To evaluate predictors of pulmonary arterial hypertension (PAH) in a prospective cohort of patients with systemic sclerosis (SSc).
To evaluate predictors of pulmonary arterial hypertension (PAH) in a prospective cohort of patients with systemic sclerosis (SSc).
Routine clinical assessments as well as measurements of the diffusing capacity for carbon monoxide/alveolar volume (DLCO/VA) ratio and N-terminal pro–brain natriuretic peptide (NT-proBNP) level were performed in a prospective cohort of 101 SSc patients who did not have PAH or severe comorbidities. After a planned 36-month followup, we evaluated the predictive value of these parameters for the development of precapillary PAH, as demonstrated by cardiac catheterization, disease progression, and death. Criteria for cardiac catheterization were a systolic pulmonary artery pressure (PAP) of >40 mm Hg on echocardiography, a DLCO value of <50% without pulmonary fibrosis, and unexplained dyspnea.
Eight patients developed PAH, 29 had disease progression, and 10 died during a median followup of 29 months. Kaplan-Meier analysis identified the following baseline parameters as being predictors of PAH: DLCO/VA ratio <70% or <60% (P < 0.01 for each comparison), elevated plasma NT-proBNP level (>97th percentile of normal; P = 0.005), echocardiographically estimated systolic PAP >40 mm Hg (P = 0.08), and erythrocyte sedimentation rate >28 mm/hour (P = 0.015). In multivariate analyses, an elevated baseline NT-proBNP level (hazard ratio [HR] 9.97 [95% confidence interval (95% CI) 1.69–62.42]) and a DLCO/VA ratio <60% (HR 36.66 [95% CI 3.45–387.6]) were predictors of the occurrence of PAH during followup. An increased NT-proBNP level together with a decreased DLCO/VA ratio of <70% was highly predictive of the occurrence of PAH during followup (HR 47.20 [95% CI 4.90–450.33]).
This prospective study identified a decreased DLCO/VA ratio and an increased NT-proBNP as predictors of PAH in SSc. Use of these markers should result in improved PAH risk stratification and allow earlier initiation of therapy.
Systemic sclerosis (SSc) is a connective tissue disease characterized by abnormalities of the vasculature, the skin and connective tissues, and the immune system. Microvascular lesions underlie many manifestations of SSc, including pulmonary arterial hypertension (PAH), which results from diffuse noninflammatory occlusive lesions of the precapillary arterioles in the lung. PAH occurs in 8–12% of SSc patients (1) and is widely recognized as a major complication of both the limited and diffuse cutaneous subtypes of SSc. Once established, severe PAH is difficult to treat and has a poor prognosis (2, 3). Early detection of PAH is challenging because its symptoms (dyspnea, fatigue, exercise intolerance) are nonspecific and overlap with those of other morbidities of SSc, including pulmonary fibrosis and cardiomyopathy. Identification of risk factors or predictors of the development of PAH in individuals with SSc would allow earlier diagnosis and institution of specific therapy for PAH at a time when it is most likely to be effective (1).
Until recently, no clinically useful predictors of the development of PAH in individuals with SSc had been identified. A retrospective case-matched controlled study by Steen and Medsger (4) showed a correlation between a progressive deterioration in the diffusing capacity for carbon monoxide (DLCO) in the lung and subsequent development of isolated PAH. Other clinical features associated with the development of PAH in that study were increased severity of both Raynaud's phenomenon and digital tip ulcers and no history of calcium-channel blocker use. Prior to the development of PAH, echocardiographic estimates of pulmonary artery pressure (PAP) were similar in the cases and controls. Another study by Schachna et al (5) suggested that older age at onset of SSc was associated with an increased risk of developing PAH.
We have previously demonstrated that elevated serum levels of N-terminal pro–brain natriuretic peptide (NT-proBNP) is a sensitive and specific diagnostic marker of an early, echocardiographically demonstrated increase in systolic PAP in patients with SSc (6). Mukerjee et al showed that, in SSc patients with PAH diagnosed by catheterization of the right side of the heart, NT-proBNP levels correlated with hemodynamics (7) and with prognosis (8). Natriuretic peptides have recently been shown to be of major interest as a diagnostic tool in various cardiovascular diseases (9). In idiopathic PAH, NT-proBNP is a marker of disease severity and is independently associated with mortality (10, 11).
