Author + information
- Received May 10, 2006
- Revision received July 13, 2006
- Accepted July 23, 2006
- Published online December 5, 2006.
- Gian Paolo Rossi, MD, FACC, FAHA⁎ (, )
- Giampaolo Bernini, MD,
- Chiara Caliumi, MD,
- Giovambattista Desideri, MD,
- Bruno Fabris, MD,
- Claudio Ferri, MD,
- Chiara Ganzaroli, MD,
- Gilberta Giacchetti, MD,
- Claudio Letizia, MD,
- Mauro Maccario, MD,
- Francesca Mallamaci, MD,
- Massimo Mannelli, MD,
- Mee-Jung Mattarello, MD,
- Angelica Moretti, MD,
- Gaetana Palumbo, MD,
- Gabriele Parenti, MD,
- Enzo Porteri, MD,
- Andrea Semplicini, MD, FAHA,
- Damiano Rizzoni, MD,
- Ermanno Rossi, MD,
- Marco Boscaro, MD,
- Achille Cesare Pessina, MD, PhD,
- Franco Mantero, MD,
- PAPY Study Investigators
- ↵⁎Reprint requests and correspondence:
Prof. Gian Paolo Rossi, DMCS Clinica Medica 4, University Hospital, Via Giustiniani 2, 35126 Padova, Italy.
Objectives We prospectively investigated the prevalence of curable forms of primary aldosteronism (PA) in newly diagnosed hypertensive patients.
Background The prevalence of curable forms of PA is currently unknown, although retrospective data suggest that it is not as low as commonly perceived.
Methods Consecutive hypertensive patients referred to 14 hypertension centers underwent a diagnostic protocol composed of measurement of Na+and K+in serum and 24-h urine, sitting plasma renin activity, and aldosterone at baseline and after 50 mg captopril. The patients with an aldosterone/renin ratio >40 at baseline, and/or >30 after captopril, and/or a probability of PA (by a logistic discriminant function) ≥50% underwent imaging tests and adrenal vein sampling (AVS) or adrenocortical scintigraphy to identify the underlying adrenal pathology. An aldosterone-producing adenoma (APA) was diagnosed in patients who in addition to excess autonomous aldosterone secretion showed: 1) lateralized aldosterone secretion at AVS or adrenocortical scintigraphy, 2) adenoma at surgery and pathology, and 3) a blood pressure decrease after adrenalectomy. Evidence of excess autonomous aldosterone secretion without such criteria led to a diagnosis of idiopathic hyperaldosteronism (IHA).
Results A total of 1,180 patients (age 46 ± 12 years) were enrolled; a conclusive diagnosis was attained in 1,125 (95.3%). Of these, 54 (4.8%) had an APA and 72 (6.4%) had an IHA. There were more APA (62.5%) and fewer IHA cases (37.5%) at centers where AVS was available (p = 0.002); the opposite occurred where AVS was unavailable.
Conclusions In newly diagnosed hypertensive patients referred to hypertension centers, the prevalence of APA is high (4.8%). The availability of AVS is essential for an accurate identification of the adrenocortical pathologies underlying PA.
Primary aldosteronism (PA) is a common cause of arterial hypertension (HT); however, more than 50 years after its discovery (1) its prevalence in newly diagnosed hypertensive patients remains uncertain. A survey of 18 recent studies showed a wide range (from 1.4% to 32%, median 8.8%) of estimates of prevalence (2). This is because most studies were performed retrospectively in a selected cohort of patients and used heterogeneous criteria for selecting the patients and diagnosing PA (3–15). Importantly, it should also be acknowledged that PA can only be conclusively diagnosed in the patients who are cured or improved by removal of an aldosterone-producing adenoma (APA) or unilateral adrenal hyperplasia (16). The other major PA subtype, idiopathic hyperaldosteronism (IHA), should not be treated surgically and therefore cannot stand on this retrospective criterion. Furthermore, differentiation of IHA from low-renin primary (essential) HT is arbitrary (7,17). Thus, establishing the prevalence of PA is a challenging task, which requires: 1) the prospective investigation of newly diagnosed consecutive unselected hypertensive patients, 2) the use of predefined state-of-the-art diagnostic criteria, and 3) an unequivocal diagnosis that is feasible only retrospectively in APA treated with adrenalectomy. A study of the unselected hypertensive patients, seen by general practitioners, is hard to perform and to justify for cost-effectiveness reasons; furthermore, it can be hampered by the difficulty of achieving a conclusive diagnosis of PA and of its underlying cause, which requires the expertise and facilities that are usually available only at specialized centers. The only attempt in this direction best testifies this difficulty because it furnished data only on the prevalence of an elevated aldosterone/renin ratio (ARR) and not of PA (18).
