Exhaled breath and urinary volatile organic compounds (VOCs) for cancer diagnoses, and microbial-related VOC metabolic pathway analysis: a systematic review and meta-analysis

Abstract

Background:

The gradual evolution of the detection and quantification of volatile organic compounds (VOCs) has been instrumental in cancer diagnosis. The primary objective of this study was to assess the diagnostic potential of exhaled breath and urinary VOCs in cancer detection. As VOCs are indicative of tumor and human metabolism, our work also sought to investigate the metabolic pathways linked to the development of cancerous tumors.

Materials and Methods:

An electronic search was performed in the PubMed database. Original studies on VOCs within exhaled breath and urine for cancer detection with a control group were included. A meta-analysis was conducted using a bivariate model to assess the sensitivity and specificity of the VOCs for cancer detection. Fagan’s nomogram was designed to leverage the findings from our diagnostic analysis for the purpose of estimating the likelihood of cancer in patients. Ultimately, MetOrigin was employed to conduct an analysis of the metabolic pathways associated with VOCs in relation to both human and/or microbiota.

Results:

The pooled sensitivity, specificity and the area under the curve for cancer screening utilizing exhaled breath and urinary VOCs were determined to be 0.89, 0.88, and 0.95, respectively. A pretest probability of 51% can be considered as the threshold for diagnosing cancers with VOCs. As the estimated pretest probability of cancer exceeds 51%, it becomes more appropriate to emphasize the ‘ruling in’ approach. Conversely, when the estimated pretest probability of cancer falls below 51%, it is more suitable to emphasize the ‘ruling out’ approach. A total of 14, 14, 6, and 7 microbiota-related VOCs were identified in relation to lung, colorectal, breast, and liver cancers, respectively. The enrichment analysis of volatile metabolites revealed a significant enrichment of butanoate metabolism in the aforementioned tumor types.

Conclusions:

The analysis of exhaled breath and urinary VOCs showed promise for cancer screening. In addition, the enrichment analysis of volatile metabolites revealed a significant enrichment of butanoate metabolism in four tumor types, namely lung, colorectum, breast and liver. These findings hold significant implications for the prospective clinical application of multiomics correlation in disease management and the exploration of potential therapeutic targets.

Introduction

Highlights

• The pooled sensitivity, specificity, and the area under the curve for cancer screening utilizing exhaled breath and urine volatile organic compounds (VOCs) were determined to be 0.89, 0.88, and 0.95, respectively.

• A pretest probability of 51% can be considered as the threshold for diagnosing cancers with VOCs. As the estimated pretest probability of cancer exceeds 51%, it becomes more appropriate to emphasize the ‘ruling in’ approach. Conversely, when the estimated pretest probability of cancer falls below 51%, it is more suitable to emphasize the ‘ruling out’ approach.

• A total of 14, 14, 6, and 7 microbiota-related VOCs were identified in relation to lung, colorectal, breast, and liver cancers, respectively.

• The enrichment analysis of volatile metabolites revealed a significant enrichment of butanoate metabolism in four tumor types, namely lung, colorectum, breast, and liver.

Cancer poses a global public health concern and remains the leading cause of mortality worldwide¹. The identification of cancer at an early stage, coupled with effective treatment, has the potential to significantly enhance survival rates among patients². Metabolomics has emerged as a valuable methodology for characterizing the metabolic profiles of diseases, aiming to identify biomarkers for disease screening and early detection³. In order to address these challenges, the field of volatolomics, which involves the analysis of volatile organic compounds (VOCs) originating from both host and microbial metabolism and detectable in diverse biological matrices, presents a promising advancement in cancer biomarker-focused strategies^4–6.

Several methodologies have been employed in the examination of VOCs in order to ascertain cancer biomarkers. Gas chromatography coupled with mass spectrometry (GC-MS) has been extensively utilized for the detection of VOCs associated with cancer⁷. Although GC-MS is informative, its analysis procedure is time-consuming and necessitates highly skilled personnel. Electronic nose (e-Nose) is another analytical method that can extract complete information from exhaled breath or urine components instead of focusing on the identification of specific biomarkers. The fundamental component of the e-Nose apparatus, the sensor array, can be constructed using a range of sensors, such as nanomaterials, metal oxide semiconductor (MOS) sensors, and other specialized sensors capable of detecting specific VOCs^8,9. The analysis of exhaled breath using e-Nose technologies has been the subject of extensive research in the field of oncology. However, it is important to note that the accuracy of e-Nose results can be significantly affected by both endogenous and exogenous factors. Therefore, further exploration is required to enhance its precision¹⁰.

VOCs present in the human volatilome originate from both endogenous biochemical processes and environmental exposures, and are released through various biological matrices, including exhaled breath and urine¹¹. Breath, being the most prevalent source of VOCs, has been subject to extensive scrutiny regarding its potential association with lung cancer¹². Numerous studies in the literature have demonstrated that the presence of VOCs in breath may serve as a reflection of the biochemical condition of the human body, thereby enabling disease diagnosis. In addition to breath, more than 10% of VOCs associated with humans and tumors can be detected in urine¹¹. The collection of urine presents advantages over breath collection and storage, and urine can be conveniently stored frozen for extended periods, facilitating the tracking of subjects. These characteristics render urine an optimal sample for the investigation of VOC profiling¹³.

The objective of this present study was to conduct a comprehensive review and meta-analysis of prior research on VOCs found in exhaled breath and urine, with the purpose of detecting cancer. As VOCs are indicative of tumor and human metabolism, our work also sought to investigate the metabolic pathways linked to the development of cancerous tumors.

Materials and methods

This study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement (Supplemental Digital Content 1, http://links.lww.com/JS9/B586) (Supplemental Digital Content 2, http://links.lww.com/JS9/B587) and AMSTAR-2 guidelines^14,15 (Supplemental Digital Content 3, http://links.lww.com/JS9/B588).

Literature search strategy

We conducted a comprehensive literature search using the PubMed database to determine the overall diagnostic performance of VOCs in detecting tumors. The following search terms were used: (volatile organic compounds OR VOCs) AND [Neoplasms (Mesh) OR cancer OR carcinoma OR tumor OR neoplasm]. The search included studies published up until 30 June 2023, without any lower date limit.

Eligibility and exclusion criteria

The inclusion criteria for diagnostic studies were as follows: (1) utilization of either e-Nose technology or mass spectrometry; (2) cancer screening; and (3) examination of exhaled breath or urine. Conversely, studies were excluded if they met any of the following criteria: (1) not conducted on human subjects; (2) analysis of alternative biofluids such as blood or feces; and (3) inadequate provision of information required for diagnostic value calculations.