The aim of this prospective cohort study was to identify the most informative risk factors for the development of precapillary PAH in SSc patients. We also evaluated factors that might predict an increase in the global severity of the disease or death as secondary outcomes.
We prospectively studied 101 consecutive SSc patients who were hospitalized for routine followup of the disease during a 9-month period from September 1, 2003 to May 31, 2004. SSc was classified as limited cutaneous or diffuse cutaneous subtype according to the criteria described by LeRoy et al (12). All patients were evaluated for renal, cardiac, and pulmonary involvement. Eligible patients were those who had been on a stable treatment regimen for a minimum of 3 months, with a prednisone dosage ≤10 mg/day. Exclusion criteria were as follows: PAH diagnosed by catheterization of the right side of the heart; symptoms of heart failure, including class III or IV dyspnea as defined by the New York Heart Association (NYHA), venous distension, or dependent edema; severe comorbidities, such as hepatic or kidney failure, cancer, or gangrene; inadvisability of stopping vasodilator therapy; and pregnancy.
For the baseline assessment, patients were asked to stop taking calcium-channel blockers 3 days before admission, as previously reported (6). This washout period corresponds to 5 times the half-life of calcium-channel blockers. Laboratory studies obtained on the morning of hospital admission included the plasma NT-proBNP concentration, complete blood cell count, Westergren erythrocyte sedimentation rate (ESR), C-reactive protein (CRP) level, serum creatinine concentration, and tests for anticentromere and anti–topoisomerase I antibodies.
The left ventricular ejection fraction (LVEF) was determined by 2-dimensional echocardiography according to the Simpson method. Systolic PAP was estimated by Doppler echocardiography at rest, based on the tricuspid and/or pulmonary regurgitation, adding 10 mm Hg for auricular pressure, as recommended. Echocardiography was performed by the referent cardiologist of each center, all of which are tertiary centers for SSc care. Pulmonary involvement was assessed by chest radiography, computed tomography, and measurement of forced vital capacity (FVC) and the DLCO/alveolar volume (DLCO/VA) ratio.
The predefined parameters studied as risk factors for the development of PAH were a DLCO/VA <70% of the normal value (13), a NT-proBNP level >97th percentile of normal values, and a systolic PAP >40 mm Hg as estimated by Doppler echocardiography, as used in our previous preliminary study (6).
At baseline and during followup, an echocardiographically estimated systolic PAP of >40 mm Hg was used as a screening threshold for PAH. Additional clinical criteria were a DLCO <50% of predicted in the absence of pulmonary fibrosis and unexplained dyspnea. Thromboembolic disease was excluded by performing spiral contrast computed tomography and D-dimer measurements, when applicable according to age. Patients identified by the above combination of noninvasive tests and clinical assessments were offered catheterization of the right side of the heart. These criteria are similar to those reported by Mukerjee et al (2). Confirmed PAH was defined as a resting mean PAP ≥25 mm Hg, with a pulmonary capillary wedge pressure of ≤15 mm Hg measured at catheterization of the right side of the heart (14). No exercise measurements were performed.
Once included in the study cohort, the patient was followed up according to his or her practitioner's routine. The practitioner was blinded to the baseline NT-proBNP value. Patients were seen at 3–6-month intervals as indicated by their disease severity and were hospitalized at least annually for clinical evaluation, including echocardiography and pulmonary function tests. Global severity of disease was assessed using the revised version of the Medsger severity scale (15), and a change of class was considered an exacerbation of disease. The final study observation was planned as the last observation of the second semester of the year 2006 (or the last available observation) or the last observation at the time of death from any cause, with analyses and completion of the database performed during the first trimester of the year 2007.
