Prognostic value of B-type natriuretic peptide in elderly patients with aortic valve stenosis: the COFRASA–GENERAC study
- Claire Cimadevilla1,
- Caroline Cueff1,
- Guillaume Hekimian1,
- Monique Dehoux2,
- Laurent Lepage3,
- Bernard Iung1,
- Xavier Duval4,
- Virginie Huart5,
- Florence Tubach6,
- Alec Vahanian1,
- David Messika-Zeitoun1,7
- 1Department of Cardiology, Assistance Publique—Hôpitaux de Paris, Bichat Hospital, Paris, France
- 2Department of Biochemistry, Assistance Publique—Hôpitaux de Paris, Bichat Hospital, Paris, France
- 3Department of Cardiac Surgery, Assistance Publique—Hôpitaux de Paris, Bichat Hospital, Paris, France
- 4Centre d'Investigation Clinique 007, Assistance Publique—Hôpitaux de Paris, Bichat Hospital, Paris, France
- 5Centre de Resources Biologiques, Assistance Publique—Hôpitaux de Paris, Bichat Hospital, Paris, France
- 6Department of Epidemiology, Biostatistic and Clinical research, Assistance Publique—Hôpitaux de Paris, Bichat Hospital, Paris, France
- 7INSERM U698 and University Paris 7, Bichat Hospital, Paris, France
- Correspondence to Dr David Messika-Zeitoun, Cardiovascular division, AP-HP, Bichat Hospital, 46 rue Henri Huchard, Paris 75018, France;
- Received 30 October 2012
- Revised 17 December 2012
- Accepted 19 December 2012
- Published Online First 24 January 2013
Objective Previous studies suggested an independent prognostic value of B-type natriuretic peptide (BNP) in aortic valve stenosis (AS) but were impeded by small sample sizes and inclusion of relatively selected young patients. We aimed to evaluate the relationship among N-terminal fragment of proBNP (Nt-proBNP), AS severity, symptoms and outcome in a large cohort of elderly patients with AS.
Design Observational cohort study, COhorte Française de Retrecissement Aortique du Sujet Agé (clinicalTrial.gov number-NCT00338676) and GENEtique du Retrecissement Aortique (clinicalTrial.gov number-NCT00647088).
Setting Single-centre study.
Patients Patients older than 70 years with at least mild AS.
Measurements A comprehensive clinical, biological and echocardiographic evaluation was performed at study entry. Asymptomatic patients were prospectively followed on a 6-months basis and AS-related events (sudden death, congestive heart failure or new onset of AS-related symptoms) collected.
Results We prospectively enrolled 361 patients (79±6 years, 230 severe AS). Nt-proBNP increased with the grade of AS severity and the NYHA class (all p<0.0001) but there was an important overlap between grades/classes. Consequently, diagnostic value of Nt-proBNP for the diagnosis of severe symptomatic AS was only modest (area under the curve of the receiver operator characteristic analysis=0.73). At 2 years, 28 AS-related events occurred among 142 asymptomatic patients prospectively followed. Nt-proBNP was associated with outcome in univariate analysis (p=0.04) but not after adjustment for age, gender and AS severity (p=0.40).
Conclusions The present study clearly highlights the limitations of Nt-proBNP for the evaluation and management of AS patients. Our results suggest that Nt-proBNP should be considered cautiously, at least as a single criterion, in the decision-making process of AS patients especially in the elderly population.
Aortic valve stenosis (AS) is the most common valvular heart disease in Western countries. AS affects 2%–7% of the population after 70 years1 ,2 and its prevalence is going to dramatically increase with the ageing of the population. According to current guidelines, surgery is recommended in symptomatic patients because of a poor survival under medical management since the symptom onset.3 ,4 However, symptoms may be difficult to assess or may be under-reported since patients may unconsciously reduce their physical activity or because elderly patients are often sedentary. On the other side, management of asymptomatic patients remains controversial due to the risk of sudden death without preceding symptoms5 and the risk of irreversible myocardial dysfunction. Therefore, identification of subgroups of asymptomatic patients with severe AS who may benefit from an early surgery is a critical clinical challenge.
