Author + information
- Received March 2, 2016
- Revision received June 3, 2016
- Accepted June 7, 2016
- Published online September 13, 2016.
- Joon Bum Kim, MD, PhDa,
- Matthew Spotnitz, MDb,f,
- Mark E. Lindsay, MD, PhDc,d,e,
- Thomas E. MacGillivray, MDb,e,
- Eric M. Isselbacher, MDc,e and
- Thoralf M. Sundt III, MDb,e,∗ ()
- aDepartment of Thoracic and Cardiovascular Surgery, Asan Medical Center, University of Ulsan College of Medicine, Seoul, South Korea
- bDivision of Cardiac Surgery, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts
- cCardiology Division, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts
- dPediatric Cardiology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts
- eThoracic Aortic Center, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts
- fHarvard T.H. Chan School of Public Health, Cambridge, Massachusetts
- ↵∗Reprint requests and correspondence:
Dr. Thoralf M. Sundt III, Massachusetts General Hospital, 55 Fruit Street, Cox 652, Boston, Massachusetts 02114.
Presented at the American Association for Thoracic Surgery Aortic Symposium Workshop 2015, Kobe, Japan, October 16, 2015.
Background Recent studies have demonstrated that many patients with acute type A aortic dissection (AD) have aortic diameters of <55 mm at presentation, prompting discussion of lowering the prophylactic surgical guidelines. However, risk of dissection at these smaller diameters is poorly defined.
Objectives The purpose of this study is to understand the risk of AD in moderately dilated ascending aortas using a large echocardiographic data set.
Methods Using an institutional echocardiography database, we identified 4,654 nonsyndromic adults (age: 68.6 ± 13.1 years; 1,003 women) with maximal ascending aortic diameters of 40 to 55 mm. We performed competing risk analysis to determine the independent risk factors of AD or aortic rupture.
Results Five hundred eighty-six individuals (12.6%) had bicuspid aortic valves (BAVs). During follow-up (14,431.5 patient-years), AD and rupture occurred in 13 and 1 patients, respectively, which demonstrated a linearized incidence of AD and/or rupture of 0.1% per patient-year. Elective ascending aortic repair was performed in 176 individuals. On multivariable analyses, independent predictors of AD and/or rupture were age (hazard ratio [HR]: 1.06; 95% confidence interval [CI]: 1.01 to 1.12; p= 0.024) and baseline aortic diameters (HR: 1.20; 95% CI: 1.05 to 1.36; p = 0.006). The presence of a BAV was not a significant factor (HR: 0.94; 95% CI: 0.10 to 8.40; p = 0.95). Estimated risks of AD and/or rupture within 5 years were 0.4%, 1.1%, and 2.9% at baseline aortic diameters of 45, 50, and 55 mm, respectively.
Conclusions Risks of AD and/or rupture were significantly correlated with the aortic diameter and age in patients with moderately dilated ascending aortas. However, the risks were low for diameters <5.0 cm when timely elective aortic repair was performed, regardless of the morphology of the aortic valve.
Dissection of the ascending thoracic aorta (aortic dissection [AD]) is a catastrophic medical illness. More than 50% of patients die during the acute phase of AD if not treated surgically, and of those treated, mortality ranges from 5% to 30% despite the significant improvements in operative strategies that have occurred during the last few decades (1–4). The risk of AD is known to correlate with increasing aortic diameter. Observational studies have suggested an inflection point at a diameter of 6 cm for the risk of AD, and accordingly, current practice guidelines have recommended elective aortic repair at 5.5 cm to prevent this lethal complication (5–7). However, recent studies have demonstrated that many patients with acute type A AD have aortic diameters of <55 mm at presentation, which has prompted discussion of lowering the threshold for prophylactic aortic repair (8,9). Furthermore, it has been suggested that the threshold should be even lower for individuals with bicuspid aortic valves (BAVs), as reflected in the most recent practice guidelines (10–13).
