Lung Function Testing in Asthma Nose, Fractional ...

Diagnostic Performance of an Electronic Nose, Fractional Exhaled Nitric Oxide, and Lung Function Testing in Asthma Paolo Montuschi, Marco Santonico, Chiara Mondino, Giorgio Pennazza, Giulia Mantini, Eugenio Martinelli, Rosamaria Capuano, Giovanni Ciabattoni, Roberto Paolesse, Corrado Di Natale, Peter J. Barnes and Arnaldo D'Amico Chest 2010;137;790-796; Prepublished online January 15, 2010; DOI 10.1378/chest.09-1836 The online version of this article, along with updated information and services can be found online on the World Wide Web at: http://chestjournal.chestpubs.org/content/137/4/790.full.html Supplemental material related to this article is available at: http://chestjournal.chestpubs.org/content/suppl/2010/04/06/chest.09-183 6.DC2.html

Chest is the official journal of the American College of Chest Physicians. It has been published monthly since 1935. Copyright2010by the American College of Chest Physicians, 3300 Dundee Road, Northbrook, IL 60062. All rights reserved. No part of this article or PDF may be reproduced or distributed without the prior written permission of the copyright holder. (http://chestjournal.chestpubs.org/site/misc/reprints.xhtml) ISSN:0012-3692

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

CHEST

Original Research ASTHMA

Diagnostic Performance of an Electronic Nose, Fractional Exhaled Nitric Oxide, and Lung Function Testing in Asthma Paolo Montuschi, MD; Marco Santonico, PhD; Chiara Mondino, MD; Giorgio Pennazza, PhD; Giulia Mantini, BEng; Eugenio Martinelli, PhD; Rosamaria Capuano, BEng; Giovanni Ciabattoni, MD; Roberto Paolesse, PhD; Corrado Di Natale, PhD; Peter J. Barnes, DM, FCCP; and Arnaldo D’Amico, PhD

Background: Analysis of exhaled breath by biosensors discriminates between patients with asthma and healthy subjects. An electronic nose consists of a chemical sensor array for the detection of volatile organic compounds (VOCs) and an algorithm for pattern recognition. We compared the diagnostic performance of a prototype of an electronic nose with lung function tests and fractional exhaled nitric oxide (FENO) in patients with atopic asthma. Methods: A cross-sectional study was undertaken in 27 patients with intermittent and persistent mild asthma and in 24 healthy subjects. Two procedures for collecting exhaled breath were followed to study the differences between total and alveolar air. Seven patients with asthma and seven healthy subjects participated in a study with mass spectrometry (MS) fingerprinting as an independent technique for assessing between group discrimination. Classification was based on principal component analysis and a feed-forward neural network. Results: The best results were obtained when the electronic nose analysis was performed on alveolar air. Diagnostic performance for electronic nose, FENO, and lung function testing was 87.5%, 79.2%, and 70.8%, respectively. The combination of electronic nose and FENO had the highest diagnostic performance for asthma (95.8%). MS fingerprints of VOCs could discriminate between patients with asthma and healthy subjects. Conclusions: The electronic nose has a high diagnostic performance that can be increased when combined with FENO. Large studies are now required to definitively establish the diagnostic performance of the electronic nose. Whether this integrated noninvasive approach will translate into an early diagnosis of asthma has to be clarified. Trial registration: EUDRACT https://eudralink.emea.europa.eu; Identifier: 2007-000890-51; and clinicaltrials.gov; Identifier: NCT00819676. CHEST 2010; 137(4):790–796 Abbreviations: FEF25%-75% 5 forced expiratory flow at 25% to 75% of forced vital capacity; FeNO 5 fractional exhaled nitric oxide; GC 5 gas chromatography; MS 5 mass spectrometry; VOC 5 volatile organic compound

volatile organic compounds (VOCs) have Several been identified in exhaled breath in healthy sub-

jects by gas chromatography (GC)/mass spectrometry (MS).1,2 Identification of selective VOC patterns in exhaled breath is potentially useful as a biomarker of asthma.3 Differences between alveolar and oropharyngeal/airway air can affect the results and should be considered when analyzing VOCs in exhaled breath.4 An electronic nose is an artificial sensor system that consists of an array of chemical sensors for VOC detection and an algorithm for pattern recognition.5-7 The electronic nose discriminates between patients with

asthma and healthy subjects,3 between patients with asthma of different severity,3 between patients with lung cancer and healthy subjects,8-10 and between patients with lung cancer and COPD.9 The primary aim of this study was to compare the diagnostic performance of a prototype of an electronic nose with fractional exhaled nitric oxide (FeNO), an independent method for assessing airway inflammation, and lung function tests in patients with asthma. Secondary aims of this study were to ascertain whether an electronic nose could discriminate between patients with asthma and healthy subjects and to establish the

