Author + information
- Received February 4, 2014
- Revision received May 8, 2014
- Accepted May 13, 2014
- Published online October 21, 2014.
- Leong L. Ng, MD∗,†∗ (, )
- Jatinderpal K. Sandhu, MPhil∗,
- Hafid Narayan, MD∗,
- Paulene A. Quinn, MPhil∗,
- Iain B. Squire, MD∗,†,
- Joan E. Davies, PhD∗,†,
- Joachim Struck, PhD‡,
- Andreas Bergmann, PhD‡,
- Alan Maisel, MD§ and
- Donald J.L. Jones, PhD#
- ∗Department of Cardiovascular Sciences, University of Leicester, Glenfield Hospital, Leicester, United Kingdom
- †National Institute for Health Research Leicester Cardiovascular Biomedical Research Unit, Glenfield Hospital, Leicester, United Kingdom
- ‡Sphingotec, GmbH, Hennigsdorf, Germany
- §Coronary Care Unit and Heart Failure Program, San Diego Veterans Affairs Medical Center, San Diego, California
- #Department of Cancer Studies and Molecular Medicine, Leicester Royal Infirmary, University of Leicester, Leicester, United Kingdom
- ↵∗Reprint requests and correspondence:
Dr. Leong L. Ng, Department of Cardiovascular Sciences, Clinical Sciences Wing, Glenfield Hospital, Leicester LE3 9QP, United Kingdom.
Background Pro-substance P (ProSP) is a stable surrogate marker for labile substance P, which has pro-inflammatory effects, increases platelet aggregation and clot strength, and reduces fibrinolysis.
Objectives This study assessed whether ProSP was associated with poor prognosis after acute myocardial infarction (AMI) to identify novel pathophysiological mechanisms.
Methods ProSP was measured in 1,148 AMI patients (825 men, mean age 66.2 ± 12.8 years). Endpoints were major adverse cardiac events (composite of death, reinfarction, and heart failure [HF] hospitalization), death/reinfarction, and death/HF. GRACE (Global Registry of Acute Coronary Events) scores were compared with ProSP for death and/or reinfarction at 6 months.
Results During 2-year follow-up, there were 140 deaths, 112 HF hospitalizations, and 149 re-AMI. ProSP levels were highest on the first 2 days after admission and related to estimated glomerular filtration rate, age, history of diabetes, ischemic heart disease or hypertension, Killip class, left ventricular wall motion index, and sex. Multivariate Cox regression models showed ProSP level was a predictor of major adverse events (hazard ratio [HR]: 1.30; 95% confidence interval [CI]: 1.10 to 1.54; p < 0.002), death and/or AMI (HR: 1.42; 95% CI: 1.20 to 1.68; p < 0.0005), death and/or HF (HR: 1.38; 95% CI: 1.14 to 1.67; p < 0.001). ProSP levels with GRACE scores were independent predictors of 6-month death and/or reinfarction (p < 0.0005 for both). ProSP-adjusted GRACE scores reclassified patients significantly (overall category-free net reclassification improvement of 31.6 (95% CI: 14.3 to 49.0; p < 0.0005) mainly by down-classifying those without endpoints.
Conclusions ProSP levels post-AMI are prognostic for death, recurrent AMI, or HF, and they improve risk prediction of GRACE scores, predominantly by down-classifying risk in those without events.
- acute coronary syndrome(s)
- B-type natriuretic peptide
- GRACE score
- myocardial infarction
- substance P
Over the last few decades, treatment of acute myocardial infarction (AMI) has improved substantially with the introduction of thrombolysis, percutaneous coronary intervention, beta-blockers, angiotensin-converting enzyme inhibitors, angiotensin receptor antagonists, statins, antiplatelet agents (such as aspirin and adenosine diphosphate receptor antagonists), and aldosterone antagonists. Although patient prognosis has improved, there remains a need to identify new pathways that may further improve outcomes after infarction.
