Metabolite biomarkers in lung cancer: unlocking the potential of body fluid analysis for early detection and prognosis—a narrative review

Abstract

Background and Objective

Lung cancer is a leading cause of cancer-related mortality, due to delayed diagnosis and the complexity of selecting the optimal treatment method, given the genetic diversity and heterogeneity of the disease. Traditional invasive techniques, such as tissue biopsy, carry risks of severe complications and are often costly. Therefore, there is increasing interest in non-invasive alternatives, particularly liquid biopsy. This review aims to propose promising circulating metabolite biomarkers for lung cancer and their clinical applications.

Methods

A PubMed search [2014–2024] was conducted, focusing on fluid-based, non-invasive samples such as blood, urine, pleural effusion, and bronchoalveolar lavage fluid. Only English-language articles relevant to lung cancer metabolomics were included.

Key Content and Findings

Analysis of altered metabolites in lung cancer patients revealed significant metabolic pathway enrichments. Upregulated pathways included arginine biosynthesis and alanine, aspartate, and glutamate metabolism, while downregulated pathways involved valine, leucine, and isoleucine biosynthesis and fatty acid metabolism. Metabolite biomarker changes across multiple body fluids were identified when comparing lung cancer patients to healthy controls. In blood, choline, serine, and threonine levels were reduced, whereas tryptophan and tyrosine were elevated. Pleural effusion exhibited decreased oleic acid and ceramide, urine showed increased creatine riboside (CR) and N-acetylneuraminic acid (NANA), and bronchoalveolar lavage fluid (BALF) revealed reductions in glycine and glycerol, highlighting distinct metabolic alterations associated with lung cancer.

Conclusions

Circulating metabolite biomarkers offer a promising, non-invasive approach for early lung cancer detection and personalized treatment strategies. Their accessibility and safety provide a viable alternative to invasive biopsies. Further clinical validation is essential to integrate these biomarkers into practice.

Introduction

Background and rationale

Lung cancer remains a major global health burden, responsible for over 2.2 million new cases and 1.8 million deaths annually, with late-stage diagnosis contributing significantly to its high mortality (1,2). While advancements in genetic profiling and targeted therapies have improved outcomes, challenges persist due to the complexity and heterogeneity of the disease (3). Genetic mutations alone rarely explain tumor progression, and resistance to targeted therapies often arises from new mutations through tumor evolution (4,5). Recent research has increasingly focused on identifying non-invasive biomarkers for early lung cancer detection and prognosis.

Conventional diagnostic methods, including imaging and tissue biopsy, are invasive, costly, and can impose significant physical discomfort on patients (6-8). This has led to increasing interest in non-invasive diagnostic alternatives, such as blood-based metabolomic profiling, which offers insights into the metabolic reprogramming of cancer cells (9,10). Metabolomics captures dynamic metabolic changes that reflect both intrinsic tumor biology and interactions with the tumor microenvironment. Specific metabolites, such as lactate, methanol, and proline, have been identified as biomarkers that help differentiate lung cancer patients from healthy individuals with high accuracy (11,12). For instance, elevated lactate levels are linked to the Warburg effect, where cancer cells preferentially convert glucose to lactate even in the presence of oxygen, fueling tumor growth (11).

In addition to diagnostic applications, metabolomics plays a key role in staging, predicting therapeutic responses, and estimating cancer risk (13). Lung cancer subtypes—such as adenocarcinoma (AD), squamous cell carcinoma (SCC), and small cell lung cancer (SCLC)—exhibit distinct metabolic profiles, enabling the development of subtype-specific biomarkers (14,15). Metabolite panels have also proven effective in distinguishing early from advanced-stage lung cancer, helping clinicians guide treatment strategies and monitor disease progression (15,16). The predictive power of metabolomics has been further enhanced through advanced machine learning models, which can identify complex metabolic patterns and predict treatment outcomes (17).

