Table of Contents  

Marzluf, Krajc, and Mueller: Principles of lung cancer screening – exhaled breath analysis

Exhaled breath analysis

Exhaled breath and body scent have served as indicators for disease since ancient times when physicians had to rely on the diagnostic ability of their senses and conditions such as diabetic ketoacidosis or liver and renal failure were diagnosed by their characteristic smells. Interest in exhaled breath analysis was revived when Pauling et al.1 first identified and measured volatile organic compounds (VOCs) in human breath in the early 1970s. The analysis of possible biomarkers for disease in exhaled breath became technically feasible and, hence, at least theoretically, attractive for screening and diagnostic purposes because of its non-invasive nature and wide applicability, for example in critically ill patients2,3 or during exercise.4 This led to an extensive search for compounds and technologies that might allow for diagnosis of pulmonary diseases, as well as other diseases, from exhaled breath. Meanwhile, more than 3400 VOCs have been identified in human breath,5 but consistent diagnostic biomarkers or simple biomarker profiles for specific diseases, in particular lung cancer, are still lacking.

In general, information from exhaled breath can be extracted by analysis of VOCs and analysis of non-volatile biomarkers in exhaled breath condensate. This review exclusively focuses on the potential of VOC analysis for lung cancer screening purposes.

Volatile organic compounds

VOCs are substances released in very low concentrations (typically parts per billion or parts per trillion) from the fluid phase of tissues into the environment because of their high vapour pressure at room temperature. They can be analysed from the headspace of cancer cell lines in vitro and through skin, urine, faeces, blood and exhaled breath in vivo.6 VOCs comprise various classes of chemical compounds, such as hydrocarbons, alcohols, aldehydes, ketones, esters and nitriles and aromatic compounds.6

With respect to their origins, exhaled breath VOCs are classified as exogenous or endogenous, but little is known about specific sources of individual compounds. Exogenous VOCs are gaseous environmental components that are inhaled or adsorbed through the skin or the gastrointestinal tract. They include toxic and carcinogenic chemicals, such as those found in cigarette smoke or air pollutants, which indicate exposure of individuals to potential triggers of diseases. These exogenous VOCs are mostly lipophilic and stored in the body’s fat compartments, from where they are slowly released and can cause tissue damage leading to subsequent cancer initiation and progression. On the other hand, endogenous VOCs are products of various local or systemic metabolic processes, such as cholesterol synthesis, energy metabolism, and especially oxidative stress.7,8 Hence, changes in these processes (e.g. resulting from cancerogenesis) are likely to qualitatively and quantitatively influence the production of VOCs. Furthermore, the degradation of a large amount of VOCs is cytochrome P45 dependent,9 and modifications of VOC clearance might occur as a consequence of disease- or medication-related interferences with catabolic enzymes. Owing to these and similar mechanisms, VOC patterns may distinctively differ between healthy subjects and lung cancer patients. However, there is a high variation in VOC patterns even between healthy individuals, with only 27 of the approximately 3400 known compounds consistently found in all subjects.5 It is therefore very likely that certain patterns of VOCs rather than single VOC biomarkers will be diagnostic for lung cancer.

Exhaled breath sampling and storage

The techniques and materials used for sampling of exhaled breath are of utmost importance to the final result of VOC analysis.10 Many different procedures are described in the literature according to the requirements of the analytical method to be used and the characteristics of possible target compounds, such as concentrations. Although task forces have started to address this issue, standardizations and consensus guidelines are still missing because of the manifold factors that influence VOCs in exhaled breath.11,12

Some of the technical issues to be considered are the characteristically low concentration of VOCs in exhaled breath with the need for optimized sample collection, in addition to highly sensitive measurement instruments, and the fact that a significant number of VOCs detected in exhaled breath can also be found in ambient air. Methods to address the last problem include simultaneous collection of room air with exhaled breath and mathematical correction for room air VOCs,13,14 using a breath collection apparatus with a VOC filter for inhalation to avoid breath contamination with ambient VOCs when collecting exhaled breath,15 or supplying clean air to the patient before breath sampling.16 However, these methods cannot account for exogenous VOCs previously inhaled by the patient, which are slowly released and might interfere with the endogenous disease-related VOC pattern (e.g. smoking). Furthermore, oral contamination originating from diet or bacterial flora must be considered. To account for dietary effects, patients can theoretically be put on a special diet for a few days before breath collection.1 Since patient compliance might be problematic, fasting for a number of hours seems reasonable.12

Another technical aspect is the breathing manoeuvre performed and the portion of exhaled air sampled. Forced or tidal breathing might contain contaminations from the upper airways and mouth,1719 whereas sampling only the last fraction of the exhaled breath provides alveolar air, which probably provides a better reflection of systemic metabolic processes.20 Bajtarevic et al.14 used mixed alveolar breath in order to also include compounds directly released by the lung.

If the collected breath sample is not directly analysed (e.g. by mass spectrometry), appropriate storage is important to preserve the sample without VOC loss or contamination with compounds released by the storage device. The commonly used bags or vials do not completely fulfil this requirement, even though the release of volatiles in non-permeable Tedlar® (DuPont, Wilmington, DE, USA) or Mylar® (DuPont Teijin Films, Wilton, UK) bags is minimal.14,21 Another attractive method is preconcentration on sorbent tubes followed by thermal desorption,13,22,23 which allows for reduction of the sample volume and increase of VOC concentrations. A simpler and less expensive variant is solid-phase microextraction, which is frequently used today.14,20 These techniques are especially useful in gas chromatography–mass spectrometry (GC–MS) experiments, since preconcentration is a prerequisite for most VOCs with very low concentrations. Excellent overviews about sampling systems used can be found in articles by Krilaviciute et al.24 and Rattray et al.21

Analysis of volatile organic compounds

At present, there are two complementary technical approaches to VOC analysis, namely chemical analytical and sensor array techniques.

Chemical analytical techniques aim to detect, identify and quantify individual compounds based on their physicochemical properties. The gold standard is GC–MS, which is highly sensitive and able to identify chemical compounds at low concentrations. However, GC–MS is bulky, expensive, requires preconcentration of VOCs from exhaled breath samples, does not allow for real-time analysis and requires highly trained personnel, which makes it unsuitable for population-based screening. Other methods that overcome some of the disadvantages of GC–MS include proton transfer reaction MS (PTR–MS), selected ion flow tube MS (SIFT–MS) and ion mobility spectrometry (IMS).21,24 PTR–MS requires no preconcentration and is technically easier to handle than GC–MS, but suffers from the disadvantage that compounds with the same mass-to-charge ratio cannot be separated.14,25 SIFT–MS allows for real-time analysis and absolute quantification, but is expensive and not suitable for point-of-care analysis.26 IMS can be integrated into a portable device but is unable to identify unknown compounds.2729 Even though some of these techniques have the potential to be used as screening tools (e.g. IMS and PTR–MS), they are not yet ready for broad application in exhaled breath VOC analysis. More detailed overviews of spectrometric methods and their application in lung cancer diagnosis can be found in Rattray et al.21

The second approach is based on the idea of mimicking the olfactory sense of mammals, deploying arrays of sensors with different sensitivities for VOCs that detect a specific ‘smellprint’, i.e. a VOC pattern, of the sample to be analysed. For obvious reasons these sensor arrays are dubbed electronic noses (e-noses). Different types of sensors are used, such as carbon polymers (e.g. Cyranose®, Sensigent, Baldwin Park, CA, USA),15,3032 quartz microbalance sensors,3336 colorimetric sensors,18,37 surface acoustic wave,38 gold nanoparticle sensors3942 and metal oxide semiconductor field-effect transistor (MOSFET)/metal oxide semiconductor (MOS) sensors.43 Excellent reviews on the design and use of e-noses can be found in the literature;14,24,4446 hence we give only a short summary of their characteristics. Compared with spectrometric-based techniques, e-noses have similar sensitivities for exhaled breath VOCs but are simpler to use, smaller, cheaper and have great potential for screening purposes. A methodological disadvantage is that they are not able to identify individual compounds. Technical issues include sensor drift, limited lifetimes of some sensors and decreased sensitivity in the presence of water vapour or a single compound at a high concentration.