B-type natriuretic peptide (BNP) is a peptide secreted by cardiomyocytes in response to pressure overload.6 Previous studies have shown that plasma levels of BNP/N-terminal fragment of proBNP (Nt-proBNP) are elevated in AS, increase with AS severity and symptomatic status7–12 and may provide independent prognostic information.7–9 13–15 However, previous studies were impeded by a small sample size and enrolment of highly selected middle-aged patients not representative of the elderly population we often face in clinical practice. In addition, prognostic thresholds of BNP/Nt-proBNP values selected in these previous studies were within the normal range. Therefore, we aimed to evaluate in a large prospective cohort of elderly patients with AS the relationship among Nt-proBNP, AS severity, symptoms and outcome.
Patients 70 years or older with at least mild degenerative AS (on bicuspid or tricuspid valve leaflet) enrolled between November 2006 and April 2010 in two ongoing prospective studies, COFRASA (COhorte Française de Retrecissement Aortique du Sujet Agé, clinicalTrial.gov number NCT 00338676) and GENERAC (GENEtique du Retrecissement Aortique, clinicalTrial.gov number NCT00647088) aiming at evaluating the determinants of AS occurrence and progression constituted our study population. Asymptomatic patients were enrolled in COFRASA and symptomatic patients in GENERAC. Exclusion criteria were age below 70 years, AS due to rheumatic valve disease or radiotherapy, previous infective endocarditis, more than mild associated valvular disease and severe respiratory or renal insufficiency (creatine clearance≤30 ml/min). Clinical, biological measurements including Nt-proBNP and echocardiographic evaluations were performed at study entry for all patients. Asymptomatic patients were contacted every 6 months and seen at our clinical research centre every year. The study was approved by our regional ethic committees and all patients gave a written informed consent.
Medical history, cardiovascular risk factors and medications were prospectively recorded. A physical examination including blood pressure measurement and ECG was performed at study entry. Body mass index (BMI) was calculated for each patient and obese patients were defined by a BMI>30 kg/m2. Assessment of symptoms was performed by experienced cardiologists blinded to Nt-proBNP values. The extent of dyspnoea was assessed based on the New York Heart Association (NYHA) classification. Asymptomatic patients had to be free of dyspnoea, angina and syncope. Coronary artery disease (CAD) was defined as a history of angina, coronary angioplasty, coronary artery bypass or myocardial infarction. Occurrence of AS-related events (sudden death, congestive heart failure or new onset of symptoms (dyspnoea, angina or syncope)) in asymptomatic patients at study entry was prospectively recorded.
A comprehensive Doppler echocardiography was performed for all patients at baseline. Assessment of AS severity was based on peak velocity, mean gradient and aortic valve area (AVA) calculated using the continuity equation.16 Mild AS was defined by an AVA between 1.5 and 2 cm2, moderate AS by an AVA between 1 and 1.5 cm2, and severe AS by an AVA<1 cm2. Cine loops of apical 4-, 3- and 2-chamber, parasternal long-axis, and short-axis views were obtained and the left ventricle ejection fraction (LVEF) was determined visually or using the Simpson method. LVEF was considered normal if ≥ 50%.
All blood samples were taken at inclusion under identical conditions, at 08:00, after 12 h fast within 24–48 h of the echo and immediately processed. Nt-proBNP was measured by chemiluminescent immunoassay (Dimension Vista, Siemens). Nt-proBNP values were considered normal if less than 300 pg/ml, intermediate between 300 and 700 pg/ml and high above 700 pg/ml as recommended in this elderly population.17 ,18 Creatine clearance was calculated using the modified diet in renal disease formula.
Continuous variables were expressed as mean±SD, median (95% CI) or number of patients (per cent). Comparisons between symptomatic and asymptomatic patients and comparisons of Nt-proBNP values between grades of AS severity or NYHA classes were performed using t test, one-way analysis of variance, χ2 or Wilcoxon/Kruskall–Wallis test as appropriate. Correlation between Nt-proBNP values (a natural logarithm transformation was used due to a non-Gaussian distribution) and haemodynamic parameters of AS severity were evaluated using linear regression. Diagnostic value of Nt-proBNP for the diagnosis of severe symptomatic AS was analysed overall and in the subgroup of severe AS. Area under the curve (AUC) of the receiver operator characteristic analysis was calculated and sensitivity, specificity, positive and negative predictive values were determined for various thresholds. Event-free survival (composite endpoint of AS-related events) was assessed using the Kaplan–Meier analysis. Comparison of event-free survival according to Nt-proBNP level was performed by means of log-rank test. Cox proportional-hazard analyses evaluated the predictive value of Nt-proBNP for event-free survival in univariate, and after adjustment for age, gender and AS severity. A p value < 0.05 was considered statistically significant.
