Developing a Screening Test for Barrettʼs Esophagus Using Electronic Nose Device Analysis of Exhaled Volatile Organic Compounds: 2017 Presidential Poster Award

Abstract

Introduction: The prognosis for advanced esophageal adenocarcinoma (EAC) remains dismal. However, while persons with Barrett's esophagus (BE) (the only known precursor of EAC) undergo surveillance, over 90% of EACs are diagnosed outside of surveillance programs. This suggests current screening and surveillance programs are inadequate. Currently available BE screening modalities are expensive and intrusive. In an effort to develop a highly acceptable, less expensive screening method, we piloted the use of a non-invasive, highly portable electronic nose device to detect BE from breath samples. Here we present the development of a screening test for Barrett's esophagus using a second generation electronic nose device. Methods: Patients provided 5-minute breath samples into a second generation Aeonose electronic nose (Aeonose, eNose Company, Zutphen, NL). Breath samples were collected from two groups: Group 1 (Group BE) was subjects with untreated BE (≥ 1 cm of columnar mucosa from the gastroesophageal junction with histopathologic confirmation of intestinal metaplasia); Group 2 (Group no BE) were subjects undergoing EGD who did not have BE and had no history of BE. The Aeonose device analyzes a breath sample over 5-minutes. During this time the patient's breath is circulated over a metal-oxide sensor array. VOC's in the breath sample cause reduction-oxidation reactions in the sensor array. These reactions cause changes in the conductance of circuits that run through the sensor and in the temperature of the sensor array. The Aeonose device measures these changes over time to create a 3-D digital signature of each breath sample. These digital signatures are uploaded into an artificial neural network in to perform pattern recognition analysis between breath samples from Group BE versus Group no BE. Results: Breath samples were obtained from a total of 58 individuals; 27 in Group BE and 31 in Group No BE. Optimal multivariate modeling based on our VOC signatures demonstrated a predictive model for BE with performance characteristics of 89% sensitivity, 71% specificity, 79% accuracy, and AUC=0.81 (Figure 1). The positive predictive value was 0.73 and negative predictive value was 0.88. The Matthews Correlation Coefficient was 0.60. The recruitment rate was 99% for breath testing.FigureConclusion: The second-generation Aeonose electronic-nose provides good discrimination between persons with BE and those without BE. The high sensitivity, low cost, non-invasive nature of electronic nose testing, and high rate of patient usage suggest the Aeonose device can be used as a screening tool to identify patients with BE for further surveillance. Further work to refine the model and improve the tests efficiency is ongoing. A validation study is also in progress.