Author + information
- Received March 1, 2017
- Revision received June 14, 2017
- Accepted June 30, 2017
- Published online September 13, 2017.
- Maciej Kubala, MDa,∗ (, )
- Laurence Guédon-Moreau, MDb,
- Frédéric Anselme, MD, PhDc,
- Didier Klug, MD, PhDb,
- Geneviève Bertaina, MDa,
- Sarah Traullé, MDa,
- Otilia Buiciuc, MDa,
- Arnaud Savouré, MDc,
- Momar Diouf, PhDa and
- Jean-Sylvain Hermida, MD, PhDa
- aCentre Hospitalier Universitaire, Amiens, France
- bCentre Hospitalier Régional Universitaire, Lille, France
- cCentre Hospitalier Régional Universitaire, Rouen, France
- ↵∗Address for correspondence:
Dr. Maciej Kubala, Service de Rythmologie et Stimulation Cardiaque, Centre Hospitalier Universitaire, Avenue Laennec, 80054 Amiens, France.
Objectives The aim of this study was to evaluate the impact of frailty in the elderly on response to cardiac resynchronization therapy (CRT).
Background CRT has been shown to improve symptoms and outcome of patients with congestive heart failure (HF) and impaired left ventricular ejection fraction (LVEF). The impact of frailty on the results of CRT is unknown.
Methods Frailty defined as <14 of 17 points using the ONCODAGE (Outil de dépistage gériatrique en oncologie) G8 score was assessed before device implantation in candidates for CRT who were >70 years of age. The primary endpoint was the response to CRT, defined as an improvement of >5% of the LVEF and the absence of hospitalization for HF or cardiovascular death at 9 months.
Results Ninety-two of 151 included patients (61%) were frail, and 89 (59%) were responders. Frailty was more frequent in nonresponders: 45 of 62 (73%) versus 47 of 89 (53%) (p = 0.014) and was identified as an independent predictor of nonresponse to CRT (R = 0.30; 95% confidence interval: 0.02 to 0.59; p = 0.039). Frailty was associated with a higher cumulative probability of hospitalization for HF (log-rank p = 0.032) and of all-cause death (log-rank p = 0.033). A G8 score <10.25 correlated with hospitalization for HF or death at 9 months (area under the curve: 0.75; 95% confidence interval: 0.63 to 0.87; cutoff <10.25; 77% sensitivity, 63% specificity).
Conclusions Frailty is as an independent predictor of nonresponse to CRT. Frail patients implanted with CRT devices have a higher risk of hospitalization for HF and mortality. Routine comprehensive geriatric assessment at the time of screening for device therapy should be recommended to optimize management. (Frailty Score Assessment for Elderly Patients Undergoing Cardiac Resynchronization Therapy [FRAILTY]; NCT02369419)
Cardiac resynchronization therapy (CRT) has been shown to improve symptoms and outcome of patients with congestive heart failure (HF) and impaired left ventricular ejection fraction (LVEF) (1–3); however, ≈30% of patients with CRT implanted do not benefit from their device (4,5). Appropriate selection of patients should take into account the presence of factors that are associated with nonresponse. Several demographic, electrocardiographic, and echocardiographic characteristics have been identified that are associated with nonresponsiveness to CRT (6–9). Because the prevalence of HF increases with age, and life expectancy is increasing, the proportion of elderly patients referred for CRT is growing (10,11). Clinical and echocardiographic benefits of CRT have been confirmed in the elderly (12,13); however, the advanced age population is heterogeneous, and frailty has been identified as a pivotal condition in predicting outcome (14,15).
Frailty is a complex condition that includes age-associated deficits in multiple organs with a functional impairment, adverse drug reactions, hospitalization, and heightened vulnerability to stressors (16). Frailty scores systematically evaluate geriatric conditions including physical activity, muscle weakness, slow walking speed, unintentional weight loss, exhaustion, and cognitive and behavioral impairment. Long-term survival in elderly patients hospitalized for HF depends on frailty and association of noncardiovascular comorbidities (17). Frailty has been identified as a risk factor for hospital admission for HF, and routine comprehensive geriatric assessment at the time of HF diagnosis has been recommended by several authors (14,18). The importance of frailty in patients with cardiovascular disease is systematically considered in the establishment of the optimal management strategy (14,19).
