Author + information
- Received November 6, 2015
- Revision received March 2, 2016
- Accepted March 3, 2016
- Published online December 1, 2016.
- Maria Sanz-de la Garza, MDa,∗ (, )
- Gonzalo Grazioli, MDa,
- Bart H. Bijnens, PhDb,
- Sebastian I. Sarvari, MD, PhDa,c,
- Eduard Guasch, MD, PhDa,
- Carolina Pajuelo, MDa,
- Daniel Brotons, MD, PhDd,
- Enric Subirats, MD, PhDe,
- Ramon Brugada, MD, PhDf,
- Emma Roca, PhDe and
- Marta Sitges, MD, PhDa
- aCardiology Department, Institut d′investigacions Biomèdiques August Pi i Sunyer, Hospital Clínic, Barcelona, Spain
- bInstitució Catalana de Recerca i Estudis Avançats, Universitat Pompeu Fabra, Barcelona, Spain
- cDepartment of Cardiology, Oslo University Hospital, Rikshospitalet, Oslo, Norway
- dCatalan Sports Council, Barcelona, Spain
- eDepartment of Medical Science, University of Girona, Girona, Spain
- fCardiology Service, Hospital Trueta, Department of Medical Sciences, University of Girona, Institute of Biomedical Research Girona, Spain
- ↵∗Reprint requests and correspondence:
Dr. María Sanz-de la Garza, Cardiology Department, Institut d'investigacions Biomèdiques August Pi i Sunyer, Hospital Clínic, Villarroel 136, 08036 Barcelona, Spain.
Objectives This study sought to understand and characterize the acute atrial response to endurance exercise and the influence of the amount of exercise performed.
Background Endurance exercise seems to be recognized as a risk factor for developing atrial arrhythmia. Atrial geometrical and functional remodeling may be the underlying substrate.
Methods Echocardiography was performed in 55 healthy adults at baseline and after a 3-stage trail race: a short race (S) (14 km), n = 17; a medium race (M) (35 km), n = 21; and a long race (L) (56 km), n = 17. Analysis consisted of standard, speckle-tracking assessment of both the left ventricle (LV) and right ventricle (RV) and both the left atrium (LA) and the right atrium (RA): a-wave strain (Sa) and strain rate (Ra) as a surrogate for atrial contractile function and s-wave strain (St) and strain rate (SR) as reservoir function.
Results After the race, RA reservoir function decreased in group M (Δ% SRs: −12.5) and further in group L (Δ% SRs: −15.4), with no changes in group S. RA contractile function decreased in group L (Δ% SRa: −9.3), showed no changes in group M (Δ% SRa: +0.7), and increased in group S (Δ% SRa: +14.8). A similar trend was documented in LA reservoir and contractile function but with less pronounced changes. The decrease in RA reservoir after the race correlated with the decrease in RV global longitudinal strain (GLS) (Δ% RVGLS vs. RASt and RASRs: +0.44; p < 0.05 and +0.41, respectively; p < 0.05).
Conclusions During a trail-running race, an acute exercise-dose dependent impairment in atrial function was observed, mostly in the RA, which was related to RV systolic dysfunction. The impact on atrial function of long-term endurance training might lead to atrial remodeling, favoring arrhythmia development.
Recent data have shown that endurance training is associated with an increased risk for atrial fibrillation (AF) and atrial flutter (1–3). A dose-response relationship between endurance training load and AF also has been demonstrated (4–6). The pathophysiology of potential exercise-related AF is still poorly understood, but exercise-induced atrial remodeling seems to be one of the major contributing factors (1,7). Classically, atrial remodeling has been described in terms of biatrial dilation (8,9) and normal atrial function among athletes (8,10). However, recent studies have shown reduced atrial deformation in the contractile and reservoir phases in athletes at rest compared to those in sedentary controls (11,12). The high cardiac output required during endurance exercise training has been shown to impose a high degree of stress on all myocardial structures, which seems to be proportionate to the amount of exercise performed (13). Following Laplace’s law, this increase in wall stress should be especially important for heart cavities with thin-wall structure: the right ventricle (RV) and both the left atrium (LA) and right atrium (RA). Indeed, various studies have shown an acute RV performance impairment after high-intensity endurance exercise (13–15). Impairment in LA function (16) and indirect evidence of myocardial edema after completing long-distance endurance races has been reported (17). However, limited data are available for the acute response of the atria (particularly the RA) to high-intensity endurance exercise and to the impact of different amounts of exercise.
