eNose in the diagnosis of gastric cancer

Gastric cancer (GC), a common malignancy, is the fifth most prevalent cancer worldwide and constitutes the fourth leading cause of cancer-related deaths. Early diagnosis of GC is crucial for improving patient survival rates. However, some patients are asymptomatic in the early stage and often progress to advanced stages at the time of clinical diagnosis. In addition, imaging techniques used clinically for the screening and diagnosis of GC, such as barium meal and abdominal computed tomography, are beset by disadvantages such as high costs, radiation damage, and low sensitivities and specificities. Although endoscopy with pathological biopsy is the most reliable method for diagnosing GC and plays a crucial role in disease screening, its widespread use is limited by invasiveness and technical expertise requirements. Exhaled VOC analysis is a cost-effective and easy-to-operate noninvasive assay with no adverse physical effects on patients. Breath VOCs have been extensively studied nationally and internationally, revealing their potential as a detection method for GC (Polaka et al. 2022). Breathomics is a branch of metabolomics that functions as a diagnostic aid by identifying and quantifying specific VOCs or VOC patterns associated with various diseases or physiological conditions resulting from disease-induced changes in metabolic processes (Daniel and Thangavel 2016; Einoch Amor et al. 2019). Currently, breath analysis is successfully used to diagnose lung, breast, gastric, prostate, colorectal, ovarian, head and neck, kidney, and bladder cancers (Peng et al 2010; Yang et al. 2021). Sensor-based gas chromatography has found significant application prospects in scientific research and clinical practice for the early detection of GC because of its low cost, noninvasiveness, high accuracy, and ease of operation (Miekisch et al. 2004; Amal et al. 2014; Haddad et al. 2020; Zhang et al. 2021; Xiang et al. 2021; Gouzerh et al. 2022). The application of eNose technology can help in the early detection of pre-cancerous lesions and ultimately reduce GC mortality. eNose performance depends on its sensitivity, accuracy, specificity, and predictive values (Shreffler and Huecker 2020; Yang et al. 2021). With advancements in technology, breath analysis methods may become critical for the detection of GC, providing more intensive information on the progression, specific VOCs, origins, and biochemical mechanisms underlying GC. VOCs contain valuable information regarding biochemical metabolism in cancer (Van Der Schee et al. 2018). Some compounds are associated with specific cancers and can be used to distinguish patients from healthy individuals (Xiang et al. 2021). Aldehydes and ketones are slightly soluble in blood and can be identified in the breath a few minutes after their release from tissues (Haick et al. 2014).

Oxidative stress, a major source of non-branched hydrocarbons in the body, causes lipid peroxidation of polyunsaturated fatty acids in cell membranes, leading to the generation of saturated alkanes, C3-C11 hydrocarbons such as ethane and pentane (Okunieff et al. 2005; Vousden and Ryan 2009). These cells are affected by pre-cancerous lesions or cells elsewhere in the body due to systemic oxidative stress. Since different cell types have different cell membranes, the VOCs emitted also differ.

Schuermans et al. (2018) investigated a breath test based on a miniature metal oxide gas sensor on exhaled breath samples to distinguish between patients with GC and healthy individuals. They used discriminant factor analysis (DFA) pattern recognition to develop a prediction model. The baseline attributes differed significantly only by age, with a mean age of 37 and 57 years for the healthy and patient groups, respectively (p = 0.000). Weight loss was the only symptom that showed a significant difference (p = 0.040). The study included 16 patients and 28 controls, of whom 13 were true positives and 20 true negatives. The sensitivity, specificity, and accuracy of the receiver operating characteristic (ROC) curve were 81%, 71%, and 75%, respectively. The positive and negative predictive values were 62% and 87%, respectively. This preliminary study suggested that eNose can be used to diagnose GC based on exhaled gases, showing promise as a predictive tool for GC screening.

