Article Text

other Versions

Original article
MR-proANP and MR-proADM for risk stratification of patients with acute chest pain
  1. Stergios Tzikas1,
  2. Till Keller2,
  3. Francisco M Ojeda3,
  4. Tanja Zeller3,
  5. Philipp S Wild1,4,
  6. Edith Lubos3,
  7. Jan Kunde5,
  8. Stephan Baldus3,6,
  9. Christoph Bickel7,
  10. Karl J Lackner8,
  11. Thomas F Münzel1,
  12. Stefan Blankenberg3
  1. 1Department of Medicine 2, University Medical Center Mainz, Mainz, Germany
  2. 2Division of Cardiology, Department of Medicine III, Goethe University Frankfurt, Frankfurt, Germany
  3. 3Clinic for General and Interventional Cardiology, University Heart Center Hamburg, Hamburg, Germany
  4. 4Center for Thrombosis and Hemostasis, University Medical Center Mainz, Mainz, Germany
  5. 5BRAHMS GmbH, Henningsdorf, Germany
  6. 6Department of Internal Medicine III, University of Cologne, Cologne, Germany
  7. 7Department of Internal Medicine, Federal Armed Forces Hospital, Koblenz, Germany
  8. 8Institute for Clinical Chemistry and Laboratory Medicine, University Medical Center Mainz, Mainz, Germany
  1. Correspondence to Dr Till Keller, Division of Cardiology, Department of Medicine III, Goethe University Frankfurt, Theodor Stern Kai 7, Frankfurt 60590, Germany; keller{at}


Objective To evaluate mid-regional pro-adrenomedullin (MR-proADM) and mid-regional pro-atrial natriuretic peptide (MR-proANP) as prognostic biomarkers in a representative ‘real world’ cohort of patients with suspected acute coronary syndrome (ACS).

Design Prospective observational multicentre cohort study.

Setting Chest pain units of three major hospitals in Germany from 2007 to 2008.

Patients Patients presenting with signs and symptoms suggestive of an ACS.

Main outcome measures Primary end point was death or non-fatal myocardial infarction (MI), and secondary end point was death, non-fatal MI, stroke, need for coronary revascularisation, and hospital admission for cardiovascular cause or acute heart failure within 6 months after enrolment.

Results 1386 patients (male/female=920/466) were enrolled. Follow-up information was available for 97.8% of patients (median follow-up time 183 days). Forty-three patients reached the primary end point, and 132 the secondary end point. Patients who reached a primary end point had significantly higher MR-proANP (271 vs 101 pmol/l, p<0.001) and MR-proADM (0.86 vs 0.59 nmol/l, p<0.001) concentrations than those who did not. Cox regression analysis revealed a 2.55-fold risk of death or non fatal MI (95% CI 1.48 to 2.46, p<0.001) for an increment of the log-transformed MR-proANP concentration by 1 SD after adjustment for cardiovascular risk factors, and a 1.91-fold risk (95% CI 1.48 to 2.46, p<0.001) for MR-proADM. Both peptides could result in significant reclassification of patients when added to the Global Registry of Acute Coronary Events risk score, with an overall net reclassification improvement of 41.2% for MR-proADM and 35.7% for MR-proANP.

Conclusions MR-proADM and MR-proANP are predictors of future cardiovascular events in patients presenting with acute chest pain and might facilitate the choice of treatment in those patients complementary to established risk scores.

Statistics from


The appropriate treatment for patients presenting with acute chest pain and possible acute coronary syndrome (ACS) is based on timely diagnosis and proper risk stratification aided by biomarkers.1 An increasing number of emerging biomarkers have been identified to play an important role in the pathophysiology and natural course of ACS, but it is crucial to validate their specific predictive value in the particular setting of acute chest pain before implementing them as diagnostic or prognostic tools.2 Examples of such biomarkers are the mid-regional peptides of atrial natriuretic peptide (ANP) and adrenomedullin (ADM).