In recent years, atrial strain and strain rate analysis by 2-dimensional (2D) speckle-tracking echocardiography has emerged as a novel method to evaluate LA and RA functions (18). The assessment of atrial function by strain and strain rate has been used as a predictor of AF recurrence in various clinical situations (19). Recently, the utility of speckle-tracking echocardiography also has been demonstrated in the evaluation of atrial function in athletes (8,12).
In this study, we aimed to comprehensively evaluate atrial adaptation and remodeling after intense endurance exercise. The study had 3 objectives: 1) to analyze atrial performance in response to different cumulative intensities of exercise; 2) to analyze the influence of ventricular function on atrial performance; and 3) to characterize the different atrial responses among individuals. The principal hypothesis of our study was that, after high-intensity endurance exercise, an impairment in RV function would be presented, as was suggested before, but potentially also in both atria, particularly the RA, directly influenced by RV performance. In addition, we presumed that RV and atrial responses to endurance exercise were going to be influenced by the amount of exercise performed, but we also expected high interindividual variability in relation with individual exercise-adaptive mechanisms.
Volunteers were recruited 8 months before a trail-running event in Catalonia, “La Gran Volta a la Cerdanya.” Of 1,484 runners registered for the event at the time of recruitment, 716 agreed to participate. The competition included 3 different races: a 14-km race with 500 m cumulative altitude, a 35-km race with 1,500 m of cumulative altitude, and a 56-km race with 3,400 m of cumulative altitude. Participants were assigned to each distance group according to their prior training, defined as total self-reported hours of endurance training per week: <3 h/week for the short-distance race (group S), 3 to 10 h/week for the medium distance (group M), and >10 h/week for the long-distance (group L). We randomly selected 20 healthy males from each distance group. The 60 selected participants underwent a 3-month standardized training protocol adjusted for each race distance and guided by a personal trainer. Cardiovascular disease was excluded by a complete cardiovascular screening that included a detailed medical history and physical examination, 12-lead surface electrocardiogram, a treadmill stress test, and transthoracic echocardiography at rest.
The study protocol complied with the Declaration of Helsinki and was approved by the Ethics Committee of our institution, and all participants provided written informed consent.
Echocardiography was performed 24 h before the race and within the first hour of arrival at the finish line. All echocardiographic images were acquired with a commercially available ultrasound system (Vivid Q, GE Medical, Milwaukee, Wisconsin). Images were acquired from the parasternal (long- and short-axis) and apical (4-, 3- and 2-chamber) views. Three cardiac cycles for each acquisition were digitally stored in a cinematic loop format for off-line analysis with commercially available software (version 113, EchoPac, GE Medical). The left ventricle (LV) and RV were assessed according to American Society of Echocardiography and European Association of Cardiovascular Imaging guidelines (20). Maximum and minimum volumes and atrial volume before the atrial contraction (Vpa) were calculated using the biplane summed discs method (20) for both LA and RA. Then, pump, reservoir, and conduction volumetric functions were calculated using previously described formulas (10). All dimensions were indexed for body surface area according to the DuBois formula. RV and LV diastolic functions were assessed from ventricular inflows with peak early (E) and atrial (A) flow velocities and annular velocities of the tricuspid and mitral lateral annulus (e′, a′, respectively) derived from color-coded tissue Doppler acquisitions from the 4-chamber apical view (20). Most participants did not have an adequate tricuspid regurgitation signal for the estimation of pulmonary artery systolic pressure (PASP). Therefore, we used the time to peak-to-ejection time (TP/ET) ratio of the power-wave Doppler signal in the RV outflow tract as a surrogate measurement (21), where higher ratios implied lower PASP and lower ratios implied higher pressures (19). LV and RV stroke volumes were calculated using quantitative Doppler as the product of LV and RV outflow tract area and velocity time integral (22).
