Author + information
- Received December 12, 2017
- Revision received April 13, 2018
- Accepted May 24, 2018
- Published online October 1, 2018.
- Øyvind H. Lie, MDa,b,c,
- Christine Rootwelt-Norberg, MDa,b,
- Lars A. Dejgaard, MDa,b,c,
- Ida Skrinde Leren, MD, PhDa,b,c,
- Mathis K. Stokke, MD, PhDa,b,c,
- Thor Edvardsen, MD, PhDa,b,c,d and
- Kristina H. Haugaa, MD, PhDa,b,c,d,∗ ()
- aDepartment of Cardiology, Oslo University Hospital, Rikshospitalet, Oslo, Norway
- bCenter for Cardiological Innovation, Oslo University Hospital, Rikshospitalet, Oslo, Norway
- cInstitute of Clinical Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway
- dInstitute for Surgical Research, University of Oslo, Oslo, Norway
- ↵∗Address for correspondence:
Dr. Kristina H. Haugaa, Department of Cardiology, Oslo University Hospital, Rikshospitalet, Sognsvannsveien 20, 0372 Oslo, Norway / P.O. Box 4950 Nydalen, 0424 Oslo, Norway.
Objectives This study aimed to identify clinical, electrocardiographic (ECG) and cardiac imaging predictors of first-time life-threatening ventricular arrhythmia in patients with arrhythmogenic cardiomyopathy (AC).
Background The role of clinical, electrocardiographic, and cardiac imaging parameters in risk stratification of patients without ventricular arrhythmia is unclear.
Methods We followed consecutive AC probands and mutation-positive family members with no documented ventricular arrhythmia from time of diagnosis to first event. We assessed clinical, electrocardiographic, and cardiac imaging parameters according to Task Force Criteria of 2010 in addition to left ventricular (LV) and strain parameters. High-intensity exercise was defined as >6 metabolic equivalents.
Results We included 117 patients (29% probands, 50% female, age 40 ± 17 years). During 4.2 (interquartile range [IQR]: 2.4 to 7.4) years of follow-up, 18 (15%) patients experienced life-threatening ventricular arrhythmias. The 1-, 2-, and 5-year incidence was 6%, 9%, and 22%, respectively. History of high-intensity exercise, T-wave inversions ≥V3, and greater LV mechanical dispersion were the strongest risk markers (adjusted hazard ratio [HR]: 4.7 [95% confidence interval (CI): 1.2 to 17.5]; p = 0.02, 4.7 [95% CI: 1.6 to 13.9]; p = 0.005), and 1.4 [95% CI: 1.2 to 1.6] by 10-ms increments; p < 0.001, respectively). Median arrhythmia-free survival in patients with all risk factors was 1.2 (95% CI: 0.4 to 1.9) years, compared with an estimated 12.0 (95% CI: 11.5 to 12.5) years in patients without any risk factors.
Conclusions History of high-intensity exercise, electrocardiographic T-wave inversions ≥V3, and greater LV mechanical dispersion were strong predictors of life-threatening ventricular arrhythmia. Patients without any of these risk factors had minimal risk, whereas ≥2 risk factors increased the risk dramatically. This may help to make decisions on primary preventive implantable cardioverter defibrillator (ICD) therapy.
Arrhythmogenic cardiomyopathy (AC, otherwise known as arrhythmogenic right ventricular cardiomyopathy [ARVC]) is an inheritable heart disease caused by dysfunctional cardiac desmosomes, resulting in fibrofatty replacement and subsequent structural and functional alterations of the right ventricle (RV) and left ventricle (LV). The disease is characterized by high risk of life-threatening ventricular arrhythmia and cardiac arrest in presumed healthy young persons. The management of patients with AC presenting with life-threatening ventricular arrhythmia is well established, with placement of implantable cardioverter-defibrillators (ICDs) as a recommended strategy (1). Management of patients presenting without life-threatening events is less well defined, and risk stratification and selection of patients to receive primary preventive ICDs is challenging.