Concerns have been raised regarding this more aggressive approach to prophylactic surgery for the dilated ascending aorta (14,15). Surgical therapy should only be offered when the risk of the natural history of the condition exceeds surgical risk. Although calculating these risks requires an understanding of the true number of individuals at risk (the denominator), most surgical studies on acute AD and all autopsy studies capture only the adverse events (the numerator) of these populations, and accordingly, the true risk–benefit ratio of surgical intervention for the moderately dilated ascending aorta has only been sparsely evaluated (9,11,13). Therefore, we sought to understand the risk of AD in the moderately dilated ascending aorta using a large echocardiographic data set. We further explored the expansion rate in proportion to baseline diameter, with particular attention to whether it was different among individuals with BAVs.
Study subjects and data collection
We used the institutional echocardiography database of the Massachusetts General Hospital to identify adult individuals (age 17 years or older) who were identified by transthoracic echocardiography to have ascending aortic dilatation of 40 to 55 mm in maximum diameter between November 2001 and February 2014. This cohort was not only confined to those who were encountered in the aortic clinic, but it also included all individuals encountered at Massachusetts General Hospital. Evaluation of the ascending aorta was a routine part of echocardiographic examinations during the study, with each of the largest internal diameters of tubular and sinus portions of the ascending aorta routinely obtained through the parasternal long-axis view and recorded.
Exclusion criteria were documented connective tissue disorders (Marfan, Loeys-Dietz, and Ehlers-Danlos syndromes), inflammatory aortic diseases, presentation with AD and/or rupture, or a history of ascending aortic surgery. Those with critical illness or those who were scheduled to receive elective ascending aortic surgery were excluded to allow for obtaining adequate follow-up information. In addition, patients who underwent heart transplantation or implantation of ventricular assist devices were excluded.
Because the aim of the study was to evaluate the prognosis of unrepaired ascending aortic aneurysms, the primary endpoints were AD or rupture in the ascending aorta. During the follow-up interval, elective repair of the ascending aorta was generally aggressive and in accordance with practice guidelines. Because the decision to intervene depends significantly on the attitudes and opinions of the care providers, this was considered a secondary endpoint. Follow-up clinical information up to 5 years from the index echocardiographic assessment was obtained through January 2015 by a retrieval of longitudinal data from Partners Health Care system. This centralized clinical data registry, which covers the largest health care system in Massachusetts, maintains all clinical information of individuals encountered within the system (16).
R statistical software (version 3.2.2, R, Vienna, Austria) was used for data analyses (17). Categorical variables were presented as frequencies and percentages, and were tested by the chi-square test or Fisher exact test for intergroup comparisons. Continuous variables were expressed as mean ± SD or the median with range, and were compared using Student unpaired t test or the Mann-Whitney U test. Kaplan-Meier methods were used to delineate the conditional probability of primary or secondary endpoints, and log-rank tests were used to compare intergroup differences in the rates.
Adverse aortic events (AD or rupture), elective aortic surgery, and deaths from other causes were regarded as competing risks in survival analysis in this study. To adequately address the competing nature of different endpoints in survival models, we performed multivariable competing risk analysis, as suggested previously (18,19).
Variables with a p value of ≤0.20 in univariable competing risk analyses were candidates for the multivariable models. Multivariable analyses involved a stepwise backward elimination technique, and only variables with a p value of <0.10 were used in the final model. We used a logistic regression model to estimate risks of “within 5-year aortic events” based on significant predictors revealed in the multivariable model.
In addition, in the interest of evaluating predictors of the rate of aortic expansion, we identified a subset of individuals who had subsequent echocardiographic assessments of the native ascending aorta over ≥12-month intervals. The aortic expansion rate was calculated based on the differences in aortic diameter between the baseline and latest follow-up measurements divided by the interval between the two, which is presented as millimeter per year (20–22).
All baseline parameters were examined in a univariable linear regression model to evaluate their associations with the aortic expansion rate. Then, multivariable linear regression analyses were conducted that involved only variables with p values <0.20 in the univariable models, and a stepwise method was used to leave covariates with p values <0.10 in the final model. Finally, significant predictors of aortic expansion revealed in the multivariable linear model were also examined in a nonparametric Loess regression model. To investigate the association between blood pressure and aortic expansion rates, we assessed all blood pressure measurement data obtained between the index and latest echocardiography examination, and the time-weighted averages of systolic, diastolic, and mean blood pressures, and pulse pressure were obtained based on intervisit periods for blood pressure measurements. These time-weighted average data were used in regression models. All reported p values were 2-sided, and a value of p < 0.05 was considered statistically significant.