Myocardial deformation imaging
Myocardial deformation of both ventricles and both atria was evaluated from 2D echocardiographic grayscale images using speckle tracking (2Dstrain, EchoPac, GE Medical). For atrial strain determination, the focal point was positioned at the mid atrial level in an apical 4-chamber view, whereas for ventricular evaluation, the focal point was moved to the mid segments of both ventricles; all images were optimized to guarantee optimal endocardial delineation. Particular care was taken to ensure acquisitions with frame rates between 60 and 80 frames/s. The regions of interest for both atria were manually indicated by tracing around the endocardial surface of the lateral wall, superior wall, and atrial septum. LA and RA strains were calculated with the reference point set at the onset of the P wave of the surface electrocardiography; the software divided the atrial wall into 6 segments, and the average was taken for analysis. Figure 1 shows the different strain and strain rate parameters. The peak negative strain and the peak negative strain rate during atrial systole represent atrial contraction (LASa, LASRa and RASa, RASRa, respectively). The sum of the absolute values of the negative and positive strain peaks (LASt and RASt) and the positive strain rate (LASRs and RASRs) represent the atrial reservoir function (23). LV global longitudinal strain (GLS) was measured as the average of the LV longitudinal strain peaks obtained in 2-, 3- and 4-chamber views. RVGLS was measured as the average of 6 RV segments (3 RV-free wall [FW] and 3 interventricular septum segments) and RV-FW peak systolic strain as an average of the 3 RV-FW segments (23).
Evaluation of dehydration
Hydration status was estimated by measuring body mass water, weight, hematocrit, and serum sodium before and after the race (Online Methods 1.1).
For statistical analysis, see Online Statistical Analysis 1.2.
Baseline population characteristics
Among the 60 individuals included in the study, 5 participants were excluded due to incomplete follow-up, and 1 was reassigned from group S to group M because he trained between 3 and 4 h/week and performed the medium distance training protocol. Therefore, the final group was composed of 55 men: 17 in group S, 21 in group M, and 17 in group L.
Participants in the 3 groups were of similar ages and had similar systolic and diastolic blood pressures, but body mass index was significantly higher in group S than in groups M and L. According to the allocation method, participants in groups M and L had more training load, despite similar years of training (Table 1).
Baseline ventricular and atrial performance
Table 1 lists the baseline characteristics of the overall study population grouped according to race distance. At baseline, RA and LA maximum volumes were gradually enlarged across the 3 groups, from short- to long-distance group. No significant differences in baseline atrial function parameters were observed among the 3 groups (Table 1).
At rest, no significant differences were observed in ventricular dimensions and function, except for a trend toward gradually increasing RV volume from the short- to the long-distance group (Table 1).
Ventricular performance after the race
Table 2 shows ventricular dimensions and function at baseline and after the race in the whole study population and in each of the 3 distance groups. After the race, the TP/ET ratio, an inverse surrogate for PASP, decreased progressively as distance increased, implying a progressive increase in PASP with increasing amount of exercise performed. Globally, RV dilation and a decrease in RV function were observed after the race. The RV was significantly dilated in the M and L groups; RVGLS and right ventricular free-wall longitudinal strain (RVFWLS) were decreased in group M and even more in group L. There were no significant changes in group S.
No changes in LV systolic function and dimensions were observed in any of the 3 groups. E and e′ mitral values were decreased in groups M and L, without changes in group S, whereas A and a′ mitral values were increased in all 3 groups but to a greater extent in group S, resulting in similar E/A and e′/a′ ratios (Table 2).
Atrial performance after the race
Atrial dimensions and deformation at baseline and after the race for the 3 distance groups are shown in Table 3. After the race, LA contractile function assessed by strain and strain rate globally increased without any change in the other atrial function parameters but with a decrease in LA volume (Table 3). However, when results were analyzed according to race distance, LA reservoir function increased slightly in group S, but a decreasing trend was documented in the M and L groups. LAS and LASR during the contractile phase increased in group S, remained unchanged in group M, and showed a trend to decrease in group L. Similar findings in LA function where shown by volumetric data (Online Table 3). LA volume decreased in all 3 groups after the race, but the extent increased progressively with the amount of exercise performed.
Globally, RAS and RASR during reservoir phase significantly decreased after the race, without changes in RA contractile function and size (Table 3). Nevertheless, remarkable differences were found among groups. RA reservoir function decreased in group M and even more in group L, with no significant changes in group S. RA contractile function decreased in group L, remained unchanged in M, and increased in group S. Similar results in RA function were observed by volumetric data (Online Table 3). RA volume did not change significantly in any of the 3 groups.