Discrimination between healthy and lung cancer VOC patterns and evaluation of diagnostic performance of exhaled breath analysis require sophisticated data processing algorithms and statistical methods. These consist of, firstly, methods such as principal component analysis to reduce data dimensionality, and, secondly, methods to build a classification model, including partial least square discriminant analysis (PLS-DA),3336,43,47 linear canonical discriminant analysis,30,48 random forest,18,48 support vector machine,15,41,48 multinominal linear regression analysis,37 fuzzy logic,49 weighted digital analysis50 and artificial neuronal networks.38,43,51 Validation of the classification models is mostly performed by ‘leave-one-out’ cross-validation or random sample split into a training and validation set in the above-cited articles.

Results for exhaled breath analysis studies on lung cancer

Numerous studies testing the potential of exhaled breath analysis for discriminating lung cancer from healthy or benign states have been published so far, with increasing numbers in the last decade. Table 1 lists trials using chemical analytical techniques, including VOCs identified as possibly lung cancer related and the performance of classification models. Table 2 summarizes methods and results of sensor array studies on exhaled breath analysis. Both philosophies demonstrate promising results for differentiating lung cancer from benign diseases or healthy controls. Except for two cases (hexanal and nonanal tested as single biomarkers),52 sensitivity ranges from 51% to 100%, with the majority being > 80%. Similarly, specificity was higher than 70%, with values up to 100%, in all but four studies. Accuracy reached more than 80% in more than five studies. Nearly all studies using sensor array techniques and approximately half of those performed with chemical analytical methods were validated utilizing either ‘leave-on-out’ cross-validation or random sample split, dividing the patients into a training set and a validation set. Most studies show similar good results as the National Lung Screening Trial for lung cancer,66 which reports a sensitivity of 93.8% and a specificity of 73.4% for lung cancer screening by low-dose computerized tomography. One study, by Hubers et al.,31 demonstrated that, by combining exhaled breath analysis by e-nose and the analysis of sputum DNA RASSF1A hypermethylation, a sensitivity of 100% for lung cancer diagnosis could be reached. Based on these data, exhaled breath analysis holds a great potential for the development of a useful screening tool for lung cancer, perhaps in combination with other diagnostic methods, such as computerized tomography or sputum analysis.


Studies on exhaled breath analysis in lung cancer with chemical analytical techniques

