Author + information
- Received May 26, 2010
- Revision received July 28, 2010
- Accepted July 30, 2010
- Published online October 1, 2010.
- Mackram F. Eleid, MD,
- Giuseppe Caracciolo, MD,
- Eun Joo Cho, MD,
- Robert L. Scott, MD,
- D. Eric Steidley, MD,
- Susan Wilansky, MD,
- Francisco A. Arabia, MD,
- Bijoy K. Khandheria, MD and
- Partho P. Sengupta, MD, DM⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Partho P. Sengupta, Division of Cardiovascular Diseases, University of California Irvine, 101 The City Drive, City Tower, Irvine, California 92668
Objectives The aim of this study was to explore the temporal evolution of left ventricular (LV) mechanics in relation to clinical variables and genetic expression profiles implicated in cardiac allograft function.
Background Considerable uncertainty exists regarding the range and determinants of variability in LV systolic performance in transplanted hearts (TXH).
Methods Fifty-one patients (mean age 53 ± 12 years; 37 men) underwent serial assessment of echocardiograms, cardiac catheterization, gene expression profiles, and endomyocardial biopsy data within 2 weeks and at 3, 6, 12, and 24 months after transplantation. Two-dimensional speckle-tracking data were compared between patients with TXH and 37 controls (including 12 post–coronary artery bypass patients). Post-transplantation mortality and hospitalizations were recorded with a median follow-up period of 944 days.
Results Global longitudinal strain (LS) and radial strain remained attenuated in patients with TXH at all time points (p < 0.001 and p = 0.005), independent of clinical rejection episodes. Failure to improve global LS at 3 months (≥1 SD) was associated with higher incidence of death and cardiac events (hazard ratio: 5.92; 95% confidence interval: 1.96 to 17.91; p = 0.049). Multivariate analysis revealed gene expression score as the only independent predictor of global LS (R2 = 0.53, p = 0.005), with SEMA7A gene expression having the highest correlation with global LS (r = −0.84, p < 0.001).
Conclusions Speckle tracking–derived LV strains are helpful in estimating the burden of LV dysfunction in patients with TXH that evolves independent of biopsy-detected cellular rejection. Failure to improve global LS at 3 months after transplantation is associated with a higher incidence of death and cardiac events. Serial changes in LV mechanics correlate with peripheral blood gene expression profiles and may affect the clinical assessment of long-term prognosis in patients with TXH.
Heart transplantation is an effective therapy for patients with end-stage heart failure, with 1-year post-transplantation survival approaching 90% (1). However, most grafts succumb slowly over time, with only 50% of patients surviving at 10 years. Although cardiac retransplantation remains a therapeutic option, the availability of donor hearts remains limited, and the success of cardiac retransplantation depends on the duration of survival of the first allograft (2). Standard measures of left ventricular (LV) systolic function, such as ejection fraction and volumes, are commonly used to monitor graft function, but tend to be stable over time and do not correlate with biopsy-proven rejection or time after transplantation (3). The assessment of LV strain by speckle tracking has recently emerged as an angle-independent, noninvasive variable for characterizing regional and global LV function from routinely acquired grayscale cardiac ultrasound images (4,5), and shows excellent correlation with LV strain measured by sonomicrometry and cardiac magnetic resonance (6–8). Although speckle-tracking strain has been used in a growing number of clinical situations (9), its utility in the serial assessment of LV mechanical function in transplanted hearts remains uncertain.
Considerable uncertainty exists regarding the range and prognostic significance of the LV ejection fraction in transplanted hearts (TXH) (10–13). Factors that determine LV function in the post-transplantation setting include the alloimmune response, development of transplant vasculopathy, and a variety of nonimmune risk factors such as post-surgical sympathetic denervation, hyperlipidemia, hypertension, diabetes mellitus, hyperhomocysteinemia, advancing age, post-transplantation infections, and donor variables including age, LV hypertrophy, and gender mismatch (1,14–19). The use of gene expression profiling of peripheral blood specimens for characterizing host immune responses (20) in combination with clinical and echocardiographic assessments of LV function has been suggested as a conservative approach to the care of heart transplant recipients (21). However, there is little information to link clinical variables and genetic expression profiles with serial changes in LV function seen in patients with TXH.
