You are here

Article Text

Article menu

Download PDFPDF

Endoscopy News
Detection of Barrett’s oesophagus through exhaled breath using an electronic nose device
  1. Yonne Peters1,
  2. Ruud W M Schrauwen2,
  3. Adriaan C Tan3,
  4. Sanne K Bogers2,
  5. Bart de Jong1,
  6. Peter D Siersema1
  1. 1 Gastroenterology and Hepatology, Radboudumc, Nijmegen, Gelderland, The Netherlands
  2. 2 Gastroenterology and Hepatology, Ziekenhuis Bernhoven, Uden, Noord-Brabant, The Netherlands
  3. 3 Canisius Wilhelmina Hospital, Nijmegen, The Netherlands
  1. Correspondence to Professor Peter D Siersema, Gastroenterology and hepatology, Radboudumc, Nijmegen, Gelderland 6525 GA, Netherlands; peter.siersema{at}radboudumc.nl

Abstract

Timely detection of oesophageal adenocarcinoma (OAC) and even more so its precursor Barrett’s oesophagus (BO) could contribute to decrease OAC incidence and mortality. An accurate, minimally-invasive screening method for BO for widespread use is currently not available. In a proof-of-principle study in 402 patients, we developed and cross-validated a BO prediction model using volatile organic compounds (VOCs) analysis with an electronic nose device. This electronic nose was able to distinguish between patients with and without BO with good diagnostic accuracy (sensitivity 91% specificity 74%) and seemed to be independent of proton pump inhibitor use, the presence of hiatal hernia, and reflux. This technique may enable an efficient, well-tolerated, and sensitive and specific screening method to select high-risk individuals to undergo upper endoscopy.

  • barrett's oesophagus
  • oesophageal cancer
  • screening
  • breath tests
  • endoscopy

https://doi.org/10.1136/gutjnl-2019-320273

Statistics from Altmetric.com

    Request Permissions

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

    Study background

    The incidence of oesophageal adenocarcinoma (OAC) has risen by sixfold over the last few decades.1 The majority of patients with OAC present with advanced disease, resulting in poor survival rates.2 Although Barrett’s oesophagus (BO) is a known precursor of OAC, >90% of patients with OAC never had prior endoscopy and OAC is usually diagnosed outside BO surveillance.3 Upper endoscopy, the current standard for BO detection, is invasive and expensive and therefore not suitable for population-based screening. Analysis of volatile organic compounds (VOCs) in exhaled breath may be a promising technique to detect undiagnosed cases of BO in the population.4 VOCs are gaseous end products resulting from physiologic metabolic processes in the body, pathophysiologic processes such as inflammation or oxidative stress-related activity, or external factors such as medication use or changes in the microbiome.5 Therefore, VOC profiles may serve as a biomarker for various diseases. Electronic noses represent a potential technique for real-time, high throughput exhaled VOC pattern analysis. In this multicentre, proof-of-principle study, we assessed the accuracy of exhaled VOCs analysis using an electronic nose device for detection of BO compared with endoscopy and biopsy as the reference standard.

    Methods

    Adult patients undergoing a clinically indicated upper endoscopy were invited to provide a 5 min breath sample using an electronic nose immediately before a scheduled endoscopy. VOC analysis by an electronic nose device is based on pattern recognition and resembles mammalian olfaction (figure 1).6 7 The Aeonose (the eNose company, Zutphen, the Netherlands) consists of three metal-oxide sensors and uses chemical to electrical interfaces to measure subtle VOC profiles of different diseases in exhaled breath. Data were analysed by an artificial neural network in a supervised fashion to identify data classifiers to extract breath print differences between patients with BO, gastro-oesophageal reflux disease (GORD), and controls. Leave-10%-out cross-validation of data was performed after training the artificial neural network to make sure the prediction model generated was disease-specific. More details on the methods and the electronic nose technology can be found in the online supplementary file.8 9