The integration of metabolomics into clinical practice offers significant potential for overcoming challenges in lung cancer management. By providing real-time metabolic insights, it complements genetic and imaging-based diagnostics, addressing issues of tumoral heterogeneity and drug resistance. Metabolomics also opens new avenues for therapeutic interventions, with research showing that targeting metabolic pathways—such as those involved in lipid metabolism or amino acid synthesis—can enhance treatment efficacy, particularly for resistant tumors (18,19). As research expands, metabolomics is poised to play an increasingly central role in personalized lung cancer care, improving early detection, precise staging, and treatment monitoring, ultimately enhancing patient outcomes and quality of life.

Objective

In this paper, we reviewed the metabolites and metabolomics-based approaches for diagnosis, treatment response, prognosis, and risk stratification in lung cancer based on body fluids, including blood, urine, sputum, bronchoalveolar lavage fluid (BALF), and pleural fluid. This review aims to identify promising metabolite biomarkers with potential for practical clinical application. We present this article in accordance with the Narrative Review reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-325/rc).

Methods

We conducted a literature search on PubMed, for articles published between 2014 and 2024. The following keywords were used: ‘lung cancer’ ‘biomarker’ and ‘metabolite’, combined with each of ‘blood’, ‘urine’, ’sputum’, ‘bronchoalveolar lavage fluid’, ‘pleural effusion’. There were 249 articles, 171 from blood, 62 from urine, 6 from sputum, 3 from bronchoalveolar fluid, and 7 from pleural effusion. Only articles written in English were considered. After reviewing the titles, abstracts, and result data of 249 studies, we excluded articles that focused on non-metabolite biomarkers, including those investigating tissue samples, circulating tumor cells, DNA, RNA, proteins, and extracellular vesicles (Table S1). Finally, total 74 articles were reviewed for this paper. We also reviewed whether each included study reported or statistically addressed potential confounding factors such as diet, medications, comorbidities, smoking status, sex, and ethnicity, and summarized this information in the corresponding tables.

Lung cancer metabolomics in blood

Blood is one of the most widely used liquid samples in clinical medical situation. It is easily collected, and cost-efficient. A variety of metabolites can be identified in blood samples, including serum and plasma, which provide information about diseases and patient conditions.

Blood metabolites as diagnostic marker

There are lots of well-known blood biomarkers for lung cancer diagnosis. Proteins such as carcinoembryonic antigen, neuron-specific enolase, progastrin-releasing peptide, and circulating tumor DNA, including mutations of Kirsten Rat Sarcoma Viral Oncogene Homolog (KRAS), epidermal growth factor receptor (EGFR), v-Raf Murine Sarcoma Viral Oncogene Homolog B1 (BRAF), are valuable biomarkers (20). Recently, along with the development of metabolomics, metabolites have emerged as promising biomarkers for cancer diagnosis.