The goals of this study were to: 1) understand the temporal evolution of LV longitudinal, circumferential, and radial mechanics as a marker of graft function in patients with TXH; and 2) identify a potential link between gene expression profiles and serial changes in LV mechanics for characterizing the role of chronic immune and inflammatory activation in cardiac allograft dysfunction.
The Mayo Clinic Institutional Review Board approved this study. Between October 20, 2005, and September 18, 2008, 51 consecutive patients (mean age 53 ± 12 years; 37 men) who underwent heart transplantation at our institution were enrolled in the study. Serial clinical and 2-dimensional (2D) echocardiographic data were obtained at 2 weeks and at 3, 6, 12, and 24 months in 45, 48, 44, 31, and 7 patients, respectively, after transplantation. Clinical data including demographics, comorbid conditions, laboratory results, cardiac events, rehospitalization, and death were recorded for each patient. Cardiac events were defined as episodes of heart failure, rejection, or acute coronary syndromes requiring hospitalization. Hospitalizations, cardiac events, and all-cause mortality were recorded from the start of enrollment (October 20, 2005) through April 25, 2010. The median length of follow-up was 944 days.
Twenty-five healthy subjects with similar age and gender to the study population (mean age 47 ± 15 years; 12 men) who had normal echocardiographic results served as controls. To compare post-operative LV mechanics in patients with TXH with those of nontransplanted post–cardiac surgery patients, 12 patients with normal LV function undergoing coronary artery bypass surgery at our institution (mean age 53 ± 9 years; 9 men) had LV mechanics analyzed on echocardiography at least 1 month (mean 125 ± 119 days) after their surgery to serve as post-operative controls.
Echocardiographic studies were performed on commercially available ultrasound equipment (Acuson Sequoia, Siemens Medical Solutions USA, Mountain View, California; and Vivid 7, GE Healthcare, Milwaukee, Wisconsin) according to the standard method recommended by the American Society of Echocardiography (22), and digital images were obtained at optimal frame rates (≥30 frames/s). Images were stored in digital cine loop format (ProSolv Cardiovascular Solutions, Indianapolis, Indiana) for offline analysis by vendor-customized 2D Cardiac Performance Analysis software (TomTec Imaging Systems, Munich, Germany).
2D Cardiac Performance Analysis is a speckle tracking–based analysis tool that is an extension of Velocity Vector Imaging software (Siemens Medical Systems, Erlangen, Germany), which has been previously validated with sonomicrometry (8,23) and magnetic resonance imaging (24,25). 2D Cardiac Performance Analysis, similar to Velocity Vector Imaging, determines myocardial motion from a user-defined tracing along the endocardial border. Both user-defined endocardial and automated subepicardial borders are traced throughout 1 cardiac cycle for calculating myocardial velocity, longitudinal strain (LS), and radial strain (RS) for both endocardial and subepicardial regions.
LS from endocardial and subepicardial regions was obtained from 6 LV segments in apical 4-chamber views. Circumferential strain (CS) and RS were obtained from 6 segments in short-axis views of the left ventricle at the level of the papillary muscle. In addition to strains obtained at the segmental level, we also calculated global LS, CS, and RS as an average of strain values obtained from the 6 segments in respective views. Assessment of LV strain was regarded as suboptimal when either: 1) speckle tracking could not be obtained for at least 4 of the 6 myocardial segments in apical 4-chamber or short-axis views; or 2) a theoretically unacceptable value or values were obtained.
Peripheral blood mononuclear cell gene expression analysis
At approximately 1-month intervals after transplantation, whole venous blood in 8-ml aliquots from each subject was transferred into a Vacutainer CPT Cell Preparation Tube (Becton Dickinson, Franklin Lakes, New Jersey) with sodium citrate. Peripheral blood mononuclear cells were then isolated using density gradient centrifugation and frozen in lysis buffer within 2 h of phlebotomy. These samples were sent to XDx, a Clinical Laboratory Improvement Amendments–certified laboratory in Brisbane, California, where ribonucleic acid was subsequently purified from the peripheral blood mononuclear cell lysate. Using real-time polymerase chain reaction, cell messenger ribonucleic acid cycle threshold levels of 20 genes were evaluated, including 11 genes representing multiple diverse molecular pathways (IL1R2, FLT3, ITGAM, ITGA4, PF4, C6orf25, PDCD1, MIR, WDR40A, SEMA7A, and RHOU) and 9 control genes used for reproducibility and standardization (26). Cycle threshold levels were then integrated into a cumulative gene expression score on a scale of 0 to 40 for each sample (19,26).