    Supplemental material

    Figure 1
    Figure 1

    The electronic nose concept resembles mammalian olfaction. Volatile organic compounds (VOCs) are presented to a metal-oxide sensor array. VOCs interact with these sensors based on their chemical characteristics. Combinations of individual sensor measurements generate a digital signal, which can be analysed by pattern recognition and artificial neural networks. Adapted from Santos et al. 7

    Results

    Breath samples were obtained from 513 individuals, resulting in a patient acceptability rate of 91.4% (figure 2). No adverse events were observed during breath analysis. After exclusions, 402 patients in total were included; 129 patients with BO, 141 patients with GORD (50 with reflux oesophagitis (35.5%)), and 132 patients in the control group. In the latter group, 76 (57.6%) patients had a normal upper gastrointestinal tract on endoscopy or only the presence of a hiatal hernia. Table 1 summarises the main characteristics of all patients enrolled in this study.

    Figure 2
    Figure 2

    Study flowchart. BO, Barrett’s oesophagus; GORD, gastro-oesophageal reflux disease; IM, intestinal metaplasia.

    View this table:
    Table 1

    Characteristics of the total study cohort (n=402)

    VOC profiles of patients with BO differed significantly from GORD patients and healthy controls (figure 3A). The cross-validated prediction model demonstrated the ability to detect BO with 91% (95% CI 84% to 95%) sensitivity, 74% (95% CI 69% to 79%) specificity, and an area under the curve (AUC) of 0.91 (95% CI 0.87 to 0.94). To limit the number of false-positive results (type 1 errors), a different cut-off point in the receiver operating curve can be selected with a corresponding specificity of the electronic nose for BO of 86% at the expense of test sensitivity which decreases to 71% (figure 3B). In addition, breath prints of BO patients could be differentiated from GORD patients with 64% (95% CI 55% to 72%) sensitivity and 74% (95% CI 66% to 81%) specificity (AUC 0.73, 95% CI 0.67 to 0.79). Sensitivity for detecting BO increased with length of the BO segment (M ≥3 cm) to 96% (95% CI 90% to 99%). To assess the robustness of the model and the effect of confounding factors, sensitivity analyses were performed including only patients using proton pump inhibitors (PPIs) for at least 1 month before study enrolment (n=300) and only patients with a hiatal hernia (n=205). Both analyses still showed a fair to good diagnostic accuracy for discriminating BO from GORD patients and controls (table 2).

    Figure 3
    Figure 3

    Scatter plot and receiver operating characteristic (ROC) curve of the cross-validated disease-specific model for detecting Barrett’s oesophagus (BO) based on 129 breath tests of patients with BO and 273 breath tests of gastro-oesophageal reflux disease (GORD) and control patients. (A) Scatterplot with optimal cut-off value. Values between −1 and +1 indicate the degree to which a subject was classified as positive (maximum value +1) or negative (minimum value −1) for BO. (B). ROC curve for detecting BO. AUC,area under the curve.

    View this table:
    Table 2

    Diagnostic performance of 10-fold cross-validated VOC analyses of breath samples of Barrett’s oesophagus, GORD patients and controls

    Comments

    The limitations of the currently recommended screening method for BO, upper endoscopy, have initiated the development of novel, minimally invasive screening techniques that can be performed in a primary care setting.10 11 This proof-of-concept study indicates that exhaled breath VOCs hold great potential for the non-invasive identification of patients most likely to have BO, thus at higher risk of developing OAC. Hence, it could be possible to reduce mortality from OAC in patients at risk, who otherwise would not have undergone upper endoscopy.

    The findings of our study provide evidence that patients with BO have VOC breath prints different from those without BO. We showed as a proof of concept that it is possible to non-invasively detect the presence of BO by VOC breath analysis using an electronic nose device in patients with and without GORD with a sensitivity of 91% and a specificity of 74%. The accuracy increases with BO segment length. Models were still able to differentiate between BO and controls in sensitivity analyses, which indicates that VOC analyses were only minimally affected by confounding factors such as PPI use, GORD symptoms, and the presence of a hiatal hernia. Furthermore, the observed slight decrease in accuracy could also be attributed to the relatively low number of patients included in these sensitivity analyses.

