Are Volatile Organic Compounds Accurate Markers in the Assessment of Colorectal Cancer and Inflammatory Bowel Diseases? A Review

Simple Summary Early diagnosis is crucial for reducing colorectal cancer-related mortality in both the general population and inflammatory bowel disease. Volatile organic compound (VOC) analysis is a promising alternative to the gold standard procedure, endoscopy, for early detection and surveillance of colorectal diseases. This review aimed to provide a general overview of the most recent evidence in this area on VOC testing in breath, stool, and urine samples. Abstract Colorectal cancer (CRC) is one of the leading causes of cancer-related death in the Western world. Early detection decreases incidence and mortality. Screening programs based on fecal occult blood testing help identify patients requiring endoscopic examination, but accuracy is far from optimal. Among the alternative strategies, volatile organic compounds (VOCs) represent novel potentially useful biomarkers of colorectal cancer. They also represent a promising tool for the screening of both intestinal inflammation and related CRC. The review is focused on the diagnostic potential of VOCs in sporadic CRC and in inflammatory bowel diseases (IBD), which increase the risk of CRC, analyzing future clinical applications. Despite limitations related to inadequate strength of evidence, differing analytical platforms identify different VOCs, and this unconventional approach for diagnosing colorectal cancer is promising. Some VOC profiles, besides identifying inflammation, seem disease-specific in inflammatory bowel diseases. Thus, breath, urine, and fecal VOCs provide a new and promising clinical approach to differential diagnosis, evaluation of the inflammatory status, and possibly the assessment of treatment efficacy in IBD. Conversely, specific VOC patterns correlating inflammatory bowel disease and cancer risk are still lacking, and studies focused on this issue are strongly encouraged. No prospective studies have assessed the risk of CRC development by using VOCs in samples collected before the onset of disease, both in the general population and in patients with IBD.


Introduction
Colorectal carcinoma (CRC), with an estimated incidence of 1.4 million patients/year, is one of the most common malignancies worldwide [1]. Prevention and early diagnosis reduce mortality, both in sporadic CRC and inflammatory bowel disease (IBD)-related CRC with a five-year survival approaching 90% [1]. Thus, identification and treatment of preneoplastic lesions in the general population and dysplasia in IBD are crucial. As bowel symptoms and fecal tests are non-specific [2], endoscopic screening represents the gold standard for early diagnosis of CRC [3]. However, being invasive, endoscopy is not Volatile organic compounds (VOCs) are produced during physiological and pathological processes including colonic inflammation and neoplasia. VOCs can also originate from exogenous sources such as food and drugs and microbiome metabolism. Once produced, VOCs are released into the gut and detectable in the intraluminal content or feces, or released into the bloodstream and then reach the pulmonary alveoli or renal tubules where they are excreted and therefore measurable in the exhaled breath and urine, respectively. IBD = inflammatory bowel disease; CRC = colorectal cancer.
The origin of individual VOCs is still a matter of debate, but largely depends on the interaction between diet and microbiota. The relationship is complex and bidirectional in normal subjects as gluten-free diet [17], ketogenic diets [18], low FODMAP diet [19] as well as alcohol intake [20] influence both the microbiome and VOC production. However, it is not known to what extent the various components of the diet favor the production of VOCs more than non-volatile organic anions such as SCFA and vice versa.
The VOC pattern results from complex interactions that differ in physiologic and pathologic conditions. The metabolic processes are indeed affected by mucosal changes, necrosis included, and the ensuing alteration of the intestinal microbiome. It has thus been anticipated that VOC analysis may prove helpful in different clinical settings including diagnosis and monitoring of disease course. Analysis of VOCs, being non-invasive, inexpensive, and well-tolerated, may represent an alternative to endoscopy [21]. A growing interest is thus focusing on conditions in which repeated endoscopy is required such as colorectal cancer screening, follow-up of benign preneoplastic lesions in the general population, and monitoring the disease course or identifying patients at high risk for dysplasia in IBD.
However, examining the potential clinical role of VOC analysis is complex as studies have been carried out using different analytical techniques. VOC analysis uses two main techniques: gas chromatography-mass spectrometry (GC-MS) and electronic noses. The first provides a chemical analysis of specific compounds, while the second needs to be "trained" to recognize specific breath patterns using machine learning [22,23].
An eNose consists of different arrays of electronic chemical sensors combined with an appropriate pattern-recognition system [24]. VOCs react on the surface of the sensors, inducing a change in conductivity that is detected by transducers and converted into electrical signals [24]. Several distinct eNose technologies have been developed including metal-oxide sensors (Aeonose, PEN3) [24], electrochemical and optical sensors (WOLF) [15], and conducting-polymer sensors (Cyranose 320) [25]. The different arrays and technologies, however, may result in a wide variety of chemical class coverage and accuracy. The identification and analysis of molecules by GC-MS are accurate but complex and timeconsuming. Thus, eNoses have recently been preferred in experimental studies.
This narrative review aims to provide a general overview of published evidence on VOCs analysis in CRC and IBD and critically analyze their clinical potential as well as future perspectives.