Numerous studies have been conducted to discover strong diagnostic metabolite biomarkers for lung cancer from blood samples. These studies are summarized in Table 1. One previous study demonstrated high diagnostic accuracy for lung cancer using a five-metabolite model, with a specificity of 77.5% and sensitivity of 76.9%. Serum metabolic profiles of healthy individuals and non-small cell lung cancer (NSCLC) patients were analyzed using proton nuclear magnetic resonance (¹H-NMR) spectroscopy, and logistic regression analysis identified five significantly altered metabolites: lactate, methanol, glutamine, choline, and threonine (11). Another study discovered three metabolites—bisphenol A, retinol, and L-proline—using ultra-performance liquid chromatography coupled with quadrupole time-of-flight tandem mass spectrometry (UPLC-Q-TOF-MS/MS), each of which effectively discriminated lung cancer and healthy group, with area under the curve (AUC) of 0.93, 0.72, 0.95, respectively. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed disruptions in vitamin digestion and the biosynthesis of amino acids in lung cancer patients. These metabolites are capable of inducing metabolic disturbances, as shown in the analysis (12). Blood metabolites provide valuable insight into lung cancer subtype and stage, which is crucial for developing an optimal treatment plan. Since each subtype—AD, SCC, SCLC, and others—originates from different cell types and exhibits unique genetic mutation patterns and signaling pathways, they exhibit distinct metabolic profiles. A study using ¹H-NMR and ultra-performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS/MS), combined with stepwise discriminant analysis and multilayer perceptron, developed a 5-biomarker model (including xanthine and S-adenosyl methionine) that classified AD, SCC, and SCLC with accuracies of 97.1%, 86.5%, and 90.9%, respectively (14). Metabolic patterns also change along with the progression of lung cancer. Using LC-MS, a panel of 5 metabolites—palmitic acid, heptadecanoic acid, 4-oxoproline, tridecanoic acid and ornithine—were able to discriminate between early (stage I/II) and advanced (stage III/IV) lung cancer, achieving an AUC >0.72 (15). A study on lung cancer staging compared plasma metabolites between patients with multiple primary lung cancer and intrapulmonary metastasis, identifying seven metabolites—phosphatidylethanolamine (PE) (38:5), PE (40:5), decanoylcarnitine, octanoylcarnitine, L-histidine, prostaglandin F2 alpha (PGF2α) ethanolamide, and undecanoylcholine—that differentiated the two lesions with AUCs ranging from 0.78 to 0.99. The observed metabolic differences may reflect the alterations in metabolism associated with the process of metastasis (16). Recently, machine learning models also have been widely utilized. A support vector machine multivariate model based on eight metabolites—ethylmalonic acid, maltose, glycerol, 3-phosphoglyceric acid, taurine, glutamic acid, glycolic acid and D-arabinose—revealed the exceptional discrimination performance (AUC =0.999) for the detection of NSCLC (17). In a recent study, metabolic profiling of NSCLC patients and healthy controls identified the top 8 metabolites—pyruvate, tyrosine, glucose, tryptophan, arginine, methionine, cholesterol, and palmitic acid—which demonstrated strong diagnostic capability. The combination of the 8 metabolites achieved AUC >0.999. The metabolic pathways of aromatic amino acids and long-chain fatty acids were found to be associated with these metabolites (21). With a larger participant group of 193 NSCLC patients and 243 healthy populations, using ultra-high-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS), a panel of 10 metabolic biomarkers—hypoxanthine, linoleic acid, 2,4-dihydroxybenzoic acid, 11,12-epoxy-icosatrienoic acid, 16-hydroxyhexadecanoic acid, testosterone sulfate, choline, piperine, cholic acid, and glycoursodeoxycholic acid—successfully diagnosed NSCLC (AUC =0.95) (22).

Table 1. Blood metabolite biomarkers for lung cancer diagnosis.