    To date, only three studies—one using VOC pattern recognition and two using mass spectrometric analysis—have been reported on VOC analysis in BO or OAC, showing a good accuracy for detecting either of them.12–14 Although evidence for VOC pattern recognition for BO is limited, analysis of VOCs in exhaled breath has been shown to be a potential diagnostic biomarker for various types of cancer, but also for the detection of several gastrointestinal diseases.15 16 Hence, exhaled breath seems to contain biomarkers with clear diagnostic potential, a finding we reproduced in this study aiming at the early detection of BO. The sensitivity and specificity of breath testing for BO are comparable to existing technologies such as mammography for breast cancer screening, faecal occult blood testing for colorectal cancer, and the cytosponge for BO.17–19

    The exact biochemical mechanism of exhaled VOCs and the connection with cancer and related neoplastic lesions is still under debate. It is thought that differences in patterns of VOCs may involve changes at a genetic and cellular level causing metabolic alterations of the BO mucosa, but also changes in the local microenvironment.5 20 21 Oesophageal microbiome changes associated with BO may also be a contributing factor to the production of VOCs.22

    VOC analysis based on pattern recognition of disease-specific VOC profiles is only one of the methods used for breath sampling. Other methods focus on identification and quantification of specific individual VOCs for diseases of interest using chemical analytical techniques such as gas chromatography-mass spectrometry. Although the quality is high, chemical analytical techniques are expensive, time-consuming and require well-trained personnel for its operation, analysis and maintenance.23 Therefore, electronic nose technology seems better suited for population-based screening than chemical analytical techniques, because it is fast, potentially cost-effective, and allows for real-time breath analysis results. Furthermore, the electronic nose type used in this study offers the opportunity for transferring calibration models.9 This implies that once a calibration model for BO has been developed, it can easily be transferred to other electronic nose devices and therefore enables large-scale application.

    A limitation of electronic nose technology is the inability to identify individual VOC biomarkers at a molecular level. Other limitations include the inability to correct for the influence of various exogenous factors and unmeasured confounding variables that could have influenced VOC patterns. Furthermore, sensor drift over time, although adjusted for in data analysis, may cause reproducibility issues.24 Future studies should focus on standardisation of all phases of VOC analysis based on characteristics most optimal for screening purposes, including optimisation of sensor arrays most sensitive to detect BO-specific VOCs, sampling protocols, and data analysis adjusted for technical and (patho)physiological confounders.25 Despite the promising results of this cross-validated training model, it is important to bear in mind that future research is needed to externally validate the diagnostic accuracy on independent patient cohorts that have not been previously exposed to the artificial neural network. Hence, a large external validation study is currently ongoing to test the developed training model on new patients and to assess reproducibility and intra-individual variation of breath prints. Finally, the case–control setting of this study makes it possible to obtain accurate test accuracy data but has limitations regarding generalisability to a primary care population consulting for GORD symptoms. Therefore, a multi-site validation study of breath testing for patients with chronic GORD symptoms in primary care is in preparation.

    In conclusion, this portable electronic nose is able to detect the presence or absence of BO in patients with and without GORD with good diagnostic accuracy. Given the high tolerability, high acceptability and low costs, breath testing may be a promising approach to be used for non-invasive screening for BO in a primary care setting.