Xu et al. (2013) demonstrated that a nanomaterial-based sensor effectively distinguished between patients with GC (n = 37) and nonmalignant gastric disease (n = 93) by developing a DFA model and analyzing 130 breath samples. The sensitivity and specificity for differentiating between GC and benign gastric disease were 89% and 90%, respectively. The sensitivity and specificity for distinguishing early GC (stages I and II) from advanced GC (stages III and IV) were 89% and 94%, respectively. The sensitivity and specificity for distinguishing benign ulcers from less severe gastric diseases (including 32 cases without an anomaly on gastroscopy and 29 with anomalies on gastroscopy but without an ulcer) were 84% and 87%, respectively.

Amal et al. (2016) collected 968 breath samples from 484 patients (including 99 with GC) for two analyses. The first sample was analyzed using gas chromatography–mass spectrometry with a multiple-corrected t-test (p < 0.017), whereas the second one was subjected to a cross-reactive nanoarray combined with pattern recognition. For the latter, the randomly selected training set comprised 70% of the sample, and the remaining 30% formed the validation set. The presence or absence and the risk level of pre-cancerous lesions were stratified using the Operative Link on Gastric Intestinal Metaplasia (OLGIM) assessment staging system. Patients with OLGIM stages III and IV were considered be at high risk. Based on the gas chromatography–mass spectrometry results, patients with cancer and high-risk patients had a unique composition of breath fingerprints. Eight significant VOCs were detected in the exhaled gases (p = 0.017). In contrast, nanoarray analysis revealed that the sensitivity, specificity, and accuracy for identifying patients with GC and controls (OLGIM stages 0–IV) were 73%, 98%, and 92%, respectively. The sensitivity, specificity, and accuracy of classification were 97%, 84%, and 87%, respectively, when comparing GC with OLGIM stages 0–II, and 93%, 80%, and 90%, respectively, when comparing GC and OLGIM stages III–IV. However, the sensitivity, specificity, and accuracy for the combination of OLGIM stages I–II, III–IV, and heterogeneous hyperplasia were 83%, 60%, and 61%, respectively. Consequently, nanoarray analysis may serve as a noninvasive screening and monitoring technique for GC and relevant pre-cancerous lesions.

eNoses in the diagnosis of Barrett’s esophagus

Barrett's esophagus (BE) is a precancerous esophageal lesion, with the potential to develop into an adenocarcinoma upon malignant transformation. Thus, awareness of this condition and its early detection, appropriate treatment, and follow-up should be more widespread. Screening and monitoring for BE are aimed at early detection and reducing the mortality rate of esophageal adenocarcinoma. BE is characterized by the replacement of squamous epithelia in the distal esophagus with metaplastic (intestinal-type) epithelia due to gastroesophageal reflux. Endoscopy with pathological biopsy is the gold standard for diagnosing and monitoring BE. The annual incidence rate of esophageal adenocarcinoma in patients with BE is 0.3–0.6%, and the prevalence rate is approximately 1–2% in the general population (Boeckxstaens et al. 2014; Hayeck et al. 2010; Dumoulin et al. 2022). As most patients with BE are asymptomatic, its true prevalence rate may be underestimated. Changes that occur during monitoring intervals depend on the extent of atypical hyperplasia, and endoscopic eradication therapy is limited to patients with BE and confirmed atypical hyperplasia. The current guidelines recommend endoscopic screening and monitoring based on various risk factors; however, these factors are limited by invasiveness, availability of experienced specialists, and the physical, psychological, and economic burden on the patient. Transnasal endoscopy is a less invasive approach with similar limitations, such as the need for trained specialists and high costs.

In contrast, non-endoscopic methods require minimal intervention, can be performed in the consultation room, and are potentially a more desirable option for large-scale public screening and monitoring. The analysis of VOCs in exhaled gases may be a promising technique for detecting undiagnosed BE. Relevant studies have been reported but are inadequate, necessitating further research for confirmation.