Substance P (SP) and the neurokinins (NK) belong to the tachykinin family and are widely distributed in the central and peripheral nervous system (1). Although levels of SP are low in the myocardium, SP still can affect the heart via its role in nociception, inflammation, plasma extravasation, platelet and leukocyte aggregation in post-capillary venules, and leukocyte chemotactic migration through vessel walls (1). As for NK receptors, they are mainly present in coronary vessels and intracardiac ganglia, but not on the ventricular or atrial myocardium (2). One direct action on the NK1 receptor in coronary arteries may be nitric oxide–mediated vasodilation (1), although this effect may be impaired in patients with coronary artery disease (3), leading to a dominant NK2-mediated vasoconstriction. Both SP and neurokinin A are negatively inotropic and chronotropic, acting via cholinergic neurons (2), whereas NK1 antagonists improve inotropy and lusitropy in rat AMI models. SP attenuates the positive inotropic effect of norepinephrine (4) and, via the NK1 receptor, has been implicated in myocardial stunning post-AMI in guinea pigs (5).
SP is also expressed in platelets, where it has a pro-aggregatory effect (6). Furthermore, NK1 receptor inhibition reduces thrombus formation. Administration of an NK1 receptor antagonist reduced fibrinous adhesion formation and increased tissue plasminogen activator messenger ribonucleic acid and activity, suggesting that SP has a role in fibrinolysis (7). Also, SP strengthens clots formed in blood, an effect that may be mediated via leucocytes, the magnitude of which depends on full-length or truncated NK1 receptor isoform expression (8).
Myocardial (9) and pulmonary (10) SP has been observed to be increased in animal models of AMI, although investigations in humans have been hampered by the very short half-life of SP (12 min) (11), so there are no large studies examining the role of SP in AMI. The recent development of an assay for stable pro-tachykinin A (Pro-substance P [ProSP]), a surrogate for labile SP (12), has enabled studies on the role of this tachykinin system in human disease. We therefore investigated the potential role of SP in AMI by measuring ProSP and studying its association with major adverse cardiac events (MACE) such as death, heart failure (HF), or reinfarction.
We studied 1,148 ST-segment elevation myocardial infarction (STEMI) and non-STEMI patients admitted to University Hospitals of Leicester National Health Service trust between August 1, 2004 and April 30, 2007. This observational cohort study complied with the Declaration of Helsinki and was approved by the local ethics committee; patients gave written informed consent. AMI was diagnosed if a patient had a cardiac troponin I level above the 99th percentile with ≥1 of the following: chest pain lasting >20 min or diagnostic serial electrocardiographic changes consisting of new pathological Q waves or ST-segment and T-wave changes (13). Patients with known malignancy, renal replacement therapy, or surgery in the previous month were excluded. Estimated glomerular filtration rate (eGFR) was calculated from the simplified Modification of Diet in Renal Disease formula (14). All patients received standard medical treatment and revascularization at the discretion of the attending physician.
Blood samples (anticoagulated with ethylenediaminetetraacetic acid and aprotinin) were drawn after 15 min bed rest, immediately after diagnosis, and within 36 h of symptom onset. Plasma was stored at –80°C until assayed in a single batch for blinded determination of plasma ProSP and N-terminal pro-B-type natriuretic peptide (NT-proBNP).
Transthoracic echocardiography was performed in 895 patients (77.9%) during index admission, using either a Sonos 5500 or IE 33 instrument (Philips Medical Systems, Reigate, United Kingdom). A 16-segment left ventricular wall motion index score was performed on the basis of the American Society of Echocardiography method (15). In suitable patients, left ventricular ejection fraction was calculated using the biplane method of disks formula. Left ventricular systolic dysfunction was defined as either a left ventricular ejection fraction <40% or a left ventricular wall motion index >1.8.
GRACE (Global Registry of Acute Coronary Events) scoring
On the basis of an international observational database of acute coronary syndrome patients, GRACE scores can be calculated on initial presentation to predict in-hospital mortality (16) or for 6-month MACE, defined as death and/or reinfarction (17). We used GRACE scores on discharge for comparison with later outcomes.
The NT-proBNP assay was on the basis of a noncompetitive assay as previously published (18). Troponin I was measured using the Centaur cTnI Ultra immunoassay (Siemens Healthcare Diagnostics, Munich, Germany), which has a coefficient of variation of 10% at 0.03 μg/l with a 99th percentile of 0.04 μg/l. An assay for stable ProSP has been previously reported in detail (12) and was modified as follows: a mouse monoclonal anti-ProSP antibody (against amino acid sequence 21 to 36 of ProSP) was used to coat polystyrene tubes. Polyclonal antibodies against amino acids 3 to 22 of the ProSP sequence were labelled with methyl-acridinium ester and served as the detector antibody. Standards (ProSP peptide; amino acids 1 to 37 of ProSP) and samples (50 μl) were incubated in tubes with the detector antibody (200 μl). After equilibration, tubes were washed and bound chemiluminescence was detected with a luminometer LB952T/16 (Berthold, Wildbad, Germany). The lower detection limit of the immunoassay was 4.4 pmol/l.