The decrease in RV function by RVGLS and RVFWLS correlated with a decrease in RA reservoir function (RASt) (r = +0.43 and r = +0.36, respectively; p < 0.05), and the changes in LVGLS correlated with the change in LASt (r = 0.30; p < 0.05), demonstrating the influence of systolic ventricular function on atria reservoir function.
Analyzing individual atrial response to exercise
In order to better understand the atrial response to exercise, we analyzed the relationship between the change in atrial reservoir and contractile functions before and after the race for the LA and RA in each participant (Figure 2). In the RA, and less pronounced in the LA, 3 patterns of atrial response to exercise were documented, depending on the amount of exercise performed, as follows: an increase in both reservoir and contractile function in athletes running 14 km, a decrease in reservoir function and increase in contractile function in athletes running 35 km, and a decrease in both reservoir and contractile function in 55-km runners. However, high interindividual variability was documented, showing athletes running 35 km who had decreases in both reservoir and contractile functions (42% RA, 17% LA) and athletes running 55 km who were able to increase contractile function in response to the decrease in reservoir function (17% RA; 23% LA) or even increase both of the atrial functions (0% RA; 17% LA).
Reproducibility of the study
Intraobserver and interobserver intraclass correlations were performed in 10 subjects at baseline and post-race measurements. For baseline measurements, intraobserver and interobserver intraclass correlations were 0.98 and 0.96, respectively, for LVGLS; 0.93 and 0.91 for RVGLS; 0.92 and 0.91 for RVFWLS; 0.96 to 0.95 and 0.91 to 0.91 for LAS and LASR, respectively, during reservoir phase; 0.93 to 0.92 and 0.90 to 0.91 for LAS and LASR, respectively, during contractile phase; 0.95 to 0.96 and 0.92 to 0.91 for RAS and RASR, respectively, during the reservoir phase; and 0.92 to 0.93 and 0.90 to 0.89 for RAS and RASR, respectively, during contractile phase. Intraobserver and interobserver intraclass correlation for post-race measurements were 0.93 and 0.90, respectively, for LVGLS; 0.90 and 0.88, respectively, for RVGLS; 0.89 and 0.89, respectively, for RV-FWLS; 0.91 to 0.90 and 0.89 to 0.88 for LAS and LASR, respectively, during reservoir phase; 0.90 to 0.90 and 0.90 to 0.87 for LAS and LASR, respectively, during contractile phase; 0.91 to 0.90 and 0.89 to 0.88 for RAS and RASR, respectively, during the reservoir phase; and 0.90 to 0.90 and 0.89 to 0.87 for RAS and RASR, respectively, during contractile phase.
The current study provides a comprehensive evaluation of RA and LA performance before and after endurance exercise. The study has 4 key findings: 1) an acute, exercise-dose dependent impairment of both atrial and RV function was observed; 2) atrial reservoir function was the first to be affected by the increase in exercise load as the result of the impairment in RV function; 3) the highest exercise load also induced acute impairment in atrial contractile function in both atria; and 4) high individual variability was documented in atrial response to exercise in individuals performing the same amount of exercise.
During endurance exercise, a marked increase in cardiac output is required for long periods of time. This volume overload seems to particularly involve the RV, a cavity that works at much lower pressures at rest than the LV but at dramatically higher pressure during exercise (24). Accordingly, a progressive increase in PASP with increasing exercise load (24,25) leads to increased RV wall stress (26) and potentially to impaired RV function. Indeed, several studies have shown RV dilation and dysfunction after an endurance race without impairment of LV function (13,15,27). In line with these studies, we confirmed reduced RV performance (RV dilation and reduced RV function by GLS and FWLS) in medium- and long-distance runners after the race, but we found no significant changes in LV dimensions and function.
Atrial reservoir function is influenced by ventricular systolic function (28). In our study, RA reservoir function was decreased in medium-distance runners and to a greater extent in long-distance runners, following the same pattern of RV function and confirming a dose-response relationship between exercise load and deterioration in the performance of the right side of the heart.