Given the complexity of CRT system implantation and the cost of devices and hospitalizations for HF, better estimation of indications for CRT in the elderly is needed. In this context, the prognostic relevance of frailty for hospital readmissions for HF and its effect on the results of CRT remain unclear. The goal of frailty assessment would be to identify frail patients whose management should be optimized before implantation or for whom intervention therapy is inappropriate and who should be treated with conventional medical therapy. The aim of this study was to assess the impact of frailty on the response to the CRT.
This is a prospective multicenter study including patients hospitalized in cardiology units of the Amiens, Lille, and Rouen University hospitals between July 2011 and July 2015. The protocol was registered at ClinicalTrials.gov (NTC02369419). We included patients older than 70 years of age with New York Heart Association (NYHA) functional class II to IV, LVEF ≤ 35%, QRS duration >120 ms, and left bundle branch block (LBBB) in sinus rhythm or in atrial fibrillation (AF) on optimal medical therapy (20). Patients with a conventional pacemaker indication and >95% of ventricular pacing, LVEF ≤35%, and NYHA functional class II to IV were also included. Patients with a high degree of atrioventricular block and LVEF >35% who required de novo CRT with an expected high percentage of ventricular pacing were excluded. After receiving detailed written information, all patients provided informed consent for CRT device implantation and follow-up. The study was approved by the local research ethics committee.
Before implantation, all patients underwent a clinical evaluation, an assessment of NYHA functional class, a 12-lead electrocardiogram (ECG), and a transthoracic echocardiogram. Ischemic heart disease was defined as 1 or more clinically documented (Q wave or enzyme-positive) prior myocardial infarctions or prior coronary artery bypass graft surgeries or percutaneous coronary interventions (balloon or stent angioplasty) (21).
LVEF was evaluated by 2-dimensional transthoracic echocardiography with the Simpson biplane method, magnetic resonance imaging, or angiographic or radionuclide methods. Associated aortic valvulopathy was defined as the presence of moderate (valvular area <1.5 cm2) to severe (valvular area <1 cm2) aortic stenosis or an aortic regurgitation of grade ≥2. All mitral regurgitation of grade ≥2 as evaluated using the proximal velocity surface area method was considered.
QRS duration was evaluated in lead II using the 12-lead ECG (25 mm/s) measurements (GE Marquette Mac 5000, GE Healthcare, Chicago, Illinois). In patients with permanently paced ventriculograms, the QRS duration was measured from the onset of the spike to the end of the QRS complex.
Typical LBBB was defined as QRS duration of ≥140 ms (men) or 130 ms (women), QS or rS in leads V1 and V2, and mid-QRS notching or slurring in ≥2 of leads V1, V2, V5, V6, I, and aVL (22). Atypical LBBB was defined as nonspecific intraventricular conduction delay and QRS widening >150 ms without typical features of LBBB or right bundle branch block (20,23).
Pre-implantation frailty assessment
Frailty was assessed in all patients before CRT implantation by the attending cardiologist using the ONCODAGE (Outil de dépistage gériatrique en oncologie) G8 score, which includes 8 items on a scale of 1 to 17 (Figure 1) (24). The G8 score was chosen because it has been approved as one of the standard valuable screening methods of comprehensive geriatric assessment for predicting outcome (24,25). A score cutoff <14 points as identified in previous studies was required for a diagnosis of frailty (24–26).
Neuropsychological problems evaluated by G8 score including dementia and depression were examined as follows. The cognition was first assessed by the cognitive disorders examination evaluating spatial orientation with a 5-question scale giving 1 point to each correct answer (27). Cognition disorder was ruled out if the score was ≥4. In case of <4 points, a complete Mini-Mental State Examination was carried out with <10 points required for a diagnosis of severe cognition disorder and a score between 10 and 19 points for a diagnosis of moderate cognition disorder (28). The search for mood disorders was first made using the 4-question Mini-Geriatric Depression Scale, with 1 point given for each affirmative answer and a result <1 required to exclude mood disorder (29,30). If the score was ≥1, mood disorder was evaluated with the Center for Epidemiologic Studies Depression Scale (CES-D) questionnaire, which has a scale of 1 to 30 points (Online Figure 1) (31). A CES-D score >21 points was required for a diagnosis of severe mood disorder, with a score between 16 and 21 required for a diagnosis of moderate mood disorder (31).