790

Original Research

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

best sampling protocol (alveolar vs oropharyngeal/ airway air) for electronic nose analysis. Materials and Methods Study Subjects Twenty-seven white patients with intermittent or mild persistent asthma and 24 healthy subjects were studied (Table 1). Among study subjects, seven patients with asthma and seven healthy subjects participated in a study with GC/MS used for MS fingerprinting, an independent technique for assessing betweengroup discrimination. Patients with asthma were recruited from the Allergy Outpatient Clinic, Istituto Dermopatico dell’Immacolata, IDI, Rome, Italy. Diagnosis and classification of asthma were based on clinical history and examination and pulmonary function parameters according to current guidelines.11 Patients had intermittent asthma with symptoms equal to or less often than twice a week (step 1) or mild persistent asthma with symptoms more often than twice a week (step 2), FEV1 of ⱖ 80% of predicted value and reversibility of ⱖ 12% to salbutamol, or a positive provocation test result with methacholine or exercise. Patients with asthma had atopy as confirmed by positive skin prick test results in response to common aeroallergens. All atopic subjects had a clinical history of atopy. They were not taking any regular medication, but used inhaled short-acting b2-agonists as needed for symptom relief. Healthy subjects had no history of asthma or atopic disease, negative skin prick test results, and normal spirometry. Subjects had no history of smoking, no upper respiratory tract infections in the previous 3 weeks, and were excluded from the study if they had used corticosteroids or antiinflammatory drugs for asthma in the previous 4 weeks. Study Design The type of study was cross-sectional. Subjects, recruited from March 2007 to September 2008, attended on one occasion for clinical examination, FeNO measurement, electronic nose analysis, lung function tests, and skin prick testing. Informed consent was obtained from patients. The study was approved by the Ethics Committee of the Catholic University of the Sacred Heart, Rome, Italy. Manuscript received August 3, 2009; revision accepted December 7, 2009. Affiliations: From the Department of Pharmacology (Dr Montuschi), Faculty of Medicine, Catholic University of the Sacred Heart; the Department of Electronic Engineering (Drs Santonico, Martinelli, Paolesse, Di Natale, and D’Amico; and Mss Mantini and Capuano), University of Tor Vergata; Faculty of Engineering (Dr Pennazza), University Campus Bio-Medico; and the Department of Immunodermatology (Dr Mondino), Istituto Dermopatico dell’Immacolata, IDI, Rome, Italy; the Department of Drug Sciences (Dr Ciabattoni), Faculty of Pharmacy, University “G. d’Annunzio,” Chieti, Italy; and the Airway Disease Section (Dr Barnes), Imperial College, School of Medicine, National Heart and Lung Institute, London, England. Funding/Support: Supported by Merck, Sharp, and Dohme, and Catholic University of the Sacred Heart academic grant 2008-2009. Correspondence to: Paolo Montuschi, MD, Department of Pharmacology, Faculty of Medicine, Catholic University of the Sacred Heart, Largo F. Vito, 1, 00168 Rome, Italy; e-mail: [email protected] © 2010 American College of Chest Physicians. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (www.chestpubs.org/ site/misc/reprints.xhtml). DOI: 10.1378/chest.09-1836 www.chestpubs.org

Table 1—Subject Characteristics Characteristic

Healthy Subjects

No. Age, y Sex, male (female) FEV1, L FEV1, % predicted FVC, L FVC, % predicted FEF25%-75%, L FEF25%-75%, % Atopyc Smoking history

24 33 6 3 12 (12) 3.9 6 0.2 108.2 6 2.2 4.7 6 1.3 109.2 6 2.2 4.2 6 1.2 98.1 6 5.0 No No

Patients With Asthma 27 39 6 3 12 (15) 3.4 6 0.2a 98.1 6 2.5b 4.5 6 1.3 102.6 6 4.6 3.1 6 1.3b 75.3 6 4.9b Yes No

Data are expressed as No. or mean 6 SEM. FEF25%-75% 5 forced expiratory flow at 25% to 75% of forced vital capacity. aP , .05 compared with healthy subjects. bP , .01 compared with healthy subjects. cPresence of atopy was confi rmed by skin prick testing for common aeroallergens. Patients with asthma had no corticosteroids or other antiinflammatory drugs for asthma within 4 wk.

Pulmonary Function and FENO Measurement Spirometry was performed with a Pony FX spirometer (Cosmed; Rome, Italy). The best of three consecutive maneuvers were chosen. FeNO was measured with the NIOX system (Aerocrine; Stockholm, Sweden) with a single-breath online method at constant flow of 50 mL/s according to American Thoracic Society guidelines.12 FeNO measurements were obtained before spirometry. Collection of Exhaled Breath Exhaled breath was collected from each subject at 8.30 am. No food or drinks were allowed at least 2 h prior to collection of exhaled breath. Two procedures for collecting exhaled breath were followed to study the differences between total exhaled breath and alveolar air and to establish the best protocol for the electronic nose analysis. In the first sampling procedure (Fig 1A), subjects were asked to inhale to total lung capacity and to exhale into a mouthpiece connected to a Tedlar bag through a three-way valve. The total exhaled breath was collected. In the second sampling procedure (Fig 1B), subjects were asked to repeat the maneuver. The sampling system was designed considering a dead space volume of 150 mL. The first 150 mL were collected into a separate Tedlar bag and discarded, whereas the remaining exhaled breath, principally derived from the alveolar compartment, was analyzed. Electronic Nose A prototype electronic nose designed at the University of Rome Tor Vergata was used.13,14 The instrument contains an array of eight quartz microbalance gas sensors coated by molecular films of metalloporphyrins.15 Sensors detect the amount of molecules absorbed in a sensitive film through the changes of resonant frequency that is proportional to the absorbed mass.15 The frequency shifts are composed in patterns and analyzed by pattern recognition algorithms.16 Samples were analyzed immediately after their collection. Ambient VOCs were subtracted from measures and results were automatically adjusted for ambient VOCs. GC/MS GC/MS used for MS fingerprinting of VOCs17 was performed (1) to confirm between-group differences in VOC patterns CHEST / 137 / 4 / APRIL, 2010

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

791

multidimensional data in a bidimensional plane and feed-forward neural network to classify electronic nose, FeNO, and spirometry data. A feed-forward neural network is a biologically derived classification model21 that is formed by a number of processing units (neurons), organized in layers. The total datasets have been divided in training (27 measures) and testing (24 measures) set. To test the presence of any drift in sensors, the first data collected were used for training and the remaining data for testing. Statistical Analysis