    References

    1. ↵
      2. Arnold M ,
      3. Laversanne M ,
      4. Brown LM , et al
      . Predicting the future burden of esophageal cancer by histological subtype: international trends in incidence up to 2030. Am J Gastroenterol 2017;112:1247–55.doi:10.1038/ajg.2017.155
      OpenUrl
    2. ↵
      2. Smyth EC ,
      3. Lagergren J ,
      4. Fitzgerald RC , et al
      . Oesophageal cancer. Nat Rev Dis Primers 2017;3.doi:10.1038/nrdp.2017.48
    3. ↵
      2. Peters Y ,
      3. Al-Kaabi A ,
      4. Shaheen NJ , et al
      . Barrett oesophagus. Nat Rev Dis Primers 2019;5.doi:10.1038/s41572-019-0086-z
    4. ↵
      2. Haick H ,
      3. Broza YY ,
      4. Mochalski P , et al
      . Assessment, origin, and implementation of breath volatile cancer markers. Chem Soc Rev 2014;43:1423–49.doi:10.1039/C3CS60329F
      OpenUrlCrossRefPubMed
    5. ↵
      2. Filipiak W ,
      3. Mochalski P ,
      4. Filipiak A , et al
      . A compendium of volatile organic compounds (VOCs) released by human cell lines. Curr Med Chem 2016;23:2112–31.doi:10.2174/0929867323666160510122913
      OpenUrl
    6. ↵
      2. Buck L ,
      3. Axel R
      . A novel multigene family may encode odorant receptors: a molecular basis for odor recognition. Cell 1991;65:175–87.doi:10.1016/0092-8674(91)90418-X
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      2. Santos JP ,
      3. Lozano J ,
      4. Aleixandre M
      . Electronic noses applications in beer technology. Brewing Technology 2017.
    8. ↵
      2. Kort S ,
      3. Brusse-Keizer M ,
      4. Gerritsen J-W , et al
      . Data analysis of electronic nose technology in lung cancer: generating prediction models by means of Aethena. J Breath Res 2017;11:026006.doi:10.1088/1752-7163/aa6b08
    9. ↵
      2. Bruins M ,
      3. Gerritsen JW ,
      4. van de Sande WWJ , et al
      . Enabling a transferable calibration model for metal-oxide type electronic noses. Sensors and Actuators B: Chemical 2013;188:1187–95.doi:10.1016/j.snb.2013.08.006
      OpenUrl
    10. ↵
      2. Fitzgerald RC ,
      3. di Pietro M ,
      4. Ragunath K , et al
      . British Society of |Gastroenterology guidelines on the diagnosis and management of Barrett's oesophagus. Gut 2014;63:7–42.doi:10.1136/gutjnl-2013-305372
      OpenUrlAbstract/FREE Full Text
    11. ↵
      2. Spechler SJ ,
      3. Sharma P ,
      4. Souza RF , et al
      . American Gastroenterological Association technical review on the management of Barrett's esophagus. Gastroenterology 2011;140:e18–52.doi:10.1053/j.gastro.2011.01.031
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      2. Kumar S ,
      3. Huang J ,
      4. 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.doi:10.1097/SLA.0000000000001101
      OpenUrlCrossRefPubMed
    13. ↵
      2. Markar SR ,
      3. Wiggins T ,
      4. Antonowicz S , et al
      . Assessment of a noninvasive exhaled breath test for the diagnosis of oesophagogastric cancer. JAMA Oncology 2018;4:970–6.doi:10.1001/jamaoncol.2018.0991
      OpenUrl
    14. ↵
      2. Chan DK ,
      3. Zakko L ,
      4. Visrodia KH , et al
      . Breath testing for Barrett’s esophagus using exhaled volatile organic compound profiling with an electronic nose device. Gastroenterology 2017;152:24–6.doi:10.1053/j.gastro.2016.11.001
      OpenUrlCrossRef
    15. ↵
      2. Krilaviciute A ,
      3. Heiss JA ,
      4. Leja M , et al
      . Detection of cancer through exhaled breath: a systematic review. Oncotarget 2015;6:38643–57.doi:10.18632/oncotarget.5938
      OpenUrl
    16. ↵
      2. Garner CE ,
      3. Smith S ,
      4. de Lacy Costello B , et al
      . Volatile organic compounds from feces and their potential for diagnosis of gastrointestinal disease. FASEB J 2007;21:1675–88.doi:10.1096/fj.06-6927com
      OpenUrlCrossRefPubMedWeb of Science
    17. ↵
      2. Gøtzsche PC ,
      3. Jørgensen KJ , Cochrane Breast Cancer Group