Weight loss was defined as unintentional loss of at least 1 kg of weight during the past 3 months preceding the implantation attributed to decreased appetite or food intake. To avoid situations in which the weight loss could be attributed to increased use of a diuretic agent, patients with changes in dose of diuretic agents during the past 3 months were not enrolled. Anorexia, as defined by G8 score, was a moderate or severe decrease in food intake over the past 3 months due to loss of appetite or chewing or swallowing difficulties (25).
Implantation and device programming
All enrolled patients underwent device implantation with standard transvenous techniques. Implantable cardioverter-defibrillator (ICD) therapy was chosen according to the American College of Cardiology/American Heart Association/Heart Rhythm Society 2008 Guidelines for Device-Based Therapy of Cardiac Rhythm Abnormalities (32). In the absence of a history of sustained ventricular tachycardia or ventricular fibrillation, the choice between a CRT defibrillator (CRT-D) and a CRT pacemaker (CRT-P) device was left to the discretion of the electrophysiologist. Very high-risk patients (defined by blood urea nitrogen 50 mg/dl or serum creatinine >2.5 mg/dl) with expected attenuation of the efficacy of an ICD were implanted with CRT-P devices (33). The right ventricular lead was positioned in the septum or, failing that, in the apex. The left ventricular (LV) lead was preferably placed in a posterolateral or lateral cardiac vein. Stimulation of basal LV segments was preferably programmed (34). The sensed atrioventricular delay was programmed at 100 ms with a 30-ms extension in patients in sinus rhythm. The VV delay was initially programmed at 0 ms. Pacing rate was programmed at 50/min in patients in sinus rhythm and at 70/min in patients in AF using a rate-adaptive mode. Anteroposterior and left anterior oblique 45° chest radiographs were systematically performed before discharge.
Clinical ECG evaluation and device testing were performed at 3- and 9-month visits and echocardiographic evaluation at 9 months. All adverse events were documented: cardiovascular (sudden cardiac death or death of HF) or noncardiovascular deaths, with the causes specified. In case of nonimprovement of clinical and hemodynamic status (based on NYHA functional class and clinical examination) at the 3-month visit or a hospitalization for HF, atrioventricular and VV delays were adjusted according to the optimal aortic ejection volume obtained by Doppler evaluation (35).
Definition of response to CRT
We chose a combined endpoint based on echocardiographic improvement of LVEF of >5% and the absence of major clinical events in relation to heart disease, including cardiovascular death or hospitalization for HF. The diagnosis of HF was based on symptoms and responsiveness to intravenous decongestive therapy observed by experienced cardiologists unaware of the study protocol.
Primary and secondary endpoints
The primary endpoint was the impact of frailty as assessed by G8 score on the response to CRT in patients older than 70 years of age implanted with biventricular devices. The secondary endpoints were the effects of frailty on deaths of any cause, hospitalizations, and the rate of redo procedures after CRT device implantation.
For all statistical analysis, we used the SPSS software package version 9.0 (SPSS Inc., Chicago, Illinois). Discrete variables were reported as percentages and continuous variables as mean ± SD. Standardized differences were estimated using Cohen’s effect size, with values of 0.2, 0.5, and 0.8 considered low, medium, and high, respectively.
Differences between groups were tested with the chi-square test or Fisher exact test. For continuous variables, Student t test or Mann-Whitney U test were used. A 2-sided p < 0.05 was considered statistically significant. The absolute change in LVEF between baseline and 9-month follow-up was evaluated by paired-sample t test.
Event-free survival was compared by the Kaplan-Meier method and log-rank statistic. To evaluate potential predictors of nonresponsiveness to CRT, only variables with p < 0.04 on univariate analysis were included in the multivariate logistic regression model giving odds ratios and 95% confidence intervals (CIs). A generalized linear mixed model was used to exclude a multicenter effect. Given a 30% expected rate of nonresponders to CRT and the unknown proportion of frail patients as evaluated by G8 score in this population, the number of subjects needed to obtain a significantly powered result was estimated at 150 using nonparametric tests (4,5). The α and β risks were set at 5% and 20%, respectively. Receiver-operating characteristic analysis was performed to determine the optimal cutoff value of the G8 score associated with hospitalizations for HF and mortality (36,37).
A total of 155 of the 161 initially recruited patients (96%) underwent successful CRT device implantation. Four patients were lost to follow-up; therefore, 151 patients were included in the analysis. The baseline characteristics of all included patients are shown in Table 1.
