Feasibility of volatile organic compound in breath analysis in the follow-up of colorectal cancer: A pilot study

Abstract

Background

Colorectal carcinoma (CRC) has a worldwide incidence of 1.4 million patients and a large share in cancer-related mortality. After curative treatment, the risk of recurrence is 30–65%. Early detection may result in curative treatment. However, current follow-up (FU) examinations have low sensitivity ranging from 49 to 85% and are associated with high costs. Therefore, the search for a new diagnostic tool is justified. Analysis of volatile organic compound in exhaled air through an electronic nose (eNose) is a promising new patient-friendly diagnostic tool. We studied whether the eNose under investigation, the Aeonose™, is able to detect local recurrence or metastases of CRC.

Methods

In this cross-sectional study we included 62 patients, all of whom underwent curative treatment for CRC in the past 5 years. Thirty-six of them had no metastases and 26 had extraluminal local recurrence or metastases of CRC, detected during FU. Breath testing was performed and machine learning was used to predict extraluminal recurrences or metastases, and based on the receiver operating characteristics (ROC)-curve both sensitivity and specificity were calculated.

Results

The eNose identified extra luminal local recurrences or metastases of CRC with a sensitivity and specificity of 0.88 (CI 0.69–0.97) and 0.75 (CI 0.57–0.87), respectively, with an overall accuracy of 0.81.

Discussion

This eNose may be a promising tool in detecting extraluminal local recurrences or metastases in the FU of curatively treated CRC. However, a well-designed prospective study is warranted to show its accuracy and predictive value before it can be used in clinical practice.

Introduction

Colorectal carcinoma (CRC) is one of the most common malignancies and a major cause of cancer-related mortality in developed countries. The worldwide incidence is estimated at almost 1.4 million patients and close to 700 thousand patients die each year as a result of CRC [1]. After curative treatment, the risk of recurrence is 30–65%, depending mostly on the stage of the primary tumour [2,3].

Nowadays, treatment for recurrent local and metastatic disease is increasingly successful and improves survival [3]. There is yet no consensus on the best method for follow-up (FU) [[2], [3], [4]]. In general, patients curatively treated for CRC undergo surveillance examinations for metastases for up to 5 years after treatment. Early detection of non-symptomatic recurrence or metastases increases the likelihood of a curative treatment and reduces morbidity [2]. Most guidelines recommend combining measurement of Carcino-Embryonic-Antigen (CEA) every 3–12 months, liver imaging by means of ultrasound or computed tomography (CT) and routinely chest X-ray or chest CT [3]. Sensitivity of these tests is generally low. For instance, the sensitivity of CEA for the detection of local recurrences is 60% and sensitivity of ultrasound and abdominal CT scan for the detection of liver metastases is 57% and 68%, respectively [5]. These tests lack diagnostic accuracy and cause significant expenses to society as a whole. Also, with progress in treatment of secondary cancers of CRC, schemes of FU are under debate. The ‘CEAwatch’ trial [6], compared a normal FU protocol, with a protocol in which frequency of CEA measurements was intensified to every 2 months. Imaging was performed in patients with two subsequent rises in CEA. Intensified FU with CEA resulted in detecting recurrences in an earlier state with less time-to-detection and more curable options. The COLOFOL Randomized Clinical Trial, compared 5-year overall mortality and CRC-specific mortality rates for patients with an intensive FU-schedule with patients with a less intensive FU-schedule. They concluded that there was no difference in both 5-year overall mortality and cancer-specific mortality between both groups [7].

Currently, FU methods and schemes are widely debated, justifying the search for better diagnostic tools to detect metastases after curative resection of CRC.

Analysis of exhaled air might be a suitable diagnostic and screening tool for early detection of CRC recurrence. Each individual has been shown to have a personal ‘breathprint’, comparable to a fingerprint, which is a reflection of health. These breathprints consist of volatile organic compounds (VOCs), which are gaseous products of metabolism [[8], [9], [10]]. Uninhibited growth, necrosis and alteration of the intestinal microbioma are but some of the mechanisms that arise when colorectal carcinoma is present. These mechanisms are known to alter breathprints [11,12]. Small pilot studies have identified specific VOC changes in patients with CRC [[12], [13], [14]].

There are two main techniques in assessing VOCs in exhaled air. One is by chemical analysis of specific compounds by means of gas chromatography with mass spectometry (GC-MS), while pattern recognition with machine learning techniques such as electronic nose devices (eNose) may be able to identify integrated VOC-profiles [[8], [9], [10]]. Using GC-MS, alterations in VOCs were found in patients with malignancies, such as CRC and pancreatic cancer [14,15]. eNoses have been shown to be able to be trained to recognise specific breath patterns, using machine learning [13,[16], [17], [18], [19], [20], [21], [22]].

Clinical studies have shown the potential of eNose in detecting CRC by analysing exhaled air [12,13,[16], [17], [18]]. Van Keulen et al. even suggest that advanced adenomas have specific breath patterns, which can be distinguished from CRC [12].

Furthermore, a future role for eNose is shown in detecting many other tumours (i.e. lung-, prostate-, gastric- and head and neck cancer) [[19], [20], [21], [22]]. In 2019, a systematic review and meta-analysis was conducted in order to assess the accuracy of VOC-based tests in cancer diagnostics. Although standardisation of breath collection methods and validation studies need to be done before actual implementation in common-practice, the authors believed that analysis of breathprints holds a promising diagnostic tool. They pointed out possible value of breath analysis as a screening tool, amongst others [23].

To date, limited studies have been conducted in order to determine the potential of eNose in detecting recurrent cancer. The diagnostic accuracy in differentiating patients with locoregional recurrent or metachronous head and neck squamous cell carcinoma from healthy patients was reported to be 83% [24]. Furthermore, Markar et al. [14] recently published data suggesting an association between propanal, a specific VOC and recurrence of CRC. Using Selected Ion Flow-Tube Mass Spectometry they were able to identify recurrent CRC with a sensitivity and specificity of 71.4% and 90.0% respectively. Although this study used Mass Spectometry, the results strongly imply that changes occur in breathprints in persons with recurrence of CRC. Electronic nose devices might be able to recognise these changed patterns.

Electronic nose devices have been shown to detect CRC, while Mass Spectometry techniques identified concentrations of propanal to be variable amongst patients with secondary CRC. Combining those study results, we hypothesised eNoses could also be a tool for detecting recurrent disease after treatment of CRC. If so, eNose is a noninvasive, low-cost diagnostic tool, with minimal burden for the patient, preferable over GC-MS technique or common FU examinations. Our aim was to determine whether the eNose is capable of detecting extra luminal local recurrences or metastases of CRC in patients that were treated for CRC in the past 5 years. To our knowledge, this is the first study using eNose in the follow-up of CRC.