The primary composite endpoint was MACE, defined as all-cause mortality, HF hospitalization, or recurrent AMI within 2 years of the index event. Hospitalization for HF was defined as a hospital readmission for which HF was the primary reason requiring treatment with high-dose diuretic agents, inotropes, or intravenous nitrate. Reinfarction was determined using the universal definition for MI (13). Other secondary composite endpoints were death and/or reinfarction and death and/or HF readmission. The endpoint of death and/or recurrent AMI at 6 months was used in analyses involving the GRACE score as this time point was used in developing the risk score. Endpoints were obtained by reviewing the local hospital databases and patient records, the Office of National Statistics Registry, and phone calls to patients. We achieved 100% follow-up.
Statistical analyses were performed on SPSS (version 20, IBM Corp., Armonk, New York) and Stata (version 12.1, StataCorp LP, College Station, Texas). Biomarker levels were log10 transformed and hazard ratios for these refer to 1 SD increment of the log10 transformed biomarker. GRACE scores were used as the original scores. Nonparametric tests were employed for data analysis (chi-squared test, Kruskal-Wallis test, and Spearman [rs] correlations). Independent predictors of ProSP levels were assessed using general linear models, with coefficients and p values reported for 2,000 bootstrap samples. To assess prognostic value of biomarkers, a “base” model was generated using Cox survival analysis, which included variables that were significantly (p < 0.10) associated with any of the study endpoints on univariate analysis (age, sex, previous history of ischemic heart disease, hypertension or diabetes, Killip class, eGFR, and biomarkers [log troponin I and log NT-proBNP]). ProSP was added to this base model to evaluate its added prognostic value. A second “comparative” Cox model was used to assess the relative prognostic power of NT-proBNP, ProSP, and the GRACE score. To demonstrate independence from clinical variables and NT-proBNP or the GRACE score with and without NT-proBNP, the added value of ProSP was evaluated on the basis of the likelihood ratio chi-square test for nested regression models. The additional prognostic value of ProSP in the base and comparative Cox models was further evaluated by reclassification analysis with calculation of category-free net reclassification improvement as described by Pencina et al. (19). We constructed classification trees using chi-square Automatic Interaction Detection (analysis performed using SPSS), which detects which biomarker has the strongest interaction with the dependent variable in stepwise analysis.
The characteristics of the study population are shown in Table 1, according to ProSP quartiles. Patients with higher ProSP levels were older, female, and more likely to have histories of hypertension, ischemic heart disease, diabetes, or HF. Higher ProSP levels also were associated with higher GRACE scores, NT-proBNP, and glucose levels, and lower ejection fractions and eGFR.
Spearman correlation analysis (rs) showed ProSP was significantly correlated to age (0.521), eGFR (–0.555), diastolic blood pressure (–0.178), NT-proBNP (0.428), wall motion score index (0.173), and heart rate (0.172) (all p < 0.0005). ProSP was not correlated to troponin or peak creatine kinase levels. A general linear model with 2,000 bootstrap samples showed eGFR, age, and Killip class > 1 as independent predictors of ProSP level (Table 2).
Day curves for ProSP
Sequential plasma sampling over 5 days was available for 110 patients, of whom 29 had a MACE within 2 years. Figure 1 demonstrates the plasma profile along with a general linear model with repeated measures that show significant changes in ProSP over time (p < 0.001) and higher levels in those with MACE (p < 0.03). In post-hoc testing, ProSP levels on day 1 were higher than on days 3, 4, or 5 (p < 0.001, 0.004, and 0.002, respectively; Bonferroni corrected for multiple comparisons). ProSP levels on days 1 and 2 were similar. There was no statistically significant interaction of the time profile of ProSP with MACE.