A large-scale multicenter study carried out prospectively at referral centers could minimize the chances of selection biases and furnish firm evidence on the accuracy of diagnostic tests for PA (19). Hence, at the end of 2001 a Working Group of the Italian Society of Arterial Hypertension started nationwide a prospective survey aimed at determining the prevalence of APA in newly diagnosed hypertensive patients referred to specialized centers for HT in Italy. This article reports on the main results of this investigation that was named the PAPY (PA Prevalence in Hypertensives) study, whereas data on early renal involvement have been reported elsewhere (20).
The study design followed the recommendations of the STARD (Statement for Reporting Studies of Diagnostic Accuracy) (21). The protocol (Fig. 1)was approved by the ethical committee of the University of Padua, and an informed written consent was obtained from each participant. To minimize the chance of any selection bias, it was decided a priori to enroll consecutive patients with a new (within the previous 6 months) diagnosis of HT who had been referred by their family doctor to specialized centers for the diagnosis and treatment of HT nationwide in Italy. A prior diagnosis of any secondary forms of HT and the patient’s refusal to participate in the study were exclusion criteria. To ensure homogeneity across centers, it was decided a priori to exclude from the analysis centers that eventually would have recruited fewer than 20 patients.
Patients were enrolled after the diagnosis of HT was confirmed by established criteria based on office blood pressure measurement by mercury sphygmomanometer (22). The patients had to be prepared for the screening test as follows: treatment with spironolactone, canrenone, or potassium canrenoate was withdrawn for at least 6 weeks (Fig. 1). Other agents, including diuretics, beta-blockers, angiotensin-converting enzyme inhibitors, and angiotensin II type 1 (AT-1) receptor antagonists, were withdrawn for at least 2 weeks. A long-acting calcium channel blocker and/or doxazosin were allowed as treatment whenever necessary for minimizing the risks of uncontrolled HT. On the day of the test, Na+and K+excretion was measured by 24-h urine collection. The completeness of the urine collection was verified with measurement of urinary creatinine excretion.
After an overnight fast and 1 h quiet rest in the sitting position, between 7 amand 9 amthe Na+and K+, plasma renin activity (PRA), aldosterone, and cortisol were measured at baseline and again 60 min after captopril administration (50 mg orally). The blood pressure levels were measured at baseline and after captopril as previously described.
The aldosterone (in ng/dl)/PRA (in ng/ml/h) ratio, at baseline and after captopril, and a score estimating the probability of PA based on a validated multivariate logistic discriminant analysis (LDF) score (7) were calculated.
A saline infusion test (2 l saline over 4 h with the patients kept in the supine position) (23), with measurement of PRA, aldosterone, cortisol, and serum K+at baseline and again after the infusion, was performed in all patients with: 1) an ARR ≥40 baseline, and/or 2) an ARR ≥30 after captopril (24), and/or 3) an LDF score ≥0.50; as well as 1 in every 4 patients not meeting such criteria. The need for further testing was decided based on the ARR at baseline and the captopril test results, regardless of the saline test results. It was decided that this test should not affect the further diagnostic workup to avoid any bias in the evaluation of the performance of the saline as a confirmatory test for PA.
A high-resolution computed tomography (CT) scan with 3-mm slices and systematic use of a contrast medium, and/or magnetic resonance (MR) imaging, was required in all patients with an ARR ≥40 baseline and an ARR ≥30 after captopril, or an LDF score ≥0.50. The adrenal nodular enlargements, identified with CT and/or MR imaging, were classified by diameter into 3 classes: <10 mm, 11 to 20 mm, and >20 mm. The kappa value between radiologists in classifying these lesions was 0.954 (p < 0.001). At the centers where adrenal vein sampling (AVS) (25) was available, it was performed in all of those selected for CT and/or MR as previously described. Dexamethasone-suppressed adrenocortical scintigraphy (26) was used to show a lateralized aldosterone excess production whenever AVS was unavailable. The AVS was performed without ACTH stimulation because the latter was shown not to improve the diagnostic accuracy (27). Only bilaterally selective AVS were judged to provide an accurate diagnosis. The criteria for considering AVS selective and for lateralization of aldosterone secretion were previously established as described in detail (25).