In 2020, Peters et al. (2020) obtained breath samples from 513 patients and observed no adverse events. Overall, 402 patients were included in the study, with 129 diagnosed with BE, 141 with gastroesophageal reflux disease [including 50 (35.5%) with reflux esophagitis], and 132 in the control group. In the control group, 76 patients (57.6%) had a normal upper gastrointestinal tract or hiatal hernia on endoscopy. The investigators developed and cross-validated a BE prediction model to analyze the VOCs. This eNose could differentiate between patients with and without BE with good diagnostic accuracy [sensitivity, 91%; specificity, 74%; area under the ROC curve (AUC), 0.91] and seemed to be independent of the use of proton pump inhibitors, hiatal hernia, and reflux. Therefore, eNose may be an efficient, well-tolerated, sensitive, and specific screening method, allowing high-risk individuals to be selected for upper gastrointestinal endoscopy.

eNoses in the diagnosis of colorectal cancer

Colorectal cancer (CRC) is the third most frequent malignancy and the foremost contributor to cancer-related mortality. Endoscopic biopsy remains the primary diagnostic method for gastrointestinal tumors. Patients may remain asymptomatic at early or advanced stages of CRC (Chow et al. 1996; Pan and Morrison 2011; Desmond et al. 2019; Park et al. 2020). Because the symptoms of early-stage CRC are nonspecific, the diagnostic rate of endoscopy is suboptimal, and cancer screening is expensive, painful, and unsuitable. Therefore, there is an urgent need for convenient, noninvasive, and low-cost diagnostic methods for the early diagnosis and screening of cancer. Fecal occult blood tests, serum biomarkers, and intestinal barium contrast X-ray angiography are commonly used diagnostic methods for CRC. The fecal occult blood test is currently the most widely used and evaluable method for screening. However, its clinical value is limited because of its high false-positive and false-negative rates. Because of their poor accuracy, serum biomarkers of intestinal tumors, such as carcinoembryonic antigen and cancer antigen 19–9, do not fulfill the expected diagnostic role. While barium contrast X-ray angiography can depict the lesion’s overall location, size, and anatomical relationship with the entire organ, it is radioactive and cumbersome. Therefore, noninvasive biomarkers for the diagnosis of intestinal cancers are needed. The overall declining trend in CRC-related mortality rates is likely due to increased screening, early detection, and improved treatment regimens (Huang et al. 2022). Recently, VOCs have been considered as potential biomarkers of CRC. Therefore, using biomarkers for early detection, diagnosis, and staging is crucial for cancer treatment (Majumdar et al. 1999; Haick et al. 2014; Ogunwobi et al. 2020; Chung et al. 2022).

Peng et al. (2010) first attempted to detect CRC by analyzing the presence of VOCs during exhalation. The results showed that a nanosensor array could differentiate between patients with colon cancer and healthy controls and the breathing conditions in patients with different cancer types, regardless of age, sex, lifestyle, and other confounding factors.

de Meij et al. (2014) used an eNose to assess the odors of disease-specific VOCs in fecal gases to distinguish patients with CRC or advanced adenoma from healthy controls. Stool samples were collected from patients scheduled for elective colonoscopies. The patterns of VOCs in the fecal gases of patients with histopathologically confirmed CRC, those with histopathologically confirmed advanced adenoma, and controls (no anomaly on colonoscopy) were detected using eNose. The CRC and advanced adenoma detection performance was evaluated using ROC curves and calculating the sensitivity and specificity. A total of 157 stool samples (40 from patients with CRC, 60 from those with advanced adenoma, and 57 from healthy controls) were analyzed using eNose. The distribution of stool VOCs in patients with CRC differed significantly from that in the control group [AUC ± 95% confidence interval (CI) 0.92 ± 0.03; p < 0.001; sensitivity, 85%; specificity, 87%].