In line with previous studies (13,15,27), LV systolic function showed no changes after the race in any of the 3 groups, but a decrease in LV filling was documented in medium- and long-distance runners. One could speculate that the decrease in LV filling velocities after the race was a result of dehydration; however, no correlation was found between Δ%E and e′ mitral velocities and dehydration parameters (Online Table 1). We believe that the decrease in LV filling after the race is actually related to the decrease in LA preload, which in turn is due to RV dysfunction, as demonstrated by earlier studies (16). Indeed, a progressive decrease in LA volume with increasing amount of exercise performed was observed after the race, while significant RV dilation and only a slight nonsignificant decrease in RA volume were documented in medium- and long-distance runners. These results suggest that blood is stagnating in the right side of the heart as a result of RV and RA dysfunction.
In our study, we also documented impairment in atrial contractile function, particularly in the RA in long-distance runners. When ventricular afterload is increased, or in case of ventricular systolic dysfunction, atrial reservoir function is the first to be affected (29,30). In contrast, atrial contractile function is maintained or even increased until advanced stages of the disease (29). We also observed that increased RV afterload and dysfunction caused a progressive impairment in RA reservoir function in medium- and long-distance runners, whereas RA contractile function decreased exclusively in long-distance runners, mimicking the steps of atrial response to increased afterload and ventricular dysfunction. Despite unchanged LV systolic function after the race, a trend toward decreased LA contractile function was observed. Decrease in LA contractile function correlated with a decrease in RV function (correlation between Δ% LASa and Δ% LASRa vs. RVGLS: r = +0.43 [p < 0.05]; r = +0.39 [p < 0.05], respectively; vs. RVFWLS: r = +0.33 [p < 0.05]; r = +0.31 [p < 0.05], respectively), suggesting that the decrease in LA preload could play a role in the impairment of LA contractile function after the race.
Although our study confirmed a dose-response relationship between atrial response to exercise and exercise load, there was great variability among individuals performing the same amount of exercise. In most long-distance runners, the race caused acute impairment in atrial reservoir and contractile function; however, this acute deleterious response was also observed in some athletes running medium distance, suggesting the presence of individuals with worse cardiac adaptation to exercise. In contrast, a small group of long-distance runners were able to increase atrial contractile function in response to the decrease in reservoir function, indicating a group of individuals with better cardiac adaptation to exercise. The underlying mechanisms to these different adaptations remain to be clarified.
The influence of training load in acute cardiac adaptation to exercise has shown controversial results in previous studies (14,15). Similar to the findings reported by Trivax et al. (27), in our population, acute atrial and ventricular response to exercise were independent of training load (Online Table 2).
Previous reports have demonstrated that impairment in reservoir and contractile atrial function assessed by strain and strain rate is related to the development of atrial fibrillation and atrial flutter (31) and could predict patient evolution after AF ablation (19). Furthermore, reduced atrial strain during the reservoir phase has also been related to atrial fibrosis (32). These results were confirmed in a “marathon” rat model, in which high-intensity training induced biatrial dilation and fibrosis and susceptibility to atrial arrhythmia (7).
Most studies have demonstrated that AF is triggered and sustained in the LA. However, the most important causative factors for AF in the general population are related to LV disease, and thus the LA is primarily affected. It is important to note that exercise-induced AF patients are underrepresented in these studies. Some studies have suggested that RA size provides prognostic value after AF ablation procedures (33). Moreover, some groups have suggested that ablation of fractionated electrograms in the RA improves outcomes of some patients undergoing AF ablation (34).
In our study, an acute impairment in atrial contractile and reservoir was demonstrated, especially in the RA. Whether these results persist or are reversible remains to be determined. However, our results clear up atrial response to exercise and suggest that the impact of repeated races on atrial function might be responsible for the development of an atrial substrate for arrhythmias in the long term.
This observational study was carried out in a relatively small sample of male participants, limiting its statistical power. We divided study population in 3 distance groups on the basis of previous training loads; thus, acute cardiac performance response to exercise would be not only influenced by exercise duration but also by individual factors when studying different cohorts of runners. Despite the possible confounding nature of using 3 cohorts of runners, we consider that this study design would be the most reasonable approach in order to guarantee that all runners were equally prepared for the race. Echocardiography examinations were performed only before and immediately after the race, without additional follow-up studies; therefore, we cannot determine whether these changes were permanent or transient. We were unable to assess PASP more accurately by using the regurgitant tricuspid flow because most of the participants showed minimal tricuspid regurgitation, with inadequate Doppler signals. Finally, we documented high individual variability in cardiac adaptation to exercise but could not determine the underlying mechanisms involved in these different adaptations.