Figure 1. Exhaled breath sampling for electronic nose analysis. (A) Sampling of total exhaled breath. (B) Sampling of alveolar exhaled air. The sampling system was designed considering a dead space volume of 150 mL. The first part of exhalation (150 mL) was discarded, whereas the remaining exhaled breath principally derived from the alveolar compartment was collected for electronic nose analysis. detected with an electronic nose, and (2) to ascertain whether exhaled breath samples were stable within 48 h from collection. With MS fingerprinting, the sum of mass spectra of VOCs detected by GC was considered as a sample pattern that was then analyzed like the electronic nose patterns.17 Total exhaled breath was collected (Fig 1A). Samples were immediately absorbed onto sterile 10 3 10 cm2 gauze pads that were sealed in 20-mL headspace glass vials with crimped seal with polytetrafluoroethylene/silicone septa (Supelco; Bellefonte, PA).15 Samples were put on ice, transferred to the laboratory, and stored at 210°C until GC/MS analysis that was performed within 48 h from sample collection. Before solid-phase microextraction, samples were kept at room temperature for 9 h. VOCs were extracted from the vials using a divinylbenzene/carboxen on a polydimethylsiloxane 50/30 m fiber (Supelco; Sigma-Aldrich; St. Louis, MO). Exposure was performed at room temperature for 15 h by piercing the silicon septum and inserting the fiber into the headspace of the vial. Solid-phase microextraction fiber was then removed from the vial and transferred to the injector of the gas chromatograph for thermal desorption. VOCs adsorbed in the fiber were desorbed in the injection port of the GC for 3 min at an inlet temperature of 250°C in the splitless mode. Mass spectra were obtained using electron ionization. The ionization occurred with a kinetic energy of the impacting electrons of 70 eV.18 Mass spectra and reconstructed chromatograms (total ion current) were acquired in the full scan mode in the mass range m/z 50-100.15 Skin Testing

FeNO values were expressed as medians and interquartile ranges (25th and 75th percentiles). Spirometry values were expressed as mean 6 SEM. Unpaired t test and Mann-Whitney U test were used for comparing groups for normally distributed and nonparametric data, respectively. Correlation was expressed as a Pearson coefficient. Significance was defined as a value of P , .05.

Results Electronic Nose The best results were obtained when electronic nose analysis was performed on alveolar air (Tables 2, 3). Diagnostic classification with 95% CIs is shown in Table 4. The diagnostic performance was determined with the test datasets in terms of the number of correct identifications of asthma diagnosis based on current guidelines.11 Diagnostic performance for the electronic nose, FeNO, lung function tests, and their combinations is shown in Table 2 and Table 3 and is related to the best performances obtained with the neural network for each specific case. The combination of electronic nose analysis of alveolar air and FeNO had the highest diagnostic performance for asthma (95.8%). Electronic nose analysis of alveolar air was able to discriminate between patients with asthma and healthy subjects in 87.5% of cases, a diagnostic performance that was higher than that of FeNO (79.2%), spirometry (70.8%), and combination of FeNO and spirometry (83.3%). There was no correlation between the electronic nose, FeNO, and lung function testing data in either the asthma or the healthy control group. GC/MC

Atopy was assessed by skin prick tests for common aeroallergens (mixture for house dust mite [Dermatophagoides pteronyssinus and Dermatophagoides farinae], grass pollen [cocksfoot and timothy], tree pollen [birch, ash tree, olive tree, oak, and cypress], weed pollen [Ambrosia artemisifolia and Parietaria officinalis], animal danders [cat and dog allergens], and fungal allergens [Aspergillus species and Alternaria alternata]) (Stallergenes; Antony, France).19 A positive skin test response was defined as a wheal with a mean diameter (mean of maximum and 908 midpoint diameters) of at least 3 mm greater than that produced with a saline control.

MS fingerprints of VOCs in exhaled breath in patients with asthma were different from those in healthy subjects (Fig 2), confirming that the electronic nose used in this study discriminates between patients with asthma and controls. MS fingerprinting performed within 48 h from sampling was similar to that obtained immediately after collection (data not shown), indicating that samples are stable for at least 48 h.

Multivariate Data Analysis

FENO and Lung Function Tests

The analysis of patterns requires multivariate statistical algorithms.20 We used principal component analysis to represent

Patients with asthma had higher median FeNO values than healthy subjects (37. 6 [26.0-61.5] ppb vs

792

Original Research

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

Table 2—Classification Matrices of the Feed-Forward Neural Network Classifier in Training and Testing Phase for Data Related to Electronic Nose, Fractional Exhaled Nitric Oxide, and Spirometry Techniques

Healthy Subjects Patients With Asthma

E-nose (alveolar exhaled air) Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma E-nose (total exhaled air) Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma FeNO Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma Spirometry Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma

11 1a

1b 14

9 0

3b 12

12 0

0 15

11 5a

12 0

1b 7

0 15

10 3a

2b 9

11 0

1b 15

9 4a

3b 8

E-nose analysis was performed in total exhaled air and alveolar exhaled air. E-nose 5 electric nose; FeNO 5 fractional exhaled nitric oxide. aFalse negatives. bFalse positives.

13.4 [10.0-19.9] ppb, P , .0001, respectively) (Fig 3). Both study groups had normal FEV1 values (Table 1). Patients with asthma had lower absolute (P 5 .032) and percentage of predicted FEV1 values (P 5 .004) than controls (Table 1). Absolute (P 5 .003) and percentage of predicted forced expiratory flow between 25% and 75% of forced vital capacity (FEF25%-75%) (P 5 .002) were lower in patients with asthma than in healthy subjects (Table 1). Discussion The original aspects of our study are: (1) the comparison between an electronic nose and FeNO, in addition to lung function tests; (2) the comparison between total and alveolar exhaled air; (3) the number of study subjects (27 patients with intermittent and persistent mild asthma and 24 healthy controls); (4) the MS fingerprinting based on GC/MS analysis; and (5) the analysis of data based on a neural network that included a training and test analysis performed in two separate datasets for stringent quality control. www.chestpubs.org

Table 3—The Classification Matrices of the FeedForward Neural Network Classifier in Training and Testing Phase for Data Related to Combination of Electronic Nose, FENO, and Spirometry Techniques

Healthy Subjects Patients With Asthma

FeNO and e-nose (alveolar exhaled air) Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma FeNO and e-nose (total exhaled air) Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma Spirometry and e-nose (alveolar exhaled air) Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma Spirometry and e-nose (total exhaled air) Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma FeNO and spirometry Training Healthy subjects Patients with asthma Test Healthy subjects Patients with asthma

12 0

0 15

11 0

1b 12

12 0

0 15

12 3a

0 9

12 0

0 15

9 1a

12 0 9 1a

3b 11

0 15 3b 11

12 0

0 15

10 2a

2b 10

Electronic nose analysis was performed in total exhaled air and alveolar exhaled air. See Table 2 for expansion of abbreviations. aFalse negatives. bFalse positives.