The utmost care was taken in drawing blood by avoiding a tourniquet, fist clenching, and any condition that might artificially increase serum K+(28). The Na+and K+levels were measured in serum and urine by the standard method; hypokalemia was defined as serum K+≤3.5 mEq/l. The glomerular filtration rate was measured using the abbreviated equation (29). Plasma renin activity was measured by radioimmunoassay with commercial kits (Ren CTK, Sorin Biomedica Saluggia, Italy, in 10 centers; or Angiotensin I RIA CT, Radim, Pomezia, Italy, in the remainder). For both kits, the intra-assay and interassay coefficient of variation was within 8% and 10%. Normal range sitting, at rest, and on a normal Na+diet was 0.65 to 2.64 ng/ml/h. The assay for aldosterone was performed with a commercial kit (Aldosterone Mirya, Techogenetics, Cassina de Pecchi, Italy). Normal range was 10 to 150 pg/ml supine and 30 to 320 pg/ml upright on a normal Na+diet; both intra-assay and interassay coefficients of variation for this assay were <5.6%; the cross-reactivity of the antibody for aldosterone for the other adrenal steroids was <0.001%. Cortisol was measured with a commercial kit (Cortisol bridge, Adaltis, Casalecchio di Reno, Italy); the intra-assay and interassay coefficients of variation for this assay were below 6% and 10 %, respectively. The cross-reactivity of the antibody for steroids was negligible. A survey of hormone values shows no differences across centers; sample centralization in a single laboratory (Padua) for a quality control project is ongoing.
The Working Group reached the consensus that a positive result of the ARR baseline had to be corroborated by the result of the captopril test. Hence, the diagnosis of PA required evidence of autonomous excess aldosterone production, which was defined as an ARR ≥40 at baseline plus an ARR ≥30 after captopril, or a logistic discriminant function ≥0.50 (7,24). For the calculation of the ARR the PRA values <0.2 ng/ml/h were arbitrarily set to 0.2. On completion of the diagnostic workup described in Figure 1, an Adjudication Committee (G.P.R., F.M.) was responsible for determining whether patients had PA or not. Moreover, the diagnosis of APA required the following strict predefined criteria: 1) lateralization of aldosterone secretion at AVS, or evidence of lateralized uptake of 131I-norcholesterol at dexamethasone-suppressed adrenocortical scintigraphy, 2) surgery, 3) pathology, and more importantly, 4) outcome of adrenalectomy as assessed at follow-up. To this end, we required demonstration of normokalemia and cure or improvement of HT at least 120 days after adrenalectomy according to the American Heart Association criteria for evaluation of outcome after correction of renal ischemia (30). Cure was defined as a systolic blood pressure <140 mm Hg and diastolic blood pressure <90 mm Hg without medications, and improvement as a systolic and diastolic blood pressure <140/90 mm Hg, respectively, on the same or a reduced number of medications and/or a reduced number of defined daily doses (31). By exclusion, patients with a biochemical diagnosis of PA without evidence for a lateralized aldosterone excess were held to have IHA.
Quantitative variables were first tested for normal distribution, and transformations were undertaken for variables that showed a skewed distribution until a normal distribution was attained. One-way ANOVA followed by a Bonferroni test was used to compare quantitative variables between groups. The distribution of categorical variables was investigated by chi-square analysis. Significance was set at p < 0.05.
The sensitivity, specificity, and accuracy of each test were calculated (32). The diagnostic odds ratio, that is the odds of positivity among diseased patients divided by the odds of positivity among nondiseased patients, was also used as measure of the diagnostic accuracy (19). When a test provides no diagnostic evidence, the diagnostic odds ratio is 1.0. The area under the receiver operator characteristics curve was also used to measure and compare the accuracy of the tests (32), for example, serum K+, ratio of urinary K+excretion to serum K+, ARR at baseline and after captopril, and the logistic discriminant function score, for prediction of the conclusive diagnosis of PA or no PA. This strategy was also used to determine the cutoff values that provided the best combination of sensitivity and false-positive rate for each test. The MedCalc software (version 126.96.36.199, MedCalc Software, Mariakerke, Belgium) was used for these purposes.