The diagnosis of AC is based on a complex set of criteria according to the revised Task Force Criteria (TFC) of 2010 (2), in which electrocardiogram (ECG) and cardiac imaging are central elements. Advances in echocardiography have provided sensitive and accurate tools for detecting cardiac abnormalities (3,4), which have been described as markers of ventricular arrhythmia in patients with AC (5). Echocardiographic strain parameters are recommended as part of a multimodality imaging approach in AC (6), but prospective studies on prognostic value are lacking. Furthermore, implementation of cascade genetic testing has shifted the AC population toward more asymptomatic mutation-positive family members. These individuals are at lower risk of arrhythmic events than probands but should still be risk stratified and considered for primary preventive ICD (7). Prognostic markers in this population are largely undefined. We aimed to explore incidence and predictors of first-time life-threatening ventricular arrhythmia in a primary prevention AC cohort.
Study design and population
Patients diagnosed with AC at Oslo University Hospital, Rikshospitalet, Norway, between 1997 and 2017, were evaluated consecutively for inclusion in a primary prevention cohort study. Probands underwent genetic testing as described previously (8), and cascade genetic screening was performed in family members of mutation positive probands. Mutation negative probands were only included if they fulfilled a definite AC diagnosis by TFC (2) and their family members were not included.
Life-threatening ventricular arrhythmia was defined as sustained ventricular tachycardia—runs of consecutive ventricular beats >100 beats for >30 s (9)—documented on 12-lead ECG, Holter or ICD monitoring, cardiac arrest, or appropriate shock therapy from a primary preventive ICD. Nonsustained ventricular tachycardia (NSVT) was defined as consecutive runs of ≥3 ventricular beats >100 beats/min for <30 s (9). Syncope at or prior to diagnosis was recorded. Exercise habits at inclusion were recorded by standardized interviews. The reported exercise activities were assigned intensity levels on the basis of the Compendium of Physical Activity of 2011 (10). High-intensity exercise was defined as physical activity >6 metabolic equivalents performed regularly during a minimum of 3 consecutive years before inclusion (11).
Patients with life-threatening event at or prior to first contact and patients with cardiopulmonary comorbidity were excluded. Time to first life-threatening ventricular arrhythmia was recorded prospectively from time of diagnosis of AC or from time of first echocardiography on a GE Vivid 7 or newer scanner. End of observation was time of event, cardiac transplantation, death, or last clinical follow-up by October 1, 2017. Written informed consent was given by all included patients. The study complied with the declaration of Helsinki and was approved by the Regional Medical Ethics Committee of South-Eastern Norway.
Twelve-lead ECG was obtained at inclusion (MAC 1200 or MAC 5000, GE Medical Systems, Milwaukee, Wisconsin, or CS-200, Schiller, Baar, Switzerland). In accordance with TFC, we recorded extent of T-wave inversions (TWI), presence of epsilon waves and increased terminal activation duration (2,12) (Figure 1, lower panel). Signal-averaged ECG (SAECG) was performed at inclusion and evaluated according to TFC 2010 (2).
All patients were examined with echocardiography at inclusion (Vivid 7, E9 or E95, GE, Vingmed, Horten, Norway), and datasets were analyzed offline (EchoPac 201, GE, Vingmed) by 2 independent observers blinded to clinical and ECG data. LV ejection fraction (LVEF) was assessed by the biplane Simpson’s method. The LV global longitudinal strain (GLS) was derived from speckle tracking analyses on 2-dimensional (2D) gray-scale image loops with >50 frames/s from the 3 apical views and expressed as the average peak systolic strain in a 16-segment LV model (13). LV mechanical dispersion was defined as the standard deviation of time from Q/R on surface ECG to peak negative strain in 16 LV segments (4) (Figure 1, upper left panels).
The RV function was assessed by the RV fractional area change (FAC), tricuspid annular plane systolic excursion (TAPSE), and RV longitudinal strain (RVLS), defined as the average peak systolic strain in 3 free-wall RV segments (Figure 1, upper right panel). RV mechanical dispersion (RVMD) was defined as the standard deviation from the time of Q/R on ECG to peak negative strain in 6 RV segments, including 3 septal segments (5). RV dimension measures included proximal diameter of RV outflow tract (RVOT) from parasternal short axis view and RV basal diameter (RVD) from RV-focused apical 4-chamber view (14). The Task Force Criteria (2) were ascertained retrospectively in patients included before 2010.
Cardiac magnetic resonance
Cardiac magnetic resonance imaging (CMR) was performed as previously reported (15) on clinical indication in a subset of patients at inclusion and evaluated for the presence of diagnostic criteria according to TFC (2). In addition, we assessed RV fat infiltration, LV volumes and EF, and late gadolinium enhancement (LGE).