Blood samples (10 ml) were obtained with the patient at rest, and were collected into tubes containing EDTA as previously described (6). Samples were centrifuged at 3,000g for 10 minutes within an hour of collection. The resulting plasma samples were stored at −80°C until used. The NT-proBNP concentration was determined with an Elecsys NT-proBNP sandwich immunoassay on an Elecsys 2010 instrument (Roche Diagnostics, Basel, Switzerland). The investigators performing the assays (DB and OGE) were blinded to the results of the clinical evaluations. The analytical range extended from 20 pg/ml to 35,000 pg/ml. The intraassay coefficient of variation was 2.5% for a concentration of 175 pg/ml and 2% for a concentration of 1,070 pg/ml; the interassay coefficient of variation was 3.2% and 2.7%, respectively. In this assay, there is no detectable cross-reactivity with atrial natriuretic peptide (ANP), NT-proANP, or BNP. The 97th percentiles of normal values, as determined by the manufacturer (2,264 healthy subjects), according to sex and age group, are as follows: in women, 130 pg/ml for those <45 years, 249 pg/ml for those 45–54 years, 287 pg/ml for those 55–64 years, 301 pg/ml for those 65–74 years, and 738 pg/ml for those ≥75 years; and in men, 86 pg/ml for those <45 years, 121 pg/ml for those 45–54 years, 210 pg/ml for those 55–64 years, 376 pg/ml for those 65–74 years, and 486 pg/ml for those ≥75 years.
The association between baseline levels of the predefined parameters and the following prespecified end points were evaluated after the prospective followup: occurrence of precapillary PAH on catheterization of the right side of the heart as the primary outcome, and as secondary outcomes, death from any cause and an increase in disease severity on the Medsger severity scale.
To evaluate its association with clinical outcomes, NT-proBNP was considered a categorical variable, and associations were assessed according to the normal values for age provided by the manufacturer. Survival was evaluated using the Kaplan-Meier method. Cox regression analysis was used to evaluate the association between the baseline values and the predefined outcomes of interest. The log-rank test was used to determine differences in survival probabilities. Statistical significance was assumed when a null hypothesis could be rejected at P < 0.05. Statistical analysis was performed with the use of R software (version 2.0.1 for Windows; Lucent Technologies, Murray Hill, NJ).
The physicians providing clinical care did not know the results of the baseline NT-proBNP levels. Moreover, the cardiologists who performed catheterization during the followup were not aware of the results of parameters measured at baseline.
The mean ± SD age of the 101 SSc patients (81 women) was 54.3 ± 13.4 years, and the mean ± SD disease duration was 6.0 ± 6.7 years. At baseline, 15 of the study patients had undergone catheterization of the right side of the heart to rule out precapillary PAH. The baseline characteristics are shown in Table 1. None of the patients had renal insufficiency. An additional 35 consecutive SSc patients who were hospitalized during the inclusion period could not be entered into the study because they met some of the exclusion criteria.
|Patients not developing PAH during study (n = 93)||Patients developing PAH during study (n = 8)||P|
|NT-proBNP, pg/ml||127 ± 135||412.6 ± 304.1||0.0003|
|No. with values >97th percentile for age||16||6||0.001|
|No. with lcSSc/dcSSc||52/41||3/5||NS|
|MRSS||11.0 ± 10.4||13.4 ± 8.4||NS|
|No. with MRSS >14||24||3||NS|
|No. with digital ulcers||16||3||NS|
|Systolic PAP by Doppler echocardiography, mm Hg||31.2 ± 5.9||38.2 ± 9.4||0.001|
|No. with systolic PAP >40 mm Hg||6||3||0.0001|
|LVEF, %||66.5 ± 7.4||67.4 ± 8.1||NS|
|No. with LVEF <55%||2||1||NS|
|FVC, % predicted||94.0 ± 18.7||80.7 ± 17.4||NS|
|No. with FVC <75%||13||3||NS|
|No. with fibrosis on CT scan||34||7||NS|
|DLCO/VA, % predicted||80.3 ± 16.2||58.1 ± 18.5||0.002|
|No. with DLCO/VA <70%||19||7||0.002|
|No. with DLCO/VA <60%||11||4||0.02|
|ESR, mm/hour||18.1 ± 15.2||33.7 ± 26.6||0.008|
|No. with ESR >28 mm/hour||14||4||0.03|
|Serum creatinine, μmoles/liter||78.9 ± 14.4||81.2 ± 18.1||NS|
|Hemoglobin, gm/dl||13.1 ± 2.3||12.6 ± 2.1||NS|
|No. with ANA titer ≥1:160||74||8||NS|
|No. with ACAs/anti–topo I antibodies||21/30||1/5||NS|
|No. taking low-dose oral prednisone||26||3||NS|
|No. taking ACE inhibitors||16||2||NS|
After a mean followup of 28 months (median 29 months), 19 of the 101 patients were identified by noninvasive screening as potentially having PAH: 13 because of a systolic PAP >40 mm Hg on echocardiography and 6 because of unexplained dyspnea or a DLCO <50%. None of these 19 patients declined the examination, and precapillary PAH was confirmed by catheterization of the right side of the heart in 8 of them (mean PAP 50 ± 14 mm Hg, mean ± SD). The PAP was in the normal range in 8 of the 19 patients, and 3 of the 19 exhibited postcapillary PAH.