Materials and Methods
A systematic electronic search of the literature (without language or data publication restriction) was performed using Medline/PubMed, Embase, Scopus, and Web of Science up to February 2021. The search included a combination of medical subject headings (MeSH) and keywords as follows: Three authors (FV, MV, SF) independently reviewed the abstract and full text to assess the eligibility. Conflicts were resolved by consensus, referring to the original articles. The reference section of each relevant publication was also screened for other publications. Prospective and retrospective cohort studies, case-control studies, and analytical crosssectional studies were considered. In vitro and animal studies were excluded.
Out of 368 articles, 281 records were excluded based on their titles and abstracts, 42 for specific reasons, and 45 full text included ( Figure 2). The specific VOCs identified in the included studies are reported in Table 1 for breath exhaled compounds, and in the caption of each table for fecal and urinary VOCs.
Altomare et al. [33] reported that a profile of 15 VOCs identified with GC-MS could discriminate between CRC patients and healthy controls with an accuracy exceeding 80%. The same group recently reported that the use of new advanced breath samplers, capturing only the alveolar breath fraction and excluding environmental contaminants, improved the overall accuracy of 14 exhaled VOCs with a sensitivity and a specificity of 90 and 93%, respectively (Table 1) [26].
These results are in line with those obtained by a group using a GC-MS, which reported sensitivity of 85%, specificity of 94%, and accuracy of 91% [30].
Conversely, an Italian study using a commercial e-nose (PEN3™ eNose-Airsense Analytics GmbH, Schwerin, Germany), was unable to discriminate between CRC patients and healthy controls despite the higher detection of methane, methane derivatives, organic and aromatic compounds, in CRC patients [29]. More recently, however, good accuracy was reported using a different eNose (Aeonose™-The eNose Company, Zutphen, the Netherlands) [23], which showed an AUC of 0.84, with a sensitivity of 95% and specificity of 64%, respectively. Interestingly, the authors [23] also reported promising results in the detection of advanced adenomas with an AUC of 0.73, 79% sensitivity, and 59% specificity. The eNose was, however, unable to differentiate between CRC and advanced adenomas.
The different results obtained using different eNoses might derive from lower sensitivity to CRC-associated VOCs of PEN3™ eNose sensors than that of Aeonose™ sensors. There are, however, no studies directly comparing the accuracy in detecting CRC-associated breathprints of the two devices.
It has also been reported that breath analysis of VOCs can discriminate patients with adenomas from healthy controls [23,29,30] and that advanced adenomas show specific breath patterns, differing from those expressed in CRC [23]. However, these results must be confirmed, more so in the case of the eNoses, as each sensor is activated by several VOCs belonging to the same chemical class [23,29].
A single study [31] reported that the pattern of exhaled VOCs in CRC patients undergoes modification following cancer resection. The 31 identified VOCs performed very well in discriminating CRC patients before and after surgery and operated patients versus healthy controls. Interestingly, the VOC pattern exhaled by disease-free patients after surgery differed from that of healthy controls, probably to different metabolic pathways induced by surgery-and chemotherapy-induced oxidative stress [31]. The study, carried out in a small series of patients, could not identify changes in the VOC pattern in patients with recurrent or metastatic disease [31].
Conversely, a different research group [28] reported that levels of propanal reverted to normal following surgery, but subsequently increased in the case of CRC recurrence. Using this VOC, the sensitivity and specificity for the identification of recurrence were respectively 71.4% and 90.9% at a threshold of 28 ppbv [28].
Despite an overall good diagnostic accuracy, studies were carried out using different methodologies and identified different VOCs. Thus, the results are not suitable for direct comparison of diagnostic accuracy.
As far as aldehydes are concerned, Altomare and coworkers [33] using GC-MS, identified 15 VOCs significantly differing between CRC and the controls. All of them were nonanal and decanal, but the methodology employed did not allow for the measurement of the more volatile short-chain aldehydes. Propanal, a short-chain compound belonging to this class, has been recently reported to be statistically different between CRC and the controls [28]. Conversely, Amal was unable to identify CRC-specific aldehydes [30].
A specific and reproducible pattern of VOCs to be used in clinical practice are still far from being available.