Data extraction and quality assessment

The data extraction process involved the participation of two reviewers (Q.W. and X.L.), with one initially extracting the data and the other conducting a double-check. The following data of each study were collected: author, nation, publication year, cancer type and stage, control group, sample size, analysis methodology (exhaled breath or urine, analytical platform and statistical method), results of true positive (Tp), false positive (Fp), false negative (Fn), and true negative (Tn) (either found or calculated from data in original published studies), and identified unique VOCs (if provided). The evaluation of study quality was conducted by two researchers who utilized the QUality Assessment of Diagnostic Accuracy Studies (QUADAS-2) checklist tool. Studies with a total score of nine points or higher were selected, and any discrepancies were resolved through consensus.

Pathway analysis

MetOrigin is an all-encompassing platform specifically designed for discerning the origins of metabolites by means of an integrated database search¹⁶. Subsequently, a compilation of metabolites exhibiting significant alterations, along with their corresponding Kyoto Encyclopedia of Genes and Genomes and Human Metabolome Database identification numbers, was uploaded, leading to the execution of the Simple MetOrigin Analysis.

Statistical analysis

Diagnostic parameters for VOCs were computed using the following formulas in each study: sensitivity=Tp divided by the sum of Tp+Fn, specificity=Tn divided by the sum of Tn+Fp, positive likelihood ratio (PLR)=sensitivity divided by one minus specificity (1-specificity), negative likelihood ratio (NLR)=one minus sensitivity (1-sensitivity) divided by specificity, and diagnostic odds ratio=Tp multiplied by Tn divided by Fn multiplied by Fp. Additionally, their 95% CI were determined. The bivariate model was adjusted to calculate the area under the curve (AUC). Fagan’s nomogram was designed to leverage the findings from our diagnostic analysis for the purpose of estimating the likelihood of cancer in patients. In order to identify the potential determinants of accuracy estimates, we endeavored to investigate the underlying causes of variability among the studies included in our analysis, specifically focusing on those with a quantified I ² value exceeding 50%. Given the significance of the threshold effect as a contributing factor to heterogeneity, we conducted an assessment using Spearman’s correlation coefficient, whereby a negative correlation (P<0.05) would indicate the presence of said threshold effect. In the absence of a threshold effect but with the presence of significant heterogeneity, additional meta-regression analysis and subgroup analysis were conducted to investigate alternative factors contributing to heterogeneity in the studies included¹⁷. Additionally, Deeks funnel plot symmetry test was employed to directly assess publication bias¹⁸. The statistical analyses were performed using Stata 17.0 (StataCorp), with all tests being two-sided and a P-value less than 0.05 considered as statistically significant.

Results

Study characteristics

The literature search resulted in the identification of 85 publications that satisfied the predetermined inclusion criteria (Table 1)^19–103. Out of these, five studies provided data for 10 separate comparisons, resulting in the incorporation of a cumulative total of 90 studies (with a combined number of cases in tumor and control groups amounting to 6077 and 8004, respectively) in our research. The process of study selection was visually depicted in Figure 1 through a detailed flowchart. The included studies examined various types of cancer, including lung cancer (n=38), colorectal cancer (n=11), urologic neoplasms (n=9 (prostate cancer=4, bladder cancer=3, renal cancer=2)), breast cancer (n=8), gastroesophageal cancer (n=7), head and neck cancer [n=7 (head and neck squamous cell carcinoma=5, thyroid cancer=2)], hepatocellular carcinoma (n=4), pancreatic cancer (n=2), ovarian cancer (n=1), cervical cancer (n=1), malignant pleural mesothelioma (n=1), and malignant mesothelioma (n=1). The majority of the included studies conducted comparisons between patients diagnosed with cancer, often exhibiting a combination of histologic subtypes, and either a healthy control group or patients with benign conditions that impacted the same organ. These studies typically encompassed patients at both early and advanced tumor stages, although 34 studies did not provide information on tumor stage. Frequently employed analytical platforms in this study encompassed e-Nose (n=41) and GC-MS (n=25). Among the e-Noses, the Cyranose 320 (n=8) and Aeonose (n=6) were the most prevalent. Additionally, a diverse assortment of custom-made e-Noses were utilized. The predominant sensor type employed was MOS (n=15). MOS was utilized in Aeonose, PEN3, and various e-nose prototypes for the detection of VOCs. Other reported sensor types included nanomaterial-based sensors (n=8) and quartz microbalance sensor (n=1).

Table 1.

Characteristics and outcomes of all 85 publications included in the qualitative analysis.