Reference Year Technique Potential diagnostic VOCs for lung cancer Classification Validation N (LC) N (controls) Sensitivity (%) Specificity (%) Accuracy (%) Other statistics Remarks
Gordon et al.53 1985 GC–MS Acetone, methyl-ethyl-ketone, n-propanol Linear discriminant analysis 12 17 100 100 100 22 peaks (compounds not listed)
12 17 93 Model using 3 VOCs out of a possible 22
O’Neill et al.54 1988 GC–MS Propenal, acetone, 2-butanone, phenol, benzaldehyde, acetophenone, nonanal, ethylpropanoate, methylisobutenoate 8 Components present at > 90% occurrence level; 19 other compounds at this level represent environmental pollutants
Phillips et al.22 1999 GC–MS Styrene, 2,2,4,6,6-pentamethyl-heptane, 2-methyl-heptane, decane, propyl-benzene, undecane, methyl-cyclopentane, 1-methyl-2-pentyl-cyclopropane, trichlorofluoro-methane, benzene Forward stepwise discriminant analysis Cross-validation ‘leave-one-out’ 60 48 71.7 66.7 69.4 Controls: patients with abnormal chest radiograph and lung cancer not confirmed; results for cross-validation. 22 VOCs identified; only first 10 compounds contributing to the model listed here
9 48 100 81.3 Results for stage I only
Phillips et al.19 2003 GC–MS Butane, 3-methyl-tridecane, 7-methyl-tridecane, 4-methyl-octane, 3-methyl-hexane, heptane, 2-methyl-hexane, pentane, 5-methyl-decane Forward stepwise discriminant analysis Cross-validation ‘leave-one-out’; two validation sets (metastatic lung cancer, negative on bronchoscopy) 67 41 89.6 82.9 87 Patients with primary lung cancer and healthy controls; predictive model built from nine VOCs
67 41 85.1 80.5 83.3 Patients with primary lung cancer and healthy controls; results for leave-one-out cross-validation
15 66.7 Patients with metastatic lung cancer; validation set
91 37.4 Abnormal chest radiograph, negative for lung cancer on bronchoscopy; validation set
Poli et al.20 2005 GC–MS Isoprene, 2-methyl-pentane, pentane, ethyl-benzene, xylenes (total), trimethyl-benzene, toluene, benzene, heptane, decane, styrene, octane, pentamethyl-heptane Multinominal linear regression analysis 36 110 72.2 93.6 Controls: 25 mild to stable COPD patients, 35 asymptomatic smokers and 50 non-smokers
Chen et al.55 2007 GC Styrene, decane, isoprene, benzene, undecance, 1-hexene, hexanal, propyl-benzene, 1,2,4-trimethyl-benzene, heptanal, methyl-cyclopentane ≥ 1 of 11 characteristic VOCs 29 20 86.2 70 79.6 Controls: 13 healthy subjects, 7 chronic bronchitis
Steeghs et al.25 2007 PTR–MS Isoprene Bootstrapped stepwise forward logistic regression 11 57 AUC = 0.81 Controls: ‘healthy smokers’; 2 compounds at masses 25 and 69 amu identified. 69 corresponds to isoprene, but no correspondence for 25 known; noise?
Phillips et al.49 2007 GC–MS 1,5,9-Trimethyl-1,5,9-cyclododecatriene, 2,2,4-trimethyl-pentan-1,3-dioldiisobutyrate, 4-ethoxy-benzoic acidethyl ester, 2-methyl-propanoic acid 1-(1,1-dimethylethyl)-2-methyl-1,3-propanediyl ester, 10,11,-dihydro-5H-dibenz-(B,F)-azepine, 2,6-bis(1,1-dimethylethyl)-2,5-cyclohexadiene-1,4-dione, 1,1-oxybis-benzene, 2,5-dimethyl-furan, 2,2-diethyl-1,1-biphenyl, 2,4-dimethyl-3-pentanone, trans-caryophyllene, 1,1,3-trimethyl-3-phenyl-1H-indene, 1-propanol, 4-methyl-decane, 1,2,-benzenedicarboxylic acid diethyl ester, 2,5-dimethyl-2,4-hexadiene Fuzzy logic Random sample split 2:1 193 211 84.6 80 AUC = 0.88 Controls: negative chest CT; model based on 16 VOCs for maximal performance
Phillips et al.50 2008 GC–MS See citation Weighted digital analysis Cross-validation by random sample split 2:1 193 211 84.5 81 AUC = 0.90 Controls: negative Chest CT; model based on 30 VOCs (listed in citation)
Bajtarevic et al.14 2009 GC–MS 2-Butanone, benzaldehyde, 2,3-butanedione, 1-propanol ≥ 1 of 4 characteristic VOCs 65 31 52 100 Model on 4 VOCs not found in healthy controls. Total of 53 VOCs exclusively in lung cancer
Additional: 3-hydroxy-2-butanone, 3-butyn-2-ol, 2-methyl-butane, 2-methyl-2-butene, acetophenone, 1-cyclopentene, methyl-propyl-sulfide, tetramethyl-urea, n-pentanal, 1-methyl-1,3-cyclopentadiene, 2,3-dimethyl-2-butanol ≥ 1 of 15 characteristic VOCs 65 31 71 100 Model on 15 VOCs not found in healthy controls
Additional: 1,2,3,4-tetrahydro-isoquinoline, 3,7-dimethyl-undecane, cyclobutyl-benzene, butyl-acetate, ethylenimine, n-undecane ≥ 1 of 21 characteristic VOCs 65 31 80 100 Model on 21 VOCs not found in healthy controls
PTR–MS Isoprene, acetone, methanol 220 441 p < 0.01 for isoprene and acetone, p = 0.11 for methanol Controls: 84 smokers, 86 ex-smokers, 271 non-smokers; isoprene, acetone and methanol significantly lower in lung cancer patients than controls
Ligor et al.56 2009 GC–MS 1-Propanol, 2-butanone, 3-butyn-2-ol, benzaldehyde, 2-methyl-pentane, 3-methyl-pentane, n-pentane, n-hexane ≥ 1 of 8 characteristic VOCs 65 31 51 100 Model on 8 selected VOCs not found in controls. Total of 80 VOCs exclusively in lung cancer
Westhoff et al.27 2009 IMS Linear discriminant analysis Cross-validation ‘leave-one-out’ 32 54 100 100 100 23 VOC peaks identified, chemical compounds not analysed
Poli et al.57 2010 GC–MS Propanal, n-butanal, n-pentanal, n-hexanal, n-heptanal, n-octanal, n-nonanal Discriminant analysis Cross-validation ‘leave-one-out’ 40 38 90 92.1 Only straight aldehydes C3–C9 analysed; controls: asymptomatic non-smokers
Fuchs et al.52 2010 GC–MS Pentanal 12 24 75 95.8 Controls: 12 smokers, 12 healthy subjects; C1–C10 analysed; no significant differences found for C1–C4, C7 and C10
Hexanal 12 24 8.3 91.7
Octanal 12 24 58.3 91.7
Nonanal 12 24 33.3 95.8
Song et al.58 2010 GC–MS 1-Butanol 43 41 95.3 85.4 AUC = 0.940 Controls: healthy; discrimination by ROC curves
3-Hydroxy-2-butanone 43 41 93 92.7 AUC = 0.964
Darwiche et al.28 2011 IMS 2-Butanol(?), nonanal 19 Comparison of breath samples from cancer-affected and contralateral non-affected lung in the same patient by bronchoscopy. Patients serve as own controls. Two peaks significantly higher (one probably 2-butanol) and three peaks (one nonanal) lower in cancer-affected than in non-affected lung
Kischkel et al.59 2012 GC–MS Butane, pentane 15 Comparison of breath samples from cancer-affected and contralateral non-affected lung in the same patient during surgery with one-lung ventilation. Patients serve as own controls
Wang et al.60 2012 GC–MS See citation Linear discriminant analysis Cross-validation ‘leave-one-out’ 85 158 96.5 97.5 97.1 Controls: 70 benign lung disease, 88 healthy subjects; 23 VOCs (listed in citation)
Buszewski et al.61 2012 GC–MS Butanal, 2-butanone, ethyl acetate, ethyl benzene, 2-pentanone, 1-propanol, 2-propanol 29 44 p < 0.001 for all compounds Controls: non-smokers and smokers; concentration of listed seven compounds higher in lung cancer than in controls
Fu et al.62 2014 FT–ICR–MS 2-Butanone, 2-hydroxyacetaldehyde, 3-hydroxy-2-butanone, 4-hydroxyhexenal ≥ 2 of 4 characteristic VOCs 97 32 89.8 81.3 87.6 Controls: 88 healthy smokers and non-smokers and 32 patients with benign nodules for VOC analysis; only benign nodules vs. lung cancer tested in model
Bousamra et al.63 2014 FT–ICR–MS 2-Butanone, 3-hydroxy-2-butanone, 2-hydroxyacetaldehyde, 4-hydroxyhexenal ≥ 2 of 4 characteristic VOCs Validation set (VOCs analysed from different set of subjects) 107 40 87.9 77.5 85 Controls: 88 healthy subjects to find lung cancer VOCs that are increased compared to healthy; 40 benign pulmonary disease; sensitivity, specificity and accuracy based on lung-cancer diagnosis by ≥ 2 of 4 characteristic VOCs
Handa et al.29 2014 IMS n-Dodecane, butanol, 2-hexanol, cyclohexanon, iso-propylamin, ethylbenzol, hexanal, heptanal, 3-methyl-1-butanol Decision tree 50 39 76 100 Decision tree starting with n-dodecane
Rudnicka et al.51 2014 GC–MS Acetone, isoprene, ethanol, 1-propanol, 2-propanol, hexanal, dimethylsulfide Artificial neuronal network Random sample split 50:25:25% 108 145 74 73 AUC = 0.97 Controls: 121 healthy, 24 other lung diseases
Wang et al.47 2014 GC–MS Propanoic acid, caprolactam PLS-DA 18 Comparison of breath samples from cancer-affected and contralateral non-affected lung in the same patient during surgery with one-lung ventilation. Patients serve as own controls. PLS-DA classification included 30 VOCs in total
Capuano et al.36 2015 GC–MS Ethanol, 2-butanone, thiophene, 4-heptanone, butanoic acid ethyl ester, (acetyloxy)-acetic acid, cyclohexanone, 4-methyl-1-(1-methylethyl)-bicyclo[3.1.0]hexane-didehydro deriv., 2,2-dimethyl-hexanal, 1,1-diethoxy-3-methyl-butane, 1-(1-ethoxyethoxy)-pentane, 2,2,6-trimethyl-octane, 2-ethyl-1-hexanol, undecane, thymol, 2-methyl-1-decanol. 3,7-dimethyl-decane PLS-DA Cross-validation ‘leave-one-out’ 20 76 Comparison of breath samples from cancer-affected and contralateral non-affected lung in the same patient by bronchoscopy. Only cancer patients included, classification into ‘affected’ and ‘non-affected’ lung. For results from sensor array study see Table 2
Li et al.48 2015 FT–ICR–MS 2-Butanone, 4-hydroxy-2-hexenal, 3-hydroxy-2-butanone, hydroxyacetaldehyde, 4-hydroxy-2-nonenal, 2-pentanone/pentanal PLS-DA, support vector machines, random forest, linear discriminant analysis, quadratic discriminant analysis Random sample split 70:30% 85 119 Best AUC for single markers: 0.962 (LC vs. non-smokers), 0.946 (LC vs. smokers), 0.901 (LC vs. benign nodules) Only carbonyl compounds analysed. Controls: 40 healthy non-smokers, 40 healthy smokers, 34 patients with benign nodules
26 11 100 64 89 Test data set: LC vs. benign nodules
26 14 100 86 95 Test data set: LC vs. healthy smokers
26 12 96 100 97 Test data set: LC vs. healthy non-smokers
26 37 96 84 89 Test data set: LC vs. all
In addition to sensor array study: for details see Table 2
Machado et al.64 2005 GC–MS Isobutane, methanol, ethanol, acetone, pentane, isoprene, isopropanol, dimethylsulfide, carbon disulfide, benzene, toluene 8
Chen et al.38 2005 GC Styrene, decane, isoprene, benzene, undecance, 1-hexene, hexanal, propyl-benzene, 1,2,4-trimethyl-benzene, heptanal, methyl-cyclopentane 20 21 Controls: 15 healthy and 7 chronic bronchitis; 11 VOCs concentrations were smaller in control samples than lung cancer samples
Peng et al.40 2010 GC–MS 1-Methyl-4-(1-methylethyl)benzene, toluene, dodecane, 3,3-dimethyl-pentane, 2,3,4-trimethyl-hexane, and 1,1‘-(1-butenylidene)bis-benzene 30 22 VOCs with no overlap in abundance between cancer patients and healthy controls
Hakim et al.41 2011 GC–MS 3-Methyl-hexane, 2,4-dimethyl-heptane, 4-methyl-octane, 2,6,6-trimethyl-octane, 3-methyl-nonane; ammonium acetate, p-xylene 25 40 Controls: healthy
Peled et al.65 2012 GC–MS 1-Octene 28 10 Controls: benign pulmonary nodules; only 1-octene showed significantly higher concentrations in malign than benign cases
Broza et al.42 2013 GC–MS 2-Methyl-1-pentene, 2-hexanone, 3-heptanone, styrene, 2,2,4-trimethyl-hexane 11 34 VOCs present in > 90% of all pre- and post-surgery samples; 5 VOCs decreased significantly after surgery; controls not included in GC–MS study

amu, atomic mass unit; AUC, area under the curve; COPD, chronic obstructive pulmonary disease; CT, computerized tomography; FT–ICR–MS, Fourier transform–ion cyclotron resonance–mass spectrometry; LC, lung cancer; N, number of subjects; ROC, receiver operating characteristic.