The mean time to development of PAH after study inclusion was 18.6 months (median 17 months). Five of the 8 patients who developed PAH during the study were women, and 3 were men. The mean ± SD age of this group was 59.9 ± 15.6 years, with a mean ± SD disease duration of 5.9 ± 5.7 years. The characteristics of this group are shown in Table 1.
Ten patients died during the followup period. Causes of death were PAH in 5, interstitial lung disease in 2, stroke in 1, and lung cancer in 2. None of the patients died of heart failure. The mean and median time to death after development of PAH (n = 5 patients) was 13.8 months and 11 months, respectively, despite the use of the best standard of care in all patients.
With respect to progression of the disease, 29 of the 101 patients had an increase of 1 or more categories on the Medsger severity scale.
Univariate analyses identified a decrease in the DLCO/VA ratio, an increase in the NT-proBNP level, or echocardiographically estimated systolic PAP as predictors of confirmed PAH. For the NT-proBNP level, the baseline increase in the 8 patients who developed PAH was 1.6–2.6 times the value of the 97th percentiles for their respective age groups. Kaplan-Meier analyses of freedom from precapillary PAH are shown in Figures 1 and 2, and hazard ratios (HRs) are shown in Table 2. A data-driven analysis based on the mean ± SD DLCO/VA value of 58.1 ± 18.5 at baseline in the group who developed PAH was performed, using a threshold value of 60%; this value was also a potent predictor of PAH (Table 2).
|Model, variable||Univariate analysis||Multivariate analysis|
|P||HR (95% CI)||P||HR (95% CI)|
|DLCO/VA <70%||0.0043||21.3 (2.5–181.3)||0.014||18.81 (1.7–206.8)|
|High NT-proBNP||0.0048||10.1 (1.96–51.72)||0.053||6.35 (0.94–82.8)|
|Systolic PAP >40 mm Hg||0.0078||1.08 (1.63–30.87)||0.54||0.40 (0.02–7.79)|
|ESR >28 mm/hour||0.015||5.6 (1.35–23.01)||0.15||6.19 (0.49–76.9)|
|Model with lower DLCO/VA threshold|
|DLCO/VA <60%||–||–||0.0023||36.66 (3.45–387.6)|
|High NT-proBNP||–||–||0.012||9.97 (1.69–62.42)|
|Systolic PAP >40 mm Hg||–||–||0.17||6.61 (0.41–106.4)|
|ESR >28 mm/hour||–||–||0.14||5.56 (0.53–58.32)|
|Model with DLCO/VA <70% plus high NT-proBNP|
|DLCO/VA <70% and high NT-proBNP||–||–||0.00062||47.20 (4.90–450.33)|
|ESR >28 mm/hour||–||–||0.14||7.86 (0.48–126.46)|
|Systolic PAP >40 mm Hg||–||–||0.38||0.25 (0.01–5.87)|
A receiver operating characteristic curve was generated to assess the sensitivity and specificity of the DLCO/VA for detecting patients with PAH. The diagnostic value of a single DLCO/VA value is reflected by an area under the curve of 0.82. The sensitivity and specificity of a DLCO/VA of <70% in predicting PAH were 87.5% and 79.5%, respectively, and the sensitivity and specificity of a DLCO/VA of <60% were 62.5% and 88.1%, respectively. The sensitivity and specificity of an elevated NT-proBNP level in predicting PAH were 75% and 83%, respectively.
The acute-phase reactants ESR and CRP were also predictive of the development of PAH. These values were collinear; thus, only the ESR was considered in further analyses.
The 3 predefined criteria and the ESR were analyzed in a Cox regression model. In the initial model, using the predefined criteria, a decrease in the DLCO/VA ratio of <70% of the normal value remained a significant predictor of PAH (HR >18.81, P = 0.014), and an elevated NT-proBNP level was close to significance (HR >6.35, P = 0.053) (Table 2). In an alternate data-driven model using a cutoff value of 60% for the DLCO/VA, both a decreased DLCO/VA value and an increased NT-proBNP level were significant predictors of PAH (Table 2). Finally, the combination of a DLCO/VA value <70% and an above-normal NT-proBNP level characterized 9 patients, 6 of whom developed PAH, indicating a sensitivity and specificity of 75% and 97%, respectively (Table 2).