Values were expressed as mean ± SD, frequencies (%), or median (interquartile range [IQR], and compared by unpaired Student’s t-test, chi-square test, Fisher’s exact test, or Mann-Whitney U test, as appropriate. Incidence of life-threatening ventricular arrhythmia was calculated by dividing number of events by number of eligible patients at 1, 2, and 5 years of follow-up. The ability of a parameter to predict life-threatening arrhythmia was expressed by Harrell’s C-statistic derived from univariable Cox regression. Multivariable Cox regression was performed with maximum 3 covariates, retaining the 3 parameters with the highest Harrell’s C-statistic in case of multiple possible confounders: 1) clinical characteristics: high-intensity exercise, probands status and history of syncope; 2) electrocardiography: major TWI and sex; 3) echocardiography: LV mechanical dispersion, RVLS, and sex; and 4) CMR: LVEDVi, RVEF, and sex. The optimal cutoff values of continuous variable were defined as the value generating the highest Harrell’s C-statistic. The parameters from the 3 separate categories' clinical characteristics, ECG, and cardiac imaging with the highest Harrell’s C-statistic were added to a stepwise Cox regression model. The incremental value of added parameters in the prediction models were assessed by reclassification analysis by integrated diagnostic improvement (IDI) and continuous net reclassification improvement (NRI), and the C-statistics for prediction models were compared. The Akaike information criterion (AIC) was estimated for the risk models. Kaplan-Meier analyses of estimated ventricular arrhythmia-free survival for the risk factors were conducted, dichotomizing continuous risk markers at the optimal cutoff value. Competing-risk regression was conducted to assess the impact of competing endpoints. Statistical analyses were performed using Stata/SE 14.2 (StataCorp LLC, Texas, added packages: somersd, nri); p values were 2-sided, and values <0.05 were considered significant.
Clinical characteristics and incidence of life-threatening ventricular arrhythmia
Of 188 patients presenting at our center with definite, borderline, or possible AC diagnosis, 7 (4%) were excluded due to concomitant heart disease, 61 (34%) were excluded due to life-threatening ventricular arrhythmia at or prior to presentation, and 3 (2%) were lost to follow-up. Therefore, 117 eligible patients (34 [29%] probands, 83 [71%] mutation-positive family members) were included, of whom 27 (5) and 78 (14) have been reported previously. The most common presenting symptoms of probands were syncope (n = 18, 53%), palpitations (n = 13, 38%), and chest pain (n = 12, 35%).
During 4.2 years (IQR: 2.4 to 7.4 years) of follow-up, 21 (18%) patients received primary preventive ICD. Life-threatening ventricular arrhythmia occurred in 18 (15%) patients (14 probands, and 4 family members) after 2.0 years (IQR: 0.5 to 3.5 years) of follow-up (Table 1); as documented sustained ventricular tachycardia in 11 (6 on clinical 12-lead ECG, 4 below therapy zone on ICD recordings, and 1 on Holter-recording) and as shock therapy from primary preventive ICD in 7. The 1-, 2-, and 5-year incidence was 6%, 9%, and 22%, respectively, and the incidence rate was 3.1 (95% confidence interval [CI]: 1.9 to 5.0) per 100 patient years. Analyses of probands separately showed 1-, 2-, and 5-year incidence of 21%, 25%, and 57%, respectively, and an incidence rate of 10.9 (95% CI: 6.3 to 18.8) per 100 patient years. Four (5%) family members experienced life-threatening ventricular arrhythmias after 3.8 years (IQR: 1.3 to 8.1 years) of follow-up. Separate analyses of family members showed 1-, 2-, and 5-year incidence of 0%, 3%, and 4%, respectively, and an incidence rate of 0.9 (95% CI: 0.3 to 2.5) per 100 patient years (p < 0.001 compared with probands). Two patients were censored by noncardiac death and 1 by cardiac transplantation.
Clinical predictors of life-threatening ventricular arrhythmia
Male sex, history of high-intensity exercise, absence of AC-related mutation, proband status, and previous syncope were predictors in univariable analyses of clinical markers (Table 1). High-intensity exercise was more prevalent among male than among female patients (53% vs. 21%; p = 0.001).