Studies on exhaled breath analysis in lung cancer with sensor array techniques

Reference Year Technique Classification Validation N (LC) N (controls) Sensitivity (%) Specificity (%) Accuracy (%) Other statistics Remarks
DiNatale et al.33 2003 Quartz microbalance PLS-DA Cross-validation ‘leave-one-out’ 35 27 100 100 100 Controls: 18 healthy, 9 lung cancer post surgery; no misclassification for lung cancer vs. both other, but 1 healthy misclassified as post-surgery and 5 post-surgery classified as healthy
Machado et al.15 2005 Cyranose® 320 Support vector machine Validation set 14 62 71.4 91.9 Controls: 30 healthy, 12 COPD, 11 asthma, 7 pulmonary hypertension, 2 resected lung cancer. Training set included 14 lung cancer patients and 45 controls (20 healthy, 19 α1-antitrypsin deficiency, 6 chronic beryllium disease)
Chen et al.38 2005 Surface acoustic wave Artificial neuronal network Validation set 5 5 80 80 80 Model built on 11 VOCs with higher concentration in lung cancer (20 patients) than in controls (21 patients, 15 healthy and 7 chronic bronchitis); compare Table 1
Mazzone et al.18 2007 Colourimetric Random forest Random sample split 70:30% 49 94 73.3 72.4 Controls: 21 healthy, 20 sarcoidosis, 20 pulmonary arterial hypertension, 15 idiopathic pulmonary fibrosis, 18 COPD
Dragonieri et al.30 2009 Cyranose® 320 Linear canonical discriminant analysis Cross-validation 10 10 90 Controls: healthy
10 10 85 Controls: COPD
Barash et al.39 2009 Gold nanoparticles Principal component analysis, cluster analysis 40 56 Controls: healthy; cluster separation 100%, no overlap
D’Amico et al.34 2010 Quartz microbalance PLS-DA Cross-validation ‘leave-one-out’ 28 36 85 100 93.75 Controls: healthy never-smokers
28 28 92.8 78.6 85.7 Controls: 16 COPD, 3 interstitial lung disease, 5 bronchitis, 4 pleurisy
Peng et al.40 2010 Gold nanoparticles Principal component analysis, cluster analysis 30 22 Very good separation between ‘healthy’ and ‘cancer’ based on principal component analysis plots; first two principal components account for > 88% of total variance in data. Breast, colorectal and prostate cancer also included
Hakim et al.41 2011 Gold nanoparticles Support vector machine Cross-validation 25 40 100 92 Controls: healthy; head-and-neck cancers also included. Data for only lung cancer. 100% sensitivity and specificity for discrimination between lung cancer and head-and-neck-cancer
Peled et al.65 2012 Two single-wall carbon nanotubes with polycyclic aromatic hydrocarbons, 16 spherical gold nanoparticles Discriminant factor analysis Cross-validation ‘leave-one-out’ 49 19 86 ± 4 96 ± 4 88 ± 3 AUC = 0.986 Controls: benign pulmonary nodules
Santonico et al.35 2012 Quartz microbalance PLS-DA Cross-validation ‘leave-one-out’ 20 10 85 85 85 Controls: benign tracheal stenosis follow-up
Mazzone et al.37 2012 Colorimetric Logistic regression analysis Bootstrapping 83 137 70 86 AUC = 0.811 83 NSCLC vs. controls; controls: 67 at risk negative at screening, 70 indeterminate lung nodules
9 137 89 85 AUC = 0.890 9 SCLC vs. controls
Broza et al.42 2013 Gold and platinum nanoparticles Discriminant factor analysis Cross-validation ‘leave-one-out’ 12 5 100 80 94 Controls: benign nodules; results shown for comparison of LC and control pre-surgery
12 5 65 57 63 Comparison LC vs. control post-surgery
12 83 75 80 Comparison LC pre- vs. post-surgery
Schmekel et al.43 2014 MOSFET and MOS PLS-DA, artificial neuronal network Cross-validation ‘leave-one-out’ 12 10 Correlation coefficient = 0.954 (PLS-DA), correlation coefficient = 0.976 (artificial neural network) Comparison between LC with survival < 12 months and healthy controls
22 Correlation coefficient = 0.86 (PLS-DA), correlation coefficient = 0.97 (artificial neural network) Comparison between LC with survival < 12 months (12 patients) and LC with survival > 12 months (10 patients)
Hubers et al.31 2014 Cyranose® 320 Logistic regression analysis Validation set 18 8 94.4 12.5 64.3 Controls: cancer-free individuals, mainly with COPD; learning set: 20 lung cancer patients, 32 controls (sensitivity 80%, specificity 48.4%, accuracy 60.8%)
Capuano et al.36 2015 Quartz microbalance PLS-DA Cross-validation ‘leave-one-out’ 20 10 93 Breath samples included from cancer-affected and contralateral non-affected lung by bronchoscopy. Controls: patients with other lung disease than malignant. Classification result independent from sampling side (affected or non-affected lung)
McWilliams et al.32 2015 Cyranose® 320 Discriminant factor analysis Random sample split 2:1 25 166 AUC = 0.803 Controls: high-risk smokers without cancer

NSCLC, non-small-cell lung cancer; SCLC, small-cell lung cancer.

However, there are several limitations to the studies published so far and various challenges to be addressed. Patient numbers were mostly small, lung cancer histology and stages were mixed and some patients had already undergone treatment with chemotherapy.34 Control groups were variably defined, including healthy subjects, smokers and non-smokers, and patients with benign lung diseases (e.g. COPD, idiopathic pulmonary fibrosis, sarcoidosis, beryllium disease, pulmonary hypertension, benign nodules, etc.; see Tables 1 and 2). The preconditioning of subjects, such as fasting, diet or abstinence from nicotine and alcohol, as well as sampling and storage of breath samples, varied among the trials because there are no universal recommendations to control for these variables. Repeated sampling from one subject was not consistently performed. Furthermore, as discussed above, there was considerable variation among technical and statistical methods implemented, which makes it difficult to compare single studies.

A large number of VOCs belonging to different chemical classes, including hydrocarbons, alcohols, aldehydes, ketones, esters, nitriles and aromatic compounds, have been proposed as potential lung cancer biomarkers, but no single compound or VOC pattern is consistently reported by all or at least a majority of studies. In addition to huge inter- and intraindividual variations that have to be suspected from data on healthy persons, the large variation of compounds and even VOC classes in these trials can partly be explained by the use of different adsorbents for VOC extraction from breath samples and restriction to certain chemical classes by some studies (e.g. straight aldehydes C3–C954, aldehydes C1–C1052). Even though sensor array techniques do not allow for direct comparison of VOC patterns, it should be kept in mind that VOC patterns detected by chemical analytical and sensor array methods are probably not comparable because of different needs for pretreatment of samples. Preconcentration by solid-phase microextraction and thermal desorption will change VOC patterns detected by GC–MS compared with other methods not requiring preconcentration.