A total of 361 patients were prospectively enrolled. Baseline characteristics of the population are presented in the table 1. Briefly, 205 (57%) were male subjects; mean age was 79±6 years. Most of the patients were in sinus rhythm (295 patients, 82%), 47 patients (13%) were in atrial fibrillation (AF) and 19 (5%) had a paced rhythm. History of CAD was present in 110 patients (30%). Mean LVEF was 61±9% (median 65%, range (35–67)) and LVEF was <50% in 36 patients (10%). In all, 34 (9%) patients had bicuspid aortic valve; 44 (12%) patients had mild AS, 87 (24%) moderate AS and 230 (64%) patients had severe AS. Among the 230 patients with severe AS, 34 (15%) were asymptomatic and 196 patients (85%) were symptomatic (189 (96%) presented with dyspnoea, 42 (21%) with angina and 16 (8%) with a history of syncope). All symptomatic patients had severe AS.
In the entire population, compared with asymptomatic patients, symptomatic patients were older (80±6 vs 78±6 years, p=0.0004), more frequently female subjects (53% vs 32%, p<0.0001), in AF (17% vs 8%, p=0.02), had a lower creatine clearance (69±25 vs 73±20 ml/min, p=0.03) and a lower LVEF (59±11% vs 63±6%, p=0.02). There was no difference with regard to history of CAD, hypertension, diabetes mellitus, BMI or use of cardiovascular medication except for diuretics, which, as expected, were more frequently given in symptomatic patients (61% vs 43%, p=0.0006).
Nt-proBNP and AS severity
Overall, mean Nt-proBNP serum level was 1412±3682 pg/ml (median 349 (30–8354)). Nt-proBNP increased with the grade of AS severity as defined using AVA (p<0.0001), mean gradient (p<0.0001) or peak velocity (p<0.0001) but correlations between Nt-proBNP and haemodynamic parameters were only modest (R=-0.43, R=0.34 and R=0.34, respectively, all p<0.0001) (figure 1). In addition, Nt-proBNP values were widely scattered with a large overlap between patients with mild, moderate and severe AS. As an illustration, medians (95% CI) for AVA were 147 (16 to 2746), 168 (27 to 1685) and 599 (48 to 13063) pg/ml, respectively. Correlations between BNP and left ventricular mass or valvulo-arterial impedance were also modest (R=0.46, p<0.0001 and R=0.29, p=0.05, respectively).
Nt-proBNP and symptoms: diagnostic value
NT-proBNP was higher in symptomatic patients than in asymptomatic patients overall (2250±4830 vs 431±628 pg/ml, p<0.0001) (figure 2A) and as in the subgroup of patients with severe AS (AVA<1 cm2) (2307±4755 vs 678±728 pg/ml, respectively, p=0.03) (figure 2B). However, Nt-proBNP values were also widely scattered with a large overlap between asymptomatic and symptomatic patients both overall and in the subgroup of severe AS. In addition, among the 196 symptomatic patients (all with severe AS), 63 patients (32%) had normal Nt-proBNP values (<300 pg/ml), 20% intermediate values and 48% high Nt-proBNP (>700 pg/ml).
Consequently, Nt-proBNP had a poor sensitivity and specificity for the diagnosis of severe symptomatic AS in the entire cohort (N=361) and AUC of the receiver operator characteristic analysis was only 0.73 (figure 3A). The threshold value of Nt-proBNP providing the best sum of sensitivity plus specificity was 597 pg/ml (53% sensitivity and 79% specificity, respectively). In the subgroup of patients with severe AS (N=230), diagnostic performance of Nt-proBNP was even worse (AUC=0.62, 39% sensitivity and 87% specificity using the threshold value of 1401 pg/ml which provided the highest sum of sensitivity plus specificity) (figure 3B). Exclusion of obese patients (N=78), of patients with LVEF<50% (N=36) or in AF (N=47) did not improve the diagnostic value of Nt-proBNP (AUC of 0.73, 0.71 and 0.73, respectively). The per cent of patients with severe symptomatic AS referred for surgery with normal Nt-proBNP values was also similarly high excluding obese patients (32%), patients with LVEF<50% (37%), in AF (38%) or all three (43%).