Fecal VOCs
The efficacy of fecal VOCs in CRC has so far been reported in six studies only, using different approaches (Table 3).
Comparable results were reported using SIFT-MS, which differentiated patients affected by CRC and advanced adenoma from healthy controls with an accuracy of 75% (72% sensitivity and 78% specificity) [47]. More recently, the eNose Scent A1 performed even better in a relatively large cohort of patients (86 CRC and 71 controls) ( Table 3) [48].
Only two studies used GC-MS and were able to identify specific VOCs [49,50]. Bosch et al. [50] reported that fecal VOCs could discriminate CRC patients from the controls with an AUC of 0.96. Similar accuracy was observed in separating patients with advanced adenomas from the controls, but not between adenoma patients and CRC [50]. Interestingly, the VOC profiles reverted to normal following polypectomy [50].
The other study used gas chromatography coupled to a sulfur chemiluminescence detector and collected VOCs during defecation. However, despite good sensitivity (90%), the overall accuracy (75%) was not superior to that observed by Bosch in stored feces [49].
Finally, some authors [51][52][53] have reported promising results in the detection of CRCrelated metabolomic alterations also using different techniques such as nuclear magnetic resonance spectroscopy.
Available data indicate that stool VOCs have acceptable performance in discriminating CRC patients from controls. The studies, however, were carried out in small series using different, non-comparable, techniques. The possibility of identifying advanced adenoma also needs further evaluation.

Urinary VOCs
Although the collection of urine samples is easily performed, only six studies investigated urinary VOCs in CRC (Table 4).
The urinary volatilomes were similar in CRC patients and spouses or first-degree relatives using FAIMS. The VOC profiles became different when the two control groups were grouped, despite far from optimal sensitivity and specificity (69%, 69%, AUC 0.72, p < 0.001). This approach minimized the influence of dietary differences in VOC profiles, but differences were not significant in subgroup analysis, possibly due to inadequate statistical power. Despite an interesting approach, no significant advantage over the currently available fecal screening test was observed [58].

Breath Exhaled VOCs
In IBD, endoscopy represents the gold standard for the diagnosis and the assessment of disease activity as well as for the detection of the related dysplasia and CRC. Being expensive and invasive, the procedure is not well accepted by many patients. Nonetheless, periodic monitoring is essential to assess the progression of mucosal inflammation [62], optimize treatment, and prevent complications. To reduce the need for endoscopy, several non-invasive biomarkers are routinely used in clinical practice, but all show limitations [5,6]. Thus, alternative strategies such as VOC analysis are needed ( Table 5).
The proposed mechanism of VOC production in IBD is largely dependent on oxidative stress causing lipid peroxidation. Several studies suggested the potential role of single breath exhaled VOCs such as pentane and hexane as markers of IBD [36,[41][42][43]. The potential role of VOC patterns including hydrogen sulfide, acetic acid, propanoic acid, and butanoic acid has also been suggested, as higher than normal concentrations are present in IBD patients [36]. The correlation of these putative markers with clinical indexes (Harvey-Bradshaw Index for CD and Simplified Clinical Colitis Activity Index for UC) is unfortunately weak [36].
Acceptable accuracy in distinguish IBD from the controls (74% sensitivity, 75% specificity, AUC 0.82), and UC from CD (67% sensitivity, 67% specificity, AUC 0.70) has also been reported [38]. The results are in line with those previously reported by Hicks et al. [39], who recorded different VOCs (Table 1) in a small cohort of patients. In this study, despite being able to distinguish both CD and UC from healthy controls (AUC 0.86 and 0.74, respectively), and CD from UC (AUC 0.83), VOC analysis did not differ significantly in active versus inactive disease [39].
Specific VOC profiles, analyzed by GC-MS, were used in a larger cohort of adult CD patients to distinguish healthy controls from active and inactive CD [40]. Two sets of six discriminatory compounds (Table 1) were identified, showing sensitivity 96%, specificity 99%, AUC 0.99, and sensitivity 96%, specificity 97%, and AUC 0.98, respectively. A panel of 10 VOCs proved to be effective in separating active from inactive disease (sensitivity 81%, specificity 80%, AUC 0.88) [40]. The reliability of results was far from optimal, as disease activity was not assessed by endoscopy, but was defined based on fecal calprotectin, c-reactive protein, and the Harvey-Bradshaw Index.
Similar results were reported by the same group in UC, using a set of 11 VOCs to discriminate clinically active from inactive disease (92% sensitivity and 77% specificity) [35]. Again, however, endoscopy was not performed, only fecal calprotectin [35].
The accuracy of the VOC pattern in the diagnosis of IBD was also assessed using eNoses [15].
No link was found between VOC levels and IBD complications [37]. Similarly, VOCs proved inadequate for identifying disease location and activity and did not correlate with laboratory parameters and therapy. The only exception was represented by the VOC patterns of patients with ileal-pouch-anal-anastomosis, which differed from that from all other groups but, again, proved unable to detect pouch inflammation [37].
Similar results have been reported in pediatric IBD. Three discriminatory VOCs were effective in separating IBD patients from healthy controls but did not correlate with the degree of inflammation [45].
These results have been confirmed by a recent Italian study, reporting good sensitivity, but low specificity in distinguishing IBD from the controls, and CD from UC [44] (Table 5). Again, no relevant differences were observed with disease activity.
In conclusion, despite differences in the sampling methods and measurement techniques, breath analysis seems to be overall effective in separating IBD patients from the normal controls. It may thus help to select patients with the suspected disease, who should undergo endoscopy. The most relevant clinical application of VOC analysis is, however, the assessment of disease activity, to avoid an unnecessary colonoscopy. In this respect, the available evidence is conflicting (Table 5) and largely based on studies flawed by the lack of endoscopy and histology.
Chronic long-standing IBD increases the risk of CRC [63]. The possibility of identifying high-grade dysplasia by non-invasive techniques by the identification of specific VOCspatterns would be of prime clinical importance. No data, however, are presently available in the literature.
The diagnostic accuracy was slightly varied with the different techniques used or differing study groups (Table 6) [68].
Again, studies were carried out in small cohorts, more so when focusing on disease activity. To our knowledge, no data are available on IBD complications, especially for IBD-related CRC.