Clinical and demographic features
On locking the database on September 30, 2004, 4 of the 18 centers had enrolled fewer than 20 patients, and therefore were excluded from the analysis. An additional 17 patients who did not have complete hormone data or showed some violation of the protocol were excluded. Thus, overall 1,180 patients were available for the analysis; their baseline data are shown in Table 1.The comparison of patients with and without PA showed not only the expected differences of PRA, aldosterone, and serum K+, but also an older age and a higher blood pressure in the PA group.
At the time of the screening test, 40% of the patients were untreated, 42% were on a calcium channel blocker or doxazosin, and 18% were on both agents. No patients experienced any problem while untreated, and no untoward effect occurred with captopril administration. There were no significant differences of PRA, aldosterone, and cortisol in either the entire group or PA or the PH patients across treatment groups, thus ruling out any systematic effect of these treatments on the results of the screening test (33,34). Likewise, there were no differences in PRA, aldosterone, and cortisol across centers, thus making unlikely a center effect on hormonal data.
At the time of the database locking, 53 patients had not completed the entire diagnostic workup; therefore, a conclusive diagnosis was attained in 95.3% (n = 1,125) of the 1,180 patients. On completion of the diagnostic workup (Fig. 1), 126 patients were diagnosed with PA. Therefore, overall the prevalence of the disease was 11.2%, without gender differences (11.7% in men, 10.6%. in women). By the aforementioned criteria, an APA was the cause of PA in 54 patients (42.8%), and IHA in the remaining 72 (57.2%). Heterogeneity in the proportion of the underlying adrenal pathologies (p < 0.001) across centers was observed. When the 2 centers with the highest prevalence that contributed most to this heterogeneity were excluded, the overall prevalence of PA and APA was 7.4% and 3.5%, respectively.
Overall 20.4% of the patients tested positive by at least 1 of the indexes derived from the baseline evaluation and the captopril test by the prespecified criteria (7,24). Table 2shows the sensitivity, specificity, accuracy, and diagnostic odds ratio (19) of each test at the prespecified cutoffs. Thus, the screening tests allowed selection of subgroups of hypertensive patients with a higher prevalence of PA. Of the indexes based on the captopril test, the ARR after captopril had a lower sensitivity and a higher specificity than the ARR at baseline, whereas the opposite occurred for the LDF score. The latter showed the highest sensitivity, and therefore the highest diagnostic odds ratio.
The comparison of the diagnostic performances of the saline suppression test and the captopril test for confirming the diagnosis of PA will be reported elsewhere.
Based on the maximum tumor diameter at CT and MR, 17% of the APA was <10 mm diameter; 28% between 10 and 20 mm; and 55% >20 mm.
Prevalence of hypokalemia by conclusive diagnosis
Spontaneous hypokalemia was found in 9.6% of the patients; however, it was observed in about one-half of the patients with APA and only 17% of those with IHA (Fig. 2).
Severity of arterial HT by conclusive diagnosis
Figure 3shows the prevalence of APA and IHA in the patients classified by blood pressure values at the screening test. High-normal blood pressure (because of ongoing treatment) was seen in about 12% of the patients; despite ongoing treatment 43% of the patients had grade 1, 31% grade 2, and 13.7% grade 3 HT (22). Increasing severity of HT was associated with an increased prevalence of both APA and IHA (chi-square = 31.4, p < 0.0001).
Receiver-operating characteristic curve analysis
By receiver-operating characteristic curve analysis, the accuracy of the ratio of 24-h urinary K+excretion to serum K+did not differ from 0.5, whereas those for ARR at baseline and after captopril and the logistic discriminant function score were significantly higher (see online supplemental table). The logistic discriminant function score showed the highest accuracy, followed by the ARR after captopril, and the ARR at baseline, which all performed better (p < 0.0001) than serum K+. The online supplemental tableshows that the cutoff values that provided the best tradeoff of sensitivity and specificity for each test differed from those prespecified based on retrospective studies (14,35).