Study ID	Nation	Tumor type	Tumor stage	Control group	No. of patients	Test sample	Analytical platform	Statistical method	Tp	Fp	Fn	Tn
Phillips et al. 199919	USA	Lung cancer	I	Cancer-free controls	9, 48	Exhaled breath	GC-MS	FSDA	9	9	0	39
Phillips et al. 2003a20	USA	Lung cancer	NR	Healthy controls	67, 41	Exhaled breath	GC-MS	FSDA	60	7	7	34
Phillips et al. 2003b21	USA	Breast cancer	NR	Healthy controls	51, 42	Exhaled breath	GC-MS	FSDA	48a	11a	3a	31a
Machado et al. 200522	USA	Lung cancer	I–IV	Cancer-free controls	14, 62	Exhaled breath	e-Nose (Cyranose 320: 32 conducting polymer sensors)	SVM	10	5	4	57
Phillips et al. 200823	USA	Lung cancer	NR	Cancer-free controls	193, 211	Exhaled breath	GC-MS	WDA	163	40	30	171
Ligor et al. 200924	Austria	Lung cancer	NR	Healthy controls	65, 31	Exhaled breath	GC-MS	NR	33	0	32	31
Song et al. 201025	China	Lung cancer	NR	Healthy controls	43, 41	Exhaled breath	GC-MS	NR	41	6	2	35
Qin et al. 201026	China	Hepatocellular carcinoma	I–IV	Healthy controls	30, 36	Exhaled breath	GC-MS	LDA	26	3	4	33
Phillips et al. 201027	USA	Breast cancer	NR	Healthy controls	14, 69	Exhaled breath	GC-MS	WDA	12a	14a	2a	55a
Dragonieri et al. 201228	Italy	Malignant pleural mesothelioma	I–IV	Asbestos exposure	13, 13	Exhaled breath	e-Nose (Cyranose 320: 32 conducting polymer sensors)	CDA	12	2	1	11
Peled et al. 201229	Israel	Lung cancer	I–IV	Benign cases	50, 19	Exhaled breath	e-Nose [18 nanomaterial sensors (GNPs and SWCNTs)]	DFA	43	1	7	18
Chapman et al. 201230	Australia	Malignant mesothelioma	Advanced stage	Healthy controls	10, 32	Exhaled breath	e-Nose (Cyranose 320: 32 conducting polymer sensors)	PCA	9b	3b	1b	29b
Altomare et al. 201331	Italy	Colorectal cancer	I–IV	Healthy controls	37, 41	Exhaled breath	GC-MS	PNN	32b	7b	5b	34b
Broza et al. 201332	Israel	Lung cancer	Early stage	Benign cases	12, 5	Exhaled breath	e-Nose (25 nanomaterial sensors: GNP and PtNPs)	DFA	12	1	0	4
Xu et al. 201333	China	Gastric cancer	I–IV	Benign gastric ulcers and less severe gastric conditions	37, 93	Exhaled breath	e-Nose [14 nanomaterial sensors (GNP and SWCNTs)]	DFA	33	9	4	84
Handa et al. 201434	Japan	Lung cancer	I–IV	Healthy controls	50, 39	Exhaled breath	IMS	NR	38	0	12	39
Arasaradnam et al. 201435	UK	Colorectal cancer	I–IV	Healthy controls	83, 50	Urine	FAIMS	FDA	73	20	10	30
Bousamra et al. 201436	USA	Lung cancer	0–II	Benign pulmonary diseases	64, 40	Exhaled breath	FT-ICR-MS	NR	53	10	11	30
Gruber et al. 201437	Israel	HNSCC	I–IV	Healthy controls	22, 19	Exhaled breath	e-Nose (6 nanomaterial sensors)	DFA	17	2	5	17
Fu et al. 201438	China	Lung cancer	I–IV	Benign pulmonary nodules	97, 32	Exhaled breath	FT-ICR-MS	NR	87	6	10	26
Leunis et al. 201439	The Netherlands	HNSCC	I–IV	Benign cases	36, 23	Exhaled breath	e-Nose (12 MOS)	LRA	32	5	4	18
Schumer et al. 201540	USA	Lung cancer	NR	Healthy controls	156, 194	Exhaled breath	silicon chip-MS	NR	146	28	10	166
Li et al. 201541	China	Lung cancer	I–IV	Benign pulmonary nodules and healthy controls	85, 119	Exhaled breath	FT-ICR-MS and GC-MS	PLS, SVM, RF, LDA, and QDA	82	19	3	100
Ligor et al. 201542	Poland	Lung cancer	III–IV	Healthy controls	123, 361	Exhaled breath	GC-MS	ANN	78b	100b	45b	261b
Kumar et al. 201543	UK	Esophageal and gastric adenocarcinoma	I–IV	Healthy controls	81, 129	Exhaled breath	SIFT-MS	LRA	71	24	10	105
Amal et al. 201544	Israel	Ovarian cancer	I–IV	Cancer-free controls	34, 34	Exhaled breath	GC-MS	DFA	30b	3b	4b	31b
Guo et al. 201545	China	Papillary thyroid carcinoma	NR	Nodular goiters	25, 39	Exhaled breath	GC-MS	PCA, PCA, and PLS-DA	23	7	2	32
Barash et al. 201546	Israel	Breast cancer	NR	Benign cases and healthy controls	169, 82	Exhaled breath	e-Nose [40 nanomaterial sensors (GNP and SWCNTs)]	DFA	132	32	37	50
Rocco et al. 201647	Italy	Lung cancer	Advanced stage	Healthy controls	23, 77	Exhaled breath	e-Nose (BIONOTE: 7 acoustic-mass sensors)	PLS-DA	20	4	3	73
Amal et al. 2016a48	Israel	Colorectal cancer	I–IV	Healthy controls	65, 122	Exhaled breath	e-Nose [6 nanomaterial sensors (GNP and SWCNTs)]	DFA	55	7	10	115
Amal et al. 2016b49	Israel	Gastric cancer	I–IV	Gastric intestinal metaplasia	30, 95	Exhaled breath	e-Nose [8 nanomaterial sensors (GNP and SWCNTs)]	DFA	22a	2a	8a	93a
Schallschmidt et al. 201650	Germany	Lung cancer	NR	Healthy controls	37, 23	Exhaled breath	GC-MS	LDA	34	1	3	22
Zou et al. 201651	China	Esophageal cancer	I–IV	Healthy controls	29, 57	Exhaled breath	PTR-MS	SDA	25	6	4	51
Gasparri et al. 201652	Italy	Lung cancer	NR	Healthy controls	70, 76	Exhaled breath	e-Nose (8 QMB sensors)	PLS-DA	57	7	13	69
Sakumura et al. 201753	Japan	Lung cancer	I–IV	Healthy controls	107, 29	Exhaled breath	GC-MS	SVM	102	3	5	26
Li et al. 201754	China	Lung cancer	NR	Healthy controls and benign cases	24, 28	Exhaled breath	e-Nose (14 sensors, MOS HWG, CCGS, EGS)	LDA (fuzzy 5-NN)	22	2	2	26
Shlomi et al. 201755	Israel	Lung cancer	Early stage	Benign nodules	16, 30	Exhaled breath	e-Nose (40 nanomaterial sensors: 26 GNP, 8 SWCNT/PAH, and 6 SWCNT/HBC)	DFA	12	2	4	28
Schuermans et al. 201856	China	Gastric cancer	NR	Healthy controls	16, 28	Exhaled breath	e-Nose (3 MOS)	ANN	13	8	3	20
Heers et al. 201857	Germany	Bladder transitional cell carcinoma	I–IV	Healthy controls	30, 30	Urine	e-Nose (Cyranose 320: 32 composite polymer sensors)	LDA	28	2	2	28
Markar et al. 201858	UK	Pancreatic cancer	NR	Cancer-free controls	25, 43	Exhaled breath	GC-MS	LRA	20b	2b	5b	41b
Widlak et al. 201859	UK	Colorectal cancer	NR	Cancer-free controls	35, 406	Urine	FAIMS	PCA	22	151	13	255
Marzorati et al. 201960	Italy	Lung cancer	Early stage	Healthy controls	6, 10	Exhaled breath	e-Nose (4 MOS)	ANN (LOOCV)	5	0	1	10
Nissinen et al. 201961	Finland	Pancreatic cancer	I–IV	Healthy controls	68, 52	Urine	FAIMS	LDA	54	11	14	41