Differences in the limits of detection of different techniques is another issue to be considered. Furthermore, some compounds listed as possible biomarkers by one or more studies are thought to originate from contamination (e.g. cyclohexanone, dimethylsulfide, branched hydrocarbons;59 carbon disulfide, styrene, ethyl-benzene, toluene, n-hexane, 2-propanol, ethanol, and isobutene14) or stress-induced changes in metabolism unrelated to the tumour (e.g. acetone, isoprene,59 n-pentane,14 branched alkanes47) by other authors. These discrepancies reflect the inherent difficulties in exhaled breath analysis and the lack of standardized procedures for collecting and analysing breath samples. However, comparison with results from in vitro studies (Table 3) performed on headspaces of cultured lung cancer cell lines exhibit similarities to VOCs found in exhaled breath studies, supporting the general feasibility of exhaled breath analysis. Barash et al.39 report that VOC patterns from breath samples and lung cancer cell lines show a reasonable overlap, suggesting that VOCs found in exhaled breath correlate with biochemical processes in cancer cells. The majority of cell lines investigated increased certain VOCs compared with culture medium and decreased others, indicating that lung cancer cells are able to produce VOCs on the one hand, but also deplete VOCs present in the culture medium because of their metabolism.6770 Only one cell line (NHI-H1666) was found to merely consume VOCs. The reasons for this deviant behaviour, as well as for variations in VOC patterns of different lung cancer cell lines, remain speculative and may be because of various influences, possibly including passage number57 and environmental contamination.


In vitro studies on VOCs from lung cancer cells

References Year Material Technique Potential diagnostic VOCs for lung cancer Classification Remarks
Chen et al.55 2007 Lung cancer tissue from 16 patients GC Isoprene, undecane 4 peaks detected in all lung cancer cell cultures; only 2 identified
Gendron et al.71 2007 6 lung cancer cell lines (L55, L65, A549, H460, M51, REN) and 2 normal cell lines (NHDF, HASM) Cyranose® 320 Canonical discrimination plots, Mahalanobis distance (> 3) All cell lines could be separated from each other by canonical discrimination plots; Mahalanobis distance < 3 for L56 and M51 vs. NHDF (fibroblasts)
Filiapak et al.67 2008 CALU-1 cell line vs. culture medium GC–MS 2,3,3-Trimethyl-pentane, 2,3,5-trimethyl-hexane, 2,4-dimethyl-heptane, 4-methyl-octane Increased in headspace of CALU-1 vs. culture medium
Acetaldehyde, 3-methyl-butanal, butyl acetate, acetonitrile, acrolein, methacrolein, 2-methyl-propanal, 2-butanone, 2-methoxy-2-methyl-propane, 2-ethoxy-2-methyl-propane, hexanal Decreased in headspace of CALU-1 vs. culture medium
Sponring et al.68 2009 NCI-H2087 cell line vs. culture medium GC–MS 2-Ethyl-1-hexanol, 2-methyl-pentane Increased in headspace of NCI-H2087 vs. culture medium
Acetaldehyde, 2-methyl-propanal, 3-methyl-butanal, 2-methyl-butanal, butyl acetate Decreased in headspace of NCI-H2087 vs. culture medium
Barash et al.39 2009 7 NSCLC cell lines (CALU-3, H1650, H4006, H1435, H820, H1975, A549) vs. culture medium GC–MS 1,2-Bis(1-methylethyl)-benzene, 3-ethyl-benzaldehyde, tricyclo[,8)]deca-3,6-diene, cyclopropylphenylmethane, acetic acid, 1-methyl-3-ethyladamantane, (1,1-dimethylpropyl)-benzene, 4,7-dimethyl-undecane, 2-(3-methylbuta-1,3-dienyl)cyclohexanone, 2-ethyl-1-hexanol, 2-ethyl-4-methyl-1-pentanol, o-xylene, ethylbenzene, p-xylene, nitric oxide Forward stepwise discrimination analysis 15 VOCs identified in headspace of NSCLC cell lines only
Gold nanoparticles sensor array Principal component analysis, cluster analysis Cluster separation between NSCLC and control medium 100%, no overlap
Sponring et al.70 2010 NCI-H1666 cell line vs. culture medium GC–MS n-Butyl acetate, methyl tert-butyl ether, ethyl tert-butyl ether, hexanal, 3-methyl-butanal, methacrolein Decreased in headspace of NCI-H1666 vs. culture medium; no release of VOCs by this cell line, only consumption
Filiapak et al.72 2010 Lung cancer cell line A549 vs. culture medium GC–MS 2-Pentanone, 2,4-dimethyl-1-heptene, methyl tert-butyl ether, ethyl tert-butyl ether Increased in headspace of A549 vs. culture medium
n-Butyl acetate Decreased in headspace of A549 vs. culture medium
Wang et al.60 2012 3 lung cancer cell lines (A549, NCI-H446, BEAS-2B) and lung cancer tissue from 18 patients GC–MS Pentadecanone, nonadecane, eicosane 3 compounds in all cell lines and tissue samples

A549, human non-small-cell lung carcinoma (adenocarcinoma); BEAS-2B, human bronchial epithelium; CALU-1, human Caucasian lung epidermoid carcinoma; CALU-3, human non-small-cell lung carcinoma (adenocarcinoma); H1435, human non-small-cell lung carcinoma (adenocarcinoma); H1650, human non-small-cell lung carcinoma (adenocarcinoma; bronchoalveolar carcinoma); H1975, human non-small-cell lung carcinoma (adenocarcinoma); H4006, human non-small-cell lung carcinoma (adenocarcinoma); H460, human non-small-cell lung carcinoma (large cell carcinoma); H820, human non-small-cell lung carcinoma (papillary adenocarcinoma); HASM, human airway smooth muscle; L55, human non-small-cell lung carcinoma (adenocarcinoma); L65, human non-small-cell lung carcinoma; M51, human metastatic squamous cell carcinoma of the lung; NCI-H1666, human non-small-cell lung carcinoma (adenocarcinoma; bronchoalveolar carcinoma); NCI-H2087, human non-small-cell lung carcinoma (adenocarcinoma); NCI-H446, human small-cell lung carcinoma; NHDF, normal human diploid fibroblasts; REN, human mesothelioma.

A few studies comparing VOCs from the ipsilateral tumour-affected and the contralateral non-affected lung by bronchoscopy or intraoperatively as well as pre- and post-surgery trials support the idea of VOCs originating from lung cancer tissue itself.20,28,36,42,47,59 Certain VOCs were increased (2-butanol,28 butane and pentane,59 propanoic acid and caprolactam47) or decreased (nonanal28) in air samples from tumour-affected lungs compared with the contralateral side. Furthermore, VOCs that were increased before surgery decreased significantly after surgery (e.g. butane and pentane59 and isoprene and decane20). Similarly, Machado et al.15 and Di Natale et al.33 showed that lung cancer patients after resection were classified as non-malignant, suggesting that altered VOC patterns before and after surgery are associated with the cancer itself. In contrast to this evidence, based on small a number of subjects, 80 patients with resected lung cancer were classified into cancer groups with an accuracy of 96.3% in the study by Phillips et al.49 The authors argue that the induction of cytochrome P450 genotype in high-risk patients increases the consumption of VOCs produced due to oxidative stress and simultaneously increases the risk of lung cancer. The changes leading to altered VOC patterns in lung cancer compared with controls might hence originate in tissues far away from the tumour site and may not be affected by tumour resection. It has to be added that VOCs identified by Phillips et al.49 do not include those found by Darwiche et al.,28 Kischkel et al.59 and Wang et al.47 These opposing results reflect our lack of knowledge about the origin of VOCs and the biochemical processes involved in their production.


In addition to the issues discussed, a number of confounders have to be taken into account, including lung cancer stages, histology, age, sex, smoking, benign lung diseases, comorbidities, diet and medication.