Urinary VOCs
Data on urinary VOCs in IBD are scarce (Table 7). VOC differences in urine samples of IBD patients and controls were first reported in 2009 using an eNose [71]. The samples of a small cohort of 48 IBD patients were subsequently analyzed with both eNose and FAIMS by the same authors. Both technologies were able to separate IBD patients from controls with 88% and 75% accuracy, respectively (p < 0.001), and active from the inactive disease [72].  GC-IMS detected significantly different VOC profiles in 10 IBD patients vs. 10 controls (p = 0.028, AUC = 0.78) [68]. This small study reported better performance of urinarycompared to fecal VOCs, as far as sensitivity was concerned (80% vs. 70%), but not for specificity (70% vs. 90%).
Urinary metabolomic profiles were studied by a combination of direct infusion liquid chromatography/tandem mass spectrometry and GC-MS assays [16]. UC patients in remission differed from IBS patients with AUC 0.99, specificity 99.9%, and sensitivity 99.8% (p < 0.001).
Additionally, in this field, some reports have suggested that NMR might prove useful in identifying metabolomic alterations [73,74].

Discussion
VOCs are the volatile fraction of metabolome, resulting from the ongoing metabolic processes of gut bacteria. These compounds can be detected in breath, urine, and feces [75] ( Figure 1). Specific VOC patterns reflect health status and are supposed to undergo diseasespecific changes [76].
Volatilome is influenced by exogenous and endogenous factors such as environmental diet and pollution, which significantly contribute to the pattern of production of VOCs [77]. Moreover, VOC profiles in CRC and IBD result from inflammation-induced dysbiosis [12,78] as well as medications directly affecting the gut microbiome such as antibiotics and proton pump inhibitors [79].
Bacterial enzymatic activities are also modulated by the microenvironmental pH. The intracolonic pH has been demonstrated to be higher than normal in CRC and lower in IBD compared to a healthy subject, thus affecting enzymatic activities and composition of bacterial metabolites [80][81][82]. The bowel preparation before endoscopy is also relevant. Thus, the relative importance of individual factors and their complex interactions need further investigation [23,83].
Nonetheless, VOC analysis represents an attractive, non-invasive method to evaluate patient health status and possibly diagnose pathological conditions. In intestinal diseases, endoscopy still represents the mainstay of diagnostic workup, but it is invasive and expansive. If feasible, this novel approach would represent an attractive alternative to endoscopy. Aside from avoiding unnecessary procedures, VOC analysis may result in a reduction in health care costs. Indeed, this novel diagnostic approach is relatively unexpansive, as fecal VOC profiling costs approximately €11 [35], 30% less than fecal occult blood testing [72]. Urinary VOC evaluation cost is slightly more expensive, twice as much as a fecal immunochemical test (28 vs. 18 £/test) [55,57].
However, VOC analysis is not standardized and is still far from clinical application. Most data refer to breath sampling, which is easy to perform and easily accepted by patients, while the analysis of fecal and urine samples, although attractive, does not represent valuable alternative strategies. Moreover, analysis of fecal samples may prove a less effective choice. Given the suboptimal compliance to the standard fecal occult blood test as a screening procedure, the same can be anticipated for fecal VOC analysis. Urine samples may be more easily accepted by patients.
Technical hurdles such as the lack of standardization in sample collection, different VOC profiles, and measurement techniques are shared by VOC analysis, irrespective of the sampling of alveolar air, stool, or urine. Urine and feces, moreover, need to be processed, stored within few hours from collection, and then be defrosted and warmed-up to obtain the headspace for analysis.