Prevalence of frailty before resynchronization therapy
Ninety-two (61%) of 151 patients in whom a CRT device was implanted had a G8 score <14 and were identified as frail. No significant difference was found between frail and nonfrail subjects according to device implantation data except for the proportion of implanted CRT-D compared to CRT-P devices, which was higher in nonfrail patients (46 of 59 [78%] vs. 47 of 92 [51%]; p = 0.001) (Table 2).
Response to CRT
After 9 months of follow-up, 89 of 151 patients (59%) were responders. Mean increase in LVEF was 9.8 ± 10.5%. Eleven patients (7%) were hospitalized for HF, and 10 (6.5%) died of cardiovascular causes. Frailty was associated with nonresponse to CRT (45 of 62 [73%] vs. 47 of 89 [53%]; p = 0.014) (Table 3). Changes in NYHA functional class, LVEF, LV end-diastolic diameter, and QRS duration at 9-month follow-up in frail and nonfrail patients are shown in Table 4 and illustrated in Figure 2.
Persistent or permanent AF (p = 0.028; R = 0.34; 95% CI: 0.04 to 0.64) and frailty (p = 0.014; R = 0.30; 95% CI: 0.02 to 0.59) were identified as independent predictors of nonresponse to CRT in multivariate analysis. Percentage of biventricular pacing was similar in patients with persistent/permanent AF and those in sinus rhythm (95 ± 7% vs. 96 ± 12%; p = 0.570) and in frail versus nonfrail patients (95 ± 11% vs. 95 ± 8%; p = 0.686).
Hospitalizations, redo procedures, and ICD therapies
Rates of hospitalizations, reinterventions, and ICD therapy are shown in Table 5. There were 35 hospitalizations related to cardiovascular disease, which included 11 (31%) for HF, 17 (48%) for a heart rhythm disorder, and 2 (6%) for acute coronary syndrome. Five other hospitalizations were caused by 1 atrioventricular node ablation, 1 pocket hematoma, 1 healing delay, 1 regressive inflammatory syndrome, and 1 atypical chest pain.
Cumulative survival without hospitalization for HF was better in nonfrail patients (98% vs. 89%; log-rank p = 0.032) (Figure 3). The 14 reinterventions included 10 LV lead repositionings, 1 additional LV lead implantation, 1 LV lead exchange, 1 device extraction for pocket infection, and 1 drainage of pocket hematoma.
Cardiovascular, noncardiovascular, and overall mortality
Fifteen of 151 patients (10%) died during the follow-up. Ten cardiovascular deaths were caused by HF. The causes of the 5 noncardiovascular deaths were colon cancer (1), esophageal cancer (1), pneumonia (1), septic shock (1), and unexplained death without preceding cardiovascular symptoms (1).
Frailty was associated with a higher cumulative probability of all-cause death (96% vs. 86%; p = 0.033). Cumulative survival curves for all-cause death, cardiovascular death, and noncardiovascular death in frail and nonfrail patients are illustrated in Figures 3 and 4.
A G8 score <10.25 correctly predicted hospitalization for HF or death (area under the curve: 0.75; 95% CI: 0.63 to 0.87; cutoff <10.25; 77% sensitivity, 63% specificity) (Figure 5).
This study establishes a strong relation between frailty and worse outcome in the population of elderly patients with HF treated with CRT. The major findings are as follows: 1) frailty, as defined by G8 score <14, is an independent predictor of nonresponse to CRT in patients >70 years of age in whom a CRT device has been implanted for conventional indications; and 2) frail patients have an increased risk of hospitalization for HF and an increased risk of all-cause death.
