O.,
- et al.
- Blanke P.,
- Dvir D.,
- Cheung A.,
- et al.
- Levine R.A.,
- Triulzi M.O.,
- Harrigan P.,
- Weyman A.E.
- Ormiston J.A.,
- Shah P.M.,
- Tei C.,
- Wong M.
- Grewal J.,
- Suri R.,
- Mankad S.,
- et al.
- Yiu S.F.,
- Enriquez-Sarano M.,
- Tribouilloy C.,
- Seward J.B.,
- Tajik A.J.
- Kono T.,
- Sabbah H.N.,
- Rosman H.,
- Alam M.,
- Jafri S.,
- Goldstein S.
- Kwan J.,
- Shiota T.,
- Agler D.A.,
- et al.
- Ring L.,
- Dutka D.P.,
- Wells F.C.,
- Fynn S.P.,
- Shapiro L.M.,
- Rana B.S.
- Pritchett A.M.,
- Mahoney D.W.,
- Jacobsen S.J.,
- Rodeheffer R.J.,
- Karon B.L.,
- Redfield M.M.