A total of 6,437 individuals met the initial enrollment criteria. Among these, 1,783 individuals were subject to exclusion criteria, leaving a final 4,654 nonsyndromic adults (age: 68.6 ± 13.1 years; 1,003 women) to form the study group. As shown in Figure 1, maximal aortic diameters were 40 to 44 mm in 4,016 (86.3%), 45 to 49 mm in 510 (11.0%), and 50 to 55 mm in 28 (2.8%) individuals. Of these, 586 individuals (12.6%) had BAVs. At baseline, those with BAVs were younger, had larger ascending aortas (p < 0.001) (Table 1), and had less of a cardiovascular risk burden, with a lower prevalence of diabetes mellitus, hypertension, cerebrovascular disease, coronary disease, peripheral artery disease, and aneurysm in the thoracoabdominal aorta compared with those with tricuspid aortic valves (TAVs). Patients with BAVs more frequently presented with moderate-to-severe aortic stenosis and showed larger ascending aortic sizes both in the sinus and tubular portions than the individuals with TAVs.
Follow-up was complete in 82.2% (3,825 of 4,654) of individuals, with a median time of 40.1 months (25th to 75th percentiles: 16.3 to 60.0 months), which included 14,431.5 patient-years of follow-up. Those who were lost to follow-up (n = 829) were significantly older (age 69.9 ± 13.5 years vs. 68.4 ± 13.0 years; p = 0.003), with a greater proportion of women (24.2% vs. 21.0%; p = 0.037) and less hypertensive individuals (36.2% vs. 50.2%; p < 0.001) compared with those with complete follow-up (n = 3,825). However, there were no significant differences in the prevalence of BAVs (11.6% vs. 13.6%; p = 0.13), aortic diameters (42.0 ± 2.5 mm vs. 42.0 ± 2.6 mm; p = 0.89), and proportions of aortic diameters ≥50 mm (2.8% vs. 2.7%; p = 0.96).
During follow-up, AD and rupture occurred in 13 and 1 patients, respectively, at a median 738 days from baseline (range: 405 to 1,800 days), demonstrating a linearized incidence of AD and/or rupture of 0.1% per patient-year for the group as a whole (Table 2). Among the 13 patients with subsequent AD, 12 presented in the acute phase of AD, whereas 1 patient was diagnosed with chronic AD of the ascending aorta that was incidentally discovered on a routine follow-up echocardiographic evaluation performed at 738 days after the index evaluation. Further details on these 14 patients who had adverse aortic events are available in Online Table 1. Figure 2 depicts cumulative incidence rates of AD and/or rupture based on baseline aortic diameter, age, and morphology of the aortic valve (AV). The aortic events rates were significantly affected by the aortic diameter (p < 0.001) and age (p = 0.029), but not by the BAV (p = 0.49) when evaluated by log-rank tests.
During follow-up, elective ascending aortic repair was performed in 176 patients; in 83 patients, this was concomitant with AV replacement, among whom the primary indication for surgical intervention was valve dysfunction in 67 patients. Among the 176 patients who underwent elective aortic repair, there were an additional 50 cases who underwent the Bentall operation (n = 37) or the aortic root reimplantation procedure (n = 13).
The composite risk of aortic events, including elective aortic surgery, was significantly affected by the baseline ascending aortic diameter (p < 0.001) and the presence of moderate-to-severe AV dysfunction (either insufficiency or stenosis; p < 0.001) (Online Figure 1).
Risk factors of adverse aortic events
On multivariable analyses, significant and independent predictors of AD and/or rupture were increasing age (HR: 1.75; 95% CI: 1.04 to 2.95; p = 0.034) and increasing baseline aortic diameters (HR: 1.20; 95% CI: 1.05 to 1.36; p = 0.006), whereas hypertension was positively associated but was not statistically significant (p = 0.086) (Table 3). In contrast, BAVs were not significantly associated with AD and/or rupture even in the univariable analyses (Table 3).
Independent predictors of elective aortic repair were larger baseline aortic diameters (HR: 1.20; 95% CI: 1.16 to 1.25; p < 0.001), moderate-to-severe AV dysfunction (HR: 4.21; 95% CI: 2.83 to 6.25; p < 0.001), and the presence of BAV (HR: 5.88; 95% CI: 4.05 to 8.52; p < 0.001) (Table 3).