ANP is derived by cleavage of its precursor pro-atrial natriuretic peptide (proANP); it is significantly more stable in the circulation than the mature peptide and has been suggested to be a more reliable analyte.3 The mid-regional part of proANP (MR-proANP), which is not so prone to enzymatic degradation as its N- and C-terminal regions, is the target of a contemporary diagnostic assay.4 MR-proANP has so far been shown to offer comparable diagnostic and prognostic performance to other natriuretic peptides in heart failure.5–7

ADM is a peptide mainly present in endothelial and vascular smooth muscle cells with multiple biological actions affecting the cardiovascular system.8 The technical difficulties in reliably quantifying ADM have been surmounted with the identification of the mid-regional proADM (MR-proADM), a peptide generated during the processing of preproADM.9 MR-proADM levels have been shown to predict adverse outcome in the recovery phase of acute myocardial infarction (AMI)10 and in patients with coronary artery disease (CAD),11 and it appears to add prognostic information beyond that of natriuretic peptides in risk stratification of patients with chronic and stable heart failure.12

However, although early results suggest the potential of MR-proANP and/or MR-proADM, considerably more data are needed before the potential of either can be considered proven. In this study, we aimed to evaluate MR-proANP and MR-proADM in a large unselected ‘real world’ cohort of patients presenting with acute chest pain suggestive of ACS. Our goal was to explore the association of these biomarkers with future cardiovascular events and test their predictive value in comparison with established predictors of cardiovascular risk such as the B-type natriuretic peptide (BNP) and sensitive troponin determination, as well as the Global Registry of Acute Coronary Events (GRACE) risk score.13


Study sample

Between January 2007 and July 2008, patients consecutively presenting with acute chest pain and therefore suspected ACS were enrolled in this study as described previously.14 Patients were enrolled at the chest pain units of Johannes Gutenberg-University Medical Center Mainz, the Federal Armed Forces Hospital Koblenz and University Hospital Hamburg-Eppendorf. Eligible patients were between 18 and 85 years of age. Exclusion criteria were major surgery or trauma within the preceding 4 weeks, pregnancy, intravenous drug misuse, and anaemia with haemoglobin level below 10 g/dl. The study complies with the Declaration of Helsinki, and the local ethics committee approved the study. Participation was voluntary and each patient gave written, informed consent.

Study protocol and definitions of final diagnosis

Blood was drawn at admission and after 3 and 6 h. A 12-lead ECG was recorded at the same time points. Among other data, the classical cardiovascular risk factors were documented at admission as described previously in detail.14

AMI was diagnosed according to the established universal definition.15 Adjudication of the final diagnosis was based on all available conventional serial troponin measurements. A detailed description of the conventional in-house troponin assays used is given in the online supplementary data. Unstable angina pectoris was diagnosed if the ECG was non-diagnostic, in-house troponin was negative, but coronary angiography revealed a culprit lesion, or ischaemia was proven in a stress test with the subsequent need of coronary intervention. Patients with excluded ACS were categorised as having non-coronary chest pain (NCCP). Two independent cardiologists made the final diagnosis on the basis of all available clinical, laboratory and imaging findings.

Cardiovascular events were registered during a follow-up period of 6 months. Patients were personally contacted by telephone and/or letter, and events were adjudicated through hospital charts, available outpatient information and data from the local civil registry office. The combined primary end point was death or non-fatal myocardial infarction (MI) (not including the index event), and the secondary end point included death, non-fatal MI (not including the index event), stroke, and subsequent need for revascularisation and hospitalisation due to a cardiovascular cause or heart failure.

Laboratory methods

Routine laboratory variables including creatinine and C-reactive protein were measured immediately after blood withdrawal by standardised methods. The estimated glomerular filtration rate (eGFR) was calculated by the abbreviated Modification of Diet in Renal Disease (MDRD) equation.16 EDTA plasma, citrate plasma and serum samples were collected at each time point and centrifuged, and aliquots were frozen at −80°C until further measurement of the investigational biomarkers.