Frailty and response to CRT
For years, research on predictors of long-term results of CRT has been based on the assessment of conventional clinical, electrocardiographic, and echocardiographic parameters. Elderly patients with decreased LVEF are at high risk for HF and are most in need of treatment; however, the benefit for an individual patient with decreased physiological reserve and comorbidities remains difficult to evaluate. A detailed evaluation of frailty, routinely performed in geriatric units, can be applied to patients affected with HF in addition to conventional examination. Frailty scoring systems have demonstrated the potential to provide a valuable assessment of mortality and preoperative risks in previous studies (37,38). Several authors have reported the association of frailty with the risk of more frequent and longer hospitalizations in the general population (39,40). The same trend to decompensate at a lower threshold has been demonstrated in patients with HF (18,41). Furthermore, frailty syndrome has been reported as common in candidates for CRT (42), which was confirmed in our study. The G8 score, validated in previous geriatric series, appeared in our study to be a valuable tool to identify frailty in a similar proportion of elderly patients with HF as described previously (36,37). The proportion of nonresponders to the CRT was similar to previous series concerning the elderly (13). The cutoff value of <14 identified patients less likely to benefit from their device. Frailty has been identified as a predictor of HF in patients with nonischemic cardiomyopathy undergoing CRT (43). Our study confirmed an increased risk of hospitalization for HF in frail patients with severe impairment of LV systolic function, expanding these results to the population with all-cause cardiomyopathies.
Separate evaluation of all geriatric conditions included in the G8 score showed that some were associated with nonresponse to CRT; however, the total score, representing a more complete and objective assessment, was identified as an independent predictor of nonresponse to CRT. Interestingly, despite a similar shortening of QRS width, frail patients had a significantly lower increase in LVEF.
The G8 test can be easily accomplished in 10 min and included in clinical practice. The score can be integrated in the discussion with the patient and taken into account while choosing the optimal therapy. Frailty has been described as a reversible condition (44). Hence, in mildly frail patients (G8 score >10.5 to 14), we believe that implantation of CRT device should be discussed individually for each patient. An etiological evaluation of decreased physiological status to rule out all underlying neoplastic processes should be systematic. Then a specific therapy to limit the progression of frailty, including measures to improve physical activity and nutritional and cognitive impairment, along with CRT, could be proposed.
This study also showed that the patients with persistent or permanent AF were less likely to respond to CRT therapy. Previous results of biventricular pacing in patients with chronic AF reported in observational studies were controversial (45–47). Patients in our series were older than in trials that showed a benefit of CRT in AF (79 years of age vs. 68 years of age in the study by Kiès et al.) (47).
Mortality and hospitalizations for HF
In this study, frail patients had a higher risk of mortality and hospitalization for HF. The total and cardiovascular mortality in our study (10% and 6.5%, respectively) was in line with the results of other series. Cardiovascular mortality of 7% was reported in the MADIT-CRT (Multicenter Automatic Defibrillator Implantation Trial–Cardiac Resynchronization Therapy) trial in younger patients (65 years of age vs. 79 years of age) and with a longer follow-up (28 months vs. 9 months) than in our series (21).
The presence of frailty in patients with HF has been described as an additional risk factor for mortality (17). Our study confirmed this finding in the population with severe LV systolic dysfunction. A G8 score with a cutoff value of <10.25 identified patients with a significantly higher risk of hospitalization for HF or death. The impact of frailty on mortality in patients undergoing CRT should be taken into account during initial screening for device therapy. We believe that a conventional drug treatment should be preferred in candidates with a G8 score <10.25.
The higher rate of hospitalization for HF results in increased health service care. In our study, a frailty assessment identified patients at higher risk of rehospitalization for HF. In this context, the G8 score can be useful to identify patients likely to be hospitalized and needing an optimized therapy.
Moreover, frailty is associated with increased vulnerability to stressors, including hospitalizations or surgery (48). This study also demonstrates that in patients undergoing less invasive interventions than cardiac surgery, frailty is also associated with higher risk of mortality. CRT device implantation in frail patients, if indicated, should involve pre-operative and post-operative surveillance optimization.
Frailty appears to be one of the missing parameters not captured by conventional evaluation before cardiac resynchronization. The influence of frailty both on results of biventricular pacing and on outcome shows the limits of device therapy. All cardiovascular deaths in our study were caused by HF, and there were no sudden deaths. The recently published DANISH (Danish Study to Assess the Efficacy of ICDs in Patients With Nonischemic Systolic Heart Failure on Mortality) study has suggested a possible lack of benefit of ICD therapy with regard to death of any cause in nonischemic cardiomyopathy among older patients (>68 years of age) (49). In this context, frailty assessment appears to be an appropriate tool to improve selection of elderly candidates for CRT. Further studies are needed to demonstrate whether frailty scores could be helpful to identify a subgroup of patients less likely to benefit from ICD therapy.