Rudnicka et al. 201962	Poland	Lung cancer	I–IV	Healthy controls	90, 91	Exhaled breath	GC-MS	DFA	72a	8a	18a	83a
Chandran et al. 201963	Australia	HNSCC	NR	Healthy controls	23, 21	Exhaled breath	SIFT-MS	NR	21	5	2	16
Mozdiak et al. 201964	UK	Colorectal cancer	NR	Healthy controls	12, 12	Urine	FAIMS	NR	12	1	0	11
Mohamed et al. 201965	Egypt	Lung cancer	I–IV	Without lung lesions	28, 20; 24, 27c	Exhaled breath; urine	e-Nose (PEN3: 10 MOS)	PCA and ANN	26; 24	2; 0	2; 0	18; 27
Diaz de Leon Martinez et al. 202066	Mexico	Breast cancer	0-IV	Healthy controls	262, 181	Exhaled breath	e-Nose (Cyranose 320: 32 conducting polymer sensors)	CDA	262	0	0	181
Chen et al. 202067	China	Lung cancer	NR	Healthy controls	24, 25	Exhaled breath	e-Nose (GO sensor)	PCA	23	1	1	24
van Keulen et al. 202068	The Netherlands	Colorectal cancer	I–IV	Normal colonoscopy cases	62, 104	Exhaled breath	e-Nose (Aeonose: 3 MOS)	ANN (Aethena)	59	37	3	67
Waltman et al. 202069	The Netherlands	Prostate cancer	I–IV	Healthy controls	32, 53	Exhaled breath	e-Nose (Aeonose: 3 MOS)	ANN	27	16	5	37
Kort 2020a70	The Netherlands	NSCLC	NR	Healthy controls	103, 84	Exhaled breath	e-Nose (Aeonose: 3 MOS)	ANN	95	41	8	43
Altomare et al. 202071	Italy	Colorectal cancer	I–IV, unknown	Cancer-free controls	82, 87	Exhaled breath	GC-MS	LRA	74	6	8	81
Gharra et al. 202072	China	Lung and gastric cancers	I–IV	Healthy controls	156, 106; 117, 109d	Exhaled breath	e-Nose (15 nanomaterial-based sensors)	DFA	156b e; 117b e	3b e; 2b e	0b e; 0b e	103b e; 107b e
Zhu et al. 202073	UK	Bladder cancer	Ta, T1-2	Healthy controls	38, 41	Urine	e-Nose (fluorescence sensitive optical array)	PLS-DA	32	5	6	36
Kort 2020b74	The Netherlands	Lung cancer	I–IV	Suspected lung cancer and healthy controls	138, 143	Exhaled breath	e-Nose (3 MOS)	ANN	130	80	8	63
Tyagi et al. 2021a75	UK	Prostate and bladder cancers	NR	Cancer-free controls	15, 36; 55, 36f	Urine	GC-IMS	XGBoost	13; 42	3; 4	2; 13	33; 32
Capelli et al. 202176	Italy	Prostate cancer	NR	Healthy controls	132, 60	Urine	e-Nose (6 MOS)	RF	108	8	24	52
Tyagi et al. 2021c77	UK	Colorectal cancer	NR	Cancer-free controls	58, 38	Urine	e-Nose (10 MOS); GC-TOF-MS	ANN	53g; 50g	17g; 7g	5g; 8g	21g; 31g
Yang et al. 202178	Taiwan, China	Breast cancer	NR	Cancer-free controls	351, 88	Exhaled breath	e-Nose (Cyranose 320: 32 conducting polymer sensors)	RF	301	3	50	85
Hong et al. 202179	China	Gastric cancer	NR	Healthy controls	29, 24	Exhaled breath	TD-SPI-MS	LRA	28	1	1	23
V A 2021a80	India	Lung cancer	NR	Healthy controls	41, 74	Exhaled breath	e-Nose (1 MOS)	XGBoost	38b	5b	3b	69b
Diaz de Leon Martinez et al. 202181	Mexico	Cervical cancer	NR	Healthy controls	12, 12	Urine	GC-MS	CAP	11	0	1	12
V A 2021b82	India	Lung cancer	I–III	Healthy controls	32, 72	Exhaled breath	e-Nose (5 MOS)	SVM	30a	3a	2a	69a
Bannaga et al. 202183	UK	Hepatocellular carcinoma	NR	Nonfibrotic cases	20, 31	Urine	GC-IMS and GC-TOF-MS	RF	12	8	8	23
Pinto et al. 202184	Portugal	Clear cell renal cell carcinoma	I–IV	Cancer-free controls	75, 75	Urine	GC-MS	PCA and PLS-DA	62	16	13	59
Long et al. 202185	China	Lung cancer	NR	Healthy controls	161, 116	Exhaled breath	GC-MS	OPLS-DA	149b e	6b e	12b e	110b e
Chen et al. 202186	China	Lung cancer	I–IV, unknown	Healthy controls	101, 134	Exhaled breath	e-Nose (11 gas sensors and 2 temperature and humidity sensors)	KPCA-XGBoost	97e	12e	4e	122e
Mohamed et al. 202187	Sudan	HNSCC	I–IV	Healthy controls	49, 35	Exhaled breath	e-Nose (Aeonose: 3 MOS)	ANN (Aethena)	43	10	6	25
Boulind et al. 202288	UK	Colorectal cancer	NR	Cancer-free controls	18, 65	Urine	GC-MS	ANN	15	12	3	53
Cheng et al. 202289	The Netherlands	Colorectal cancer	I–IV	Healthy controls	30, 84	Exhaled breath	TD-SPI-MS	PCoA	24	11	6	59
Gasparri et al. 202290	Italy	Lung cancer	I–III	Healthy controls	46, 81	Urine	GC-IMS; e-Nose (12 QMB sensors)	SVM	46b e h; 43b e h	3b e h; 1b e h	0b e h; 3b e h	78b e h; 80b e h
Scheepers et al. 202291	The Netherlands	Differentiated thyroid carcinoma	NR	Benign cases	48, 85	Exhaled breath	e-Nose (Aeonose: 3 MOS)	NR	35	26	13	59
Sukaram et al. 202292	Thailand	Hepatocellular carcinoma	A–D	Healthy controls and cirrhosis	61, 91	Exhaled breath	GC-MS	SVM	46b	16b	15b	75b
Gashimova et al. 202293	Russia	Lung cancer	NR	Young healthy controls	77, 78	Exhaled breath	GC−MS	ANN	68b e	13b e	9b e	65b e
Anzivino et al. 202294	Italy	HNSCC	NR	Allergic rhinitis case and healthy controls	15, 30	Exhaled breath	e-Nose (Cyranose 320: 32 polymer sensors)	PCA and CDA	14	4	1	26
Liu et al. 202395	China	Prostate cancer	NR	Cancer-free controls	23, 23	Urine	GC-IMS	SVM	20a	1a	3a	22a
Costantini et al. 202396	Italy	Renal cancer	I–III	Healthy controls	110, 142	Urine	e-Nose (Cyranose 320: 32 organic polymer sensors)	PCA	79	15	31	127
Li et al. 202397	China	Breast cancer	NR	healthy controls	25, 26	Urine	HS-SPME and GC-HRMS	OPLS-DA	19a	2a	6a	24a
Nakayama et al. 202398	Japan	Breast cancer	I–IV	Cancer-free controls	45, 51	Exhaled breath	SIFT-MS	MLR	39e	22e	6e	29e
Liu et al. 202399	China	Breast cancer	I–IV, unknown	Healthy controls	249, 1545	Exhaled breath	HPPI-TOFMS	RF	222a	190a	27a	1355a
Temerdashev et al. 2023100	Russia	Lung cancer	I–IV	Healthy controls	78, 84	Exhaled breath	GC-MS	ANN	69e	13e	9e	71e
Sani et al. 2023101	China	Lung cancer	Early stage	Healthy controls	291, 95	Exhaled breath	LC-MS/MS	RF	241	21	50	74
Sukaram et al. 2023102	Thailand	Hepatocellular carcinoma	A–C	Healthy controls and cirrhosis	124, 219	Exhaled breath	GC-FAIMS	XGBoost	87	25	37	194
Kort et al. 2023103	The Netherlands	NSCLC	I–IV	Healthy controls	158, 216	Exhaled breath	e-Nose (Aeonose: 3 MOS)	ANN, SVM, RF, XGBoost, and logistic regression	147b	99b	11b	117b