Probabilities of experiencing AD and/or rupture and elective ascending aortic surgery within 5 years based on baseline ascending aortic diameter estimated by logistic regression models are shown in Figure 3. For the evaluation of the primary endpoint (AD and/or rupture) in these models, data were available for 1,615 individuals who were followed for 5 years from baseline (n = 1,598) or until AD and/or rupture (n = 14). For the endpoint of aortic surgery, 1,774 individuals were followed for 5 years (n = 1,598) or until elective surgery (n = 176). The estimated probabilities of AD and/or rupture in the ascending aorta were 0.4%, 1.1%, and 2.9% at the baseline aortic diameters of 45, 50, and 55 mm, respectively. Estimated probabilities of elective aortic repair within 5 years were 6.4%, 21.2%, and 51.4% at the aortic diameters of 45, 50, and 55 mm, respectively (Figure 3).
The associations among estimated risk of AD and/or rupture within 5 years, patient age, and baseline aortic diameter are illustrated in three-dimensional plots in Figure 4, in which the risk of ascending AD and/or rupture was synergistically affected by the 2 factors. The estimated 5-year risk of AD and/or rupture was 0.6%, 1.6%, and 4.0% in those aged 40, 60, and 80 years, respectively (Figures 4B to 4D) for the initial diameter of 5.5 cm, and 0.2%, 0.6%, and 1.6%, respectively, for the initial diameter of 5.0 cm.
Aortic expansion rate
Follow-up echocardiographic assessments beyond 12 months from the index visits were available for 1,414 individuals (65.8 ± 13.4 years; 289 women), including 270 individuals (19.0%) with BAVs. Baseline mean ascending aortic diameter of this cohort was 42.4 ± 2.7 mm. Overall, data from 4,740 echocardiographic assessments were retrieved (3.4 per patient) during a median follow-up of 46.9 months (25th to 75th percentiles: 25.7 to 60.0 months) to calculate the aortic expansion rate.
On crude comparisons, the aortic expansion rates were significantly greater in individuals with BAVs compared with those with TAVs (p < 0.001) (Online Figure 2A), and the greater expansion rates of the BAV over TAV were consistent throughout all of the index aortic diameters (Online Figure 2B).
On multiple linear regression models (R2 = 0.016; p < 0.001), female sex (beta = 0.088; p = 0.001) and BAV (beta = 0.065; p = 0.013) were significant and independent risk factors of aortic expansion (Table 4), whereas the baseline aortic diameter was inversely associated with the aortic expansion rate (beta = −0.074; p = 0.005) (Figure 5A). Loess models demonstrated a superior model fitness (R2 = 0.070) than parametric models that demonstrated a mixed relationship between baseline aortic diameter and aortic expansion rate; a positive correlation existed only in larger (>50 mm) aortas (Figure 5B). Systolic, diastolic, and mean blood pressures during follow-up (average 11.9 measurements per patient) were not predictive of aortic expansion rates (p = 0.52 to 0.95) (Online Figure 3).
The risk of AD in individuals with moderate dilation of the ascending aorta is clinically important, but poorly quantified. This study confirms the increased risk of AD and/or rupture with increasing aortic diameter among individuals with moderately dilated (40 to 55 mm) ascending aortas; however, this risk is quite low. Furthermore, these data suggest that the presence of a BAV had no significant impact on the risk of AD and/or rupture, although it did appear as a risk factor (along with female sex) for aortic expansion. Baseline aortic diameter showed a mixed relationship with the aortic expansion rate.
The correlation between aortic diameter and risk of aortic catastrophe is consistent with the findings of previous studies (6,7,23–26). Based on these observations, practice guidelines used aortic diameter as the principal driver of prophylactic thoracic aortic repair, with a recommended threshold of 55 mm for nonsyndromic individuals. This threshold has been challenged by more recent studies that demonstrated that a significant proportion of patients present with acute AD at smaller aortic diameters (<55 mm). The rationale for earlier intervention has been suggested by a recent study from the IRAD (International Registry of Acute Aortic Dissection) (9). Among 591 patients who presented with acute type A AD in this IRAD study, 59% had an ascending aortic diameter of <5.5 cm at presentation. This finding could suggest that current surgical guidelines for elective aortic aneurysm repair might fail to prevent many acute ADs; therefore, the premise for the guideline should be reexamined. For patients with BAVs in particular, a recent statement from the American College of Cardiology/American Heart Association in 2016 clarified that elective repair of the 4.5- to 5.5-cm sized ascending aorta is a reasonable treatment option (Class IIa recommendation), whereas replacing an ascending aorta of ≥5.5 cm in diameter is indicated as class I recommendation (27).