Prognostic value of Nt-proBNP in asymptomatic patients
Among the 165 asymptomatic patients, 12 patients with severe AS underwent a prophylactic surgery in the following weeks after their initial evaluation because of physicians’ or patients’ preference and were excluded from the present analysis. In all, 11 patients were either lost of follow-up or refused to remain in the study and follow-up data were available for 142 patients (93%).
Mean follow-up duration was 1.8±0.8 years. An AS-related event occurred in 28 patients (12 during the first year of follow-up and 16 during the second year); 27 patients developed symptoms—24 dyspnoea and three angina—and one patient presented a sudden death. This last patient had a severe AS (AVA=0.7 cm2, mean gradient=71 mm Hg), and only mildly elevated Nt-proBNP (403 pg/ml). Five patients also died from non-AS-related cause. Among the 27 patients who developed symptoms (excluding the patient who presented a sudden death), a follow-up echocardiography was performed in 19 patients and showed an AS progression in all but three. A surgical valve replacement was performed in 18, a transcatheter aortic valve replacement in six and three remained under medical therapy because of comorbidities or patients’ preference. Event-free survival was 96% at 1 year and 84% at 2 years. Nt-proBNP values at baseline were lower in patients who remained asymptomatic than in patients who developed symptoms (319±431 vs 633±670 pg/ml, p=0.001). Survival curves showed that patients with baseline measurement of Nt-proBNP less than 300 pg/ml had a better outcome than patients with Nt-proBNP values between 300 and 700 pg/ml or above 700 pg/ml (91%, 77% and 59% event-free survival at 2 years, respectively, p=0.005) (figure 4A). In univariate analysis, as a continuous variable, Nt-proBNP was significantly associated with outcome (p=0.04). However, in multivariate analysis, after adjustment for age, gender and AVA, Nt-proBNP was not predictive of outcome (p=0.40). Thus, among the 28 patients who developed symptoms, 11 patients (39%) had normal Nt-proBNP values, seven patients (25%) had intermediate Nt-proBNP values and only 10 (36%) had elevated Nt-proBNP values. A total of 25 patients had high Nt-proBNP values at baseline and only 9 (36%) developed symptoms during the 2 years of follow-up. Adding LVEF or/and AF into the model did not improve the predictive value of Nt-proBNP (p=0.06, p=0.87 and 0.44, respectively).
Subgroup analysis in patients with moderate to severe AS
Among the 142 asymptomatic patients prospectively followed, 102 had moderate or severe AS (mean AVA 1.15±0.25 cm2, mean gradient 29±17 mm Hg, mean peak velocity 3.36±0.83 m/s, mean LVEF 62±7%). All the 28 AS-related events occurred in this subgroup of patients. Event-free survival at 1 and 2 years were significantly different according to Nt-proBNP values (96% and 87% respectively for patients with normal Nt-proBNP values, 91 and 67% respectively for patients with intermediate Nt-proBNP values and 90 and 50% respectively for patients with high Nt-proBNP values; p=0.04 and 0.01 respectively) (figure 4B). Nt-proBNP values were lower in patients who remained asymptomatic than in patients who became symptomatic at 2 years (344±501 vs 633±670 pg/ml, p=0.002). However, in patients with moderate or severe AS, Nt-proBNP was not predictive of outcome neither in univariate nor in multivariate analysis after adjustment for age, gender and AVA (p=0.12 and p=0.43, respectively). Furthermore, Nt-proBNP was also not predictive of outcome only considering age and gender in the model. When LVEF was also added into the model, Nt-proBNP did not reach the statistical significance (p=0.06). Inclusion of AF with or without LVEF into the model also did not improve the predictive value of Nt-proBNP (p=0.97 and 0.30, respectively). Also, in the subgroup of patients with moderate to severe AS, in sinus rhythm with normal LVEF (N=84, 20 events), Nt-proBNP was not predictive of outcome (p=0.06). Nt-proBNP was remeasured after the occurrence of symptoms in 12 patients (mean 601±494 pg/ml, median 492 (29–1474)) and was normal in four, intermediate in four and high in only four patients.
In the present study, in a large prospective cohort of elderly patients with AS, we showed that Nt-proBNP increased with AS severity and symptomatic status but that there was a wide overlap between groups and consequently that Nt-proBNP had both a poor sensitivity and a poor specificity for the diagnosis of severe symptomatic AS. In asymptomatic patients, Nt-proBNP was associated with AS-related events in univariate but not in multivariate analysis after adjustment for age, gender and AS severity. The present study raises caution regarding the use of Nt-proBNP in the decision-making process of AS patients especially in the elderly population.