Impact of AVS on subtypes of PA
The AVS was available at 5 of the 14 centers that qualified for the study and was performed in 43 patients; however, it was bilaterally selective in 40 (93%) of the patients. Hence, the hypothesis that the availability of AVS affects the prevalence of adrenocortical pathologies underlying PA could be tested by splitting the centers into those with and those without AVS. We found that at the centers with AVS significantly more cases of APA and fewer cases of IHA were diagnosed (Fig. 4).The opposite occurred at the centers where AVS was unavailable.
Compelling evidence indicates that hyperaldosteronism is associated with detrimental consequences on the cardiovascular system (36), including early renal damage (20). Removal of an APA or a unilateral autonomous hyperplasia normalizes aldosterone secretion, and warrants cure or marked amelioration of HT, and the cardiovascular changes in most patients (37). The systematic search for PA and efforts toward identifying unilateral causes of aldosterone secretion could therefore be justified, particularly if the prevalence of PA were high as suggested (2–15,38), although the cost effectiveness of this strategy has been challenged (39).
This multicenter study prospectively investigated the prevalence of APA and PA with a predefined diagnostic protocol in newly diagnosed hypertensive patients referred to specialized centers for HT nationwide. State-of-the-art tests (40) and a consensus protocol for diagnosing APA were exploited, thus allowing a conclusive establishment of the presence or absence of the disease. Because of these unique features, robust evidence on the prevalence of this potentially curable form of arterial HT and information that may affect future strategies for the screening and diagnosing PA and its underlying causes could be obtained.
Prevalence of PA
Only APA can be unequivocally diagnosed with strict criteria such as those described; at variance, no such diagnostic gold standard exists for identifying IHA (7,17). Thus, although we found that the overall prevalence of PA was 11.2%, which accords well with an earlier prediction (3,5), we would like to underline that an APA was found in 42% of the cases initially held to have PA. Thus, 4.4% of the 1,180 consecutive hypertensive patients originally screened (and 4.6% of those with a conclusive diagnosis) had an APA that was surgically removed with a long-term cure. This figure might underestimate the proportion of APA for several reasons. First, because we used a rather high ARR baseline for the initial screening, which can provide a higher specificity but a lower sensitivity than lower cutoffs, some PA patients might have been missed. Second, we unexpectedly found a high rate of spontaneous hypokalemia in the patients without PA. Because hypokalemia lowers aldosterone secretion, some of these patients might have had falsely normal aldosterone, which might have been revealed if they had been studied when normokalemic. A few of them might also have apparent mineralocorticoid excess, Liddle syndrome, familial type I hyperaldosteronism (glucocorticoid-remediable aldosteronism), or another undefined cause of primary HT with hypokalemia. Unfortunately, in this multicenter study we had no opportunity to reinvestigate these patients after correction of the hypokalemia and to dissect these monogenetic forms of HT. Finally, at centers where AVS was unavailable, IHA was diagnosed by exclusion of an APA. As discussed later, the tests used as an alternative to AVS (CT or MR and dexamethasone-suppressed adrenocortical scintigraphy) are inaccurate for diagnosing a small APA, thus suggesting that some APAs have been missed at these centers.
Generalizability of the results
Our patients were referred to a specialized HT center. We found a heterogeneity of PA prevalence across centers, which suggests occurrence of some referral bias despite all of the precautions that were taken. The relevance of our results to the practicing generalist evaluating patients with HT may therefore be questioned. The following considerations, however, are necessary. First, patients were enrolled consecutively without applying any selection criteria (such as serum K+, blood pressure levels, and/or resistance to the antihypertensive treatment). Second, patients who had any clues to a secondary form of HT were carefully excluded. Third, the anthropometric and blood pressure data (Table 1) are similar to those of the general Italian population of hypertensive individuals. Fourth, the prevalence of a high ARR in this study was even lower than the 32.4% prevalence of a high ARR found in a general practitioners study in Italy, despite the use of a higher (>50) cutoff (18). Finally, even after the exclusion of the centers with the highest prevalence of PA, the rate of APA remained around 3.5%. Thus, while retaining general validity, our data support the view that the screening for PA in the newly diagnosed hypertensive patients is justified and can open the door to a cure of arterial HT in several cases.