We compared the diagnostic performance of an electronic nose, FeNO, and lung function testing in patients with a physician-based diagnosis of asthma. The combination of electronic nose and FeNO had the best diagnostic performance for asthma (95.7%). When alveolar exhaled air was sampled, the electronic nose had a diagnostic performance of 87.5%, higher than that observed with FeNO (79.2%), spirometry (70.8%), and their combination (83.3%). Spirometry had the lowest diagnostic performance in line with a previous prospective study that compared FeNO, sputum eosinophils, and FEV1 in children with asthma.22 The present study confirms that FeNO CHEST / 137 / 4 / APRIL, 2010

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

793

Table 4—Diagnostic Classification With 95% of CIs in Training and Testing Phase for Data Related to Combination of Electronic Nose, FENO, and Spirometry Classification Rate, % Testing

Technique E-nose (alveolar exhaled air) E-nose (total exhaled air) FeNO Spirometry FeNO and e-nose (alveolar exhaled air) FeNO and e-nose (total exhaled air) Spirometry and e-nose (alveolar exhaled air) FeNO and spirometry Spirometry and e-nose (total exhaled air)

92 96 98.5 96 98.7 99.1 99.3 98.5 99.1

86.2 72.5 77.1 70.5 92.7 86.6 81.8 81.3 81.1

Electronic nose analysis was performed in total exhaled air and alveolar exhaled air. See Table 2 for expansion of abbreviations.

has a good diagnostic performance for asthma. The higher diagnostic performance of electronic nose observed in our study might reflect the fact that electronic nose analyzes patterns generated by a complex mixture of VOCs in exhaled breath, whereas the NO analyzer detects only one biomolecule. A limitation of our study is the relatively small number of subjects, which precludes definitive conclusions on asthma diagnostic performance of electronic nose, FeNO and lung function testing. Large-powered studies are required to definitively establish the diagnostic performance of these techniques in patients with a known diagnosis of asthma. We were unable to ascertain whether the electronic nose can be used as

Figure 2. PC analysis of mass spectrometry fingerprinting of patients with asthma and healthy subjects. PC 5 principal component.

Figure 3. FeNO concentrations in 27 patients with asthma (䊏) and 24 healthy subjects (䊉). Median values are shown with bars. FeNO 5 fractional exhaled nitric oxide.

a diagnostic tool for screening of patients with asthma that requires large prospective studies. MS fingerprinting confirms data obtained with the electronic nose and further supports its use for asthma diagnosis. Identification and quantitative GC/MS analysis of VOCs in exhaled breath are required to establish the pathophysiological role of single VOCs in asthma. Taken together, these data indicate that electronic nose might be useful for asthma diagnosis, particularly in combination with FeNO. Our results are consistent with a previous study showing that a different electronic nose discriminates between patients with asthma and healthy subjects3 and prospect the possibility of using different noninvasive techniques for achieving a greater asthma diagnostic performance. Because asthma is principally characterized by airway inflammation, it may seem surprising that the best results with the electronic nose were obtained when collecting alveolar air rather than total exhaled breath that includes exhaled breath from the airways. This might reflect the contribution of oropharyngeal air to total breath.4 A previous study with GC/MS has shown that for 47 VOCs very significant differences in concentrations/detection were recorded between alveolar and oropharyngeal air.4 When sampling total breath, oropharyngeal air is likely to introduce confounding factors that make electronic nose analysis less reflective of what occurs in the respiratory system. Because of significant interindividual variability in dead space volume, it is likely that mixed airways/alveolar air rather than alveolar air was collected in our study. For this reason, the results of electronic nose analysis of alveolar air could partially reflect the production of VOCs within the peripheral airways.

794

Original Research

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

The lack of correlation between electronic nose results and FeNO might indicate that these techniques reflect different aspects of airway inflammation. Formal studies to compare electronic nose analysis with independent biomarkers of airway inflammation (eg, FeNO, sputum eosinophils)12,23,24 and to ascertain whether the electronic nose could be used for assessing and monitoring airway inflammation in patients with asthma are warranted. The electronic nose is not suitable for ascertaining the cellular source of VOCs in exhaled breath that requires bronchoscopy studies. Chronic airway inflammation can modify the metabolic pathways in patients with asthma as reflected by increased concentrations of exhaled pentane, one of the end products of lipid peroxidation.25 Because patients included in our study were not on regular antiinflammatory drugs for asthma, we were unable to assess the effect of pharmacological therapy on VOC patterns in exhaled breath for which controlled studies are required. Both the present and a previous study3 aimed at ascertaining whether electronic nose could discriminate between atopic patients with asthma and healthy nonatopic subjects. For this reason, these studies were unable to assess the contribution of atopy itself to the classification based on the electronic nose. A control group of atopic subjects without asthma should be included in future studies to clarify whether, and to what extent, atopy is responsible for the selective VOC patterns in exhaled breath observed in patients with atopic asthma. When analyzing electronic nose data, validation of the statistical model used for classification is essential. In contrast to a previous study in which the same dataset was used for training the model and classification,3 we used two different datasets for training and testing that were obtained in different periods of time, reducing the possibility of introducing a temporal bias in the model. This way, the predictive capacity of the classification model is likely to be more suitable for a real-life situation. In conclusion, the electronic nose discriminates between patients with asthma and healthy subjects and its performance is increased when combined with FeNO. Large studies are required to definitively establish the diagnostic performance of the electronic nose. Whether this integrated noninvasive approach will translate into an early diagnosis of asthma has to be clarified. Acknowledgments Author contributions: Dr Montuschi: contributed to study planning, study design, measurement of FeNO, spirometry, data analysis, data interpretation, and wrote the manuscript. Dr Santonico: contributed to electronic nose analysis, data interpretation, and mass spectrometry fingerprinting. Dr Mondino: contributed to recruitment of patients and skin prick testing. www.chestpubs.org