Tumour, nodes, metastasis (TNM) stage

Most studies have included all stages of lung cancer. Since patient numbers were small in most cases, data concerning the influence of the tumour stage have to be interpreted with care. Fu et al.62 report statistically significant higher concentrations of 2-butanone in stages II–IV (51 patients) than stage I patients (34 cases) found by FT–ICR–MS, and Mazzone et al.37 could differentiate between stage I + II (41 cases) and stage III + IV (42 cases) NSCLC patients with an accuracy of 97.3% using a colorimetric sensor array. Peled et al.65 report very good discrimination between early and advanced cancer by nanoarray sensor (AUC = 0.961, accuracy 88%) but not by GC–MS. Other authors report no influence of TNM stages on VOCs or VOC patterns.14,1820,22,49,50,58


A significantly higher concentration of 4-hydroxyhexenal in squamous cell carcinoma than adenocarcinoma was reported by Fu et al.62 In the same study, these authors found pentanal to be increased only in samples from SCLC patients, not NSCLC, compared with controls. Handa et al.29 describe n-dodecane measured by IMS useful for the discrimination of adenocarcinomas into epidermal growth factor receptor (EGFR)-positive (n-dodecane increased) or -negative cases. Darwiche et al.28 report different VOC concentrations in different histological subtypes assessed by IMS. In particular, nonanal was detected in every case of squamous cell carcinoma but only one with adenocarcinoma, who later turned out to also have a laryngeal cancer of squamous cell origin. In the study by Song et al.58 GC–MS identified higher concentrations of tumour-related VOCs in adenocarcinoma than other histological subtypes. Santonico et al.35 applied quartz microbalance sensors to separate adenocarcinoma and squamous cell carcinoma with an overall accuracy of 75%. Mazzone et al.37 could discriminate between different histological subtypes using calorimetric sensor arrays, while Peled et al.65 used nanoarray sensors to successfully differentiate adenocarcinoma from squamous cell carcinoma (AUC = 0.974, accuracy 88%). However, Peled et al.65 could not discriminate between histological subtypes using GC–MS. Other authors reported no influence of histology on their results, including Phillips et al.19 and Capuano et al.36 Capuano et al.36 state that separation of adenocarcinoma from squamous cell carcinoma was not possible since the differences in VOC concentrations is obviously smaller than the resolution of their quartz microbalance sensors.


Only one study, by Batjarevic et al.,14 using PTR–MS technology identified differences with age for isoprene. No influence of age on the classification models was found by Phillips et al.,22 Mazzone et al.,18 Peng et al.40 and Fuchs et al.52


Two studies reported differences due to sex: Batjarevic et al.14 found significantly lower isoprene concentrations in breath samples from female lung cancer patients than healthy female controls (p < 0.00001), while no such difference was found in male subjects. An influence of sex on prediction models was also reported by McWilliams et al.,32 with AUC values consistently higher for males than for females. Sex did not influence discrimination of lung cancer and controls in the studies conducted by Phillips et al.,22 Peng et al.40 and Fuchs et al.52


A number of compounds have been related to smoking, such as toluene, benzene, acetonitrile, some furans, among others. Bajtarevic et al.14 recommend that these compounds not be used in GC–MS studies for differentiating between lung cancer and healthy controls. Smoking status also reduced the discriminative power between lung cancer and controls calculated by AUC by McWilliams et al.32 Smoking habit was not found to have any effect on results reported by a number of other authors implementing either chemical analytical or sensor array techniques.15,18,19,22,27,33,40,49,50

Medication and diet

No influence of either medication or diet was reported by Peng et al.40


In summary, the results from studies published so far show that the method of exhaled breath analysis by chemical analytical or e-nose technologies is principally feasible and could provide a promising tool for non-invasive, cost-effective large-scale screening. However, larger prospective and blinded studies are needed and a number of technical issues have yet to be addressed.

Canine studies

Case reports and a few small studies on dogs able to smell cancer in humans53,7375 have raised interest in training sniffer dogs to detect lung cancer using breath samples from patients. However, only a few prospective trials with divergent results have been published so far.17,61,76,77 McCulloch et al.17 used 55 lung cancer and 83 healthy control breath samples for training and blinded testing, reporting a sensitivity and specificity of 99% each, independent of the disease stage. Buszewski et al.61 compared the diagnostic ability of sniffer dogs with VOC analysis by GC–MS in 29 patients with lung cancer and 44 healthy control subjects. Dogs detected lung cancer with a sensitivity and specificity of 82.2% and 82.4%, respectively. A positive correlation between sniffer dog results and GC–MS analysis was found for ethyl acetate (r = 0.85) and 2-pentanone (r = 0.97). In the study carried out by Ehmann et al.,76 patients with COPD were included in addition to healthy control subjects and lung cancer patients, since COPD is a common comorbidity and may precede lung cancer evolution. Breath samples of 110 healthy subjects, 60 lung cancer and 50 COPD patients were used for training and testing. The overall sensitivity and specificity for the detection of lung cancer by sniffer dogs was 71% and 93%, respectively. COPD, smoking and food intake did not influence the dogs’ accuracy. Interestingly, accuracy was best in TNM stage I (100%) and worst in stage IV (63%), suggesting possible changes in VOC by inflammation or necrosis in advanced tumours.

Compared with these promising results, Amundsen et al.77 report rather discouraging data on the ability of dogs to detect lung cancer in an unselected cohort of 93 patients. Sixty-three patients were diagnosed with lung malignancies; the other 30 patients had benign diseases, with the exception of one patient with a history of urinary bladder cancer. A comparable distribution of COPD, coronary artery disease and inflammatory diseases was found in patients with and without lung malignancies, representing a typical clinical collective. The detection of lung cancer from breath samples ranged from 56% to 76% and from 8.3% to 33.3% for sensitivity and specificity, respectively, across different dogs, and included intensified training. These few studies show that even though lung cancer detection by sniffer dogs theoretically is an appealing approach, it faces a number of issues. As for the e-nose, it is still unclear which VOC biomarkers or patterns sniffer dogs identify as typical for lung cancer or other conditions. Confounders may decrease the dogs’ ability to detect cancer, even though accuracy was not influenced by smoking and diet in the study by Ehmann et al.76 Since dogs are living creatures, individual skill, training, motivation, illness and other factors may interfere with a dog’s ability to detect cancer. Furthermore, in vitro studies, where dogs have been used to detect lung cancer from cell culture experiments, show dissenting results. Although Yoel et al.78 report a sensitivity and specificity of 100%, the number of cancer cell culture samples was small [five samples of type 2 epithelial lung cancer (A549) together with five samples of melanoma and 10 samples of breast cancer cell cultures compared against pure cell culture medium]. In the study conducted by Schallschmidt et al.79 dogs could not discriminate between lung cancer cell lines [A549 and Lu7466 (lung adenocarcinoma cell line)] and control pure medium. In this study, however, gas samples from the culture headspaces were absorbed on polypropylene fleece and presented to the dogs, compared with full cell culture samples in the trial by Yoel et al.78 VOC profiling by GC–MS showed only quantitative differences between cancer and control headspace gas samples, suggesting that cancer cell metabolism decreased certain compounds but did not add any compound with the potential to act as a cancer biomarker.

In conclusion, there is a need for standardized double-blind studies with a greater number of patients and screening-like situations to answer the question whether or not dogs might be helpful in lung cancer screening.

Exhaled breath analysis – conclusion and perspective

Exhaled breath analysis is a relatively young field of research that is highly attractive for lung cancer screening because of its non-invasive character, wide applicability and cost-effectiveness. Numerous different methods of chemical analytical techniques, with GC–MS still being the gold standard, and sensor array techniques (e-noses) have been established. Furthermore, even sniffer dogs have been evaluated for their potential to detect lung cancer in human breath samples. New technological developments, in particular in the field of nanotechnology, will very likely improve the results of exhaled breath analysis in the coming years and hopefully help to overcome major shortcomings. In addition and concomitant to technological innovations standardization of breath sampling techniques, storage devices and pretreatment of samples is urgently needed. Task forces are already concerned with these challenging issues, and consensus recommendations are eagerly awaited.