^aData derived from a validated model.

^bData derived from a training model.

^cExhaled breath: lung cancer n=28, without lung lesions n=20; urine: lung cancer n=24, without lung lesions n=27.

^dLung cancer n=156, healthy controls n=106; gastric cancer n=117, healthy controls n=109.

^eDiagnostic values were estimated based on calculable data in the study.

^fProstate cancer n=15, cancer-free controls n=36; bladder cancer n=55, cancer-free controls n=36.

^geNose: Tp, Fp, Fn, Tn n=53, 17, 5, 21; GC-TOF-MS: Tp, Fp, Fn, Tn n=50, 7, 8, 31.

^hGC-IMS: Tp, Fp, Fn, Tn n=46, 3, 0, 78; e-Nose: Tp, Fp, Fn, Tn n=43, 1, 3, 80.

ANN, artificial neural network; CAP, canonical principal coordinate analysis; CDA, canonical discriminant analysis; DFA, discriminant factor analysis; FAIMS, field asymmetric ion mobility spectrometry; FDA, fisher discriminant analysis; Fn, false negative; Fp, false positive; FSDA, forward stepwise discriminant analysis; FT-ICR-MS, fourier transform-ion cyclotron resonance-mass spectrometry; GC-FAIMS, gas chromatography-field asymmetric ion mobility spectrometry; GC-HRMS, gas chromatography-high resolution mass spectrometry; GC-MS, gas chromatography-mass spectrometer; GC-TOF-MS, gas chromatography-time of flight-mass spectrometry; HNSCC, head and neck squamous cell carcinoma; HPPI-TOFMS, high-pressure photon ionization-time-of-fight mass spectrometry; HS-SPME, headspace-solid phase microextraction; IMS, ion mobility spectrometry; KPCA, kernel principal component analysis; LDA, linear discriminant analysis; LRA, logistic regression analysis; MLR, multiple logistic regression; MOS, metal oxide semiconductor; NR, not reported; NSCLC, nonsmall cell lung cancer; OPLS-DA, orthogonal partial least-squares-discriminant analysis; PCA, principal component analysis; PCoA, principal co-ordinates analysis; PLS, partial least-squares; PLS-DA, partial least-squares-discriminant analysis; PNN, probabilistic neural network; PTR-MS, proton transfer reaction-mass spectrometry; QDA, quadratic discriminant analysis; QMB, quartz microbalance; RF, random forest; SDA, stepwise discriminant analysis; SIFT-MS, selected ion flow tube mass spectrometry; SVM, support vector machine; TD-SPI-MS, single-photon ionization-mass spectrometry coupled with on-line thermal desorption; Tn, true negative; Tp, true positive; WDA, weighted digital analysis; XGBoost, eXtreme Gradient Boosting.

Figure 1.

Flow diagram showing the study selection process.

Assessment of study quality and publication bias

The evaluation results of each publication based on QUADAS-2 were presented in Supplementary Table S1 (Supplemental Digital Content 4, http://links.lww.com/JS9/B589). The Deeks test for funnel plot symmetry revealed the absence of any significant publication bias (bias=0.73, P=0.469). Figure 2 displayed the corresponding Deeks plot, also known as funnel plot.

Figure 2.

Asymmetrical funnel plots indicated no publication bias (bias=0.73, P=0.469).

Diagnostic performance

A comprehensive analysis of receiver operating characteristic data from 90 studies revealed a pooled sensitivity of 0.89 (95% CI: 0.87–0.91) and specificity of 0.88 (95% CI: 0.85–0.90) (Figs 3–4). The pooled studies exhibited substantial heterogeneity, as indicated by an I ² index of 84.70% for sensitivity and 92.51% for specificity (Fig. 3). Notably, there was no evidence of a threshold effect (Spearman correlation coefficient=0.32, P=0.10). The study findings revealed that the PLR, NLR, and AUC had values of 7.4 (95% CI: 6.0–9.1), 0.13 (95% CI: 0.10–0.15), and 0.95 (95% CI: 0.92–0.96), respectively. Additionally, Table 2 presented the diagnostic parameters for various types of cancers, including lung cancer, colorectal cancer, urologic neoplasms, breast cancer, gastroesophageal cancer, head and neck cancer, and hepatocellular carcinoma. The pretest probability was determined to be 0.25, 0.51, and 0.75, resulting in corresponding positive predictive values and negative predictive values of 71, 88, 96, and 4, 12, 28%, respectively (Fig. 5).

Figure 3.

Pooled sensitivity and specificity analyses of all 90 studies.

Figure 4.

Summary receiver operating characteristic curve analysis of all studies, the pooled area under the curve was 0.95 (95% CI: 0.92–0.96).

Table 2.

VOC-based exhaled breath and urine tests for the diagnosis of different types of tumors.

	Sensitivity (95% CI)	Specificity (95% CI)	PLR (95% CI)	NLR (95% CI)	AUC (95% CI)
Overall (n=90)	0.89 (0.87–0.91)	0.88 (0.85–0.90)	7.4 (6.0–9.1)	0.13 (0.10–0.15)	0.95 (0.92–0.96)
Lung cancer (n=38)	0.91 (0.88–0.93)	0.90 (0.86–0.93)	9.0 (6.4–12.8)	0.10 (0.07–0.14)	0.96 (0.94–0.97)
Colorectal cancer (n=11)	0.87 (0.82–0.91)	0.80 (0.70–0.87)	4.3 (2.8–6.7)	0.16 (0.11–0.23)	0.91 (0.88–0.93)
Urologic neoplasms (n=9)	0.81 (0.76–0.85)	0.86 (0.81–0.91)	6.0 (4.1–8.7)	0.22 (0.17–0.28)	0.89 (0.86–0.91)
Breast cancer (n=8)	0.91 (0.79–0.96)	0.90 (0.67–0.97)	8.8 (2.3–34.0)	0.10 (0.04–0.28)	0.96 (0.94–0.97)
Gastroesophageal cancer (n=7)	0.92 (0.80–0.97)	0.92 (0.84–0.97)	12.0 (5.3–27.1)	0.09 (0.03–0.23)	0.97 (0.95–0.98)
Head and neck cancer (n=7)	0.86 (0.79–0.91)	0.78 (0.71–0.84)	3.9 (2.9–5.4)	0.17 (0.11–0.28)	0.88 (0.84–0.90)
Hepatocellular carcinoma (n=4)	0.72 (0.65–0.78)	0.86 (0.79–0.91)	5.1 (3.4–7.8)	0.32 (0.25–0.42)	0.83 (0.80–0.86)

AUC, area under the curve; NLR, negative likelihood ratio; PLR, positive likelihood ratio.