      . Screening for breast cancer with mammography. Cochrane Database Syst Rev 2013;156.doi:10.1002/14651858.CD001877.pub5
    18. ↵
      2. Lin JS ,
      3. Piper MA ,
      4. Perdue LA , et al
      . Screening for colorectal cancer: updated evidence report and systematic review for the US preventive services task force. JAMA 2016;315:2576–94.doi:10.1001/jama.2016.3332
      OpenUrlCrossRefPubMed
    19. ↵
      2. Ross-Innes CS ,
      3. Debiram-Beecham I ,
      4. O'Donovan M , et al
      . Evaluation of a minimally invasive cell sampling device coupled with assessment of trefoil factor 3 expression for diagnosing Barrett's esophagus: a multi-center case–control study. PLoS Med 2015;12:e1001780.doi:10.1371/journal.pmed.1001780
    20. ↵
      2. Masoudi-Nejad A ,
      3. Asgari Y
      . Metabolic cancer biology: structural-based analysis of cancer as a metabolic disease, new sights and opportunities for disease treatment. Semin Cancer Biol 2015;30:21–9.doi:10.1016/j.semcancer.2014.01.007
      OpenUrlCrossRefPubMed
    21. ↵
      2. Lin EW ,
      3. Karakasheva TA ,
      4. Hicks PD , et al
      . The tumor microenvironment in esophageal cancer. Oncogene 2016;35:5337–49.doi:10.1038/onc.2016.34
      OpenUrlCrossRefPubMed
    22. ↵
      2. Yang L ,
      3. Lu X ,
      4. Nossa CW , et al
      . Inflammation and intestinal metaplasia of the distal esophagus are associated with alterations in the microbiome. Gastroenterology 2009;137:588–97.doi:10.1053/j.gastro.2009.04.046
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      2. Arasaradnam RP ,
      3. Covington JA ,
      4. Harmston C , et al
      . Review article: next generation diagnostic modalities in gastroenterology - gas phase volatile compound biomarker detection. Aliment Pharmacol Ther 2014;39:780–9.doi:10.1111/apt.12657
      OpenUrlCrossRefPubMed
    24. ↵
      2. Wilson A
      . Application of electronic-nose technologies and VOC-biomarkers for the noninvasive early diagnosis of gastrointestinal diseases. Sensors 2018;18.doi:10.3390/s18082613
    25. ↵
      2. Bikov A ,
      3. Lázár Z ,
      4. Horvath I
      . Established methodological issues in electronic nose research: how far are we from using these instruments in clinical settings of breath analysis? J Breath Res 2015;9:034001.doi:10.1088/1752-7155/9/3/034001

    Footnotes

    • Contributors YP: study concept and design, patient inclusion, data acquisition, quality control of data and algorithms, data analysis and interpretation, statistical analysis, manuscript preparation, editing, and review; RWMS: contributed to study design, patient inclusion, data acquisition, critical revision of the manuscript. AT: contributed to study design, patient inclusion, data acquisition, critical revision of the manuscript. SKB: contributed to study design, patient inclusion, data acquisition, critical revision of the manuscript. BdJ: contributed to study design, patient inclusion, data acquisition, critical revision of the manuscript. PDS: study concept and design, data interpretation, manuscript editing, and review, study supervision.

    • Funding This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. The four Aeonose devices were supplied by the eNose Company (Zutphen, the Netherlands) at no costs.

    • Competing interests The Aeonose™ devices are provided by the eNose Company (Zutphen, the Netherlands) for use in research. No additional funding for this study was received and all research was investigator initiated.

    • Patient consent for publication Not required.

    • Ethics approval The study protocol was approved by the research ethics committee of the Radboud University Nijmegen Medical Centre, the Netherlands, and thereafter by all local ethics committees of the participating centres. This study was performed in compliance with the Declaration of Helsinki and registered at the Netherlands Trial Register (Identifier NL6750 (NTR6928)).

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