Accordingly, there has been enthusiasm for lowering the threshold for elective aortic repair. The lack of population studies encompassing the denominator of individuals at risk has been regarded as the most critical limitation to determine the true benefits of a more aggressive surgical approach. The present study, although not truly population-based, aimed to more broadly assess the risks of AD and/or rupture among patients identified by echocardiography performed for any reason, with a focus on those with intermediate dilatation for whom there was a greatest disagreement. Our findings of a low risk—an incidence of only 1 per 1,000 patient-years—correlate well with a recent study on AD risks in BAV by Michelena et al. (28). In that study, the incidence of AD among individuals with BAVs during long-term follow-up was only 3.1 cases per 10,000 patient-years. This risk sets a high bar for surgical results if the treatment is to be less dangerous than the disease.
Finally, our observation that BAV did not appear to increase risk of dissection independent of diameter is also consistent with findings in a previous study by Davies et al. (26). In this study, individuals with BAVs showed higher aortic growth rate (p = 0.0102), but the incidence of AD and rupture were similar to the TAV group (p = 0.43 and p = 0.84, respectively). However, our findings and those of Davies et al. (29) challenged the current trend to set the threshold for intervention lower for those with BAVs. Based on autopsy studies that showed higher incidence of BAV among cases of AD, the most recent practice guidelines recommended lower surgical threshold for prophylactic aortic repair in BAV (10). However, our study findings contradict this recent movement. Our findings might also be biased by elective surgical intervention because the institutional approach was to be more aggressive with BAVs, as recommended in the 2010 guidelines. Accordingly, we observed more surgical procedures among patients with BAVs. To minimize the impact of the competing nature of elective aortic intervention on AD and/or rupture, we used competing risk analyses. Using this approach, no evidence for increased risk of AD and/or rupture in BAV was evident, even after an adjustment by multivariable competing risk analyses.
The seemingly low risk of AD in smaller ascending aortic aneurysms (40 to 55 mm) in the present study might actually be an overestimate of the risk in the general population. The incidence of AD from community-based studies was reported to be 1 to 3 per 100,000 patient-years (30–32). Based on our data, the linearized incidence of AD and/or rupture was approximately 20 per 100,000 patients-years at the diameter of 40 mm, which was an almost 10-fold risk compared with the general population. This risk discrepancy in small ascending aneurysms compared with the general population might in part be explained by the nature of individuals in the study. The study subjects were not selected at random from the general population, but rather underwent echocardiography at the Massachusetts General Hospital for a clinical indication. Recently, community-based evaluations of the entire aorta were conducted as a part of the Framingham Heart Study (33). A total 3,431 adults were evaluated by computed tomography for aortic diameters in that study, with resultant mean diameters of the ascending aorta of 34.1 ± 3.9 mm for men and 31.9 ± 3.5 mm for women. The largest ascending aortic diameters among these 3,431 individuals were 49.6 mm, indicating that an ascending aortic diameter >50 mm would be observed only in <0.03% of the general population. Assuming a normal distribution of the aortic size, an ascending aorta of 40 mm corresponds to the 93.45 percentile for men and 98.96 percentile for women (33). Therefore, the patients included in the present study were biased, as one would expect, in favor of those with cardiovascular pathology and were likely a higher risk group. Therefore, the risks estimated might be considered typical for a cardiovascular clinic population.
One of the limitations of current guideline recommendations is that, beyond valve phenotype or presence of a connective tissue disorder, they do not take into account patient specific details, such as age or presence of hypertension. Surprisingly, hypertension was not positively associated with either aortic expansion or aortic failure. When stratified by age, our data showed a particularly low risk of AD in younger individuals. Accordingly, in order for a prophylactic operation to be of benefit for a patient younger than 65 years of age with a 4.5-cm aorta, the operative risk would need to be much <0.3%, the estimated risk of AD within 5 years by our model (Central Illustration). This is a particularly challenging surgical threshold, especially for those patients with functionally normal or minimally abnormal BAVs for whom the decision whether to intervene on the valve is most challenging. It seems clear that a substantial minority of patients develop acute AD at diameters <5.5 cm, whereas the aortic size alone may not be sufficient to predict the risk in all cases. In this regard, other factors such as flow dynamics and innate properties of the aortic wall also need to be evaluated in future studies for clearer understanding on the contributors of aortic complications, especially in those with BAVs.