An ESR >28 mm/hour at baseline was predictive of death during the study period, by multivariate analysis (Table 3). A baseline ESR >28 mm/hour was also predictive of an increase in the Medsger score, by both univariate and multivariate analyses (Table 4).
|Variable||Univariate analysis||Multivariate analysis|
|P||HR (95% CI)||P||HR (95% CI)|
|ESR >28 mm/hour||0.0033||6.78 (1.83–24.82)||0.0084||7.05 (1.60–31.06)|
|DLCO/VA <70%||0.054||3.65 (0.94–13.9)||0.059||3.83 (0.92–15.86)|
|High NT-proBNP||0.69||1.31 (0.33–5.24)||0.39||0.42 (0.05–3.14)|
|Systolic PAP >40 mm Hg||0.062||3.82 (0.91–15.99)||0.73||1.42 (0.17–11.37)|
|Variable||Univariate analysis||Multivariate analysis|
|P||HR (95% CI)||P||HR (95% CI)|
|ESR >28 mm/hour||0.086||2.06 (0.88–4.76)||0.046||2.79 (1.01–7.82)|
|High NT-proBNP||0.034||2.22 (1.04–4.70)||0.043||2.23 (1.01–4.94)|
|DLCO/VA <70%||0.2||1.65 (0.75–3.6)||0.12||1.89 (0.82–4.31)|
|Systolic PAP >40 mm Hg||0.97||1.03 (0.29–3.53)||0.09||0.26 (0.05–1.32)|
This is the first prospective study to demonstrate that decreased DLCO and increased serum NT-proBNP levels after washout of vasodilators predict the subsequent development of PAH in patients with SSc. SSc is the most life-threatening of the connective tissue disorders, although survival rates have improved in recent years (16, 17). SSc patients with PAH are at high risk of early death, with a reported 1-year survival rate of 55% (2, 4, 18). While endothelin antagonists have been shown to improve the prognosis of patients with connective tissue disease–related PAH, even those who are treated with endothelin antagonists have a 2-year mortality risk in excess of 25% (19). In our study, 5 of 8 SSc patients with PAH died a mean of 13.8 months after the diagnosis of PAH, despite use of the best standard of care in all patients.
Considering the availability of new therapies, there is general agreement that there is a need for accurate early detection of PAH. Steen and Medsger (4) showed that patients with limited cutaneous SSc who subsequently developed isolated PAH had a progressive deterioration in the DLCO starting more than 10 years before the diagnosis of PAH. Our results confirm that a decrease in the DLCO/VA is a sensitive and specific predictor of the development of PAH, using threshold values of 70% and 60% of the predicted value.
Our results also show that elevated levels of NT-proBNP predict the occurrence of PAH. B-type natriuretic peptides (BNP and NT-proBNP) have emerged in recent years as an important diagnostic and prognostic marker of cardiac failure. Studies in patients with PAH have demonstrated that plasma BNP levels are increased in proportion to the extent of right ventricular dysfunction (20, 21). There is growing evidence that the BNP level might be a biomarker for PAH in terms of screening, diagnostic evaluation, evaluation of response to therapy, and prediction of disease severity (10). In SSc, we previously demonstrated that NT-proBNP is a diagnostic marker of an early elevation of the PAP (6), and other investigators have shown that NT-proBNP levels correlate with hemodynamics (7) and prognosis (8) in SSc patients with PAH confirmed by cardiac catheterization.
In the current study, we show that an increase in the NT-proBNP value at baseline is predictive of the subsequent development of PAH. We used the 97th percentiles of normal values provided by the manufacturer, and not the threshold values validated for established heart failure. This suggests that our patients who were considered to be free of PAH according to the gold standard method and to be free of left-sided heart dysfunction had ventricular strain or some other physiologic stress that led to the secretion of elevated levels of NT-ProBNP. Our results suggest that we can identify “pre-PAH” or “early PAH” using accurate measures of the DLCO/VA as a reflection of capillary gas exchange and of the NT-proBNP as a reflection of cardiac wall stress. Moreover, the combination of these 2 variables is a very strong predictor of the development of PAH. We recommend that SSc patients with both a DLCO <70% and an elevated NT-ProBNP should be very carefully monitored, and we suggest that such patients would constitute an appropriate target group for investigation of early therapeutic intervention for PAH in a randomized trial.