During 2-year follow-up, patients with elevated ProSP levels (log10 transformed and standardized by 1 SD) had more MACE, deaths, and rehospitalizations with HF or reinfarction (Table 1). Table 3 shows the univariate hazard ratios (HR) of various factors that affected MACE. In multivariate analysis for predicting 2-year MACE, significant independent predictors included age, Killip class >1, eGFR, and NT-proBNP. Addition of ProSP to the model (model 2 in Table 3) showed ProSP had an independent predictive value (HR: 1.30; 95% confidence interval [95% CI]: 1.10 to 1.54; p < 0.002), and the added value of ProSP as evaluated by the likelihood ratio chi-square test for nested regression models was p < 0.0001. Kaplan-Meier survival analysis visualizes the MACE rates in ProSP quartiles (Figure 2), showing quartile 4 was significantly different from all other quartiles (p < 0.0005, log rank test [Mantel-Cox]), and quartile 3 was significantly different from quartiles 4 (p < 0.0005), 2 (p < 0.022), and 1 (p < 0.001).
There was no significant benefit with the inclusion of glucose in Cox survival models for MACE (HR: 1.11; 95% CI: 0.97 to 1.26; p = NS), but there was for ProSP (HR: 1.24; 95% CI: 1.03 to 1.49; p = 0.02). Inclusion of white cell count in survival models for MACE showed no effect either (HR: 0.98; 95% CI: 0.95 to 1.01; p = NS) compared with ProSP, which did (HR: 1.27; 95% CI: 1.06 to 1.53; p = 0.01).
In other models for prediction of the secondary composite endpoints of death and/or recurrent AMI (Table 4) and death and/or HF readmission (Table 5), ProSP remained an independent predictor (p < 0.0005 and p < 0.001, respectively) of these endpoints. In both models, ProSP showed added value to the clinical variables and NT-proBNP (log likelihood chi-square test p < 0.0001 for both composite endpoints).
Comparison With GRACE Scores
The GRACE risk score (17) was originally derived for prediction of death and/or reinfarction at 6 months. We investigated the utility of the biomarkers NT-proBNP and ProSP for prediction of death and/or reinfarction as well as other composite endpoints (MACE, death, and/or HF). In univariate analysis, GRACE scores and the biomarkers NT-proBNP and ProSP predicted all composite endpoints (Table 6). In multivariate analysis for MACE, death, and/or recurrent AMI, and death and/or HF at 6 months, both NT-proBNP and ProSP demonstrated added value to the GRACE score. Moreover, ProSP showed added value to models with GRACE and NT-proBNP for all composite endpoints analyzed (p < 0.0001 for all) (Table 6).
Using receiver-operating characteristic curve analysis for death and/or reinfarction at 6 months, the area under the curve (AUC) increased from 0.69 (95% CI: 0.65 to 0.74) for GRACE scoring only to 0.72 (95% CI: 0.68 to 0.77) with the addition of ProSP (p = 0.01). Addition of NT-proBNP to the GRACE score yielded a higher area (AUC: 0.70; 95% CI: 0.65 to 0.75; p = NS). Comparison of the areas for GRACE score and NT-proBNP (AUC: 0.70; 95% CI: 0.65 to 0.75) and that of GRACE score, NT-proBNP, and ProSP (AUC: 0.72; 95% CI: 0.68 to 0.77) was not significant (p = 0.06).
Category-free reclassification analysis (with no arbitrary cutoff probabilities) was employed to calculate the net reclassification improvement for the effect of adding NT-proBNP or ProSP to the probabilities derived from the GRACE score in predicting the endpoints of death and/or reinfarction, MACE, and death and/or HF (Table 7). NT-proBNP up-classified risk in all those with events for all these composite endpoints. However, it wrongly (and significantly) up-classified risk in those without events for the death and/or reinfarction and MACE endpoints, although it correctly down-classified risk in those without events for the death and/or HF endpoint. However, ProSP correctly down-classified risk in those without events for all composite endpoints. When ProSP was added to a composite risk score composed of GRACE and NT-proBNP, ProSP down-classified risk in those without events for the endpoints of MACE and death and/or MI, but did not significantly reclassify those with the specific endpoint of death and/or HF (Table 7).