Finally, the R2 value of the final Loess regression model was 0.070, which indicated that the model only explained 7% of the variability in aortic remodeling. This finding suggested that the relationship between conventional parameters (age, sex, BAV) and aortic remodeling was weak, and that other potential parameters (e.g., innate aortic properties) should also be considered in the future studies. Ideally, follow-up assessments on aortic sizes should be performed at pre-defined time points in such studies in a prospective manner to overcome the limitations of variability in the follow-up intervals shown in the present study.
This study was not truly population-based research and might be biased by referral from the community and by the selection of the echocardiographic evaluations. Furthermore, the true natural history of the moderately dilated aorta could not be fully assessed despite the use of competing risk models, because individuals underwent elective aortic intervention at the discretion of their physicians. Follow-up (82.2%) was good, but incomplete.
Determining significant predictors of AD and/or rupture was based on 14 events, which was likely too small a number to allow adequate multivariable analyses. Therefore, studies with a larger event rate are needed to further verify the study results.
Evaluations on aortic remodeling based on follow-up imaging at various time points are challenging. For this, we used a method of dividing the absolute differences in the aortic diameters between the baseline and latest evaluations by follow-up intervals, which was most commonly used in the literature (20–22). Although this method is simple and easily reproducible, all the variations in the measurement data during follow-up were blunted by averaging the data, by which the results might not completely reflect all the measurements obtained during the observed periods.
The clinical implications of these findings suggest that prophylactic aortic repair in nonsyndromic individuals with an ascending aorta <5.0 cm should be considered sparingly, regardless of the morphology of the AV. Sex and BAV were predictive of aortic expansion over time, whereas baseline aortic diameter showed mixed associations with aortic expansion, with a positive correlation only in larger (>50 mm) aortas. Because the statistical models only explain 7% (R2 = 0.070) of the variability in aortic remodeling, further research is needed to determine more robust indicators of aortic expansion.
COMPETENCY IN MEDICAL KNOWLEDGE: The risk of dissection in the ascending aorta is positively associated with the aortic diameter and age; however, the absolute risk is low in moderately dilated aortas (40 to 55 mm).
TRANSLATIONAL OUTLOOK: Further large-scale, population-based research is needed to validate the risk of aortic dissection based on various risk variables, including aortic size.
For supplemental tables and figures, please see the online version of this article.
This study was funded by the American Association for Thoracic Surgery’s Evarts A. Graham Foundation. Dr. Kim is sponsored by the American Association for Thoracic Surgery’s Evarts A. Graham Memorial Traveling Fellowship.
Dr. Sundt has served as a consultant for Thrasos. The other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Acronyms and Abbreviations
- aortic dissection
- aortic valve
- bicuspid aortic valve
- confidence interval
- hazard ratio
- tricuspid aortic valve
- Received March 2, 2016.
- Revision received June 3, 2016.
- Accepted June 7, 2016.
- 2016 American College of Cardiology Foundation
- Pape L.A.,
- Awais M.,
- Woznicki E.M.,
- et al.
- Wang W.,
- Duan W.,
- Xue Y.,
- et al.
- Kruger T.,
- Weigang E.,
- Hoffmann I.,
- Blettner M.,
- Aebert H.
- Pape L.A.,
- Tsai T.T.,
- Isselbacher E.M.,
- et al.
- Hiratzka L.F.,
- Bakris G.L.,
- Beckman J.A.,
- et al.
- ↵The R Project for Statistical Computing. Available at: https://www.R-project.org. Accessed May 1, 2016.
- Kim J.B.,
- Kim K.,
- Lindsay M.E.,
- et al.
- Hiratzka L.F.,
- Creager M.A.,
- Isselbacher E.M.,
- et al.
- Nienaber C.A.,
- Powell J.T.