Figure 5.

Fagan nomograms for the elucidation of post-test probabilities with different pretest probabilities. (A) The pretest probability was 0.25, yielding a PPV of 71% and a NPV of 4%. (B) The pretest probability was 0.51, yielding a PPV of 88% and a NPV of 12%. (C) The pretest probability was 0.75, yielding a PPV of 96% and a NPV of 28%.

Univariable meta-regression and subgroup analysis

In order to investigate the origins of heterogeneity, we conducted meta-regression and subgroup analysis across six dimensions, as depicted in Figure 6 and Table 3. These dimensions encompassed the nation of the included patients (developing country vs. developed country), tumor type (lung cancer vs. other cancers), control group (health controls vs. other controls), sample size (<80 vs. >80 patients), test sample (VOCs from exhaled breath or urine), and analytical platform (e-Nose vs. non-e-Nose). The findings indicated that all the six parameters exhibited heterogeneity in both sensitivity and specificity (P<0.05). The subgroup analysis, as presented in Table 3, provided a summary of the results. Notably, studies conducted in developing countries, focusing on lung cancer, utilizing health controls, involving fewer than 80 patients, utilizing exhaled breath samples, and employing e-Nose demonstrated higher pooled sensitivities and specificities compared to studies conducted in developed countries, involving other types of cancers, utilizing other control groups, including more than 80 patients, utilizing urine samples, and not utilizing e-Nose technology.

Figure 6.

Univariable meta-regression and subgroup analysis for differentiating cancers.

Table 3.

Univariable meta-regression and subgroup analysis for differentiating cancers.

Parameter	Category	No. of studies	Sensitivity (95% CI)	P	Specificity (95% CI)	P
Nation	Developing country	31	0.92 (0.90–0.95)	<0.001	0.92 (0.89–0.95)	<0.001
	Developed country	59	0.86 (0.84–0.89)		0.85 (0.82–0.88)
Tumor type	Lung cancer	38	0.91 (0.88–0.93)	<0.001	0.90 (0.86–0.93)	<0.001
	Other cancers	52	0.87 (0.84–0.90)		0.87 (0.83–0.90)
Control group	Health controls	51	0.90 (0.88–0.92)	<0.001	0.90 (0.87–0.93)	<0.001
	Other controls	39	0.87 (0.83–0.90)		0.85 (0.81–0.89)
Sample size	<80	32	0.89 (0.86–0.93)	<0.001	0.91 (0.88–0.95)	<0.001
	>80	58	0.89 (0.86–0.91)		0.86 (0.83–0.89)
Test sample	Exhaled breath	69	0.89 (0.87–0.92)	<0.001	0.88 (0.85–0.91)	<0.001
	Urine	21	0.87 (0.82–0.91)		0.88 (0.83–0.93)
Analytical platform	e-Nose	41	0.91 (0.88–0.93)	<0.001	0.89 (0.85–0.92)	<0.001
	non-e-Nose	49	0.87 (0.84–0.90)		0.87 (0.84–0.91)

Origin analysis and origin-based pathway analysis

Constrained by the limited number of compounds, we conducted a focused investigation into the metabolic pathways of four specific types of tumors, namely lung, colorectal, breast, and liver cancers. Additionally, we included relevant studies on VOCs in colorectal and breast cancers^88,104–116. Following the removal of not-found and duplicate compounds, we identified 79, 102, 49, and 42 unique VOCs associated with lung cancer, colorectal cancer, breast cancer, and hepatocellular carcinoma, respectively (Supplementary Table S2, Supplemental Digital Content 5, http://links.lww.com/JS9/B590). These metabolites were deemed to exhibit significant alterations.

The metabolites of lung cancer were initially derived from various sources, including four metabolites specific to the human host, 45 metabolites originating from bacteria, and 134 metabolites from other sources such as drugs, food, the environment, and unknown origins (Supplementary Fig. S1A, Supplemental Digital Content 6, http://links.lww.com/JS9/B591). In Supplementary Figs S1B-C (Supplemental Digital Content 6, http://links.lww.com/JS9/B591), we compared the total number of metabolites and metabolic pathways belonging to the host with those belonging to the microbiota. A total of 14 metabolites (isopropanol, butane, n-butanal, butanol, methanol, methylacetate, acetone, 2,3-butandione, cyclohexane, cyclohexanone, toluene, benzaldehyde, propanal, and o-xylene) were identified as significantly specific to microbiota in four enriched metabolic pathways (butanoate metabolism (ko00650), caprolactam degradation (ko00930), toluene degradation (ko00623), and xylene degradation (ko00622)) (Table 4; Supplementary Fig. S1D, Supplemental Digital Content 6, http://links.lww.com/JS9/B591). The host metabolites did not have an enriched metabolic pathway, whereas 10 significantly metabolic pathways were enriched in the human and microbial communities (Supplementary Table S3, Supplemental Digital Content 7, http://links.lww.com/JS9/B592; Supplementary Fig. S1D, Supplemental Digital Content 6, http://links.lww.com/JS9/B591).

Table 4.

Significantly metabolic pathways enriched in the microbial communities.

	Pathway name	Pathway ID	Sig. volatile metabolites	P
Lung cancer	Butanoate metabolism	ko00650	Isopropanol; Butane; n-Butanal; Butanol; Methanol; Methylacetate; Acetone; 2,3-Butandione	8.21e-11
	Caprolactam degradation	ko00930	Cyclohexane; Cyclohexanone	4.82e-03
	Toluene degradation	ko00623	Toluene; Benzaldehyde	2.19e-02
	Xylene degradation	ko00622	Propanal; o-Xylene	2.30e-02
Colorectal cancer	Toluene degradation	ko00623	Methylbenzene; Benzaldehyde; p-Cresol	1.25e-03
	Xylene degradation	ko00622	m-Cymene; p-Xylene; p-Cymene	1.35e-03
	Butanoate metabolism	ko00650	2,3-Butanedione; Butanal; Acetone	2.07e-03
	Ethylbenzene degradation	ko00642	Ethylbenzene; 1-Phenylethanone	2.96e-03
	Dioxin degradation	ko00621	Biphenyl; Dibenzofuran	1.60e-02
	Biosynthesis of various alkaloids	ko00996	Vanillin	2.37e-02
Breast cancer	Butanoate metabolism	ko00650	Methanol; 2-Propanol; Acetone	1.69e-04
	Furfural degradation	ko00365	Furfural	2.13e-02
	Ethylbenzene degradation	ko00642	Ethylbenzene	2.71e-02
	Caprolactam degradation	ko00930	Cyclohexanone	4.48e-02
Hepatocellular carcinoma	Butanoate metabolism	ko00650	Acetone; 2-Propanol; Butanal; 2,3-Butandione; 1-Butanol	2.30e-07
	Xylene degradation	ko00622	m-Xylene; p-Xylene	7.71e-03