Troponin I (TnI) was measured using a contemporary sensitive assay on the ADVIA Centaur XP system (TnI-Ultra; Siemens Healthcare Diagnostics, Germany). The assay range is 0.006–50 ng/ml, and the coefficient of variation (CV) at 0.03 ng/ml is 10%. The reference limit based on the 99th centile for a healthy population is 0.04 ng/ml.17 BNP was assayed on the ARCHITECT i System (Abbott Diagnostics, Germany). The analytical sensitivity of the assay is ≤10 pg/ml with a range of 0–5000 pg/ml. MR-proANP was determined using a commercial fluoroimmunoassay (BRAHMS MR-proANP KRYPTOR; BRAHMS GmbH, Hennigsdorf, Germany). The lower detection limit of this assay is 6.0 pmol/l. The intra-assay CV is 10% for samples containing 23–3000 pmol/l MR-proANP and 20% for samples containing 18–22.8 pmol/l. The interassay CV is 10% at 65 pmol/l MR-proANP and 20% at 18 pmol/l MR-proANP.4 MR-proADM was measured using a commercial fluoroimmunoassay (BRAHMS MR-proADM KRYPTOR) using two polyclonal antibodies to amino acids 45–92 of proADM. The assay has an analytical detection limit of 0.08 nmol/l, and the interassay CV is <20% for values ≥0.12 nmol/l.9

Statistical analysis

Skewed variables are reported as median (IQR), and symmetric variables as mean±SD. The Mann–Whitney test was used to compare median biomarker levels among different patient groups. The Spearman correlation coefficient was used to study the association between biomarkers and continuous variables. The relation of the investigational biomarkers to the predefined end points was tested with Cox regression analysis. Three models were used: model 1, adjusted for age and gender; model 2, adjusted for age, gender and classical risk factors18 (body mass index, presence or absence of hypertension, diabetes, smoking, dyslipidaemia, family history of CAD); model 3, adjusted for the GRACE risk score.19 HRs are given for an increase of 1 SD in the tested log-transformed biomarker. Reclassification analysis was performed for the addition of the investigational biomarkers—both as continuous and as dichotomised variables (using predefined cut-offs)—to a model based on the GRACE risk score. For these analyses, the continuous net reclassification improvement (NRI) was used.20 Survival curves according to tertiles of biomarker concentration were estimated by Kaplan–Meier analysis and compared for significance with the log-rank test. The predictive value of the investigational biomarkers for future cardiovascular events was also assessed with C-indices. p<0.05 was considered significant. Analyses were carried out using R 2.14.2 (R Foundation for Statistical Computing, Vienna, Austria).


AMI was diagnosed in 299 patients (21.6%), with 93 (6.7%) having an ST segment elevation MI (STEMI) and 206 (14.9%) a non-STEMI (NSTEMI). Unstable angina pectoris was diagnosed in 184 (13.3%) patients. Baseline characteristics of the study sample stratified for diagnosis are summarised in online supplementary table S1. Follow-up information was available for 1356 (97.8%) patients with median follow-up time of 183 days. Forty-three patients reached the combined primary end point, and 132 the combined secondary end point. Patients who reached the primary end point had significantly higher MR-proADM (0.86 vs 0.59 nmol/l, p<0.001) and MR-proANP (271 vs 101 pmol/l, p<0.001) at baseline than those who did not. Baseline characteristics of the study sample stratified for prognosis are summarised in table 1.

Table 1

Characteristics of the study sample according to cardiovascular 6-month outcome

Descriptive analysis of the investigational biomarkers

The associations of the investigational biomarkers with categorical variables are described in online supplementary table S2. Regarding the association with continuous variables, MR-proADM showed a moderate positive correlation with age (r=0.6, p<0.001) and a moderate negative correlation with renal function (eGFR) (r=0.59, p<0.001). MR-proANP also showed a moderate positive correlation with age (r=0.62, p<0.001) and a moderate negative correlation with renal function (eGFR) (r=0.48, p<0.001).

Prognostic value of the investigational biomarkers

Kaplan–Meier curves and log rank test p values for event-free survival (primary end point) for the overall study cohort divided into subgroups according to tertiles of MR-proADM and MR-proANP levels are presented in figure 1. A decrease in event-free survival over tertiles was observed, with the highest event rates in the upper tertiles. Moreover, there seemed to be a threshold effect, since the risk of adverse events was relatively similar in the lower two tertiles and a clear risk increase could be observed for subjects in the upper tertile.

Figure 1

Cumulative survival according to tertiles of mid-regional pro-adrenomedullin (MR-proADM) and mid-regional pro-atrial natriuretic peptide (MR-proANP). Kaplan–Meier survival curves are presented for patients with chest pain for the combined primary end point of death or non-fatal myocardial infarction within 6 months of presentation according to tertiles of MR-proADM and MR-proANP.