Another major challenge is to obtain longitudinal data on breath VOC patterns in order to assess inter- and intraindividual variability and define reference values for possible VOC biomarkers. A number of confounders, such as age, sex, smoking status, comorbidities, medication, diet, environmental factors, among others, are assumed to potentially influence VOC patterns, but the present accumulation of data does not allow for conclusions on how VOC patterns may depend on these variables. Furthermore, metabolic processes that lead to VOC production or consumption and their changes in disease, especially carcinogenesis, have to be studied in more detail to enhance our understanding of the origin of VOCs. It is likely that the key to lung cancer detection will not involve one or a few single VOC biomarkers, but rather ‘breathomics’ – a ‘smellprint’ – consisting of a characteristic VOC profile, which will not require assessment of individual compounds but sophisticated methods of pattern recognition. Hence, further developments or innovations in statistical approaches to cope with the huge number of data involved in exhaled breath analysis by e-noses are needed to further improve sensitivity and specificity.

With the huge effort that is currently put into all aspects of exhaled breath analysis, ‘breathomics’ may have the potential to evolve into a viable screening tool for lung cancer in the near future.



Pauling L, Robinson AB, Teranishi R, Cary P. Quantitative analysis of urine vapor and breath by gas–liquid partition chromatography. Proc Natl Acad Sci USA 1971; 68:2374–6.


Bos LDJ, Weda H, Wang Y, et al. Exhaled breath metabolomics as a noninvasive diagnostic tool for acute respiratory distress syndrome. Eur Respir J 2014; 44:188–97.


Bos LDJ, Schultz MJ, Sterk PJ. Exhaled breath profiling for diagnosing acute respiratory distress syndrome. BMC Pulm Med 2014; 14:72.


Szabó A, Ruzsanyi V, Unterkofler K, et al. Exhaled methane concentration profiles during exercise on an ergometer. J Breath Res 2015; 9:016009.


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; 729:75–88.


Haick H, Broza YY, Mochalski P, Ruzsanyi V, Amann A. Assessment, origin, and implementation of breath volatile cancer markers. Chem Soc Rev 2014; 43:1423–49.


Phillips M, Cataneo RN, Greenberg J, Grodman R, Gunawardena R, Naidu A. Effect of oxygen on breath markers of oxidative stress. Eur Respir J 2003; 21:48–51.


Mitsui T, Kondo T. Inadequacy of theoretical basis of breath methylated alkane contour for assessing oxidative stress. Clin Chim Acta 2003; 333:91; author reply 93–4.


Hamuro Y, Molnar KS, Coales SJ, OuYang B, Simorellis AK, Pochapsky TC. Hydrogen-deuterium exchange mass spectrometry for investigation of backbone dynamics of oxidized and reduced cytochrome P450cam. J Inorg Biochem 2008; 102:364–70.


Miekisch W, Kischkel S, Sawacki A, Liebau T, Mieth M, Schubert JK. Impact of sampling procedures on the results of breath analysis. J Breath Res 2008; 2:026007.


Herbig J, Beauchamp J. Towards standardization in the analysis of breath gas volatiles. J Breath Res 2014; 8:037101.


Beauchamp JD, Pleil JD. Simply breath-taking? Developing a strategy for consistent breath sampling. J Breath Res 2013; 7:042001.


Phillips M. Method for the collection and assay of volatile organic compounds in breath. Anal Biochem 1997; 247:272–8.


Bajtarevic A, Ager C, Pienz M, et al. Noninvasive detection of lung cancer by analysis of exhaled breath. BMC Cancer 2009; 9:348.


Machado RF, Laskowski D, Deffenderfer O, et al. Detection of lung cancer by sensor array analyses of exhaled breath. Am J Respir Crit Care Med 2005; 171:1286–91.


Miekisch W, Schubert JK, Noeldge-Schomburg GFE. Diagnostic potential of breath analysis--focus on volatile organic compounds. Clin Chim Acta 2004; 347:25–39.


McCulloch M, Jezierski T, Broffman M, Hubbard A, Turner K, Janecki T. Diagnostic accuracy of canine scent detection in early- and late-stage lung and breast cancers. Integr Cancer Ther 2006; 5:30–9.


Mazzone PJ, Hammel J, Dweik R, et al. Diagnosis of lung cancer by the analysis of exhaled breath with a colorimetric sensor array. Thorax 2007; 62:565–8.


Phillips M, Cataneo RN, Cummin ARC, et al. Detection of lung cancer with volatile markers in the breath. Chest 2003; 123:2115–23.


Poli D, Carbognani P, Corradi M, et al. Exhaled volatile organic compounds in patients with non-small cell lung cancer: cross sectional and nested short-term follow-up study. Respir Res 2005; 6:71.


Rattray NJW, Hamrang Z, Trivedi DK, Goodacre R, Fowler SJ. Taking your breath away: metabolomics breathes life in to personalized medicine. Trends Biotechnol 2014; 32:538–48.


Phillips M, Gleeson K, Hughes JMB, et al. Volatile organic compounds in breath as markers of lung cancer: a cross-sectional study. Lancet 1999; 353:1930–3.


Phillips M, Boehmer JP, Cataneo RN, et al. Heart allograft rejection: detection with breath alkanes in low levels (the HARDBALL study). J Heart Lung Transplant 2004; 23:701–8.


Krilaviciute A, Heiss JA, Leja M, Kupcinskas J, Haick H, Brenner H. Detection of cancer through exhaled breath: a systematic review. Oncotarget 2015; 6:38643–57.


Steeghs MML, Cristescu SM, Munnik P, Zanen P, Harren FJM. An off-line breath sampling and analysis method suitable for large screening studies. Physiol Meas 2007; 28:503–14.


Kumar S, Huang J, Abbassi-Ghadi N, et al. Mass Spectrometric analysis of exhaled breath for the identification of volatile organic compound biomarkers in esophageal and gastric adenocarcinoma. Ann Surg 2015; 262:981–90.


Westhoff M, Litterst P, Freitag L, Urfer W, Bader S, Baumbach J-I. Ion mobility spectrometry for the detection of volatile organic compounds in exhaled breath of patients with lung cancer: results of a pilot study. Thorax 2009; 64:744–8.


Darwiche K, Baumbach JI, Sommerwerck U, Teschler H, Freitag L. Bronchoscopically obtained volatile biomarkers in lung cancer. Lung 2011; 189:445–52.


Handa H, Usuba A, Maddula S, Baumbach JI, Mineshita M, Miyazawa T. Exhaled breath analysis for lung cancer detection using ion mobility spectrometry. PLoS One 2014; 9:e114555.


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; 64:166–70.


Hubers AJ, Brinkman P, Boksem RJ, et al. Combined sputum hypermethylation and eNose analysis for lung cancer diagnosis. J Clin Pathol 2014; 67:707–11.


McWilliams A, Beigi P, Srinidhi A, Lam S, MacAulay CE. Sex and smoking status effects on the early detection of early lung cancer in high-risk smokers using an electronic nose. IEEE Trans Biomed Eng 2015; 62:2044–54.


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; 18:1209–18.


D’Amico A, Pennazza G, Santonico M, et al. An investigation on electronic nose diagnosis of lung cancer. Lung Cancer 2010; 68:170–6.


Santonico M, Lucantoni G, Pennazza G, et al. In situ detection of lung cancer volatile fingerprints using bronchoscopic air-sampling. Lung Cancer 2012; 77:46–50.


Capuano R, Santonico M, Pennazza G, et al. The lung cancer breath signature: a comparative analysis of exhaled breath and air sampled from inside the lungs. Sci Rep 2015; 5:16491.


Mazzone PJ, Wang X-F, Xu Y, et al. Exhaled breath analysis with a colorimetric sensor array for the identification and characterization of lung cancer. J Thorac Oncol 2012; 7:137–42.


Chen X, Cao M, Li Y, et al. A study of an electronic nose for detection of lung cancer based on a virtual SAW gas sensors array and imaging recognition method. Meas Sci Technol 2005; 16:1535–46.


Barash O, Peled N, Hirsch FR, Haick H. Sniffing the unique “odor print“ of non-small-cell lung cancer with gold nanoparticles. Small 2009; 5:2618–24.