Dr Pennazza: contributed to electronic nose analysis, data interpretation, and mass spectrometry fingerprinting. Ms Mantini: contributed to electronic nose analysis and mass spectrometry fingerprinting. Dr Martinelli: contributed to multivariate data analysis and manuscript preparation. Ms Capuano: contributed to electronic nose analysis and mass spectrometry fingerprinting. Dr Ciabattoni: contributed to manuscript preparation and revision. Dr Paolesse: contributed to biosensor manufacture. Dr Di Natale: contributed to electronic nose setup, data interpretation, and manuscript preparation. Dr Barnes: contributed to manuscript preparation and revision. Dr D’Amico: contributed to electronic nose setup, data interpretation, and manuscript preparation and revision. Financial/nonfinancial disclosures: The authors have reported to CHEST the following conflicts of interest: Dr Barnes receives research funding from and has been on Scientific Advisory Boards for AstraZeneca, Boehringer-Ingelheim, Chiesi, GlaxoSmithKline, Novartis, Pfizer, and UCB. Drs Montuschi, Santonico, Mondino, Pennazza, Martinelli, Ciabattoni, Paolesse, Di Natale, D’Amico, and Mss Mantini and Capuano have reported to CHEST that no potential conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article. Other contributions: This work was performed at the Catholic University of the Sacred Heart, Rome, Italy.

References 1. Phillips M. Method for the collection and assay of volatile organic compounds in breath. Anal Biochem. 1997;247(2): 272-278. 2. Phillips M, Herrera J, Krishnan S, Zain M, Greenberg J, Cataneo RN. Variation in volatile organic compounds in the breath of normal humans. J Chromatogr B Biomed Sci Appl. 1999;729(1-2):75-88. 3. Dragonieri S, Schot R, Mertens BJ, et al. An electronic nose in the discrimination of patients with asthma and controls. J Allergy Clin Immunol. 2007;120(4):856-862. 4. van den Velde S, Quirynen M, van Hee P, van Steenberghe D. Differences between alveolar air and mouth air. Anal Chem. 2007;79(9):3425-3429. 5. Turner AP, Magan N. Electronic noses and disease diagnostics. Nat Rev Microbiol. 2004;2(2):161-166. 6. Röck F, Barsan N, Weimar U. Electronic nose: current status and future trends. Chem Rev. 2008;108(2):705-725. 7. D’Amico A, Di Natale C, Paolesse R, et al. Olfactory systems for medical applications. Sens Actuators B Chem. 2008;130(1):458-465. 8. Machado R, Laskowski D, Deffenderfer O, et al. Detection of lung cancer by sensor array analyses of exhaled breath. Am J Respir Crit Care Med. 2005;171(11):1286-1291. 9. Dragonieri S, Annema JT, Schot R, et al. An electronic nose in the discrimination of patients with non-small cell lung cancer and COPD. Lung Cancer. 2009;64(2):166-170. 10. Di Natale C, Macagnano A, Martinelli E, et al. Lung cancer identification by the analysis of breath by means of an array of non-selective gas sensors. Biosens Bioelectron. 2003; 18(10):1209-1218. 11. National Asthma Education and Prevention Program. Expert Panel Report III. Guidelines for the Diagnosis and Management of Asthma. NIH publication 08-5847. Bethesda, MD: National Heart, Lung, and Blood Institute; 2007. 12. Recommendations for standardized procedures for the on-line and off-line measurement of exhaled lower respiratory nitric oxide and nasal nitric oxide in adults and children-1999. Am J Respir Crit Care Med. 1999;160(6):2104-2117. CHEST / 137 / 4 / APRIL, 2010

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

795

13. Di Natale C, Paolesse R, D’Amico A. Metalloporphyrins based artificial olfactory receptors. Sens Actuators B Chem. 2007;121(1):238-246. 14. D’Amico A, Di Natale C, Paolesse R, Mantini A, Macagnano A. Metalloporphyrins as basic material for volatile sensitive sensors. Sens Actuators B Chem. 2000;65(1-3):209-215. 15. D’Amico A, Bono R, Pennazza G, et al. Identification of melanoma with a gas sensor array. Skin Res Technol. 2008; 14(2):226-236. 16. Gardner J, Bartlett P. A brief history of electronic noses. Sens Actuators B Chem. 1994;18(1-3):211-220. 17. Pavòn JL, Sanchez M, Pinto C, Laespada E, Cordero M, Peña A. Strategies for qualitative and quantitative analyses with mass spectrometry-based electronic noses. Trends Analyt Chem. 2006;25(3):257-266. 18. Kusch P, Knupp G. Headspace-SPME-GC-MS Identification of volatile organic compounds released from expanded polystyrene. Journal of Polymers and the Environment. 2004;12(2):83-87. 19. Montuschi P, Mondino C, Koch P, Barnes PJ, Ciabattoni G. Effects of a leukotriene receptor antagonist on exhaled leu-

20. 21. 22. 23.

24.