In Cox regression analysis, MR-proANP treated as a continuous variable was strongly related to the event-free survival. One SD increase in the log-transformed MR-proANP levels was associated univariately with a 2.78-fold risk of death or non-fatal MI (95% CI 2.14 to 3.63, p<0.001). The strong predictive value of MR-proANP remains after adjustment for age and gender (model 1), classical risk factors (model 2) and the GRACE risk score (model 3) (table 2).

Table 2

Prognostic performance of biomarkers in patients presenting with chest pain

MR-proADM was also strongly related to the event-free survival. One SD increase in log-transformed MR-proADM levels was associated univariately with a 2.24-fold risk of death or non-fatal MI (95% CI 1.84 to 2.73, p<0.001). Cox regression multivariate analysis for the primary end point, incorporating cardiovascular risk factors and plasma MR-proADM concentrations revealed an HR of 1.91 (95% CI 1.48 to 2.46, p<0.001) for an increment in the normalised MR-proADM levels by 1 SD (table 2).

MR-proANP and MR-proADM showed in most cases a comparable predictive value to BNP and TnI for future cardiovascular events and mortality. The predictive performance for future cardiovascular events using Harrell C-indices is also shown in table 2.

In our analysis, consistent with previous reports,21 ,22 we did not consider the primary diagnosis of MI as an event but only the events recorded during the follow-up time of 6 months. When the primary MI is considered as an event in the analysis, the C-index (adjusted for all cardiovascular risk factors) is 0.73 (95% CI 0.69 to 0.76) for MR-proANP, 0.68 (95% CI 0.65 to 0.71) for MR-proADM and 0.71 (95% CI 0.68 to 0.74) for BNP. TnI in this case outperforms the other markers, with a C-index of 0.95 (95% CI 0.94 to 0.97), because of its superior diagnostic power, as we have already reported.14

Serial measurements of biomarkers

The kinetics of the investigational biomarkers according to the diagnosis are shown in online supplementary figure S1. In non-ACS patients, there is almost no change in biomarker levels within the first 6 h after presentation, whereas patients with AMI show a rise in BNP and MR-proADM levels and a decrease in MR-proANP levels. The three different sampling time points yielded no additional value regarding prognostic ability (see online supplementary table S3).

Amendment to clinical risk classification

Net reclassification for the addition of MR-proADM to the GRACE risk score placed 15 patients with events as lower risk and 29 as higher risk, while at the same time 635 patients without an event were reclassified upwards, and 653 ended up in a lower risk group. The overall NRI was 41.2%. Similar results were observed with MR-proANP (overall NRI 35.7%) and with BNP (overall NRI 42.1%) (table 3).

Table 3

Reclassification table for adding biomarkers to the GRACE score

We performed additional analyses after dichotomising the biomarker concentrations using either the medians or other predefined cut-offs to better reflect possible usage in clinical routine. To define risk categories, a cut-off of 144 GRACE score points was used as previously recommended in order to identify patients who benefit the most from early invasive treatment.23 As the predefined cut-off, we used a value of 236 pmol/l, reported to be the optimal prognostic cut-off,24 for MR-proANP, a value of 1.11 nmol/l22 for MR-proADM, and a value of 500 pg/ml,25 used to identify high-risk patients, for BNP. The prognostic performance of these cut-offs is summarised in online supplementary table S4.

Calculation of the established GRACE score, which incorporates a troponin cut-off as a marker of myocardial necrosis showed good prognostic power, with a C-index of 0.80. MR-proADM, MR-proANP and BNP all increased the C-index when added to the GRACE score to 0.81 (see online supplementary table S5).


This prospective observational multicentre cohort study demonstrates that plasma MR-proANP and MR-proADM concentrations in patients presenting with acute chest pain and a suspicion of ACS provide significant prognostic information about future cardiovascular events in the 6 months after presentation. We can therefore broaden and refine our knowledge on mid-regional peptides for cardiovascular risk prediction in the setting of acute chest pain. Our study is the first to investigate the possible merit of these two mid-regional peptides for prognosing outcome in an unselected cohort of patients with acute chest pain, using high-sensitivity troponin, BNP and the GRACE risk score as comparators.