All continuous data are reported as mean ± SD and categorical data as percentages. Student t test was used for comparisons of continuous variables between patients with histories of rejection and nonrejection. One-way analysis of variance with the post hoc Scheffé test was used for parametric comparisons of measurements obtained within 2 weeks and 3, 6, 12, and 24 months after transplantation. Pearson correlation coefficient was used to reveal relations between 2 continuous variables. Univariate simple linear regression was used to compare clinical variables with the outcome of averaged global LS at 1 year. Multiple linear regression analysis was used to evaluate the relationship between averaged endocardial and epicardial LS at 12 months and various clinical and echocardiographic variables showing significant correlations (p < 0.10) using commercially available software (SPSS version 12.0, SPSS, Inc., Chicago, Illinois). Improvement in strain at 3 months was defined as a ≥1-SD change in global LS assessed at the first post-transplantation echocardiographic study (27–29). Event rates for death, hospitalizations, and cardiac events were calculated by dividing the total number of events by the number of patient-years and compared using Poisson distribution. Survival analysis was performed using the Kaplan-Meier method. Mortality curves were generated separately for patients with and without improvement in global LS and then compared using the log-rank test. Interobserver and intraobserver variability was calculated as the absolute difference of the corresponding pair of repeated measurements in percent of their mean in each patient and then averaged for 18 randomly selected patients. A p value <0.05 was considered statistically significant.
The clinical and echocardiographic characteristics of cardiac transplantation patients at baseline are shown in Table 1. Of 51 transplant recipients, 18 (35%) experienced 1 or more episodes of biopsy-proven International Society for Heart and Lung Transplantation grade 2R or higher acute cellular rejection (30) in the first year. Differences in clinical and echocardiographic variables in patients who experienced 1 or more episodes of International Society for Heart and Lung Transplantation grade 2R or higher acute cellular rejection versus those who did not have rejection at completion of 1 year are shown in Tables 2 to 4.⇓⇓ Coronary allograft vasculopathy at 1 year after transplantation was seen in 13 patients (25%). Six patients died during the study period at 78, 87, 222, 296, 448, and 896 days after transplantation. Primary causes of death in these patients, respectively, were pseudomonas pneumonia, primary graft failure, massive myocardial infarction with invasive aspergillosis, Clostridium difficile colitis with renal failure, end-stage liver disease due to hepatitis C, and coronary allograft vasculopathy. A total of 36 cardiovascular events and 117 hospitalizations occurred.
Serial assessment of LV mechanics by speckle-tracking echocardiography was feasible in 42 of 48 patients (88%) at 3 months, 43 of 44 (98%) at 6 months, 30 of 31 (97%) at 1 year, and 6 of 7 (86%) at 2 years.
In comparison with controls, global endocardial LS and RS were markedly attenuated in patients with TXH, irrespective of the presence of early biopsy-detected rejection throughout the study period (Fig. 1). Persistent abnormalities in global LS (<1-SD increase in global LS) were seen in 31 of 42 patients (74%) at 3 months, in 34 of 43 (79%) at 6 months, and in 24 of 29 (83%) at the end of 1 year. Endocardial and epicardial CS and RS remained unchanged at all intervals of follow-up (−14.9% vs. −18.4%, p = 0.10, and −6.0% vs. −6.4%, p = 0.64, respectively).
Global LS, RS, and CS at the time of rejection were no different than in patients who did not experience rejection (Table 2). Furthermore, global LS and RS remained equally attenuated in both groups on serial follow-up (Fig. 1). Decreased endocardial CS was observed at 2 years in subjects who experienced allograft rejection compared with those who did not (−24.9 ± 7.9 vs. −12.6 ± 4.1, p = 0.045) (Fig. 1).
In comparison with controls, segmental strains were attenuated at all segments in post–coronary artery bypass patients. Both groups, however, showed higher regional strains at the LV apex than at the LV base. In contrast, LV strains in patients with TXH were attenuated in all 3 segments in comparison with controls (p = 0.015, p < 0.001, and p < 0.001 for the basal, mid-, and apical segments, respectively). On paired analysis, LS remained persistently attenuated at the LV apex at 1 year (Figs. 2 and 3), independent of episodes of rejection.