25.

kotriene E4 and prostanoids in children with asthma. J Allergy Clin Immunol. 2006;118(2):347-353. Johnson R, Wichern D. Applied Multivariate Statistical Analysis. 3rd ed. New York, NY: Prentice Hall; 1992. Bishop CM. Pattern Recognition and Machine Learning. Heidelberg, Germany: Springer; 2006. Sivan Y, Gadish T, Fireman E, Soferman R. The use of exhaled nitric oxide in the diagnosis of asthma in school children. J Pediatr. 2009;155(2):211-216. Malerba M, Ragnoli B, Radaeli A, Tantucci C. Usefulness of exhaled nitric oxide and sputum eosinophils in the longterm control of eosinophilic asthma. Chest. 2008;134(4): 733-739. Petsky HL, Kynaston JA, Turner C, et al. Tailored interventions based on sputum eosinophils versus clinical symptoms for asthma in children and adults. Cochrane Database Syst Rev. 2007; (2):CD005603. Paredi P, Kharitonov SA, Barnes PJ. Elevation of exhaled ethane concentration in asthma. Am J Respir Crit Care Med. 2000;162(4 Pt 1):1450-1454.

796

Original Research

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

DIAGNOSTIC PERFORMANCE OF ELECTRONIC NOSE, FRACTIONAL EXHALED NITRIC OXIDE AND LUNG FUNCTION TESTING IN ASTHMA 1

Paolo

Montuschi,

M.D.,

[email protected],

2

2

[email protected],

Giorgio

Pennazza,

Ph.D.,

Marco

Santonico,

[email protected],

Ph.D., 3

Chiara

Mondino, M.D., [email protected], 2Giulia Mantini, B.Eng., [email protected], 2Eugenio Martinelli, Ph.D., [email protected], 2Rosamaria Capuano, B.Eng., [email protected], 4

Giovanni

Ciabattoni,

M.D.,

[email protected],

2

Corrado

Di

Natale,

Ph.D.,

[email protected], 5Peter J. Barnes, DM, [email protected], 2Arnaldo D’Amico, Ph.D, [email protected]. 1

Department of Pharmacology, Faculty of Medicine, Catholic University of the Sacred Heart,

Rome, Italy, 2Department of Electronic Engineering, University of Tor Vergata, Rome, Italy, 3

Department of Immunodermatology, Istituto Dermopatico dell’Immacolata, IDI, Rome, Italy,

4

Department of Drug Sciences, Faculty of Pharmacy, University “G. d’Annunzio”, Chieti, Italy,

5

Airway Disease Section, Imperial College, School of Medicine, National Heart and Lung Institute,

London, United Kingdom.

This work was performed at the Catholic University of the Sacred Heart, Rome, Italy. This study was supported by Merck, Sharp, and Dohme and Catholic University of the Sacred Heart Author for correspondence: Paolo Montuschi, MD Department of Pharmacology, Faculty of Medicine Catholic University of the Sacred Heart, Largo F. Vito, 1, 00168 Rome, Italy Tel: 39-06-30156092; fax: 39-06-30156292; e-mail: [email protected]

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

INTRODUCTION Currently diagnosis, treatment and management of asthma are principally based on clinical and physiological assessment, although chronic airway inflammation is the pathophysiological hallmark of this disease. Identification of non-invasive techniques for assessing airway inflammation is a priority in asthma research. The assessment of airway inflammation is an important goal in asthma management as it can provide new insights into the pathophysiology of asthma, indicate that pharmacological treatment is required before the onset of symptoms and/or reduction in lung function, be useful in the follow-up of patients with asthma and for guiding drug intervention.1 A better knowledge of the airway inflammatory process in patients with asthma might also help identify asthma phenotypes. Non-invasive assessment of airway inflammation in patients with asthma is currently limited to the measurement of fractional exhaled nitric oxide (FENO), that is a surrogate marker of airway inflammation.2 Assessment of eosinophil counts in sputum is a direct measure of airway inflammation and has been successfully applied to management of patients with asthma.3,4 However, sputum induction is a semi-invasive technique, generally less acceptable to patients, and can be difficult to perform in subgroups of patients (e.g., children with asthma or patients with severe asthma). The electronic nose discriminates between asthmatic patients and healthy subjects,3 between patients with asthma of different severity,3 between patients with lung cancer and healthy subjects,810

and between patients with lung cancer and COPD.9 Two different electronic noses were used in

these studies.

METHODS Study subjects

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

The diagnosis and classification of asthma was based on clinical history and examination and pulmonary function parameters according to the Guidelines issued by the National Heart, Lung, and Blood Institute of the National Institutes of Health.5 Among study subjects, 7 asthmatic patients (3/4, males/females, age 32 ± 5.4 years, p = 0.97; FEV1 3.0 ± 0.2 L, p = 0.11; FEV1 96.1 ± 4.1% predicted value, p = 0.08; FVC 4.2 ± 0.3 L, p = 0.62; FVC 109.7± 2.7% predicted value, p = 0.70; FEF25%-75% 2.6 ± 0.2 L, p = 0.02; FEF25%-75% 65.0 ± 3.8% predicted value, p = 0.003) and 7 healthy subjects (3/4, males/females, age 32 ± 5.9 years, FEV1 3.6 ± 0.2 L, FEV1 106.9 ± 3.7% predicted value, FVC 4.4 ± 0.3 L, FVC 107.7 ± 4.3% predicted value, FEF25%-75% 3.7 ± 0.3 L, FEF25%-75% 92.4 ± 6.4% predicted value) participated in a study with GC/MS used for MS fingerprinting as an independent technique for assessing between-group discrimination. Healthy subjects were recruited from staff.

Pulmonary function FEV1, forced vital capacity (FVC), FEV1/FVC, forced expiratory flow between 25% and 75% of FVC (FEF25%-75%), and peak expiratory flow (PEF) were measured with a Pony FX spirometer (Cosmed, Rome, Italy) and the best of three consecutive maneuvers chosen.

Exhaled nitric oxide measurement Exhalations were repeated after 1-minute relaxation period until the performance of three FENO values varies less than 10%. FENO measurements were obtained before spirometry.