Decision tree analysis
To determine optimal cut points for biomarkers, we constructed decision trees (using ProSP and NT-proBNP levels and GRACE scores) to classify patients for event-free survival or those with an endpoint of death and/or reinfarction at 6 months. Using ProSP as an initial classifier (Figure 3), a ProSP level <72.08 pmol/l and GRACE score <137 defines a low-risk group of patients (n = 512, 44.6% of the total) who had an event rate that was 3.0%. Of these, only 1 (0.09%) had died within 30 days, and 3 patients (0.26%) had died within 6 months. ProSP levels >121.6 pmol/l defined a high-risk group of patients with a death/recurrent AMI rate of 37.7% and a death rate of 30.7% (Figure 3).
In our cohort of AMI patients, ProSP was most strongly correlated with renal function and was also influenced by age, past history of diabetes and ischemic heart disease, Killip class, wall motion index, sex, and blood pressure. There was no relation to infarct size. ProSP may therefore closely reflect a patient's renal function at AMI presentation. ProSP levels peaked at days 1 and 2 after chest pain onset, permitting an early assessment of risk.
During follow-up, ProSP was associated with cardiovascular outcomes such as death, recurrent AMI, and HF rehospitalization. Existing biomarkers such as NT-proBNP mainly predicted mortality and HF, with poorer detection of death and/or reinfarction. In contrast, ProSP provided independent prognostic information for the composite of MACE, death and/or reinfarction, and also death and/or HF. These analyses suggest that SP may potentially have a role in the pathophysiology of outcomes after AMI.
Analysis of the increment in receiver-operating characteristic AUC showed that addition of ProSP had small effects on this area. However, such analyses are relatively insensitive to the addition of novel biomarkers (20) and reclassification analyses should also be performed. For example, NT-proBNP demonstrated a small nonsignificant increase in receiver-operating characteristic AUC, whereas it showed added value in reclassification analyses for both up- and down-classifying risk in those with MACE, death and/or MI, and death and/or HF. In the reclassification analysis, ProSP demonstrated additional utility to the GRACE score, used as a standard risk classification tool in AMI, mainly by down-classifying risk in those without endpoints. Such a biomarker would be especially useful in detecting patients with low risk, which was confirmed on decision tree analysis. ProSP levels <72.1 pmol/l may define a low-risk group of patients, who potentially could be discharged from hospital earlier. The event rates in this group were very low during follow-up with only 3 deaths (0.26%) by 6 months.
Reclassification analyses also suggested that NT-proBNP could up- or down-classify risk in those assessed with the GRACE score, but ProSP only showed added value for the endpoints of MACE and death and/or MI not death and/or HF. However, because the GRACE score was derived mainly for prediction of death and/or MI, this could be a limitation when including HF as an endpoint.
Association of ProSP with poor outcomes may reflect some of the known effects of SP on physiology and pathophysiology (Central Illustration). The association with HF rehospitalization could be due to the known negative inotropic and lusitropic effects of SP that have been demonstrated in animal models of AMI (4,5). SP is expressed in monocytes and macrophages (21) and may play a role in inflammation (22) as well as leukocyte chemotaxis and egress from vessels (1,2), which may also affect myocardial function. Emerging evidence suggests SP is a mast cell secretagogue via the NK1 receptor and mast cells may play a role in adverse remodelling (23,24). In animal models of remodelling from volume overload, NK1 antagonists prevented an increase in mast cell density and myocardial tumor necrosis factor alpha (23), and remodelling also was reduced in TAC1 gene knockout mice. Mast cells are co-localized with cardiac nerves (24) and secrete proteases to activate collagenases and gelatinases, the putative mediators of remodelling. Furthermore, mast cells secrete stored renin, resulting in local activation of the renin-angiotensin system (9).
The association of SP with readmission for AMI may be due to effects of SP on platelet aggregation (6,25). Tachykinin family peptides (SP, endokinins A and B) are found in platelets (25), and receptors on platelet membranes (NK1 and NK3) may mediate a modulatory positive paracrine feedback on platelet activation. NK1 inhibition reduces thrombus formation, but SP strengthens a clot after formation (8), which is dependent on NK1 receptors on leucocytes. This effect is more marked in those patients with a full-length NK1 receptor (8). In addition, SP may reduce tissue plasminogen activator activity and expression (7), hence potentially promoting thrombosis.
On the other hand, SP also may have some potentially beneficial roles in ischemia post-conditioning (26) and in mobilization of progenitor cells that may play a role in angiogenesis within ischemic tissue (27). Our findings are hypothesis-generating for investigating the role of SP on outcomes after AMI, as it is uncertain whether beneficial effects of SP could be outweighed by their deleterious effects.