LS and clinical outcomes
A Kaplan-Meier curve for death and/or cardiovascular events for 11 patients (26%) with improvements in global LS and 31 patients (74%) who did not show improvements in global LS within the first 3 months of transplant is shown in Figure 5. Failure to improve LS within the first 3 months of transplantation was associated with a higher combined risk for death and hospitalization for heart failure or biopsy-detected rejection compared with those who showed improvements in LS (≥1 SD; hazard ratio: 5.92; 95% confidence interval: 1.96 to 17.91; p = 0.049) and a higher hospitalization rate (event rate: 1.13 vs. 0.27 per patient-year, p < 0.01).
Determinants of persistent LV dysfunction at 1 year
Forty-five clinical variables were entered into a univariate simple linear regression with outcome averaged global LS at 1 year. Seven variables showed significance or trends toward significance (Table 5). A higher gene expression score was significantly associated with higher LS averaged from endocardial and epicardial regions (R = −0.7, p < 0.001) (Fig. 4).
After multivariate analysis, only the gene expression score averaged over 1 year showed a significant relationship with global averaged LS (R2 = 0.53, p = 0.005) (Table 5). A trend toward a significant relationship was observed between LV mass and global LS at 12 months. There was no significant correlation between averaged global LS at 1 year in comparison with endomyocardial biopsy results, right-heart catheterization data, the presence of coronary allograft vasculopathy, body mass index, serum creatinine, and B-type natriuretic peptide at transplantation and 1 year.
Peripheral blood gene expression analysis
To further characterize the origin of the relationship between global LS and the gene expression score, each specific gene subcomponent of the gene expression profile was analyzed in comparison with averaged global LS at 1 year (Table 6). The highest correlation was with the SEMA7A gene (r = −0.84, p < 0.001). With increased expression (lower cycle threshold) of SEMA7A, heart transplantation recipients had increasing LS (Fig. 4).
Interobserver and intraobserver variability
The absolute intraobserver differences for endocardial LS, epicardial LS, endocardial CS, and epicardial CS and RS were 0.6 ± 2.2%, 0.1 ± 1.6%, 2.5 ± 2.1%, 2.2 ± 1.7%, and 8.3 ± 7.8%, respectively, and the corresponding intraobserver variability were calculated as 10 ± 7%, 8 ± 7%, 11 ± 10%, 25 ± 22%, and 24 ± 20%. The absolute interobserver differences for each measurement were 2.5 ± 1.3%, 1.9 ± 1.2%, 3.5 ± 2.9%, 2.2 ± 1.7%, and 9.1 ± 8.6%, respectively, and the corresponding interobserver variability were calculated as −13.6 ± 6.3%, −12.6 ± 7.9%, 16 ± 15%, 26 ± 21%, and 28 ± 29%, respectively.
To the best of our knowledge, this is the first study to describe the serial changes of LV mechanics in patients with TXH using speckle-tracking strain imaging. Our data support the use of strain imaging for assessing the burden of LV dysfunction that evolves in patients with TXH independent of biopsy-detected acute cellular rejection. Failure to improve strains was associated with a higher incidence of death and cardiac events. We further investigated the determinants of persistent allograft LV dysfunction at 1 year and found that the gene expression score used to characterize host alloimmune responses was the only independent predictor of global LV LS, independent of the presence of cardiac rejection, allograft vasculopathy, and other clinical variables. However, contrary to logical expectations, the gene expression score was positively correlated with LV LS, and this relationship appeared to be driven primarily by SEMA7A gene expression. We also observed a trend toward a significant relationship between LV mass and global LS at 12 months, as shown in Table 5.
LV function in patients with TXH
The burden of LV dysfunction that evolved in patients with TXH, independent of biopsy-detected cellular rejection in the present study, is consistent with a previous investigation in which the scintigraphic evidence of myocardial injury exceeded that seen in endomyocardial biopsy specimens (31). Indeed, the individual variability of the course of rejection is striking: some patients acquire tolerance to grafts as early as 3 months after transplantation; in others, this process might not develop until the third or fourth year after the operation. Up to one-third of patients who undergo transplantation who survive more than 5 years show smoldering rejection and never develop tolerance to their grafts (32). Several groups have reported detection of rejection-related myocardial damage through myocardial uptake of indium-111-labeled monoclonal antimyosin antibodies (33–36). In these studies, hearts that showed decreases of antibody uptake during the first 3 months after transplantation appeared to be free from severe rejection-related complications during the first year, whereas persistent antimyosin uptake during the first 3 months indicated a higher risk for such complications during that interval (34). Correspondingly, in our study, those subjects without improvements in global LS at 3 months after transplantation had a higher rate of complications, including death, rejection, and heart failure.
