Author + information
- Received April 4, 2018
- Revision received August 23, 2018
- Accepted August 30, 2018
- Published online January 7, 2019.
- Haotian Gu, PhDa,
- Sahrai Saeed, MD, PhDb,
- Andrii Boguslavskyi, PhDc,
- Gerald Carr-White, PhDd,
- John B. Chambers, MDd,∗ and
- Phil Chowienczyk, BSca,∗∗ ()
- aBritish Heart Foundation Centre, King’s College London, London, United Kingdom
- bHaukeland University Hospital, Bergen, Norway
- cClinical Research Facilities, Guy's and St Thomas's Hospital, London, United Kingdom
- dCardiothoracic Centre, Guy's and St Thomas’ Hospital, London, United Kingdom
- ↵∗Address for correspondence:
Dr. Phil Chowienczyk, Department of Clinical Pharmacology, St Thomas's Hospital, London SE1 7EH, United Kingdom.
Objectives This study investigated the prognostic value of first-phase ejection fraction (EF1) in patients with aortic stenosis (AS), a condition in which left ventricular dysfunction as measured by conventional indices is an indication for valve replacement.
Background EF1, the ejection fraction up to the time of maximal ventricular contraction may be more sensitive than existing markers in detecting early systolic dysfunction.
Methods The predictive value of EF1 compared to that of conventional echocardiographic indices for outcomes was assessed in 218 asymptomatic patients with at least moderate AS, including 73 with moderate, 50 with severe, and 96 with “discordant” (aortic area <1.0 cm2 and gradient <40 mm Hg) AS, all with preserved EF, followed for at least 2 years. EF1 was measured retrospectively from archived echocardiographic images by wall tracking of the endocardium. The primary outcome was a combination of aortic valve intervention, hospitalization for heart failure, and death from any cause.
Results EF1 was the most powerful predictor of events in the total population and all subgroups. A cutoff value of 25% (or EF1 of <25% compared to ≥25%) gave hazard ratios of 27.7 (95% confidence interval [CI]: 13.1 to 58.7; p < 0.001) unadjusted and 24.4 (95% CI: 11.3 to 52.7; p < 0.001) adjusted for other echocardiographic measurements including global longitudinal strain, for events at 2 years in all patients with asymptomatic AS. Corresponding hazard ratios for all-cause mortality in the total population were 17.5 (95% CI: 5.7 to 53.3) and 17.4 (95% CI: 5.5 to 55.2) unadjusted and adjusted, respectively.
Conclusions EF1 may be potentially valuable in the clinical management of patients with AS and other conditions in which there is progression from early to late systolic dysfunction.
Aortic stenosis (AS) is the most common form of primary heart valve disease in economically developed countries, with a prevalence of moderate or severe disease of approximately 3% in a population ≥75 years of age (1). It is often asymptomatic, but when symptoms develop and/or there is a reduction in left ventricular ejection fraction (LVEF), survival without intervention is poor (2). Symptoms or reduced EF are thus Class I indications for aortic valve replacement (AVR) in severe AS, whereas the management of asymptomatic severe AS remains less certain (3–5), particularly for patients in whom measurements of AS severity by valve area and pressure gradient are “discordant” (6,7). Mortality soon after the onset of symptoms before surgery can be performed is up to 15% (8). An objective measurement that predicts impairment in EF and/or imminent onset of symptoms would, thus, be invaluable in selecting high-risk asymptomatic patients with preserved EF for more frequent clinical monitoring or early surgical intervention.
The biophysics of myocyte contraction suggest that, when systolic function is impaired early in systole, a mechanism may exist to preserve overall LVEF at the expense of a slower sustained contraction (9–11). A novel measurement of early EF is the first-phase EF (EF1) (Figure 1), the EF up to the time of maximal ventricular fiber shortening. This may be a fundamental characteristic of LV function that determines subsequent events in systole through mechanosensing (10). In a pilot study, the present authors observed a marked impairment of EF1 in AS, despite preserved overall EF (Haotian Gu, John Chambers, and Phil Chowienczyk, unpublished data, August 2016). EF1 was inversely related to the severity of AS, suggesting it could be a sensitive prognostic marker of future systolic function and subsequent events in AS. The present study examined the value of EF1 in predicting the onset of symptoms requiring valve intervention or major events in a cohort of patients with moderate to severe AS. EF1 values were compared with conventional measurements of valve and LV functions.