The relationship of frailty with HF remains challenging given the similar outcome of both entities. Association of frailty with HF resulting in exacerbation creates a vicious circle that leads to higher mortality and increased risk of hospitalization. This bidirectional relationship has not been fully understood, and in our opinion, both subclinical HF and pre-frailty should be targeted earlier.
The G8 score used previously in oncogeriatric evaluations has lower specificity than other scores to detect cognitive disorders, although the G8 score is easier to use in clinical practice and is of higher value to evaluate nutrition impairment. In this study, the impact of biventricular pacing on reversibility of frailty in patients who responded positively to CRT was not studied.
Frailty is as an independent predictor of nonresponse to CRT in the elderly. Frail patients have a higher risk of hospitalization for HF and mortality. Routine comprehensive geriatric assessment at the time of screening for device therapy should be recommended to optimize management. A G8 score cutoff value of 10.25 can be proposed for better selection of candidates for CRT. Further evaluation and better understanding of the bidirectional effects of frailty and HF is needed.
COMPETENCY IN MEDICAL KNOWLEDGE: Frailty assessment in the elderly before CRT for conventional indications identifies a subgroup of frail patients less likely to benefit from their devices. Frail patients identified using the G8 score have a higher risk of hospitalization for HF and mortality.
TRANSLATIONAL OUTLOOK: The G8 score cutoff value of 10.25 can be used for better selection of candidates for CRT. In mildly frail patients with G8 score >10.5 to 14, specific measures to improve reversible reduced mobility and nutritional and cognitive impairment can be proposed along with CRT. The easily evaluable G8 test can be included in routine practice in addition to conventional clinical evaluation at the time of screening for device therapy.
For a supplemental figure, please see the online version of this paper.
This work was supported in part by Biotronik (for data processing). Dr. Kubala has received honoraria from LivaNova and Boston Scientific. Dr. Guédon-Moreau has received consulting fees from LivaNova, Medtronic, St. Jude Medical, Boston Scientific, and Biotronik. Dr. Anselme has received honoraria and consulting fees from LivaNova, Medtronic, St. Jude Medical, Boston Scientific, and Biotronik. Dr. Klug has received consulting fees from Livanova, Medtronic, St Jude Medical, Boston Scientific, and Biotronik. Dr. Hermida has received consulting fees from St. Jude Medical and LivaNova. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- confidence interval
- cardiac resynchronization therapy
- cardiac resynchronization therapy pacemaker
- heart failure
- implantable cardioverter-defibrillator
- left bundle branch block
- left ventricular
- left ventricular ejection fraction
- New York Heart Association
- Received March 1, 2017.
- Revision received June 14, 2017.
- Accepted June 30, 2017.
- 2017 American College of Cardiology Foundation
- Linde C.,
- Leclercq C.,
- Rex S.,
- et al.
- Gold M.R.,
- Thébault C.,
- Linde C.,
- et al.
- Bordachar P.,
- Lafitte S.,
- Reuter S.,
- et al.
- Roger V.L.,
- Go A.S.,
- Lloyd-Jones D.M.,
- et al.,
- on behalf of the American Heart Association Statistics Committee and Stroke Statistics Subcommittee
- Chaudhry S.I.,
- Wang Y.,
- Gill T.M.,
- Krumholz H.M.
- Chaudhry S.I.,
- McAvay G.,
- Chen S.,
- et al.
- Brignole M.,
- Auricchio A.,
- Baron-Esquivias G.,
- et al.
- Zareba W.,
- Klein H.,
- Cygankiewicz I.,
- et al.,
- on behalf of the MADIT-CRT Investigators
- Kenis C.,
- Decoster L.,
- Van Puyvelde K.,
- et al.
- Epstein A.E.,
- DiMarco J.P.,
- Ellenbogen K.A.,
- et al.
- Goldenberg I.,
- Vyas A.K.,
- Hall W.J.,
- et al.
- Kydd A.C.,
- Khan F.Z.,
- Watson W.D.,
- Pugh P.J.,
- Virdee M.S.,
- Dutka D.P.
- Stortecky S.,
- Schoenenberger A.W.,
- Moser A.,
- et al.
- Mlynarska A.,
- Mlynarski R.,
- Biernat J.,
- Sosnowski M.,
- Golba K.S.
- Kiès P.,
- Leclercq C.,
- Bleeker G.B.,
- et al.
- Køber L.,
- Thune J.J.,
- Nielsen J.C.,
- et al.,
- for the DANISH Investigators