To overcome the uncertainty in assessing LV mechanical function in patients with TXH, several novel echocardiographic markers have been described. However, the enthusiasm in using these indexes, including Doppler- and tissue Doppler–derived indexes of LV mechanical performance, in patients with TXH has been marred by discordant results in different investigations (37,38). Little is known, however, about the spectrum of changes in LV mechanical function in patients with TXH using speckle-tracking strain echocardiography. A previous study suggested that regional strains were attenuated in the basal and mid-lateral walls of the LV in patients with TXH with more than International Society for Heart and Lung Transplantation grade 1B rejection; however, considerable overlap was seen, particularly for strain values obtained from the LV septum, where LS was equally attenuated in all patients with TXH (39,40). Our study suggests that patients with TXH have a substantial burden of LV regional and global dysfunction within the first 2 years, and these unique cumulative changes in LV mechanical function evolve independent of changes due to clinically detectable transplant rejection. Global LS has now emerged as an accurate predictor of clinical outcomes, including all-cause mortality, in the nontransplantation population (41). In our study of cardiac transplantation patients, changes in global LS similarly were associated with death and cardiac events, including rejection and heart failure.
Gene expression profiles correlate with LV mechanical function
The gene expression score is a 20-gene real-time polymerase chain reaction assay that produces a gene expression signature that has been shown to correlate with biopsy-defined rejection (20). In the Cardiac Allograft Gene Expression Observational study (26), to date the largest and most systematic investigation of gene expression score for the diagnosis of heart graft rejection, the use of a diagnostic score threshold favoring negative predictive value was estimated to have a negative predictive value of 99.6%, a positive predictive value of 6.8%, sensitivity of 88.9%, and specificity of 69.4% for detecting grade ≥3A rejection. An interesting feature of the Cardiac Allograft Gene Expression Observational data was that grade 1B endomyocardial biopsy samples were associated with gene expression scores similar to those of grade ≥3A samples and significantly higher than those of grade 2 samples (as well as grades 1A and 0) (26,42). Considering the diversity of genes incorporated into the gene expression score, it is conceivable that their individual and cumulative activations may serve a variety of purposes beyond adaptive immunological activation.
Upon examination of specific genetic subcomponents of the gene expression score, we found that SEMA7A had the strongest correlation with averaged global LS at 1 year. Semaphorins constitute a large family of secreted and membrane-tethered cell signaling molecules with functions in neural development, macrophage and T-cell activation pathways, cardiac growth, and vascular development (43,44). Although semaphorins were initially incorporated into the AlloMap gene expression score for their possible role in acute cellular rejection, the present study raises the possibility that their other pathways may be associated with graft function. Analysis from the Cardiac Allograft Gene Expression Observational study revealed that SEMA7A expression increased in a time dependent manner beyond 6 months after transplantation. The exact reason for this was not clear, and it was suggested that intrinsic or extrinsic factors that vary by time after transplantation may alter the peripheral molecular signatures of peripheral blood mononuclear cells in heart transplant recipients. Several possibilities exist for explaining the relationship of SEMA with LV longitudinal mechanics in patients with TXH. Notably, semaphorins modulate sympathetic reinnervation patterning (43) and cardiac regional contractility (45,46), variables that may be associated with improved exercise tolerance in patients with cardiac transplantation at 1 year (47,48).
After cardiac transplantation, significant efforts are directed toward the identification and prevention of acute cellular rejection and coronary allograft vasculopathy in heart transplant recipients (14). Our data emphasize the need to recognize the presence of substantial LV dysfunction in patients with TXH that evolves independent of clinically detectable acute cellular rejection or allograft vasculopathy that correlates with long-term clinical outcomes. Although LV ejection fraction is widely used in transplantation practice to make clinical decisions, our data suggest that the assumptions in the range and variability of LV function on the basis of ejection fraction may not be valid. Rather, regional and global LV mechanics quantified by novel techniques such as speckle-tracking echocardiography may serve as robust biomarkers. The novel relationship between gene expression profile, LV mechanics, and the impact on overall graft survival needs further exploration. Particularly, the role of SEMA7A gene requires future consideration as a potential molecular target for gene therapy (49) in chronic graft dysfunction.