We calculated the ARR at baseline and 2 additional indexes, the ARR after captopril and the logistic discriminant function score, that are viewed as confirmatory tests (40). These indexes outperformed spontaneous hypokalemia and the ratio of 24-h urinary K+excretion to serum K+when compared by receiver-operating characteristic curve analysis (see online supplemental table). This was because a substantial proportion of our patients without PA had hypokalemia and conversely, most patients with confirmed PA had normokalemia (Fig. 2). Thus, nowadays the normokalemic cases of PA seem to be the rule rather than the exception, in accord with most studies (7,41), albeit not all of the retrospective studies (10). The widespread use of low-sodium and/or high-potassium diets along with the more diffuse prescription of potassium-sparing antihypertensive drugs, such as angiotensin-converting enzyme inhibitors and AT-1 receptor antagonists, which were not permitted in this study, are likely explanations for the current common occurrence of normokalemic cases of PA. Clinically, this result implies that if hypokalemia were to be used as a screening test, the majority of the PAs would have been missed (7,14,40,41).
Tests for identifying the underlying adrenal pathology
The use of high-resolution CT and MR scans has shown that the diameter of APA is generally smaller than 20 mm and often smaller than 10 mm (25). In this study 45% of the APAs discovered by CT or MR were smaller than 20 mm, and 17% were smaller than 10 mm. This finding accords well with the 12.4-mm mean diameter of a recent report (42). Because these small tumors can escape detection with all imaging techniques, it is not surprising that they could be accurately identified with CT or MR only in about one-half of the cases (42). Considering the high prevalence of arterial HT (22) and incidentally discovered adrenal masses (43), of which only 1.6% are APAs (44), the concurrence by chance of primary HT and nonfunctioning adrenal tumors is, according to conditional probability calculation (Bayes rule), about 3.6%.
We and others proposed that AVS is crucial to show lateralization of aldosterone secretion from the affected adrenal gland in the commonly occurring small-sized APA (25,42,45). Hence, the lack of performing AVS might affect the proportion of PA attributed to APA or IHA. This hypothesis, which has never been tested in a large multicenter study, was confirmed by comparing the diagnoses made at the centers that could perform AVS with those made at the centers where this test was unavailable: the former centers diagnosed more APAs than the latter centers, where more cases of PA had to be attributed to IHA. Therefore, an accurate diagnosis of the adrenocortical pathology underlying PA requires AVS. This test is technically demanding and not universally available, but can be performed at a very low (0.5% to 1%) risk of major complications in experienced hands (42,46).
Almost 5% of the newly diagnosed hypertensive patients who are referred to specialized centers for HT have a surgically curable APA. However, this figure might underestimate the true prevalence of the tumors because of the lack of systematic use of AVS. An additional 6.4% of case had IHA, thus resulting in an overall prevalence of PA of about 11.2%. Given the high prevalence of arterial HT, screening for PA is warranted in newly diagnosed hypertensive patients. The biochemical identification of PA represents a compelling indication for the search of a unilateral adrenal cause of aldosterone excess, which is feasible with AVS and is mandatory before undertaking adrenalectomy.
A Prospective Study of the Prevalence of Primary Aldosteronism in 1,125 Hypertensive Patients
Supported by grants from FORICA (The Foundation for Advanced Research in Hypertension and Cardiovascular Diseases) to Dr. Rossi and from the Societá Italiana dell’Ipertensione Arteriosa.
- Abbreviations and Acronyms
- aldosterone-producing adenoma
- aldosterone/renin ratio
- adrenal vein sampling
- computed tomography
- idiopathic hyperaldosteronism
- logistic discriminant analysis
- magnetic resonance
- primary aldosteronism
- plasma renin activity
- Received May 10, 2006.
- Revision received July 13, 2006.
- Accepted July 23, 2006.
- American College of Cardiology Foundation
- Rossi E.,
- Regolisti G.,
- Negro A.,
- et al.
- Schwartz G.L.,
- Turner S.T.
- Bossuyt P.M.,
- Reitsma J.B.,
- Bruns D.E.,
- et al.
- Rossi G.P.,
- Sacchetto A.,
- Chiesura-Corona M.,
- et al.
- Abdelhamid S.,
- Blomer R.,
- Hommel G.,
- et al.
- ↵The National Kidney Foundation Kidney Disease Outcomes Quality Initiative (NKF KDOQI). Available at: http://www.kdoqi.org. Accessed July 8, 2006.
- Rundback J.H.,
- Sacks D.,
- Kent K.C.,
- et al.,
- American Heart Association
- Rossi G.P.,
- Sacchetto A.,
- Pavan E.,
- et al.