Peng G, Hakim M, Broza YY, et al. Detection of lung, breast, colorectal, and prostate cancers from exhaled breath using a single array of nanosensors. Br J Cancer 2010; 103:542–51.


Hakim M, Billan S, Tisch U, et al. Diagnosis of head-and-neck cancer from exhaled breath. Br J Cancer 2011; 104:1649–55.


Broza YY, Kremer R, Tisch U, et al. A nanomaterial-based breath test for short-term follow-up after lung tumor resection. Nanomedicine 2013; 9:15–21.


Schmekel B, Winquist F, Vikström A. Analysis of breath samples for lung cancer survival. Anal Chim Acta 2014; 840:82–6.


Adiguzel Y, Kulah H. Breath sensors for lung cancer diagnosis. Biosens Bioelectron 2015; 65:121–38.


Montuschi P, Mores N, Trové A, Mondino C, Barnes PJ. The electronic nose in respiratory medicine. Respiration 2013; 85:72–84.


Wilson AD. Advances in electronic-nose technologies for the detection of volatile biomarker metabolites in the human breath. Metabolites 2015; 5:140–63.


Wang C, Dong R, Wang X, et al. Exhaled volatile organic compounds as lung cancer biomarkers during one-lung ventilation. Sci Rep 2014; 4:7312.


Li M, Yang D, Brock G, et al. Breath carbonyl compounds as biomarkers of lung cancer. Lung Cancer 2015; 90:92–7.


Phillips M, Altorki N, Austin JHM, et al. Prediction of lung cancer using volatile biomarkers in breath. Cancer Biomark 2007; 3:95–109.


Phillips M, Altorki N, Austin JHM, et al. Detection of lung cancer using weighted digital analysis of breath biomarkers. Clin Chim Acta 2008; 393:76–84.


Rudnicka J, Walczak M, Kowalkowski T, Jezierski T, Buszewski B. Determination of volatile organic compounds as potential markers of lung cancer by gas chromatography–mass spectrometry versus trained dogs. Sensors Actuators, B Chem 2014; 202:615–21.


Fuchs P, Loeseken C, Schubert JK, Miekisch W. Breath gas aldehydes as biomarkers of lung cancer. Int J Cancer 2010; 126:2663–70.


Gordon RT, Schatz CB, Myers LJ, et al. The use of canines in the detection of human cancers. J Altern Complement Med 2008; 14:61–7.


O’Neill HJ, Gordon SM, O’Neill MH, Gibbons RD, Szidon JP. A computerized classification technique for screening for the presence of breath biomarkers in lung cancer. Clin Chem 1988; 34:1613–18.


Chen X, Xu F, Wang Y, et al. A study of the volatile organic compounds exhaled by lung cancer cells in vitro for breath diagnosis. Cancer 2007; 110:835–44.


Ligor M, Ligor T, Bajtarevic A, et al. Determination of volatile organic compounds in exhaled breath of patients with lung cancer using solid phase microextraction and gas chromatography mass spectrometry. Clin Chem Lab Med 2009; 47:550–60.


Poli D, Goldoni M, Corradi M, et al. Determination of aldehydes in exhaled breath of patients with lung cancer by means of on-fiber-derivatisation SPME-GC/MS. J Chromatogr B Anal Technol Biomed Life Sci 2010; 878:2643–51.


Song G, Qin T, Liu H, et al. Quantitative breath analysis of volatile organic compounds of lung cancer patients. Lung Cancer 2010; 67:227–31.


Kischkel S, Miekisch W, Fuchs P, Schubert JK. Breath analysis during one-lung ventilation in cancer patients. Eur Respir J 2012; 40:706–13.


Wang Y, Hu Y, Wang D, Yu K, Wang L. The analysis of volatile organic compounds biomarkers for lung cancer in exhaled breath, tissues and cell lines. Cancer Biomarkers 2012; 11:3233.


Buszewski B, Ligor T, Jezierski T, Wenda-Piesik A, Walczak M, Rudnicka J. Identification of volatile lung cancer markers by gas chromatography–mass spectrometry: comparison with discrimination by canines. Anal Bioanal Chem 2012; 404:141–6.


Fu X-A, Li M, Knipp RJ, Nantz MH, Bousamra M. Noninvasive detection of lung cancer using exhaled breath. Cancer Med 2014; 3:174–81.


Bousamra M, Schumer E, Li M, et al. Quantitative analysis of exhaled carbonyl compounds distinguishes benign from malignant pulmonary disease. J Thorac Cardiovasc Surg 2014; 148:1074–80; discussion 1080–1.


Machado RF, Laskowski D, Deffenderfer O, et al. Detection of lung cancer by sensor array analyses of exhaled breath. Am J Respir Crit Care Med 2005; 171:1286–91.


Peled N, Hakim M, Bunn PA, et al. Non-invasive breath analysis of pulmonary nodules. J Thorac Oncol 2012; 7:1528–33.


Church TR, Black WC, Aberle DR, et al. Results of initial low-dose computed tomographic screening for lung cancer. N Engl J Med 2013; 368:1980–91.


Filipiak W, Sponring A, Mikoviny T, et al. Release of volatile organic compounds (VOCs) from the lung cancer cell line CALU-1 in vitro. Cancer Cell Int 2008; 8:17.


Sponring A, Filipiak W, Mikoviny T, et al. Release of volatile organic compounds from the lung cancer cell line NCI-H2087 in vitro. Anticancer Res 2009; 29:419–26.


Filipiak W, Ruzsanyi V, Mochalski P, et al. Dependence of exhaled breath composition on exogenous factors, smoking habits and exposure to air pollutants. J Breath Res 2012; 6:036008.


Sponring A, Filipiak W, Ager C, et al. Analysis of volatile organic compounds (VOCs) in the headspace of NCI-H1666 lung cancer cells. Cancer Biomark 2010; 7:153–61.


Gendron KB, Hockstein NG, Thaler ER, Vachani A, Hanson CW. In vitro discrimination of tumor cell lines with an electronic nose. Otolaryngol Head Neck Surg 2007; 137:269–73.


Filipiak W, Sponring A, Filipiak A, et al. TD-GC-MS analysis of volatile metabolites of human lung cancer and normal cells in vitro. Cancer Epidemiol Biomarkers Prev 2010; 19:182–95.


Williams H, Pembroke A. Sniffer dogs in the melanoma clinic? Lancet 1989; 1:734.


Willis CM, Church SM, Guest CM, et al. Olfactory detection of human bladder cancer by dogs: proof of principle study. BMJ 2004; 329:712.


Horvath G, Järverud GAK, Järverud S, Horváth I. Human ovarian carcinomas detected by specific odor. Integr Cancer Ther 2008; 7:76–80.


Ehmann R, Boedeker E, Friedrich U, et al. Canine scent detection in the diagnosis of lung cancer: revisiting a puzzling phenomenon. Eur Respir J 2012; 39:669–76.


Amundsen T, Sundstrøm S, Buvik T, Gederaas OA, Haaverstad R. Can dogs smell lung cancer? First study using exhaled breath and urine screening in unselected patients with suspected lung cancer. Acta Oncol 2014; 53:307–15.


Yoel U, Gopas J, Ozer J, Peleg R, Shvartzman P. Canine scent detection of volatile elements, characteristic of malignant cells, in cell cultures. Isr Med Assoc J 2015; 17:567–70.


Schallschmidt K, Becker R, Zwaka H, et al. In vitro cultured lung cancer cells are not suitable for animal-based breath biomarker detection. J Breath Res 2015; 9:027103.

Add comment 

Home  Editorial Board  Search  Current Issue  Archive Issues  Announcements  Aims & Scope  About the Journal  How to Submit  Contact Us
Find out how to become a part of the HMJ  |   CLICK HERE >>
© Copyright 2012 - 2013 HMJ - HAMDAN Medical Journal. All Rights Reserved         Website Developed By Cedar Solutions INDIA