The initial origins of colorectal cancer metabolites encompassed various sources, including 35 host metabolites, 74 bacterial metabolites, and 179 metabolites from other origins such as drugs, food, environment, and unknown sources (Supplementary Fig. S2A, Supplemental Digital Content 8, http://links.lww.com/JS9/B593). Supplementary Figs S2B–C, (Supplemental Digital Content 8, http://links.lww.com/JS9/B593) illustrated the total number of metabolites and metabolic pathways attributed to the host, and provide a comparison with those associated with the microbiota. A total of 14 metabolites (methylbenzene, benzaldehyde, p-cresol, m-cymene, p-xylene, p-cymene, 2,3-butanedione, butanal, acetone, ethylbenzene, 1-phenylethanone, biphenyl, dibenzofuran, and vanillin) were identified across six enriched metabolic pathways, namely toluene degradation (ko00623), xylene degradation (ko00622), butanoate metabolism (ko00650), ethylbenzene degradation (ko00642), dioxin degradation (ko00621), and biosynthesis of various alkaloids (ko00996) (Table 4; Supplementary Fig. S2D, Supplemental Digital Content 8, http://links.lww.com/JS9/B593). These metabolites exhibited a significant association with the microbiota. Conversely, no enriched metabolic pathway specific to host metabolites was observed. Additionally, a total of 37 significantly metabolic pathways were found to be enriched in both human and microbial communities (Supplementary Table S3, Supplemental Digital Content 7, http://links.lww.com/JS9/B592; Supplementary Fig. S2D, Supplemental Digital Content 8, http://links.lww.com/JS9/B593).

The metabolites associated with breast cancer were initially derived from various sources, including 12 host metabolites, 32 bacterial metabolites, and 90 others originating from drugs, food, the environment, and unknown origins (Supplementary Fig. S3A, Supplemental Digital Content 9, http://links.lww.com/JS9/B594). Supplementary Figs. S3B–C (Supplemental Digital Content 9, http://links.lww.com/JS9/B594) illustrated the comprehensive count of metabolites and metabolic pathways associated with the host, juxtaposed with those of the microbiota. The microbiota exhibited a noteworthy specificity, with six metabolites (methanol, 2-propanol, acetone, furfural, ethylbenzene, and cyclohexanone) distributed across four enriched metabolic pathways (namely, butanoate metabolism (ko00650), furfural degradation (ko00365), ethylbenzene degradation (ko00642), and caprolactam degradation (ko00930)) (Table 4; Supplementary Fig. S3D, Supplemental Digital Content 9, http://links.lww.com/JS9/B594). Conversely, no enriched metabolic pathways were identified for host metabolites, while five significantly metabolic pathways demonstrated enrichment in both human and microbial communities (Supplementary Table S3, Supplemental Digital Content 7, http://links.lww.com/JS9/B592; Supplementary Fig. S3D, Supplemental Digital Content 9, http://links.lww.com/JS9/B594).

The metabolites of hepatocellular carcinoma were initially derived from various sources, including four host metabolites, 25 bacterial metabolites, and 72 others (such as drugs, food, environment, and unknown sources) (Supplementary Fig. S4A, Supplemental Digital Content 10, http://links.lww.com/JS9/B595). Supplementary Figs S4B–C, (Supplemental Digital Content 10, http://links.lww.com/JS9/B595) illustrated the total number of metabolites and metabolic pathways associated with the host and compare them to those of the microbiota. Among these, seven metabolites (acetone, 2-propanol, butanal, 2,3-butandione, 1-butanol, m-xylene, and p-xylene) in two enriched metabolic pathways (butanoate metabolism (ko00650) and xylene degradation (ko00622)) were found to be significantly specific to the microbiota (Table 4; Supplementary Fig. S4D, Supplemental Digital Content 10, http://links.lww.com/JS9/B595). No enriched metabolic pathway was identified for host metabolites, while eight significantly metabolic pathways were enriched in both human and microbiota (Supplementary Table S3, Supplemental Digital Content 7, http://links.lww.com/JS9/B592; Supplementary Fig. S4D, Supplemental Digital Content 10, http://links.lww.com/JS9/B595).

Discussion

Over the past decade, numerous proof-of-principle studies have provided evidence for the effectiveness of employing VOC profiles in clinical diagnostics. Notably, VOC analysis has proven successful in detecting various cancers, including lung, breast, prostate, gastric, colorectal, and liver cancers. While further validation through larger-scale studies is necessary, these findings appear to corroborate the notion that pathological mechanisms within the body can impact VOC profiles, thereby yielding distinct VOC signatures for each disease. The objective of the present study was to conduct a comprehensive review and meta-analysis of prior research investigating the use of exhaled or urinary VOCs for the purpose of cancer detection. In addition to their role as indicators of tumor metabolism, this study also aimed to investigate the metabolic pathways associated with the development of tumors.

The analysis of VOCs in exhaled breath and urine presented a promising approach for the diagnosis of various tumor types. Our study demonstrated the efficacy of VOC analysis in exhaled breath and urine for cancer diagnosis, as evidenced by an area under the receiver operating characteristic curve of 0.95 (95% CI: 0.92–0.96), a pooled sensitivity of 0.89 (95% CI: 0.87–0.91), and a pooled specificity of 0.88 (95% CI: 0.85–0.90). These parameters collectively suggested that this test showed superior diagnostic performance in distinguishing tumors from control groups. The diagnostic accuracy of LRs (PLR and NLR) is widely acknowledged in academic circles, particularly for their utility in clinical decision-making, surpassing sensitivity, specificity, or AUC. To be deemed highly useful, a diagnostic test should possess a PLR exceeding 10.0 and an NLR below 0.10, as recommended by a commonly employed heuristic for LR interpretation¹¹⁷. A higher PLR suggests proficiency in disease diagnosis, whereas a lower NLR suggests proficiency in disease exclusion. The PLR and NLR of VOCs in cancer diagnosis were determined to be 7.4 and 0.13, respectively. These findings indicate that VOCs exhibit a greater ability to rule out tumor diagnosis, while their capacity to confirm it is comparatively lower.