Nt-proBNP and AS severity
BNP is secreted predominantly in cardiac ventricles in response to increased wall stress. Previous studies have shown a significant association between BNP/Nt-proBNP values and haemodynamic parameters of AS severity.8 ,9 ,19 In the present study, one of the largest of the literature, we also found a significant correlation between Nt-proBNP and valve area, peak velocity or mean gradient. However, there was an important overlap between grades of AS severity with wide scatter ranges of Nt-proBNP values. Looking more carefully at the literature such an overlap was previously observed and correlations between BNP and AS severity were also only modest with coefficients ranging from 0.30 to 0.60 suggesting that AS severity is only one determinant of BNP.8 ,9 ,19
Nt-proBNP and symptoms
Indeed, it has been shown that BNP/Nt-proBNP was associated with symptomatic status and more specifically to NYHA class.8 ,13 BNP/Nt-proBNP values increased with NYHA class and some have suggested that BNP/Nt-proBNP was accurate to separate symptomatic from asymptomatic patients.7 ,8 However, such a good discriminative value was not confirmed by others10 ,13 and one main finding of the present study is to highlight the important inter-individual variations and overlap between NYHA class groups, frequently observed but often neglected.7 ,9 ,12 In the present study, AUC of Nt-proBNP for the diagnosis of severe symptomatic AS was only 0.73 in the entire cohort and even worse in the subgroup of severe AS (AUC of 0.62). In addition, cut-off values determined in previous studies for the diagnosis of severe symptomatic AS were very low, often in the normal range (BNP of 66 pg/ml8 or Nt-proBNP of 60 pmol/l7) and despite that, sensitivity values were as low as 78%7 or 84%.8 In the present study, the optimised Nt-proBNP threshold was in the intermediate range (597 pg/ml) and was neither sensitive nor specific. Our results are in agreement with those of Weber et al9 showing an optimised threshold of 550 pg/ml, a sensitivity and specificity of 71% and 68% respectively and an AUC of 0.73. Thus, Nt-proBNP may be elevated despite the absence of symptoms in a large number of AS patients and determinants of Nt-proBNP in this elderly population—especially its link to diastolic function parameters10—deserve further investigations. On the other side, Nt-proBNP may remain normal despite the existence of AS-related symptoms requiring surgery in almost a third of patients. One possible explanation is that BNP is measured at rest and that Nt-proBNP may only increase during exercise in those patients. Whatever the explanations, the present study showed that diagnostic value of BNP for severe symptomatic AS is only modest.
Nt-proBNP and outcome
Finally, the main theoretical interest of BNP/Nt-proBNP is not to be used as a surrogate for AS severity or symptomatic status but its potential value as a predictor of the occurrence of symptoms. Management of asymptomatic patients with severe AS remains a source of debate and identification of prognostic factors highly desirable. Several studies have suggested that BNP/Nt-proBNP could provide independent prognostic information and may guide surgical indications. However, these studies were impeded by multiple limitations, small sample size, selection bias and retrospective analysis. In addition, the proposed thresholds were very low and seemed hardly compatible with a clinical use (Nt-proBNP>80 pmol/l13 or BNP>31 pg/ml8). Furthermore, Monin et al, in a recent study, failed to identify any threshold prognostic value of BNP and suggested using a score combining AS severity, BNP and gender.14
In the present study, we prospectively evaluated the prognostic value of Nt-proBNP using appropriate thresholds values for this elderly population, at fixed time-points, using prespecified events all attributed to AS with a high certitude. All patients were enrolled in a prospective cohort representative of the AS population we are facing in routine practice, namely, elderly patients with frequent cardiovascular risk factors and comorbidities. Thus, our population was 5–15 years older than in previous studies. In this elderly population, Nt-proBNP was linked to outcome in univariate but not in multivariate analysis after only adjustment for demographic characteristics (it has been previously shown that BNP/Nt-proBNP increased with age and gender7 ,10 ,12 ,20) and AS severity. More than half of the patients developed symptoms despite normal Nt-proBNP values at study entry whereas two-thirds of the patients with high Nt-proBNP values remained asymptomatic at 2 years. It is worth noting that Nt-proBNP was measured a second time after the occurrence of symptoms in a subset of patients and was elevated in only a third of patients while it remained normal in another third (and intermediate in the last third).