ECG predictors of life-threatening ventricular arrhythmia
ECG and SAECG were performed at the same day as echocardiography. Major TWI was the only predictor on ECG (Table 2, Figure 2, mid panel). SAECG was performed in 100 (91%) patients without RBBB, of whom 48 (48%) had abnormal findings (47 had fQRSd >114 ms, 35 (35%) had RMS <20 μV, and 35 (35%) patients had HFLA >38 ms).
Cardiac imaging predictors of life-threatening ventricular arrhythmia
Four patients (3%) had their first echocardiographies performed on older scanners, and time to event was recorded from their first echocardiographies on a GE Vivid 7 or newer scanners. LV speckle tracking analyses were feasible in 110 patients (frame rate 63 [IQR: 56 to 69] and RV in 108 patients (frame rate 65 [IQR: 56 to 73]). Echocardiographic parameters of cardiac structure and function were consistently worse in patients who later experienced life-threatening ventricular arrhythmias (Table 2). This was evident both in probands and mutation-positive family members (Online Table 1A and 1B, respectively). LV mechanical dispersion and RVLS had the highest Harrell’s C-statistic (0.84 and 0.81, respectively) and were independent predictors in multivariable analysis (Table 2). Adding LV mechanical dispersion to RVLS improved risk reclassification (NRI 1.05; p < 0.001 and IDI 0.24; p < 0.001). The optimal statistical cutoff for LV mechanical dispersion was ≥45 ms and for RVLS worse than –23%. Kaplan-Meier analysis demonstrated unfavorable arrhythmic outcome in patients with LV mechanical dispersion ≥45 ms at inclusion (log-rank, p < 0.001, Figure 3, right panel). In a risk prediction model with known GLS, LV mechanical dispersion improved the risk reclassification (Online Table 2).
CMR was available in 88 (75%) patients (35% probands, 49% female, age 38 ± 16 years, 0.2 [IQR: 0.0 to 0.5] years after echocardiography) of whom 16 experienced subsequent life-threatening ventricular arrhythmias during follow-up (Table 2).
Risk prediction model
The parameter with the highest Harrell’s C-statistic from clinical characteristics, ECG, and cardiac imaging were added to a Cox regression model (Figure 3, left panel). Adding TWI to high-intensity exercise significantly increased Harrell’s C-statistic and improved risk reclassification. Adding LV mechanical dispersion to the combination of high-intensity exercise and TWI further improved C-statistic significantly and improved risk reclassification (Figure 3, left panel). Estimated survival free from life-threatening ventricular arrhythmia in patients with all 3 risk factors was only 1.2 (95% CI: 0.4 to 1.9) years compared with 12.0 (95% CI: 11.5 to 12.5) years in patients without any risk factors. All patients with 3 risk factors experienced life-threatening ventricular arrhythmias within 2 years, whereas only 1 of 44 patients without any risk factors experienced the endpoint during 218 patient years of follow-up. There was no significant increase in risk from no risk factors to 1 risk factor, but a 4-fold increase from 1 to 2 risk factors (hazard ratio [HR]: 4.1, 95% CI: 1.1 to 16.6; p = 0.04) and almost a 10-fold increase in risk from 2 to 3 risk factors (HR: 9.4, 95% CI: 2.3 to 37.8; p = 0.002) (Figure 3, right panel).
The results of the final model were similar in competing-risk regression (Online Table 3). Furthermore, the model performed well also when excluding all patients with previous syncope (Harrell C-statistic 0.79 [95% CI: 0.57 to 1.00]), and LV mechanical dispersion remained a predictor of ventricular arrhythmia (HR: 1.89, 95% CI: 1.03 to 3.48 by 10-ms increments; p = 0.04).
This is the first study following patients with AC with no previous life-threatening ventricular arrhythmia prospectively until their first event. The incidence of life-threatening ventricular arrhythmia in the total population was 22% at 5 years of follow-up, highlighting the importance of continuous follow-up and risk stratification. High-intensity exercise, TWI in ECG, and greater LV mechanical dispersion from echocardiography were the strongest risk markers, and the combination of these improved prediction of risk significantly. Patients with all 3 risk factors had almost 10-fold risk compared with those who had 2 risk factors, indicating that a primary preventive ICD may be appropriate. However, patients with no risk factors or only one risk factor had very low risk, suggesting that longer follow-up intervals are sufficient.