More accurate functional and structural analysis of pulmonary vasculature, RV and atrial geometry and function, and the relationship between possible individual protective or maladaptive mechanisms would be useful to better understand and define RV and atrial adaptation to exercise.
During a trail-running race, an acute exercise-dose dependent impairment in atrial function was observed, mostly in the RA, which was associated with RV systolic dysfunction and did not seem to be related to previous training. Whether the observed acute impairment in atrial performance persists or is reversible, remains to be clarified. However, the impact on atrial function of long-term endurance training might lead to atrial remodeling, favoring arrhythmia development.
COMPETENCY IN MEDICAL KNOWLEDGE: An acute impairment in both atrial and RV performance was documented after a trail-running race. The decrease in atrial function was more pronounced in the RA, probably due to the direct influence of RV response to exercise. A dose-response relationship was observed between the amount of exercise performed and the impairment in ventricular and atrial performance, but a high interindividual variability was also documented suggesting that there are individual adaptive mechanisms (acquired or genetic) to deal with the high cardiac output required during endurance events.
TRANSLATIONAL OUTLOOK: Understanding cardiac remodeling in response to exercise will be of help to learn when this remodeling starts to be irreversible and potentially pathological with the ultimate risk of posing a substrate for potential arrhythmias. Additionally, learning how different individuals adapt to exercise loads might enable us to adapt personalized training protocols to each individual and potentially, to early identify those individuals with a better pattern of adaptation to exercise.
For supplemental methods, statistical analysis, tables, and references, please see the online version of this article.
This study was funded in part by Generalitat de Catalunya grant FI-AGAUR 2014-2017 (RH 040991) to Dr. Sanz, and Spanish Government Plan Nacional I+D, Ministerio de Economia y Competitividad grant DEP2013-44923-P. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- free-wall longitudinal strain
- global longitudinal strain
- left ventricle end-diastolic volume
- left ventricle ejection fraction
- right ventricle end-diastolic area
- right ventricle fractional area change
- strain during contractile atrial phase
- strain rate during contractile atrial phase
- strain rate during reservoir atrial phase
- strain during reservoir atrial phase
- Received November 6, 2015.
- Revision received March 2, 2016.
- Accepted March 3, 2016.
- American College of Cardiology Foundation
- Molina L.,
- Mont L.,
- Marrugat J.,
- et al.
- Mont L.,
- Elosua R.,
- Brugada J.
- Baldesberger S.,
- Bauersfeld U.,
- Candinas R.,
- et al.
- Andersen K.,
- Farahmand B.,
- Ahlbom A.,
- et al.
- Guasch E.,
- Benito B.,
- Qi X.,
- et al.
- McClean G.,
- George K.,
- Lord R.,
- et al.
- D′Ascenzi F.,
- Pelliccia A.,
- Natali B.M.,
- et al.
- La Gerche A.,
- Burns A.T.,
- Mooney D.J.,
- et al.
- Neilan T.G.,
- Januzzi J.L.,
- Lee-Lewandrowski E.,
- et al.
- Oxborough D.,
- Shave R.,
- Warburton D.,
- et al.
- Wilhelm M.,
- Zueger T.,
- De Marchi S.,
- et al.
- Gabrielli L.,
- Bijnens B.,
- Borràs R.
- Lang R.M.,
- Badano L.P.,
- Mor-Avi V.,
- et al.
- Mor-Avi V.,
- Lang R.M.,
- Badano L.P.,
- et al.
- Bossone E.,
- Rubenfire M.,
- Bach D.S.,
- Ricciardi M.,
- Armstrong W.F.
- La Gerche A.,
- MacIsaac A.I.,
- Burns A.T.,
- et al.
- Trivax J.E.,
- Franklin B.A.,
- Goldstein J.A.,
- et al.
- Rosca M.,
- Lancellotti P.,
- Popescu B.A.,
- Piérard L.A.
- Kurt M.,
- Wang J.,
- Torre-Amione G.,
- Nagueh S.F.
- Kojima T.,
- Kawasaki M.,
- Tanaka R.,
- et al.
- Akutsu Y.,
- Kaneko K.,
- Kodama Y.,
- et al.
- Chen Y.L.,
- Ban J.E.,
- Park Y.M.,
- Choi J.I.,
- Park S.W.,
- Kim Y.H.