In addition, the VOC profile of patients with advanced adenoma was distinguishable from that of the control group (AUC ± 95% CI 0.79 ± 0.04; p < 0.001; sensitivity, 62%; specificity, 86%). These results imply that fecal gas analysis using eNose is a promising novel screening tool for the early detection of advanced neoplasia and CRC.

van Keulen et al. (2020) collected 511 breath samples. Overall, 64 patients were excluded from the study owing to unsatisfactory breath testing (n = 51), incomplete colonoscopy (n = 8), or colitis (n = 5). Patients were classified according to the most advanced lesions, viz. CRC (n = 70), advanced adenoma (n = 117), non-advanced adenoma (n = 117), hyperplastic polyps (n = 15), or no anomalies on colonoscopy (n = 125). The AUC was 0.76 for CRC and 0.71 for advanced adenoma. The AUCs for CRC and advanced adenoma obtained by blinded validation were 0.74 and 0.61, respectively; the AUCs generated by the CRC and advanced adenoma models were 0.84 (sensitivity, 95%; specificity, 64%) and 0.73 (sensitivity, 79%; specificity, 59%), respectively. This study suggested that exhaled VOCs are potential noninvasive biomarkers for detecting CRC and advanced adenoma. Future studies should include larger samples to improve the potential of VOC analysis for identifying malignant colorectal lesions.

Tyagi et al. (2021) used eNose and gas chromatography–mass spectrometry to differentiate between the CRC group and non-cancer group based on their chemical fingerprints, revealing that eNose had good sensitivity and specificity. Using a neural network classifier, eNose could distinguish between the CRC and non-cancer groups, with an AUC of 0.81, high sensitivity of 91%, and specificity of 55%. Analysis of the CRC and non-cancer groups using a random forest classifier yielded an AUC of 0.80, sensitivity of 82%, and specificity of 55%.

eNoses in the diagnosis of inflammatory bowel disease

Inflammatory bowel disease (IBD) is an idiopathic, chronic, and nonspecific inflammatory intestinal disorder that can be classified into ulcerative colitis (UC) and Crohn’s disease (CD). The etiology and pathogenesis, which involve genetic susceptibility, environmental triggers (such as diet and lifestyle), and effects on the host microbiome, remain unclear. Bacterial diversity is difficult to study, because only 50% of organisms can be successfully cultured. Although modern genomic technologies can circumvent this problem, they are costly and laborious, making them difficult to adapt to routine clinical use. Among patients with IBD, the endoscopic disease activity level is associated with poor outcomes, and endoscopy remains the most reliable test for evaluating symptomatic patients. However, endoscopy is invasive and imposes physical, psychological, and financial burdens on patients, highlighting the need for noninvasive biomarkers for IBD diagnosis.

Arasaradnam et al. (2013) first used eNose and field asymmetric ion mobility spectrometry (FAIMS) to detect VOCs in the urine of patients with IBD and successfully generated a characteristic chemical fingerprint. They recruited 62 study participants, including 48 patients with IBD (24 with CD and 24 with UC) and 14 healthy controls. The disease activity of the study participants was recorded, and urine samples were collected. The urine samples were analyzed for VOCs using eNose and FAIMS. The eNose data obtained from the experiment were analyzed, and the results showed that eNose could accurately distinguish between patients with IBD and healthy controls, with an accuracy of 0.75% (p-value < 0.001).

Tiele et al. (2019) used eNose and a commercial gas chromatography-ion mobility spectrometer to examine the breath samples of patients. The study enrolled 39 participants: 14 were diagnosed with CD, 16 with UC, and 9 served as controls. Both methods could distinguish patients with IBD from controls, with eNose technology having an AUC ± 95% CI of 0.81 ± (0.66–0.96), sensitivity of 67%, and specificity of 89%. In addition, this method could differentiate UC from CD, with eNose technology having an AUC ± 95% of 0.88 ± (0.77–0.98), sensitivity of 71%, and specificity of 88%.