Our study failed to demonstrate that echocardiographic estimation of the systolic PAP is predictive of PAH by multivariate analysis. Although echocardiographic estimation of the systolic PAP is highly recommended as a first-line screening tool for PAH, the range of right ventricular systolic pressure among healthy controls has been found to vary widely, particularly with age and body mass index (22). In addition to the possibility of being a false-positive finding, an increase in echocardiographically estimated systolic PAP may represent primary cardiac involvement in SSc (23). Indeed in a recent nationwide prospective multicenter study of early detection of PAH in France (24), 3 (9.1%) of 33 SSc patients who underwent cardiac catheterization after echocardiographic screening had postcapillary pulmonary hypertension despite the absence of echocardiographic evidence of left-sided heart dysfunction. In the current study, we found that 3 (16%) of the 19 patients assessed by cardiac catheterization had postcapillary PAH. Routine annual evaluation of patients with SSc should include determination of NYHA class, chest radiography, electrocardiography, estimate of systolic PAP by Doppler echocardiography, and measurement of the DLCO (25, 26), but this remains to be validated.
Some other new techniques, such as stress echocardiography (27) and magnetic resonance imaging (28), remain to be investigated in SSc and may be helpful for stratifying patient risk. However, measurements of the DLCO/VA and NT-proBNP are simple, cost-effective, and widely available, and are therefore strong candidates for first-line assessment of PAH risk in SSc and other connective tissue disorders.
The small number of deaths in this cohort does not allow strong conclusions to be drawn regarding predictors of mortality. However, elevated levels of acute-phase reactants seem to correlate with mortality. An elevated ESR was previously found to be a risk factor for decreased survival in a French Canadian cohort of SSc patients (13).
Our study had several limitations and, thus, can only support general conclusions. The sample size can be regarded as small, and this may explain the wide confidence intervals and may suggest that the HRs are unstable. Thus, considered as a pilot study, our findings will require confirmation in larger studies. However, as a prospective study, our study does provide reasonably accurate information that was not previously available concerning the outcome of patients with SSc. In addition, it has identified simple markers of vascular and cardiac pathology that have predictive value in this patient sample.
We assessed NT-proBNP levels after a washout of regular treatment with calcium-channel blockers, as we have previously reported (6). We assume that measurement under these conditions is more accurate, since we have previously shown that these drugs may influence NT-proBNP levels in patients with an early increase in systolic PAP (6). In light of previous studies highlighting the potential effects of calcium-channel blockers on early elevation of the systolic PAP and their postulated beneficial role in reducing the risk of PAH (4), we need to better define the role of these agents in SSc patients with early PAH. However, no hypotheses or conclusions about calcium-channel blockade can be drawn from this study.
We included in this study patients who were identified as being free of PAH by echocardiographic assessment or resting catheterization, but we cannot exclude the possibility that some patients may, at baseline, have had abnormalities on formal exercise testing. No quantitative score measuring the extent of pulmonary fibrosis was available in our study, and thus, the influence of pulmonary fibrosis on PAH risk could not be determined in our study.
In conclusion, our results suggest that the DLCO/VA and NT-proBNP values can be used to identify SSc patients who are at high risk for the development of PAH. This has important clinical implications, in that noninvasive tests may be used to identify high-risk patients who should undergo more careful evaluation, potentially including cardiac catheterization. Based on the pathogenesis of the disease and the experience from animal models, early intervention is likely to prevent or slow the progression of the disease, but this remains to be determined. Further studies assessing the use of these markers in therapeutic strategies aimed at selection of patients for early treatment are indicated.
Dr. Allanore had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Allanore, Borderie, Kahan.
Acquisition of data. Allanore, Borderie, Avouac, Meune, Hachulla, Mouthon, Guillevin, Meyer, Weber.
Analysis and interpretation of data. Allanore, Borderie, Avouac, Zerkak.
Manuscript preparation. Allanore, Zerkak, Meune, Ekindjian, Kahan.
Statistical analysis. Allanore, Zerkak.