Our findings are on the basis of a population from a single center, with 2 admitting hospitals, and should be verified in other larger populations. The rate of early revascularization in our non-STEMI population was low and may not reflect the more contemporary invasive approach of revascularization within 72 h of presentation. One advantage is that the relation of ProSP with poor outcomes would not have been confounded by higher early revascularization rates.
After AMI, circulating ProSP levels provide added value to the prognostic information determined by the GRACE score and the prognostically important biomarker NT-proBNP. The ability of ProSP to predict recurrent AMI in addition to mortality may confer clinical utility on the tachykinin system in risk stratification after AMI.
COMPETENCY IN MEDICAL KNOWLEDGE: Pro-substance P provides information beyond that of existing clinical risk assessment tools and may have prognostic value in patients with myocardial infarction as a predictor of heart failure, reinfarction, and survival.
TRANSLATIONAL OUTLOOK 1: Larger multicenter studies are needed to explore the value of Pro-substance P in clinical risk stratification among survivors of acute coronary syndromes.
TRANSLATIONAL OUTLOOK 2: The effect of neurokinin receptor inhibitors on clinical outcomes after myocardial infarction requires further investigation.
This work was supported by the John and Lucille Van Geest Foundation and the National Institute for Health Research Leicester Cardiovascular Biomedical Research Unit. Dr. Ng receives research funding from Sphingotec AG; and holds intellectual property submissions on biomarkers.
Dr. Bergmann holds ownership in Sphingotec AG, which provided the ProSP assay; and is a member of the board of directors of Sphingotec GmbH. Dr. Maisel has received consulting fees from Alere, Critical Diagnostics, Sphingotec, BG Medicine, and EFG Diagnostics; receives research funds from Alere, Abbott, BG Medical, and Brahms; receives speaking fees from BG Medical and Alere; and is a co-founder of Cardero Therapeutics and My Life Diagnostics. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Sphingotec GmbH is a mid-sized company based in Hennigsdorf, Germany; it commercializes immunoassays and has developed the ProSP assay, for which it owns patent rights. The Guest Editor for this paper is the same as the accompanying editorial comment author, Christopher B. Granger.
- Abbreviations and Acronyms
- acute myocardial infarction
- area under the curve
- confidence interval(s)
- estimated glomerular filtration rate
- heart failure
- hazard ratio(s)
- major adverse cardiac event(s)
- N-terminal pro-B-type natriuretic peptide
- Pro-substance P
- substance P
- ST-segment elevation myocardial infarction
- Received February 4, 2014.
- Revision received May 8, 2014.
- Accepted May 13, 2014.
- American College of Cardiology Foundation
- Otsuka M.,
- Yoshioka K.
- Jones S.,
- Tucker K.L.,
- Sage T.,
- et al.
- Reed K.L.,
- Fruin A.B.,
- Gower A.C.,
- Stucchi A.F.,
- Leeman S.E.,
- Becker J.M.
- Morrey C.,
- Brazin J.,
- Seyedi N.,
- Corti F.,
- Silver R.B.,
- Levi R.
- Thygesen K.,
- Alpert J.S.,
- White H.D.,
- et al.,
- for the Joint ESC/ACCF/AHA/WHF Task Force for the Redefinition of Myocardial Infarction
- Smilde T.D.,
- van Veldhuisen D.J.,
- Navis G.,
- Voors A.A.,
- Hillege H.L.
- Schiller N.B.,
- Shah P.M.,
- Crawford M.,
- et al.
- Omland T.,
- Persson A.,
- Ng L.,
- et al.
- Cook N.R.
- Ho W.Z.,
- Lai J.P.,
- Zhu X.H.,
- Uvaydova M.,
- Douglas S.D.
- Severini C.,
- Improta G.,
- Falconieri-Erspamer G.,
- Salvadori S.,
- Erspamer V.
- Meléndez G.C.,
- Li J.,
- Law B.A.,
- Janicki J.S.,
- Supowit S.C.,
- Levick S.P.
- Levick S.P.,
- Meléndez G.C.,
- Plante E.,
- McLarty J.L.,
- Brower G.L.,
- Janicki J.S.
- Amadesi S.,
- Reni C.,
- Katare R.,
- et al.