Storage significantly affects the more volatile compounds, but not necessarily the results [36,84]. The diagnostic accuracy of fecal and urinary VOCs was not significantly influenced by storage duration (20 months for fecal and 12 months for urinary VOCs) in some studies [65,85]. Other authors have reported that different sampling conditions and sample characteristics (sample mass, temperature, water content, time at room temperature before storage) may instead affect the outcome [86]. Optimal stool sampling and storage conditions have not yet been defined, but freezing 500 mg fecal samples diluted with 10 mL of tap water, thawing 10 min before analysis, and a single thaw freeze cycle showed the best accuracy for disease detection in IBD [65]. It is unclear whether the same applies to CRC.
VOCs include a large number of different chemical compounds, largely overlapping in health and disease. Thus, statistics using principal component analysis (PCA) and probabilistic neural network (PNN) analysis is required for reducing a large set of variables to small ones. This, in turn, further contributes to the variability of the results reported in different studies. Thus, the results and interpretation of data often differ in different settings despite the use of the same instrumentation and comparable series of patients, more so using different procedures based on GC-MS versus eNoses. Consequently, distinct VOC profiles have been identified in the same disease, thus far preventing the use of VOC analysis in clinical practice. Nonetheless, some stimulating information is growing in specific conditions such as IBD, IBD-related CRC, sporadic colorectal adenomas, and sporadic CRC (Figure 3). In cancer, most data are derived from small-powered pilot studies, inadequate to validate the test as a mass screening tool. In IBD, patients are mainly compared with healthy volunteers.
Despite acceptable accuracy in separating IBD patients from the controls, the methodology is still far from useful clinical use in IBD. Specific VOC patterns pinpointing active inflammation, fibrostenosing, or fistulizing phenotype in CD and cancer risk in UC are lacking.
Despite all the above-mentioned shortcomings, VOC analysis is promising, considering that recent advances render VOC analysis progressively more effective for early diagnosis of colorectal cancer. Being non-invasive, time-and cost-sparing, and safe, this approach may help improve compliance in screening programs and possibly replace fecal occult blood testing.
Combination of differing VOC panels, and improving the performance of diagnostic algorithms will enhance accuracy. Refining analytical methods will also allow more accurate detection and quantification of specific metabolites, rendering VOC analysis a suitable tool for the diagnosis and management of several diseases, especially CRC and IBD.

Conclusions
Despite all the above-mentioned shortcomings, VOC analysis could be reasonably effective in separating IBD or CRC patients from the controls (Figure 3).
Future studies with more robust methodology, carried out in suitably large cohorts of patients are needed to validate the currently available evidence. Focus studies on the interrelation between bacterial metabolism, disease, and specific VOC pattern, and the mechanism leading to their production are needed to explore the full potential of this novel tool.
Besides identifying specific VOCs pattern in adenoma, sporadic CRC, and IBD in the presence or absence of high-grade dysplasia, it shall be crucial to establish the relationship between VOCs and the underlying disease. It is still unclear whether metabolic processes leading to differing VOC panels are a consequence of disease or represent a contributing factor to its development. Furthermore, no prospective studies have assessed the risk of CRC development by using VOCs in samples collected before the onset of disease, both in the general population and in patients with IBD.

Conflicts of Interest:
The authors declare no conflict of interest.