Biomarkers	Sample	Method	LC group	Control group	Performance	Confounding factors†	Reference
5 metabolites model: lactate(↑), methanol(↑), glutamine(↓), choline(↓), threonine(↓)	Serum	1H-NMR	Discovery: NSCLC =142. Validation: NSCLC =40	Healthy =74. Healthy =13	Sensitivity =77.5%; specificity =76.9%	Age, gender, smoking status	(11)
Bisphenol A(↑), retinol(↓), L-proline(↓)	Serum	UPLC-Q-TOF-MS/MS	LC =35	Healthy =70	AUC =0.93, Sensitivity =91.4%, Specificity =95.7%; Retinol: AUC =0.72, Sensitivity =81.4%, Specificity =59.0%; L-Proline: AUC =0.95, Sensitivity =100%, Specificity =93.5%	Age, gender	(12)
5 biomarker model: 2 metabolites: xanthine (AD > SCC > SCLC), S-adenosyl methionine (AD > SCLC > SCC)	Serum	UPLC-MS/MS	LC =143 (AD =69, SCC =52, SCLC =22)	–	Subtype classification: AUC =0.976, accuracy =97.1% (AD); AUC =0.975, accuracy =86.5% (SCC); AUC =0.990, accuracy =90.9% (SCLC)	Age, gender, smoking status	(14)
5 metabolites model: palmitic acid(↓), heptadecanoic acid(↓), 4-oxoproline(↑), tridecanoic acid(↓), ornithine(↑)	Plasma	LC-MS	Discovery: LC =64. Validation: LC =34	Healthy =50. Healthy =25	AUC =0.918, sensitivity =85.5%, specificity =80.7%; AUC =0.869, sensitivity =82.9%, specificity =76.7% (LC vs. healthy)	Age, gender (matched)	(15)
5 metabolites model: palmitic acid (AD > SCLC > SCC), heptadecanoic acid, pentadecanoic acid, acylcarnitine C8:1 (AD > SCC > SCLC), ornithine (SCLC > SCC > AD)	–	–	Discovery: AD =49, SCC =8, SCLC =2, other =3, unknown =2. Validation: AD =21, SCC =6, SCLC =5, other =0, unknown =2	–	AUC >0.75, accuracy =73.1% (AD vs. SCC vs. SCLC)	Age, gender (matched)	–
5 metabolites model: palmitic acid, heptadecanoic acid, ornithine, tridecanoic acid, stearic acid (early < advanced)	–	–	Discovery: stage I =30, II =5, III =17, IV =12, unknown =0. Validation: stage I =15, II =2, III =9, IV =6, unknown =2	–	AUC >0.72, accuracy =59.9% (early vs. advanced)	Age, gender (matched)	–
PE (38:5), PE (40:5) (MPLC < IM) decanoylcarnitine, octanoylcarnitine, L-histidine, PGF2α ethanolamide, undecanoylcholine (MPLC > IM)	Plasma	Q-TOF MS	Patient =179 (SPLC =136, MPLC =24, IM =19)	Healthy =96	AUC =0.78–0.99, sensitivity =70.8–94.7%, specificity =68.4–100% (MPLC vs. IM)	Age, smoking status, comorbidities (P>0.05); gender (P=0.011*)	(16)
8 metabolites model; SVM multivariate model: maltose(↑), 3-phosphoglyceric acid(↑), taurine(↑), glutamic acid(↑), d-arabinose(↑), ethylmalonic acid(↓), glycerol(↓), glycolic acid(↓)	Plasma	GC-Q-TOF MS	Discovery: NSCLC =64. Validation: NSCLC =27	Healthy =28. Healthy =12	AUC =0.999, sensitivity =98.4%, specificity =96.4%; AUC =1, sensitivity =100%, specificity =100%	Gender (P>0.05); age, smoking status, diabetes mellitus (P<0.05*)	(17)
8 metabolites panel: pyruvate(↑), tyrosine(↑), glucose(↑), tryptophan(↑), arginine(↑), methionine(↑), cholesterol(↑), palmitic acid(↑)	Plasma	LC-MS, GC-MS	NSCLC =43	Healthy =43	AUC >0.999, sensitivity =95.3%, specificity =100%	Age, gender (P>0.05)	(21)
10 metabolites panel: hypoxanthine(↑), linoleic acid(↓), 2,4-dihydroxybenzoic acid(↓), 11,12-epoxy-icosatrienoic acid(↑), 16-hydroxyhexadecanoic acid(↓), testosterone sulfate(↓), choline(↓), piperine(↓), cholic acid(↑), glycoursodeoxycholic acid(↑)	Serum	UHPLC-MS/MS	Discovery: NSCLC =135. Validation: NSCLC =58	Healthy =170. Healthy =73	AUC =0.93, sensitivity =80%, specificity =91%; AUC =0.95, sensitivity =85.0%, specificity =88.5%	Age, gender (P>0.05); smoking status (P=0.016*)	(22)