The recently updated European Society of Cardiology Guidelines have introduced the use of natriuretic peptides in the management of patients with AS (class IIb recommendation, level of evidence C): ‘aortic valve replacement (AVR) may be considered in asymptomatic patients with severe AS, normal EF and none of the above mentioned exercise test abnormalities, if surgical risk is low’ and ‘Markedly elevated natriuretic peptide levels confirmed by repeated measurements and without other explanations’.3 The wide scattering of Nt-proBNP values across AS severity and functional status groups, the important overlap between groups and the lack of prognostic value of Nt-proBNP after adjustment for AS severity raise caution regarding its use as a clinical tool to guide follow-up or timing of surgery. Normal Nt-proBNP values were neither reassuring nor were elevated Nt-proBNP values predictive of AS-related event in multivariate analysis. Our results clearly raise caution regarding the use of Nt-proBNP in the decision-making process in AS patients especially in the elderly population and these findings may implement the future recommendations and improve the level of evidence. Whether serial Nt-proBNP measurements among the same individual may be of interest as suggested21 deserves further studies. Recent data suggest a possible prognostic value of BNP in mitral regurgitation22 but the population is 15–20 years younger than AS population and these data deserve further validation.
The present study deserves several comments. First, patients were considered asymptomatic based on clinical judgment and no exercise test was performed. Therefore, we cannot definitively exclude that some patients who claimed to be asymptomatic did not have an objective limitation of their functional capacity. However, exercise testing is recommended in ‘physically active’ patients and is often not feasible in elderly patients. Thus, the present study reflects real-life practice.1 In addition, most of the patients with elevated Nt-proBNP did not develop symptoms during a 2-year period and thus can be quite surely considered as asymptomatic. Second, ‘only’ 102 patients with moderate or severe AS were considered in the outcome study. However, our series compares well with previous studies,7 ,13 ,14 is prospective and specifically focuses on elderly patients as we routinely face in clinical practice. We, as others, also enrolled patients with moderate AS in the outcome analysis since the potential for rapid AS progression and excess mortality has been recently pointed out.23 Third, outcome was only analysed at 1 and 2 years but we do not expect any biomarker to predict clinical events for a longer time-period and recommendations are to follow these patients at least yearly. Fourth, ‘only’ 28 AS-related events occurred during the 2 years of follow-up. However, all AS-related events occurred in patients with moderate or severe AS and still represent a 25% event-rate in this subset. We did not consider aortic valve replacement as an end-point as it may be performed based on patients’ or physicians’ preference before the occurrence of class I indication for surgery. Finally, Nt-proBNP results were communicated to the physicians in charge of the patients. However, no surgery was performed based on Nt-proBNP values, aortic valve replacement was not considered as a criterion of judgment and we only considered strong clinical AS-related events confirmed by an experienced cardiologist in the outcome analysis.
Our study is the first to highlight the limits of BNP for the evaluation and management of AS patients. In a large prospective cohort of elderly patients with AS we showed that Nt-proBNP was associated with AS severity and functional status but had a poor diagnostic value for severe symptomatic AS. More importantly, in asymptomatic patients Nt-proBNP was not an independent predictor of outcome. Our results suggest that BNP should be considered cautiously, at least as a single criterion, in the decision-making process of AS patients especially the elderly population.
We would like to specially think the team of the Centre d'Investigation Clinique for their help and support.
Contributors Conception and design, or analysis and interpretation of data: ClC, MD, AV, DM-Z. Drafting the article or revising it critically for important intellectual content: CaC, GH, LL, BI, XD, VH, FT. Final approval of the version to be published: ClC, DM-Z. Contributors as being responsible for the overall content as guarantors: ClC, DM-Z.
Funding Dr Claire Cimadevilla was supported by a grant from the Assistance Publique—Hôpitaux de Paris, The COFRASA (clinicalTrial.gov number NCT 00338676) and GENERAC (clinicalTrial.gov number NCT00647088) studies are supported by grants from the Assistance Publique—Hôpitaux de Paris (PHRC National 2005 and PHRC regional 2007).
Competing interests None.
Patient consent Obtained.
Ethics approval CCP Hôtel-Dieu, Paris, France.
Provenance and peer review Not commissioned; externally peer reviewed.