The prognostic value of MR-proANP has so far been investigated in various clinical settings. The LAMP Study (Leicester AMI Peptide Study) reported that MR-proANP is a relevant prognostic biomarker in post-MI patients independently of established conventional risk factors.26 In that study, MR-proANP showed a predictive power equal to N-terminal proBNP for the prediction of death (area under the curve for both, 0.83), while there was no comparison with a novel more sensitive troponin assay. Moreover, the blood sampling was carried out in the recovery phase of AMI, 3–5 days after the first admission of the patient to the hospital, and STEMI patients were over-represented in the sample. In another study, von Haehling et al investigated a large cohort of patients with symptomatic CAD and showed that MR-proANP is able to identify high-risk patients.24 This study group suggested an optimal cut-off of 236 pmol/l for MR-proANP. We confirmed the good prognostic value of this suggested cut-off in our cohort; the increase in MR-proANP concentration above 236 pmol/l was associated with an HR of 4.69 (95% CI 2.35 to 9.38). Recently, another study group21 investigated the diagnostic and prognostic value of admission MR-proANP in a cohort of 675 patients with chest pain and compared it with conventional and high-sensitivity troponin determination. Although MR-proANP could not add diagnostically to a modern sensitive troponin assay, it may improve its predictive power. That study used a cohort with similar characteristics to the cohort in our study, but a longer follow-up period of 360 days and no serial measurements. Our reported C-index of MR-proANP (0.776) for the prediction of death non-fatal MI is consistent with that reported in the above study of Meune et al (0.750).21

More studies on the prognostic value of MR-proADM in various clinical settings are available. Khan et al used a combined primary end point of death or heart failure to evaluate the prognostic value of MR-proADM in a large cohort of patients after AMI.27 The threshold effect present in our analysis for MR-proADM with significantly worse outcome in the upper tertiles was also present in that study, which reported a value of 1.04 nmol/l for MR-proADM as a good discriminator for worse outcome. In our study, however, we also found good discriminating power for lower concentrations (0.71 nmol/l). The limitations mentioned above for the LAMP Study cohort concerning blood sampling and over-representation of STEMI patients in the sample also apply here. More recently, investigators in the LAMP II Study provided additional evidence on the prognostic utility of MR-proADM in a large cohort of NSTEMI patients, even beyond that of the GRACE score.22 Our reported C-index for MR-proADM (0.777) for the prediction of death non-fatal MI is consistent with that reported in that study for death (0.790) or Major Adverse Cardiac Event (MACE) (0.720). The reported optimal cut-off for MR-proADM of 1.11 nmol/l showed good prognostic value in our cohort also; the increase in MR-proADM concentration above 1.11 nmol/l was associated with an HR of 3.30 (95% CI 1.62 to 6.72). MR-proADM measurements on discharge did not provide additional prognostic information in that study. The value of repeated measurement of MR-proADM has also been recently investigated in patients admitted with acute dyspnoea.28 In that study, the additional blood sampling was performed within 14–48 h of presentation and/or at discharge. Once discharge values were combined with admission values, the second measurement no longer added prognostic utility. In that study, investigators used a cut-off of 1.985 nmol/l for MR-proADM for the optimal prediction of 90 days mortality in patients diagnosed with acute heart failure. A further study by von Haehling et al29 concludes that MR-proADM is an independent predictor of mortality in patients with chronic heart failure, which adds prognostic information to N-terminal proBNP. The optimal cut-off for MR-proADM in that study for the prediction of 6 months outcome was reported to be 0.73 nmol/l, which is closer to the value reported in our cohort for the third tertile (>0.71 nmol/l) associated with worse outcome. Furthermore, the biomarker levels for the upper tertile identifying patients at high risk in our study (MR-proADM>0.71 nmol/l, MR-proANP> 144 pmol/l) are also comparable with the age-dependent cut-offs calculated in a recently published study by Shah et al investigating a cohort of 560 patients with acute dyspnoea (MR-proADM>0.77 nmol/l, MR-proANP>194 pmol/l).30