A retrospective study of the predictive value of EF1 for major adverse cardiovascular events was undertaken in consecutive asymptomatic patients with at least moderate AS seen in a specialist valve clinic at Guy's and St Thomas's Hospital between January 2008 and December 2015, using anonymized data from the Guy's and St Thomas's Valve Study Group. Data included demographics, symptoms, and risk factors including smoking, hypertension, diabetes, coronary artery disease (defined as a history of angina pectoris or evidence of a coronary artery stenosis of >50% on angiography), and chronic kidney disease as well as outcome data for all patients.
All patient clinical records were entered in a hospital digital database (electronic patient record, iSOFT Group, Aldershot, UK). Outcome data were verified from hospital records. Symptoms were carefully evaluated by one of the authors (J.C.) at the time of presentation, and for patients in whom symptoms were uncertain, an exercise tolerance test was administered. Inclusion criteria consisted of asymptomatic AS of at least moderate severity, with LVEF ≥50%. Exclusion criteria included resting arrhythmia, more than moderate disease of other valves, suboptimal echocardiographic window and loss to follow-up. The primary outcome was a combination of aortic valve intervention (because of developing symptoms or EF <50%), hospitalization for heart failure and all-cause mortality. All-cause mortality was examined as a secondary outcome.
Echocardiography and first-phase ejection fraction
Transthoracic echocardiography was performed with the patients at rest, using a Vivid 7 ultrasonography platform (GE Healthcare, Andover, Massachusetts). All echocardiographic measurements were performed using standard techniques according to the recommendations of the American society of Echocardiography (12). LV mass was measured from a 2-dimensional (2D) parasternal view according to American Society of Echocardiography recommendations. LV mass index (LVMi) was calculated by indexing LVM to body surface area (BSA). Left atrial volume (LAV) was measured by tracing the left atrium border from apical 4- and 2-chamber views and left atrial volume index (LAVi) obtained by indexing LAV to BSA. Stroke volume was calculated as the difference between end-diastolic (ED) and end-systolic (ES) volumes indexed to BSA to give stroke volume index (SVi). Transaortic flow rate (TAFR) was defined as stroke volume over ejection time. Tissue Doppler measurements were obtained at levels of the lateral and septal mitral annulus to obtain an optimal spectral Doppler waveform. The E/e′ ratio was calculated as a measurement of diastolic function from the ratio of the transmitral Doppler E wave velocity to the mean of basal lateral and septal tissue Doppler e′ waves (13). Effective aortic orifice area was calculated using the continuity equation and indexed to BSA (the aortic valve area index [AVAi]). Moderate AS was defined as peak AV velocity of 3.0 to 4.0 m/s, mean pressure gradient (MPG) of 20 to 40 mm Hg, and an AVA of 1.0 to 1.5 cm2. Severe AS was defined as peak AV velocity of >4.0 m/s, an MPG of >40 mm Hg, and an AVA of <1.0 cm2. Patients with MPG <40 mm Hg and AVA <1.0 cm2 were defined as “discordant” (6,7). Current algorithms (14,15) further divide this discordant group by SVi with a cutoff point of 35 ml/m2 as a surrogate for flow. In this study, the discordant group was also divided by calculated transaortic flow rate (±200 ml/ms) (16), by SVi and EF1, and by global longitudinal strain (GLS) as further measurements of LV dysfunction, using cutoff points derived from receiver-operating characteristic (ROC) curve analysis, as described below.
Additional image analysis was performed by 2 authors who were blinded to clinical outcomes on archived images, using EchoPAC analysis software (EchoPAC, GE Healthcare). GLS was measured by speckle tracking echocardiography, by placing 6 points along the endocardial border and adjusting the width of interest to accommodate myocardial thickness from apical 4-, 2-, and 3-chamber views. LV volumes were measured using the 2D Simpson’s method from apical 4- and 2-chamber views at end-diastole, time of peak aortic valve flow, and end-systole. EF was calculated as the percentage of change of LV volume from end-diastole to end-systole. EF1 was taken as the percentage of change in LV volume from end-diastole to the time of peak AV flow (TPAVF) (Figure 1), a time that approximates the time of peak ventricular contraction in individual myocytes. EF1 was thus derived by: EF 1 = (EDV − V1)/EDV × 100%, where EDV is ED volume, and V1 is volume at time of peak AV flow.