Additionally, we have employed Fagan’s nomogram to leverage the findings from our diagnostic analysis for the purpose of estimating the likelihood of cancer in patients during routine clinical practice¹¹⁸. According to the findings depicted in Figure 5, patients exhibiting suspected cancer with a pretest probability of 25% experienced a notable increase in the likelihood of cancer to 77% upon receiving a positive test result. Conversely, a negative test result led to a reduction in the probability to 6%. Similarly, patients with suspected cancer and a pretest probability of 75% observed a substantial rise in the likelihood to 97% following a positive test result, while a negative test result resulted in a decrease to 37% probability. When evaluating a higher pretest probability, it was found that the ‘ruling in’ probability of cancer was more suitable than ‘ruling out’. Conversely, a lower pretest probability increased the probability of cancer for ‘ruling out’. The pretest probability was notably established at 51%, while the likelihood of developing cancer following a positive result was determined to be 88%. Stated differently, the probability of not having cancer subsequent to a positive result was 12%, which corresponded to the probability of being cancer-free after a negative result. This indicated that VOCs achieved comparable performance in both confirming and excluding the disease. Hence, a pretest probability of 51% can be considered as the threshold for diagnosing cancers with VOCs. As the estimated pretest probability of cancer exceeds 51%, it becomes more appropriate to emphasize the ‘ruling in’ approach. Conversely, when the estimated pretest probability of cancer falls below 51%, it is more suitable to emphasize the ‘ruling out’ approach.

In order to ascertain potential sources of heterogeneity, we conducted univariate meta-regression and meta-analysis. These analyses unveiled that various factors, such as nation, tumor type, sample size, source of VOCs, analytical platforms, and types of the control group, may contribute to the heterogeneity observed in our meta-analysis. Further analysis demonstrated a higher level of precision in detecting VOCs for cancer diagnosis in developing countries in contrast to developed countries, suggesting that the effectiveness of VOCs in cancer screening might vary depending on racial or ethnic factors. When considering the detection of VOCs, the diagnostic efficacy of utilizing urine was equally optimal in comparison to exhaled breath. The utilization of urinary VOC analysis in clinical tumor diagnosis practice offered distinct advantages over exhaled VOC analysis, thus warranting further investigation into the potential of urinary VOCs. In the context of the analytical platform, the utilization of e-Nose for the analysis of VOCs exhibited superior sensitivity and specificity in comparison to alternative platforms. Due to the heightened sensitivity of e-Noses toward alterations in both endogenous and exogenous factors, the diagnostic precision may exhibit variability across diverse research settings and patient cohorts¹¹⁹. Consequently, it becomes imperative to undertake extensive, multicenter external validation studies to explore the applicability and replicability of the findings across various research settings and patient populations.

The microbiome and its metabolites exhibit a strong correlation with human health and diseases. Present research predominantly employs statistical correlation analysis to investigate the interplay between microbial communities and all identified metabolites, thereby elucidating their association¹²⁰. It is of great importance to undertake additional measures in order to distinguish the metabolic activities attributed to the host, bacteria, or both. Otherwise, comprehending the genuine metabolic functions of microbial metabolites becomes arduous when all metabolites from the host and bacterial communities are combined. Differentiating between microbial metabolites and other sources such as the host, food, or environment, and investigating their metabolic functions and associations with the microbiome, could potentially enhance the effectiveness and precision of biomarker identification. The second aim of the current study was to investigate the metabolic pathways linked to the development of cancerous tumors, as VOCs serve as markers for both tumor presence and human metabolism. We identified a total of 14, 14, 6, and 7 VOCs associated with microbiota in lung, colorectal, breast, and liver cancers, respectively. The enrichment analysis of these volatile metabolites demonstrated a significant enrichment of butanoate metabolism in the aforementioned types of tumors. Butyric acid, a significant metabolite generated through the fermentation of intestinal flora, serves as a transport substrate in butanoate metabolism and plays a crucial role in regulating host energy homeostasis mediated by intestinal flora. Recent literature has highlighted the potential impact of butyrate on the development of diverse diseases, including cancer. Sodium butyrate has been observed to enhance the expression levels of P-gp and signal transducer and activator of transcription (STAT) 3, as well as promoted STAT3 phosphorylation and enhanced mRNA stability of ATP-binding cassette subfamily B member 1 in human lung cancer cells¹²¹. Zhang et al.¹²² indicated that butyrate induced apoptosis through the activation of mitogen-activated protein kinase signaling in colon cancer RKO cells. Sodium butyrate elicited apoptosis through the augmentation of reactive oxygen species levels, enhancement of caspase activity, and reduction of mitochondrial membrane potential in both normal breast and breast cancer cells¹²³. Butyrate exerted inhibitory effects on cell proliferation and induced apoptosis in both Hep3B and HepG2 cells. Prolonged exposure to butyrate was associated with enhanced hepatocyte growth and fibrosis in the liver. Therefore, butyrate significantly impacted the proliferation of liver cells¹²⁴.

Conclusions

The analysis of exhaled breath and urinary VOCs, particularly utilizing e-Nose technology, showed promise for cancer screening. However, it was imperative to conduct multicenter validation studies to confirm its efficacy, as well as to evaluate the performance of the optimal analytical approach and consider confounding factors. For example, before clinical implementation can be realized, the lack of standardization and reproducibility in the field of e-Nose research must be addressed. In addition, the enrichment analysis of volatile metabolites revealed a significant enrichment of butanoate metabolism in four tumor types, namely lung, colorectum, breast, and liver. These findings hold significant implications for the prospective clinical application of multiomics correlation in disease management and the exploration of potential therapeutic targets.

Ethical approval

The article is a systematic review and does not require ethical approval.

Consent

The article is a systematic review and does not require ethics committee approval and fully informed written consent.

Sources of funding

This work was supported by the Natural Science Foundation of Qinghai Province of China (No. 2022-ZJ-912).

Author contribution

M.Z., Q.H.W., X.Y.L., Y.C., J.Z.X., and D.Z.C.: study design; M.Z., Q.H.W., X.Y.L., P.Z., and R.Y.: systematic literature search; M.Z., Q.H.W., and X.Y.L.: collected data and drafted the manuscript; M.Z., Q.H.W., X.Y.L., and Y.C.: data analysis and interpretation; M.Z., Q.H.W., X.Y.L., and Y.C.: statistical analysis; M.Z., Q.H.W., X.Y.L., and Y.C.: manuscript preparation. All authors critically revised the manuscript for important intellectual content. All authors had full access to all the data and approved the final version of the manuscript, including the authorship list.

Conflicts of interest disclosure

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Research registration unique identifying number (UIN)

Research registry registration number: reviewregistry 1702.

Guarantor

Daozhen Chen is the guarantor of this article.

Data availability statement

The main data supporting the findings of this study are included within the article and its supplementary materials. Additional data are available from the corresponding authors upon reasonable request.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Supplementary Material