Although the present study has identified a relationship between LV mechanics and peripheral blood mononuclear cell gene expression scores, the heart surface was not directly examined for the presence of these markers to establish the definite pathways. Furthermore, LV torsional mechanics were not investigated in this study, because of nonavailability of optimal cross-sectional views of the LV apex in the immediate post-transplantation period. Variability in CS and RS was higher than in LS, a finding that concurs with reproducibility data from a previous study (50). This may be due to through-plane motion affecting 2D speckle-tracking data obtained in short-axis views. Future studies with 3-dimensional speckle tracking may be useful for confirming the observed relationships. The AlloMap gene expression score has been validated in the transplantation population; understanding its use and relationship with myocardial mechanics in the nontransplantation population will require further investigation.
Speckle tracking–derived LV strains are helpful in estimating the burden of LV dysfunction in patients with TXH that evolves independent of biopsy-detected cellular rejection. Failure to improve global LS at 3 months after transplantation is associated with a higher incidence of death and cardiac events. Serial changes in LV mechanics correlate with peripheral blood gene expression profiles and may affect clinical assessment of long-term prognosis in patients with TXH.
The authors thank Diane L. Kasper, RN, CCTC, for assistance with cardiac transplant database management, Amylou C. Dueck, PhD, for assistance in statistical analysis, and Xdx Laboratory, Brisbane, California, for assistance with gene expression score analysis.
Dr. Khandheria is currently affiliated with Aurora Health Care, Advanced Cardiovascular Group, Milwaukee, Wisconsin. Dr. Sengupta is currently affiliated with the Division of Cardiovascular Diseases, University of California Irvine, Irvine, California. The authors have reported that they have no relationships to disclose. Drs. Eleid and Caracciolo contributed equally to this work.
- Abbreviations and Acronyms
- circumferential strain
- longitudinal strain
- left ventricular/ventricle
- radial strain
- transplanted hearts
- Received May 26, 2010.
- Revision received July 28, 2010.
- Accepted July 30, 2010.
- American College of Cardiology Foundation
- Amundsen B.H.,
- Helle-Valle T.,
- Edvardsen T.,
- et al.
- Greenberg M.L.,
- Uretsky B.F.,
- Reddy P.S.,
- et al.
- Teo K.K.,
- Yusuf S.,
- Wittes J.,
- et al.
- Hunt S.A.,
- Haddad F.
- Bengel F.M.,
- Ueberfuhr P.,
- Ziegler S.I.,
- Nekolla S.,
- Reichart B.,
- Schwaiger M.
- Bristow M.R.
- Wilson R.F.,
- Christensen B.V.,
- Olivari M.T.,
- Simon A.,
- White C.W.,
- Laxson D.D.
- von Scheidt W.,
- Bohm M.,
- Schneider B.,
- Reichart B.,
- Erdmann E.,
- Autenrieth G.
- Lang R.M.,
- Bierig M.,
- Devereux R.B.,
- et al.
- Armstrong G.,
- Pasquet A.,
- Fukamachi K.,
- Cardon L.,
- Olstad B.,
- Marwick T.
- Kukulski T.,
- Jamal F.,
- Herbots L.,
- et al.
- Ballester M.,
- Obrador D.,
- Carrio I.,
- et al.
- Ballester M.,
- Obrador D.,
- Carrio I.,
- et al.
- Estep J.D.,
- Shah D.J.,
- Nagueh S.F.,
- Mahmarian J.J.,
- Torre-Amione G.,
- Zoghbi W.A.
- Stanton T.,
- Leano R.,
- Marwick T.H.
- Lorentz C.U.,
- Alston E.N.,
- Belcik T.,
- Lindner J.R.,
- Giraud G.D.,
- Habecker B.A.
- Carter B.D.,
- Feng N.,
- Paolocci N.
- Vassalli G.,
- Roehrich M.E.,
- Vogt P.,
- et al.