Collection of exhaled breath

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

Therefore, sampling of exhaled air was performed either by collecting total exhaled air (oropharynx, airways and alveoli) or by discarding the first 150 ml of exhaled air and collecting exhaled air principally from the alveolar compartment. Sensors are housed in a stainless steel cell, and air sample is carried into the sensor cell by an internal pump. As a reference, laboratory air filtered by a CaCl2 humidity trap was used during the experiment. The use laboratory air facilitates detection of VOCs in exhaled air. A set of valves regulates sampling of exhaled air or reference air. The result of a single measurement is the resonant frequency shifts of the QMB sensors correspondent to the exposure to the reference air and to exhaled breath sample. The frequency shifts are composed in patterns and analyzed by pattern recognition algorithms.

Collection of exhaled breath Sampling systems are shown in Figure 1. Before sampling, subjects were asked to rinse their mouths with water. The first part of exhalation (150 ml) was collected into a separate Tedlar bag and discarded, whereas the remaining exhaled air principally derived from the alveolar compartment was collected in another Tedlar bag for immediate electronic nose analysis.

Electronic nose The instrument contains an array of eight quartz microbalance (QMB) gas sensors whose chemical sensitivity is provided by molecular films of metallo-porphyrins coating the QMB surfaces.6 Metalloporphyrins coated QMBs are particularly suitable for electronic nose applications to the detection of VOC patterns in human samples (exhaled breath, urine, skin). The result of a single measurement is the resonant frequency shifts of the QMB sensors corresponding to the exposure to an exhaled breath sample Repeatability
 of
 the
 electronic
 nose
 measurements
 was
 assessed
 in
 10
 healthy
 subjects.
 One
 electronic
nose
measurement
was
performed
every
day
at
the
same
time
for
6
consecutive
days.


Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

For
each
sensors
the
standard
deviation
of
the
6
measurements
performed
on
the
same
patient
 with
the
same
protocol
is
shown
in
table
S1.

Gas chromatography/mass spectrometry analysis The GC method was initiated with an initial oven temperature of 40 °C for 5 min. The temperature was then ramped at 10 °C/min until it reached 300 °C, and then was held at 300 °C for 2 min (total run time: 33 min). Helium 6.0 grade was used as a carrier gas (Westfalen, AG, Muensler, Germany).18 A constant pressure of 70 kPa helium was used. The detector voltage was 350 V. The interface and ion source temperatures were maintained at 250 °C. The solvent cut time was 3 min.

Multivariate data analysis PCA is a simple algorithm based on the fact that each pattern is represented as a vector in an Euclidean space where each variable defines a coordinate. Each sensor signal is considered as a variable and the multivariate data have a dimension equal to the number of sensors in the array. The main aim of PCA is to calculate a novel basis of coordinates (the principal components) where the correlations between variables disappear. The representation of the data can be limited only to those coordinates carrying the largest variance. In many cases, the data set can be conveniently represented even in a bidimensional space mantaining most of the statistical properties of the data. PCA data plots are interpreted assuming that the distance between data in the principal components plane does provide a measure of the similitude among patterns, and clusters of data identify similar samples. Before the application of pattern recognition algorithms, sensors patterns have been scaled to zero-mean and unitary variance. This normalization avoids any bias introduced by the fact that sensor signals can occur in different numerical ranges. Each neuron is characterized by a non-linear transfer function (activation function) whose argument is a weighted sum of its inputs. Sensors data are processed layer by layer until to the output. The weights of the all connections are adjusted during a training with a proper learning algorithm.

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

During the training, taking advantage of the contemporary knowledge of the measures (input) and class membership (output), the learning algorithm modifies the connection weights of all neuron in order to match as much as possible the network prediction with the real membership. Such a neural network can be adapted to reproduce any input-output relationship. This property may result in classifiers that can perfectly estimate the membership of training data, but fails to associate any test pattern to its correct class. This problem, known as overfit, can be avoided by training and testing the network with independent data. This operation requires a large data set. A feed forward neural network, based on a single hidden layer of neurons with a sigmoid activation function was adopted and the network was trained with a backpropagation learning algorithm.

Sample size As the biosensors detect patterns of VOCs in each subject, planning a classical sample size calculation that is based on the identification of a quantitative primary outcome and the minimal difference of biological significance is not feasible. Using principal component analysis followed by linear canonical discriminant analysis, a previous study with a different e-nose has reported between group discrimination in 20 patients with asthma and 20 healthy subjects.8 Twenty-seven patients with asthma and 24 healthy subjects were included in this study.

RESULTS Electronic nose Fifty-one electronic nose measures from 27 patients with asthma and 24 healthy subjects were included. The neural network was trained with a set of 27 patterns (12 healthy subjects and 15 patients with asthma). Once trained, the network capability to correctly identify asthma was tested on the remaining 24 patterns. Training and test data were taken at different times. Two distinct

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

neural network classifiers were trained and tested with patterns related to total exhaled air and alveolar air.

The patterns resulting from electronic nose analysis were used to build a classification model aimed at discriminating between patients with asthma and healthy subjects. Data analysis was performed using a non-linear classifier based on feed forward neural network.

In the test analysis of electronic nose measures, 3 out of 12 healthy subjects were false positives, whereas all patients with asthma were correctly classified. When a combination of electronic nose analysis and FENO was used, 1 out of 12 healthy subjects was a false positive, whereas all patients with asthma were recognized as such.

Gas chromatography/mass chromatography analysis GC/MS was feasible in all samples. MS fingerprinting performed within 48 h from sampling was similar to that obtained when exhaled air samples were analyzed immediately after collection (data not shown) indicating that in our experimental conditions samples are stable for at least 48 hours.