Incidence of life-threatening ventricular arrhythmia
The 5-year incidence of first-time life-threatening ventricular arrhythmia of 22% in our study is in line with previous reports (16). The majority of patients with such events were probands, and the higher risk of these compared with family members is well known (7,17). The lower incidence rate in family members confirmed a more benign prognosis (18) but still the need for risk stratification and clinical follow-up of these patients.
Prognostic value of parameters at inclusion
This is the first prospective report of the harmful effect of high-intensity exercise in patients with AC, confirming the strong association between expression of disease and exercise (8,19). Syncope has also been reported as a risk marker in previous studies (1,16,18). Patients with syncope at inclusion may have had unrecognized ventricular tachycardia but were included in this primary prevention population in line with the clinical management recommendations (1). In the adjusted statistical model, syncope was not an independent predictor when combined with proband status, reflecting a covariation with many probands with syncope at inclusion. Status as mutation positive was a marker of favorable arrhythmic outcome, which is explained by the high number of mutation positive family members with mild disease without ventricular arrhythmia rather than a protective effect of pathogenic mutations. The presence of depolarization abnormalities—that is, epsilon wave and abnormal SAECG—was not associated with subsequent life-threatening ventricular arrhythmia. This is in line with the limited risk prediction value of epsilon waves (1). The prognostic value of major criterion TWI has been described previously in a longitudinal cohort study (7) and was confirmed in our study.
All assessed echocardiographic parameters were predictors of adverse arrhythmic outcome, highlighting the fact that when structural changes are evident by imaging techniques, risk of life-threatening ventricular arrhythmia is increased. LV mechanical dispersion was a strong risk marker for arrhythmic events. This is noteworthy, as conventional parameters of LV function have been reported to be spared until late stages of conventional AC phenotypes (20) and affected earlier only in a LV selective phenotype (21). Our finding underlined the fact that AC is a biventricular disease, also supported by previous longitudinal studies (22,23). LV mechanical dispersion has been reported as a marker of ventricular arrhythmias in other conditions by our group (4,24) and independent groups (25,26), although others have not reported added benefit in prediction of risk (27). These diverging results may be explained by different methodologies and heterogeneous study populations and illustrates the need for further research.
We have previously demonstrated that RV mechanical dispersion is a marker of previous arrhythmic events in smaller cross-sectional AC cohorts (5,14). RV mechanical dispersion was a good predictor of ventricular arrhythmia in the current study, but LV mechanical dispersion had higher C-statistic and was retained in the risk-prediction model. RVLS was also a strong risk predictor, and adding LV mechanical dispersion to RVLS improved risk reclassification, underlining the importance of biventricular assessment in AC patients.
As expected, lower RVEF, the presence of RV wall-contraction abnormalities, or RV aneurysms by CMR were predictors of life-threatening ventricular arrhythmia. Furthermore, increased LV end-diastolic volume was strongly associated with events, which may reflect both LV disease penetrance and physiological adaptations to exercise.
Risk prediction model and estimates of risk
The risk-prediction model improved by each added modality. This emphasized the importance of multimodality evaluation in these patients not only for diagnostic purposes but also for risk stratification. Patients with 1 or no risk factors from the final model had a relatively benign arrhythmic prognosis, illustrated by long estimated time to life-threatening ventricular arrhythmia. This indicates that persons presenting with low risk can be identified and possibly that appropriate modifications—for example, exercise restriction—can be made at an early stage.
Fulfilling 2 or more risk factors was associated with significantly worse prognosis. The worst prognosis was seen in patients with all 3 risk factors, in which all experienced life-threatening ventricular arrhythmia within 2 years. This could indicate that patients with more than 1 risk factor could benefit from a primary preventive ICD, but it should be confirmed in a separate prospective cohort and balanced against ICD complication rates and patient inconvenience. One in 3 patients with primary preventive ICD received appropriate shock therapy during follow-up, and 1 in 10 patients without ICD experienced sustained VT. This suggested that the ICD selection was reasonable but not optimal. However, cardiac arrest did not occur in any patient during follow-up, meaning that the purpose of primary preventive ICD treatment was fulfilled. This goal may be easier to obtain when considering the risk factors we have described.
This was the first prospective study highlighting the importance of assessing history of high-intensity exercise together with ECG and cardiac imaging for optimal prediction of risk in AC. Fulfilling 2 risk factors was associated with a 4-fold increased risk of life-threatening ventricular arrhythmia and should trigger consideration for primary preventive ICD. Patients with all 3 risk factors had the worst arrhythmic outcome, and primary preventive ICD may be warranted.
