American Journal of Respiratory and Critical Care Medicine

To the Editor:

Pulmonary arterial hypertension (PAH) is a devastating pulmonary-vascular disease of unknown etiology, characterized by pulmonary arterial remodeling that causes a progressive increase in vascular resistance, ultimately leading to terminal right-heart failure (1, 2). Due to the complexity of the disease (3), there is no recognized biomarker for early diagnosis. The currently available biomarkers reflect increased right ventricular pre-load, inflammation, oxidative stress or endothelial cell dysfunction and can be used at best for prognosis (4). So far, the right-heart catheterization remains the gold-standard to accurately diagnose PAH. This procedure is invasive, costly, risky to the patient, and, hence, unsuitable for widespread screening (5). Accumulating evidence indicates that early management will translate into better outcome (3). Therefore, effective methods for early PAH diagnosis are urgently needed.

The detection of patterns of disease-related volatile organic compounds (VOCs) (6) in exhaled breath using sensor arrays on the basis of cross-reactive gold-nanoparticles coated with organic ligands overcomes many constraints of conventional screening/monitoring techniques (611). Indeed, this approach may offer several advantages, including: non-invasive, easy-to-perform, direct and real-time monitoring of PAH. These devices have been tested for (early) detection of cancers (9, 11) and other diseases states (see Reference 9 and citations therein), showing an ability to distinguish not only between different types of disease, but also between early and advanced stages of the disease (10, 11).

Here, we explored the feasibility of an array of cross-reactive molecularly-modified gold-nanoparticle sensors, in conjugation with pattern recognition methods, for detecting and classifying PAH. Some of the results have been previously reported in the form of an abstract (12). We undertook a case-control study on 22 patients with idiopathic or heritable PAH from the French Referral Center of Severe Pulmonary Hypertension and 23 healthy volunteers, under approval by our Institutional Ethics Committee (CPP-IDF-VII, approval No. 11-013). The clinical data are summarized in Table 1; diagnosis and treatment of PAH are described in the online supplement. Exhaled alveolar air was collected using an “offline” method that effectively filtered exogenous contaminants and excluded the dead space volume (see the online supplement and References 10 and 11). The breath samples were analyzed, using the sensor array described in the online supplement and in References 10 and 11. Predictive models were developed, using discriminant factor analysis (DFA; see the online supplement) (10, 11). The accuracy of the prediction, for each model, was estimated through a blind test: 65–75% of the samples were used as a training set for calculating the DFA model; the remaining samples (25–35%) were used for the blind validation test, as described in the online supplement.

TABLE 1. CLINICAL AND HEMODYNAMIC DATA OF PATIENTS WITH PAH AND HEALTHY CONTROL SUBJECTS

 Patients with Pulmonary Arterial Hypertension (n = 22)Control Subjects (n = 23) 
Age, yr47.5 ± 3.338.2 ± 2.0 
Sex, M/F6/1610/13 
Heritable/idiopathic, n7/15  
NYHA functional class, n   
 I or II15  
 III7  
Disease duration, yr   
 0.1–59  
 >513  
Acute vasodilator response,* n5  
6-MWD, m489 ± 19  
mPAP, mm Hg47.23 ± 2.99  
CI, L/min/m23.48 ± 0.18  
PVR, mm Hg/L/min/m6.92 ± 0.67  
Specific PAH therapy, n   
 ERA3  
 PDE5-i1  
 ERA + PDE5-i4  
 Prostanoids1  
 PDE5-i + Prostanoids1  
 ERA + PDE5-i + prostanoids7  
 ERA + CCB3  
 CCB2  

Definition of abbreviations: 6-MWD = 6-minute walk distance; CCB = calcium channel blockers; CI = cardiac index; ERA = endothelin receptor antagonists; mPAP = mean pulmonary arterial pressure; NYHA = New York Heart Association; PDE5-I = phosphodiesterase 5 inhibitors; PVR = pulmonary vascular resistance index.

Data expressed as mean ± SEM unless otherwise indicated. Exclusion criteria were age under 18 and over 65 years, pregnancy, heavy smoking, ongoing known infectious conditions, history of recent respiratory tract infection, acute right heart failure or any underlying respiratory disease in the past 4 weeks, and/or current uncontrolled other medical condition. Control subjects were matched for age, sex, and smoking status. Patients with heritable pulmonary arterial hypertension, carrying a BMPR2 mutation, had similar therapy and NYHA class as patients with idiopathic pulmonary arterial hypertension. All patients under prostanoids received either intra-venous or sub-cutaneous treatment; none of the patients included were on inhaled prostanoids.