Intra- and inter-observer variabilities in measurements of EF1 were assessed in 18 randomly selected subjects by measurements repeated 2 months apart by 2 observers, with the coefficient of variation defined as the ±SD of differences in measurements expressed as a percentage of the mean measurement.
Continuous variables were tested for normality, and those following a Gaussian distribution were presented as mean ± SD. Other variables are median (interquartile range [IQR]). Groups were compared by using Student t-test for continuous variables or by chi-square test for categorical variables. ROC analysis was performed to examine the sensitivity and specificity of measurements for events, and the best cutoff value for predicting events was determined to maximize values of sensitivity and specificity. Kaplan-Meier curves were used to examine cumulative event rates, and differences between groups were tested using the log rank test. Univariate Cox regression analysis was performed to identify potential predictors of events. Multivariate Cox proportional hazards models were used to test the independent value of echocardiographic measurements for predicting future events. Analyses were repeated for the total population and in patients with moderate, severe, and discordant AS and in patients with EF in the lowest tertile. A 2-tailed p value of <0.05 was considered statistically significant. We assessed hazard ratios (HRs) after 2 years of follow-up because this is a time frame over which risk would inform clinical management (and ultimately all patients have events). All the above-described statistical analyses were performed using SPSS version 24 software (for Macintosh, SPSS Inc. Chicago, Illinois). Categorical net reclassification improvement (assessed using R software [version SurvIDINRI V1.1-1, Hajime Uno, R Foundation for Statistical Computing, Vienna, Austria) was used to determine the added value of EF1 to conventional indices in predicting events.
Patient characteristics and major events
A total of 333 patients meeting inclusion criteria were identified, of whom 115 were excluded from the final analysis due to arrhythmias (n = 28), more than moderate valve disease other than AS (n = 37), loss to follow-up (n = 21), and suboptimal echocardiographic image (n = 29) (Figure 2). Thus, a total of 218 asymptomatic patients, including 73 (33.5%) with moderate AS, 49 (22.5%) with severe AS, and 96 (44.0%) with discordant AS, were included in the final analysis. Causes of AS included degenerative in 82.1% of subjects, congenital (bicuspid aortic valve) in 17.4%, and uncertain in 0.5%. Most patients who developed symptoms (n = 102) during follow-up either underwent AVR (88 of 102) or died (11 of 102) while waiting for surgery, shortly after developing symptoms. Only 3 patients developed symptoms without undergoing AVR (because of unacceptable risk). In total, 143 patients (65.6%) experienced an event after a median follow-up of 33.4 months (IQR: 21.0 to 51.2 months), including 89 (40.8%) who had AVR, 17 (7.8%) who underwent transaortic AVR, 3 (1.4%) who had balloon aortic valvuloplasty, 2 (0.9%) were hospitalized due to heart failure, and 32 (14.7%) who died (before AVR or valvuloplasy). In patients with coronary artery disease, 13 had both coronary artery bypass graft and AVR, but the primary indication for surgery was AVR in all 13 patients. Event-free survival rates were 70.6%, 43.1%, and 17.0% at 2, 3, and 5 years, respectively. The proportion of patients who experienced events was related to the severity of AS (47.9%, 89.8%, and 66.7% in patients with moderate, severe, and discordant AS, respectively).
EF1 and other echocardiographic measurements
Baseline clinical characteristics and echocardiographic data of patients with moderate, severe, and discordant AS according to events are shown in Table 1. EF and GLS were not significantly different among the 3 groups. Intraobserver and interobserver coefficients of variation for EF1 were 6.7 ± 3.6% and 9.8 ± 6.1%, respectively. Absolute differences in EF1 were 2.9% and 2.1% between observers (Supplemental Figure 2). EF1 was negatively associated with AV MPG (β = −0.192; p = 0.005) with or without adjustment for age, sex, EF, GLS, LVMi, and the TPAVF/ejection time ratio. EF1 was the lowest in patients with severe AS (24.8 ± 11.7%) compared to those with discordance (27.8 ± 9.5%) and moderate AS (30.8 ± 9.2%).