To the best of our knowledge, there are no data available on serial sampling of MR-proANP and MR-proADM at the time of initial triage after presentation due to chest pain and/or possible ACS. In the present study, we prognostically investigated serial triage sampling of both mid-regional peptides and verified their different kinetics. MR-proANP shows quicker release in the circulation than BNP and MR-proADM (online supplementary figure S1), with determination directly on admission delivering robust predictive value, and further serial measurements not adding any relevant prognostic information. These results confirm data from an older study by Jernberg et al showing that the relative change in N-terminal proBNP within 6 h of admission for acute-onset chest pain suggesting ACS did not carry any prognostic information.31 Although the concentrations of the different natriuretic peptides correlate with each other, they have different modes of secretion, different modes of degradation, and, most important, different ranges, cut-off values and half-lives.32

An important aspect of our study is the validation of the utility of the GRACE risk score33 as a predictive tool. As applied in our study, it performed excellently (C-index 0.80 (95% CI 0.73 to 0.87)) in predicting death or non-fatal MI in the 6 months after admission. According to our results, this performance could be further improved by the use of novel biomarkers such as MR-proADM and MR-proANP. This is consistent with previous results on different study cohorts22 ,24 and was shown for the first time in an unselected cohort of patients with chest pain, which might have important clinical implications.

The principal symptom that initiates the diagnostic and therapeutic cascade in patients with ACS is acute chest pain. Current guidelines1 recommend an early invasive strategy, based on elevated troponin and clinical risk scores such as the GRACE score. Identifying patients at risk and treating them aggressively can lead to a reduction in death, MI and refractory ischaemia of up to 28%.23 The present results generate the hypothesis that, in the setting of ACS, MR-proANP and MR-proADM could identify patients at risk, who might benefit from the early use of invasive treatment. Future studies should investigate this hypothesis. A further challenging issue is optimal care of patients with unstable angina (troponin negative) and elevated MR-proANP or MR-proADM, and even more demanding may be the optimal management of patients with non-cardiac chest pain with elevated MR-proANP or MR-proADM.

Limitations of the study

This study was performed on a representative German sample; distribution may differ in other regions. A relatively small number of events were registered during follow-up in this intermediate risk cohort, which may have led to unsound results especially in the multivariate survival analysis. Furthermore, the small number of events does not allow reliable statistical analysis regarding subgroups of the study cohort. The clinical benefit from improved risk stratification was not assessed from the study and can only be hypothesised.


MR-proANP and MR-proADM are significant predictors of future cardiovascular events in patients presenting with acute chest pain. Their predictive value is greater than that of the established biomarkers, TnI and BNP, and complement clinical risk scoring systems such as the GRACE score. The possible benefit of using such predictors in risk stratification and triaging of patients with chest pain requires further investigation.


We thank the doctoral students Dudu Kutlu, MD, Christian Bredel, MD, Gerhard Pioro, MD, Ewa Czyz, MD, Eliana Pallazetti, MD, Irene Schönhagen, MD, Lars Lillpopp, MD and Anne Schauer (all from the Department of Medicine II, University Medical Center, Johannes Gutenberg University Mainz) for help with data acquisition. We also thank the employed staff of the biomarker laboratory (Department of Medicine II, University Medical Center, Johannes Gutenberg University Mainz) for their help with biomarker determination.


View Abstract
  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • TFM and SBl contributed equally.

  • Contributors All authors contributed significantly to the manuscript. Conception and design: ST, TK, SBl, TFM. Analysis and interpretation of data: FMO, TZ, PSW, EL, JK, SBa, CB, KJL. Drafting the article: ST, TK. Revising the article critically for important intellectual content: FMO, TZ, PSW, EL, JK, SBa, CB, KJL, TFM, SBl. Final approval of the version to be published: ST, TK, FMO, TZ, PSW, EL, JK, SBa, CB, KJL, TFM, SBl. ST and SBl had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

  • Funding This work was supported by the Johannes Gutenberg University of Mainz through the research programmes ‘Wissen schafft Zukunft’ and ‘Schwerpunkt Vaskuläre Prävention’ and by an unrestricted grant from BRAHMS and Abbott Diagnostics. The funders had no role in the design and conduct of the study, in the collection, management, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.

  • Competing interests JK is employed by BRAHMS. SBl is a member of the BRAHMS medical advisory board.

  • Ethics approval Ethics committee of the Johannes Gutenberg University of Mainz.

  • Patient consent Obtained.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.