*Acute vasodilator response is defined by a fall in Ppa of at least 10 mm Hg, reaching an absolute value of Ppa under 40 mm Hg, associated with no change or an increase in CI.

To test the sensors’ ability to detect PAH as well as subcategories of PAH, two DFA models were developed, using two different combinations of sensing-features that were extracted from the sensor array (see the online supplement). The first DFA model could identify PAH among a mixed population of patients with PAH (n = 22) and healthy control subjects (n = 23). Note that the number of input features for the DFA-model was kept low (i.e., considerably smaller than the number of samples in the smallest study group; ratio of sensing features to data samples lower than 1:6) to avoid overfitting the experimental data. The sensitivity, specificity, and accuracy of PAH identification in the blind test set were 100%, 83%, and 92%, respectively (Figure 1A). Note that the application of a blind validation tests for the DFA model further reduces the risk of random associations. It should be mentioned that all patients with PAH received (diverse) treatment (Table 1); no correlation was observed between a patient’s medication and the (mis)classification of his or her breath sample.

The second DFA model distinguished the idiopathic (n = 15) from the heritable (n = 7) patients with PAH with an accuracy of 87% in a blind test (Figure 1C), indicating a potential relationship between the breath fingerprint and a genetic mutation. It should be noted that all patients with heritable PAH but one carried an identified mutation in the bone morphogenetic protein receptor type 2 (BMPR2) gene. The patient with heritable PAH (confirmed family history of PAH) and no identified BMPR2 (or other PAH predisposing genes) mutation was the only one misclassified as idiopathic. Receiver operator curve (ROC) analysis assessed the performance of the classifications and the accuracy of the predictions; the areas under the curves were close to 1, indicating high discriminatory power of the test (Figures 1B and 1D).

The effect of some potentially important confounding factors (smoking status; age group; sex; residence area) on the two developed DFA models was studied by attempting to distinguish between, for example, smokers and nonsmokers (see detailed description in the online supplement). These tests yielded arbitrary classification in all cases, confirming the stability of both DFA models against the studied confounding factors (see Tables E1 and E2 in the online supplement).

In addition to the demonstrated separation of the PAH from the healthy population, and of idiopathic from heritable PAH, we have tested a further DFA model that distinguished well between patients with severe and less severe PAH, using the patients’ New York Heart Association (NYHA) functional class as severity marker (see the online supplement for details and Figure E1 for DFA map and ROC curve), suggesting that the sensor-array approach could also be useful for determining disease severity. However, the NYHA classification is relatively arbitrary, and other severity markers should be considered in future larger-scale clinical studies, to develop a more reliable model for PAH severity.

The main limitation of our study was the limited cohort size, which precluded definitive conclusions. When statistical analysis is performed on a small number of observations, there is always a risk of finding random associations. In this study, however, we reduced the risk of random associations by using blind validation tests. Therefore, we believe that these preliminary results are encouraging and might open new horizons in assessing PAH. The pilot study presented here has encouraged the planning of a large-scale multicenter prospective study with a more diverse study population to definitely establish the clinical usefulness of the sensor-array approach as a diagnostic tool of PAH. The planned study will address, among other things, the effects of different severity markers of PAH (e.g., NYHA functional class, 6-min-walk distance, hemodynamic parameters, B-type natriuretic peptide levels) and of the patients’ acute pulmonary vasoreactivity.

An inherent limitation of the presented sensor-array approach to breath analysis is the inability to identify specific VOCs in the breath. This makes it impossible to identify the VOC-related biological pathways of the disease. A separate chemical analysis (by gas-chromatography/mass-spectrometry) of the breath samples that were studied here is underway and will be published elsewhere.

In conclusion, our results constitute explorative evidence and a proof-of-concept that breath fingerprints could be useful for developing a non-invasive diagnostic tool for detecting and classifying PAH. Whether this noninvasive approach will translate into a tool for early diagnosis in subgroups that are at risk of developing PAH (13, 14) is a challenge for the future.

The authors thank Marie Camille Chaumais (Service de Pharmacie) and Veronique Nietzsche (Service Biomedical) and the nurses from Service des Explorations Fonctionnelles Respiratoires for facilitating installation of the collection device at Hôpital Antoine Béclère, Clamart, France. M. N. and H. H. thank Dr. U. Tisch (Technion, Haifa) for helpful discussions.