DISCUSSION Basal spirometry is an essential objective measure to assess the severity of asthma, but is of limited diagnostic utility in patients with intermittent and mild persistent asthma who have normal FEV1 values by definition.5 In view of this fact, in the multivariate data analysis, we included both FEV1

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

and FEF25%-75% values that reflect peripheral airway function and were actually reduced in the asthmatic group. Although FENO has not been evaluated prospectively as an aid in the diagnosis of asthma in adults,9 a recent prospective study has shown that FENO is useful in early diagnosis of pediatric asthma.10 To collect exhaled air from the alveolar compartment, the first 150 ml of exhalation, corresponding to dead space volume, were discarded. However, due to significant inter-individual variability in dead space volume, it is likely that mixed airways/alveolar air rather than alveolar air was collected in our study. For this reason, the results of electronic nose analysis of alveolar air should be interpreted cautiously as they could partially reflect the production of VOCs within the peripheral airways. A previous study has shown that electronic nose discriminates between patients with severe asthma who were being treated with beclomethasone or equivalent at a daily dose equal or higher than 1000 µg and healthy subjects.8 These findings indicate that the diagnostic value of electronic nose analysis is maintained in patients on high doses of inhaled corticosteroids. However, controlled studies to establish the effect of pharmacological therapy on electronic nose analysis in patients with asthma are warranted.

Electronic nose analysis is a non-invasive technique that is potentially applicable to respiratory medicine. Several methodological issues including optimization and standardization of sample collection, transfer and storage of samples, use of calibration VOC mixtures, and quantitative GC/MS analysis, need to be addressed.

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

References 1. Montuschi P. Indirect monitoring of lung inflammation. Nat Rev Drug Discov 2002;1:238242. 2. Recommendations for standardized procedures for the on-line and off-line measurement of exhaled lower respiratory nitric oxide and nasal nitric oxide in adults and children-1999: official statement of the American Thoracic Society 1999. Am J Respir Crit Care Med 1999;160:2104-17. 3. Malerba M, Ragnoli B, Radaeli A, Tantucci C. Usefulness of exhaled nitric oxide and sputum eosinophils in the long-term control of eosinophilic asthma. Chest 2008;134:733739. 4. Petsky HL, Kynaston JA, Turner C, Li AM, Cates CJ, Lasserson TJ, Chang AB. Tailored interventions based on sputum eosinophils versus clinical symptoms for asthma in children and adults. Cochrane Database Syst Rev 2007:CD005603. 5. National Asthma Education and Prevention Program: Expert panel report III. Guidelines for the diagnosis and management of asthma. MD, Bethesda: National Heart, Lung, and Blood Institute, 2007; 1-61 (NIH publication no. 08-5847). Available at: www.nhlbi.nih.gov 6. D’Amico A, Bono R, Pennazza G, Santonico M, Mantini A, Bernabei M, Zarlenga M, Roscioni C, Martinella E, Paolesse R, Di Natale C. Identification of melanoma with gas sensor array. Skin Res Techn 2008;14:226-236. 7. Kusch P, Knupp G. Headspace-SPME-GC-MS Identification of Volatile Organic Compounds Released from Expanded Polystyrene. Journal of Polymers and the Environment 2004;12:83-87. 8. Dragonieri S, Schot R, Mertens BJ, Le Cessie S, Gauw SA, Spanevello A, Resta O, Willard NP, Vink TJ, Rabe KF, Bel EH, Sterk PJ. An electronic nose in the discrimination of patients with asthma and controls. Allergy Clin Immunol. 2007;120:856-862.

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

9. Sivan Y, Gadish T, Fireman E, Soferman R. The use of exhaled nitric oxide in the diagnosis of asthma in school children. J Pediatr 2009;155:211-216. 10. GINA. Global Initiative for Asthma. Global strategy for asthma management and prevention. Update 2008. Available at: www.ginasthma.com

Table S1. Repeatability of electronic nose measurements. Sensor 1

Sensor 2

Sensor 3

Sensor 4

Sensor 5

Sensor 6

Sensor 7

Sensor 8

3Hz

4.Hz

3.2Hz

2.1Hz

3.1Hz

3.7Hz

3.1Hz

3.3Hz

Repeatability of electronic nose measurements in 10 healthy subjects. Values are SD of 6 emasurements performed in the same subject.

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians

Diagnostic Performance of an Electronic Nose, Fractional Exhaled Nitric Oxide, and Lung Function Testing in Asthma Paolo Montuschi, Marco Santonico, Chiara Mondino, Giorgio Pennazza, Giulia Mantini, Eugenio Martinelli, Rosamaria Capuano, Giovanni Ciabattoni, Roberto Paolesse, Corrado Di Natale, Peter J. Barnes and Arnaldo D'Amico Chest 2010;137; 790-796; Prepublished online January 15, 2010; DOI 10.1378/chest.09-1836 This information is current as of May 18, 2011 Supplementary Material View e-supplements related to this article at: http://chestjournal.chestpubs.org/content/suppl/2010/04/06/chest.09-1836.DC2.html Updated Information & Services Updated Information and services can be found at: http://chestjournal.chestpubs.org/content/137/4/790.full.html References This article cites 10 articles, 1 of which can be accessed free at: http://chestjournal.chestpubs.org/content/137/4/790.full.html#ref-list-1 Cited Bys This article has been cited by 1 HighWire-hosted articles: http://chestjournal.chestpubs.org/content/137/4/790.full.html#related-urls Permissions & Licensing Information about reproducing this article in parts (figures, tables) or in its entirety can be found online at: http://www.chestpubs.org/site/misc/reprints.xhtml Reprints Information about ordering reprints can be found online: http://www.chestpubs.org/site/misc/reprints.xhtml Citation Alerts Receive free e-mail alerts when new articles cite this article. To sign up, select the "Services" link to the right of the online article. Images in PowerPoint format Figures that appear in CHEST articles can be downloaded for teaching purposes in PowerPoint slide format. See any online figure for directions.

Downloaded from chestjournal.chestpubs.org by guest on May 18, 2011 © 2010 American College of Chest Physicians