Comparing patients who experienced events to those who did not, EF1 was reduced in all 3 groups; and in those with severe and discordant AS, EF1 was the only measurement of systolic function that was reduced. In patients with moderate AS, MPG was significantly greater and AVAi smaller than in those who had not had events; SVi was significantly lower, and GLS was significantly impaired in those with events than in those who had not had events. The percentages of patients with coronary artery disease and chronic kidney disease were also higher in those with events than in those who had not had events (Table 1).
Prediction of major events by EF1 and other echocardiographic measurements
ROC curve analyses of EF1, EF, GLS, MPG, AVAi, SVi, and TAFR for predicting the primary outcome at 2 years are shown in Figure 3. The area under the curve (AUC) was largest for EF1 (0.927; p < 0.0001), followed by EF, MPG, GLS, AVAi, SVi, and TAFR, and the AUC for EF1 was significantly greater than that for the other predictors (p < 0.0001). A cutoff value of 25% for EF1 had a sensitivity of 87.1% and a specificity of 90.0% for prediction of events. For GLS a cutoff-value of −15% had a sensitivity of 56.5% and a specificity of 66.7%. The incremental value of EF1 to a model including EF, GLS, LVMi, and MPG, in terms of net reclassification, was 67.3% (95% confidence interval [CI]: 50.5% to 77.7%; p < 0.001).
Kaplan-Meier analysis showed that EF1 was a strong predictor of events in the total population over the whole follow-up period (Figure 4A). When EF1 was <25%, the 1-, 2-, and 5-year event rates were 36.2%, 78.3%, and 94.2%, respectively, compared to 1.3%, 6.7%, and 77.9%, respectively, in patients with an EF1 ≥25%. Comparing tertiles of the distribution of EF1, only 2 of 72 patients (2.8%) in the upper tertile had an event at 2 years, whereas 53 of 73 patients (72.6%) in the lower tertile had events. EF1 was a stronger predictor of events than GLS (Supplemental Figure 1). EF1 had strong predictive value in all patients regardless of severity (moderate, severe, or discordant) (Figures 4B to 4D). In the discordant group, EF1 was a stronger predictor of events (p < 0.001) than GLS (p = 0.049) (Figure 5A), SVi, or TAFR (Figures 5B and 5C).
On univariate Cox regression analysis, several measurements of AS severity (MPG and AVAi) and of LV systolic function (EF, GLS, and SVi) as well as LVMi, LAVi, and TPAVF (Table 2) were related to events at 2 years in the the total population. However, EF1 was more strongly related to events, and the HR for each 1% increase in EF1 was 0.888 (95% CI: 0.865 to 0.911; p < 0.001). Heart rate and ejection time were not significantly related to events, with TPAVF the only timing variable significantly related to events. In multivariate model 1, the independent predictive values of EF1 with GLS, MPG, and LVMi were tested. In the total population, both MPG and EF1 were independently predictive of events, but EF1 was more strongly predictive than MPG (Table 2). In model 2, when all significant predictors in univariate analysis were entered, EF1 remained the most significant predictor of events (Table 2) with the HR for EF1 unchanged despite adjustment for these additional variables. This was also the case when measurements of systolic function were added sequentially in a forward step-wise analysis (Supplemental Table S2). When an alternate endpoint consisting of symptoms, hospitalization, and death was considered, HRs were unchanged (HR: 0.882; 95% CI: 0.860 to 0.906; p < 0.001 compared to HR: 0.885; 95% CI: 0.863 to 0.903; p < 0.001 when AVR was used as the endpoint).
When EF1 was considered a categorical variable (EF1 <25% vs. EF1 ≥25%), HRs were 27.7 (95% CI: 13.1 to 58.7; p < 0.001) unadjusted and 24.4 (95% CI: 11.3 to 52.7; p < 0.001) adjusted for other echocardiographic measurements including GLS for 2-year events in the total population. In patients with EF in the lowest tertile, EF1 remained the strongest predictor (HR: 0.846; 95% CI: 0.802 to 0.894; p < 0.001). Forcing heart rate or ejection time into the final model made no significant differences to the predictive value of EF1 (data not shown).
EF1 was also the strongest predictor of events in all groups with an HR similar to that in the total population. In the discordant group, EF1 was the only significant predictor of events (HR: 0.897; 95% CI: 0.866 to 0.930; p < 0.001) (Table 2), and in patients with moderate and severe AS, EF1 remained a strong predictor in both univariate and multivariate models (Supplemental Table S1).