The relatively low sample size and limited number of endpoints were the biggest limitations of the current study. The single-center design limited the external validity. Patients with primary preventive ICD had continuous rhythm monitoring and thus had greater chance of recognizing arrhythmic events than those without continuous rhythm registration. The current Task Force criteria (2) were applied retrospectively to patients diagnosed before 2010. CMR was only available in a subset of patients with a time delay and may have been subject to selection bias. Pathogenic mutations in the plakophillin-2 gene were abundant in our study population, and results may not be generalizable to AC populations with other dominating mutations. LV mechanical dispersion is dependent on temporal resolution and should only be used as a global measure. The threshold of >45 ms found to predict life-threatening ventricular arrhythmia in the current study should be confirmed in a separate cohort.
Patients with AC presenting without documented ventricular arrhythmia had 1-, 2-, and 5-year incidence of serious arrhythmic events of 6%, 9%, and 22% during follow-up. History of high-intensity exercise, T-wave inversions on ECG, and greater echocardiographic LV mechanical dispersion were strong predictors of life-threatening ventricular arrhythmia with incremental prognostic value. Patients with no risk factors had excellent prognoses, whereas fulfilling all risk factors increased the risk dramatically to 50% within 1.2 years. These findings may help decisions on primary preventive ICD treatment.
COMPETENCY IN MEDICAL KNOWLEDGE: Patients with arrhythmogenic cardiomyopathy have high risk of life-threatening ventricular arrhythmia. Management of patients presenting without arrhythmia is challenging. This study demonstrated that parameters from clinical characteristics, electrocardiogram, and cardiac imaging had independent and incremental value as predictors of first-time life-threatening ventricular arrhythmia in this patient group.
TRANSLATIONAL OUTLOOK: Recognition of patients at risk of life-threatening ventricular arrhythmia is vital in the management of patients with arrhythmogenic cardiomyopathy. An increasing proportion of patients are identified through increased awareness of disease and genetic screening. It could be helpful to consider exercise intensity, right precordial T-wave inversions, and left ventricular mechanical dispersion when selecting patients with arrhythmogenic cardiomyopathy to receive primary prevention implantable cardioverter-defibrillator therapy.
This research was supported by a public research grant from South-Eastern Norway Health Authority and the Center for Cardiological Innovation, which is supported by the Norwegian Research Council. Drs. Haugaa and Edvardsen have licensed a patent of mechanical dispersion. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- arrhythmogenic cardiomyopathy
- cardiac magnetic resonance imaging
- implantable cardioverter-defibrillator
- left ventricle
- right ventricle
- signal-averaged electrocardiogram
- Task Force Criteria
- T-wave inversion
- Received December 12, 2017.
- Revision received April 13, 2018.
- Accepted May 24, 2018.
- 2018 The Authors
- Marcus F.I.,
- McKenna W.J.,
- Sherrill D.,
- et al.
- Kolias T.J.,
- Edvardsen T.
- Haugaa K.H.,
- Grenne B.L.,
- Eek C.H.,
- et al.
- Haugaa K.H.,
- Basso C.,
- Badano L.P.,
- et al.
- Bhonsale A.,
- James C.A.,
- Tichnell C.,
- et al.
- ↵World Health Organization. Intensity of physical activity. Available at: http://www.who.int/dietphysicalactivity/physical_activity_intensity/en/ Accessed February 13, 2017.
- Edvardsen T.,
- Haugaa K.H.
- Leren I.S.,
- Saberniak J.,
- Haland T.F.,
- Edvardsen T.,
- Haugaa K.H.
- Saberniak J.,
- Leren I.S.,
- Haland T.F.,
- et al.
- Mazzanti A.,
- Ng K.,
- Faragli A.,
- et al.
- Groeneweg J.A.,
- Bhonsale A.,
- James C.A.,
- et al.
- James C.A.,
- Bhonsale A.,
- Tichnell C.,
- et al.
- Gemayel C.,
- Pelliccia A.,
- Thompson P.D.
- Sen-Chowdhry S.,
- Syrris P.,
- Prasad S.K.,
- et al.
- Candan O.,
- Gecmen C.,
- Bayam E.,
- Guner A.,
- Celik M.,
- Dogan C.
- Ersboll M.,
- Valeur N.,
- Andersen M.J.,
- et al.