1. Humbert M, Sitbon O, Simonneau G. Treatment of pulmonary arterial hypertension. N Engl J Med 2004;351:14251436.
2. O’Callaghan DS, Savale L, Montani D, Jaïs X, Sitbon O, Simonneau G, Humbert M. Treatment of pulmonary arterial hypertension with targeted therapies. Nat Rev Cardiol 2011;8:526538.
3. Galiè N, Hoeper MM, Humbert M, Torbicki A, Vachiery JL, Barbera JA, Beghetti M, Corris P, Gaine S, Gibbs JS, et al.; Task Force for Diagnosis and Treatment of Pulmonary Hypertension of European Society of Cardiology (ESC), European Respiratory Society (ERS), International Society of Heart and Lung Transplantation (ISHLT). Guidelines for the diagnosis and treatment of pulmonary hypertension. Eur Respir J 2009;34:12191263.
4. Cracowski JL. Towards prognostic biomarkers in pulmonary arterial hypertension. Eur Respir J 2012;39:799801.
5. Hoeper MM, Lee SH, Voswinckel R, Palazzini M, Jais X, Marinelli A, Barst RJ, Ghofrani HA, Jing ZC, Opitz C, et al. Complications of right heart catheterization procedures in patients with pulmonary hypertension in experienced centers. J Am Coll Cardiol 2006;48:25462552.
6. Hakim M, Broza YY, Barash O, Peled N, Phillips M, Amann A, Haick H. Volatile organic compounds of lung cancer and possible biochemical pathways. Chem Rev 2012;112:59495966.
7. Barash O, Peled N, Hirsch FR, Haick H. Sniffing the unique “odor print” of non-small-cell lung cancer with gold nanoparticles. Small 2009;5:26182624.
8. Peng G, Tisch U, Adams O, Hakim M, Shehada N, Broza YY, Billan S, Abdah-Bortnyak R, Kuten A, Haick H. Diagnosing lung cancer in exhaled breath using gold nanoparticles. Nat Nanotechnol 2009;4:669673.
9. Tisch U, Haick H. Arrays of chemisensitive monolayer-capped metallic nanoparticles for diagnostic breath testing. Rev Chem Eng 2011;26:171179.
10. Xu ZQ, Broza YY, Ionsecu R, Tisch U, Ding L, Liu H, Song Q, Pan YY, Xiong FX, Gu KS, et al. A nanomaterial-based breath test for distinguishing gastric cancer from benign gastric conditions. Br J Cancer 2013;108:941950.
11. Peled N, Hakim M, Bunn PA Jr, Miller YE, Kennedy TC, Mattei J, Mitchell JD, Hirsch FR, Haick H. Non-invasive breath analysis of pulmonary nodules. J Thorac Oncol 2012;7:15281533.
12. Cohen-Kaminsky S, Nakhleh M, Montani D, Girerd B, Garcia G, Simonneau G, Haick H, Humbert M. A proof of concept for exhaled air analysis using nano artificial nose (na-nose) in pulmonary arterial hypertension. 5th World Symposium on Pulmonary Hypertension. Nice Acropolis, Nice, France; 2013.
13. Simonneau G, Robbins IM, Beghetti M, Channick RN, Delcroix M, Denton CP, Elliott CG, Gaine SP, Gladwin MT, Jing ZC, et al. Updated clinical classification of pulmonary hypertension. J Am Coll Cardiol 2009;54:S43S54.
14. Humbert M, Sitbon O, Chaouat A, Bertocchi M, Habib G, Gressin V, Yaïci A, Weitzenblum E, Cordier JF, Chabot F, et al. Survival in patients with idiopathic, familial, and anorexigen-associated pulmonary arterial hypertension in the modern management era. Circulation 2010;122:156163.

*These authors contributed equally to the work.

Supported in part by Assistance Publique-Hôpitaux de Paris (AP-HP), the French National Institute for Health and Medical Research (INSERM), and Université Paris-Sud. D.M. is supported by the Association HTAP France.

Author Contributions: Conception and design: S.C.-K., H.H., and M.H.; patient recruitment: B.G., D.M., G.S., and M.H.; sensor-array set-up: M.N. and H.H.; collection of biological and clinical data: B.G. and D.M.; patient interview and documentation, and sample collection: S.C.-K. and F.P.; acquisition of GC-MS and sensor-array data: M.N. and H. H.; analysis and interpretation: S.C.-K., M.N., F.P., H.H., and M.H.; drafting the letter for important intellectual content: S.C.-K., M.N., H.H., and M.H. All authors approved the final letter.

This letter has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org

Author disclosures are available with the text of this letter at www.atsjournals.org.

Related

No related items
American Journal of Respiratory and Critical Care Medicine
188
6

Click to see any corrections or updates and to confirm this is the authentic version of record