Prediction of all-cause mortality
Thirty-two patients died before intervention with a median follow-up of 20.8 months (IQR: 13.7 to 34.8 months). One-year and 2-year mortality rates in all patients with a reduced EF1 (<25%) were 7.2% (n = 5) and 21.7% (n = 15) compared with 0.7% (n = 1) and 2.7% (n = 4), respectively, in patients with EF1 ≥25% (log rank 46.3; p < 0.001) (Figure 6). HR for all-cause mortality at 2 years for EF1 <25% versus EF1 ≥25% in the total population were 17.5 (95% CI: 5.7 to 53.3) unadjusted and 17.4 (95% CI: 5.5 to 55.2) adjusted for MPG, GLS, LVMi, and TAFR. TAFR had modest predictive value for all-cause mortality in univariate Cox regression analysis (HR: 0.993; 95% CI: 0.986 to 0.999; p = 0.018) but was not significantly associated with mortality when EF1 was included in the multivariate model.
The present authors hypothesized that, in patients with asymptomatic AS and preserved EF, EF1, a measurement of early systolic function up to the time of maximal AV flow, would have greater prognostic impact than measurements calculated to end-systole. We showed a striking HR of 27.7 (95% CI: 13.1 to 58.7) for an EF1 <25% in predicting a major event within 2 years of presentation. EF1 was superior to 2D GLS and other measurements of LV or aortic valve function (including newer indices of valve function derived from timing of aortic valve flow such as TPAVF ). There is a trend toward a decision for prophylactic surgery in patients with asymptomatic severe AS (18,19), and it is possible that EF1 could be used to guide this decision.
EF1 was equally strongly predictive of events in groups classified with moderate AS, severe AS, and discordant AS. Within the discordant group, current guidelines advocate the use of SVi to subdivide this group into moderate or low-flow severe AS (3,14,15). However, we found no predictive value of SVi within this group. Similarly, other measurements of LV systolic function, flow rate, and GLS did not predict events in the discordant group. This might have been because these measurements are calculated over the whole of systole and did not capture the reduction in early systolic function, which we have shown is prognostically important. A surprising finding was that EF1 was predictive of major events in patients with moderate AS. Only a minority of patients with high-gradient severe AS are thought to progress to low-flow, low-gradient severe AS (20), and it is known that LV function can be affected by coronary artery disease and other risk factors (21,22). The present study provides further corroborative evidence that LV systolic function can change independently of valve function, and it is notable that in our study patients with moderate AS and major events had a higher prevalence of CAD than those without events.
Although our primary outcome was the prediction of events in asymptomatic patients with AS, it is notable that EF1 was also predictive of all-cause mortality in the total population, so that 2-year mortality for all patients in the lowest tertile of EF1 was 21.7% compared to 2.7% for patients with EF1 in the highest tertile. The fact that TAFR was predictive of all-cause mortality when it was considered alone but not when it was added to EF1 is consistent with findings in previous studies (16,23) and, again, highlights the importance of events early in systole.
There are important limitations to our study that should be considered. There was some loss to follow-up of a relatively small number of patients. EF1 was measured by 2 trained observers retrospectively. Patients with a suboptimal echo image in which neither EF1 nor EF could be measured with confidence were excluded from the study. Measurements made in clinical use may exhibit more variability, although this will be reduced by improvements in imaging technology that will allow automatic determination of EF1. We compared EF1 with 2D rather than 3D GLS and 3D GLS is a better predictor of events in AS than 2D GLS (24). Some of the outcomes in our study may be dependent on the specific characteristics of our cohort, and prospective replication in other cohorts will be required before the measurement can be reliably used to inform clinical management. Circulating biochemical biomarkers such as brain natriuretic peptide have also been shown to predict events in asymptomatic AS (25,26) but may lack specificity. Further prospective studies will be required to determine the place of EF1 relative to other biomarkers in the risk stratification of AS, particularly in the various subgroups of AS and to determine the importance of concurrent coronary artery disease.
EF1 is a simple robust measurement of early systolic function, predicting major events in patients with asymptomatic AS and preserved EF. If the findings of the present study are replicated in other cohorts, it could be used to inform clinical management of AS. It may also be useful for identifying early systolic dysfunction under other conditions where progression to late systolic function is associated with a poor prognosis.
COMPETENCY IN MEDICAL KNOWLEDGE: In patients with AS, a reduction in EF is associated with a poor prognosis and is a strong indication for intervention. Mechanosensing within myocytes, however, may act to preserve EF so that a fall in EF is a late event in the evolution of systolic dysfunction. EF1 is a novel measurement of early systolic function that may be more sensitive than EF in detecting systolic dysfunction.
TRANSLATIONAL OUTLOOK 1: The present study is the first to demonstrate the independent prognostic value of EF1 in asymptomatic patients with preserved EF and moderate or severe AS. EF1 was superior to other measurements of LV or aortic valve function. If this is confirmed in other cohorts, EF1 could be used as a marker of increased risk to aid in the decision to perform prophylactic surgery in asymptomatic patients with severe AS.
TRANSLATIONAL OUTLOOK 2: In patients with moderate AS, a reduced EF1 appears to identify a high-risk group, and it is possible that there should be further investigation for coronary disease or treatment of other determinants of LV dysfunction, for example, systemic hypertension. This may help to better manage LV dysfunction despite only moderate AS and may also improve LV recovery when aortic valve intervention is judged to be indicated. Patients with discordant velocity/gradient and AVA measurements are hard to classify and manage. Our preliminary results suggest that EF1 classifies these patients better than conventional measurements, notably SVi.
TRANSLATIONAL OUTLOOK 3: The present study provides proof-of-concept for early systolic function as an important risk factor under other conditions that may be characterized by progression from early systolic to late systolic dysfunction.
↵∗ Drs. Chambers and Chowienczyk contributed equally to this work and are joint senior authors.
Supported by the British Heart Foundation. All authors have received support from the Department of Health through the National Institute for Health Research comprehensive Biomedical Research Centre and Clinical Research Facilities awards to Guy's and St Thomas's Hospital National Health Service (NHS) Foundation Trust in partnership with King’s College London and King’s College Hospital NHS Foundation Trust. Drs. Gu and Chowienczyk are named on a patent application that relates to first-phase ejection fraction. All other authors report they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- aortic stenosis
- area under the curve
- effective aortic valve area
- aortic valve replacement
- first-phase ejection fraction
- global longitudinal strain
- left ventricular mass
- mean pressure gradient
- stroke volume
- transaortic flow rate
- time to peak aortic valve flow
- Received April 4, 2018.
- Revision received August 23, 2018.
- Accepted August 30, 2018.
- Lancellotti P.,
- Magne J.,
- Donal E.,
- et al.
- Baumgartner H.,
- Hung J.,
- Bermejo J.,
- et al.
- Gu H.,
- Li Y.,
- Fok H.,
- et al.
- Reconditi M.,
- Caremani M.,
- Pinzauti F.,
- et al.
- Lang R.M.,
- Badano L.P.,
- Mor-Avi V.,
- et al.
- Nagueh S.F.,
- Appleton C.P.,
- Gillebert T.C.,
- et al.
- Hachicha Z.,
- Dumesnil J.G.,
- Bogaty P.,
- Pibarot P.
- Clavel M.A.,
- Dumesnil J.G.,
- Capoulade R.,
- Mathieu P.,
- Senechal M.,
- Pibarot P.
- Saeed S.,
- Senior R.,
- Chahal N.S.,
- et al.
- Gamaza-Chulian S.,
- Diaz-Retamino E.,
- Camacho-Freire S.,
- et al.
- Kang D.H.,
- Park S.J.,
- Rim J.H.,
- et al.
- Dahl J.S.,
- Eleid M.F.,
- Pislaru S.V.,
- Scott C.G.,
- Connolly H.M.,
- Pellikka P.A.
- Vekshtein V.I.,
- Alexander R.W.,
- Yeung A.C.,
- et al.
- Carroll J.D.,
- Carroll E.P.,
- Feldman T.,
- et al.
- Chahal N.S.,
- Drakopoulou M.,
- Gonzalez-Gonzalez A.M.,
- Manivarmane R.,
- Khattar R.,
- Senior R.
- Nagata Y.,
- Takeuchi M.,
- Wu V.C.,
- et al.
- Clavel M.A.,
- Malouf J.,
- Michelena H.I.,
- et al.