9 metabolites classifier: benzaldehyde(↑), hydroxypyruvic acid(↑), urea(↑), glycolic acid(↓), L-isoleucine(↓), gluconic acid lactone(↓), allyl laurate(↓), linolenic acid(↓), L-phenylalanine(↓)	Serum	GC-MS	NSCLC =31	Healthy =92	AUC =0.99, sensitivity =100%, specificity =95.0%	Age, gender, smoking status	(23)
4 metabolites classifier: LPC 16:0(↓), 18:0(↓), 18:1(↓), 18:2(↓)	Serum	LC-MS	Early NSCLC =100	Healthy =300	AUC =0.80, sensitivity =98%, specificity =10%	Age, gender (matched), smoking status	(24)
4 metabolites panel: LPE(18:1)(↑), ePE(40:4)(↑), C(18:2)CE(↓), SM(22:0)(↓)	Plasma	ESI-MS/MS	Discovery: early NSCLC =105. Validation: early NSCLC =94	Healthy =80. Healthy =67	AUC =0.823, sensitivity =81.9%, specificity =70.7%; AUC =0.808, sensitivity =78.7%, specificity =69.4%	Age, gender, ethnicity, smoking status	(25)
5 metabolites model: β-hydroxybutyric acid(↑), LPC 20:3(↑), PC ae C40:6(↓), citric acid(↑), fumaric acid(↑)	Plasma	DI-MS, RP-HPLC-MS/MS	Discovery: early NSCLC =87. Validation: early NSCLC =43	Healthy =40. Healthy =20	AUC =0.974, sensitivity =93.7%, specificity =92.2%; AUC =0.959, sensitivity =91.9%, specificity =90.0%	Age, gender (P>0.05); smoking status (P<0.001***)	(26)
9 biomarkers model (3 metabolites and 6 non-metabolite biomarkers); citrulline(↑), glycodeoxycholic acid(↑), taurochenodeoxycholic acid(↓)	Plasma	TMT-LC-MS/MS	Discovery: NSCLC =88. Validation: NSCLC =22	BPD =86. BPD =22	AUC =0.871, sensitivity =86%, specificity =78%; AUC =0.871, sensitivity =77%, specificity =86%	Age, gender (matched)	(27)
3 metabolites panel: cysteine(↓), serine(↓), 1-monooleoylglycerol(↓)	Serum	GC-MS	NSCLC =30	Benign lung tumor =6	AUC =0.956, sensitivity =100%, specificity =83.3%	Age, gender (matched)	(28)
16 metrics model: ornithine(↓), valine(↑), arginine(↓), asparagine(↓), glutamic acid(↑), serine(↓), carnitines (C16(↑), C4DC(↓), C5DC(↓), C5(↓), C22(↓), C4-OH(↓), C12(↑), C26(↓)) age, sex	Serum	LC-MS/MS	Discovery: LC =339. Validation: LC =139	Benign lung nodule =254. Benign lung nodule =116	AUC =0.81, sensitivity =74%, specificity =76%	Age, gender (P<0.05*)	(29)
Octanoylcarnitine, retinol, decanoylcarnitine (PLC > PMC)	Plasma	Q-TOF MS	LC =80; PMC =16	Healthy =48	AUC =0.96–0.99, sensitivity =87.5–100%, specificity =93.8–100% (PLC vs. PMC)	Age, gender, smoking status, comorbidities (P>0.05)	(30)

^†, this column summarizes whether confounding variables (e.g., age, gender, smoking status, comorbidities) were reported or statistically controlled; *, P<0.05; ***, P<0.001. ↑, upregulated in LC; ↓, downregulated in LC. AUC, area under the curve; BPD, benign pulmonary disease; C, cholesterol; CE, cholesteryl ester; DI-MS, direct infusion-mass spectrometry; AD, adenocarcinoma; ePE, ether-linked phosphatidylethanolamine; ESI-MS/MS, electrospray ionization-tandem mass spectrometry; GC-MS, gas chromatography-mass spectrometry; GC-Q-TOF MS, gas chromatography coupled with quadrupole time-of-flight mass spectrometry; IM, intrapulmonary metastasis; LC, lung cancer; LC-MS, liquid chromatography-mass spectrometry; LC-MS/MS, liquid chromatography-tandem mass spectrometry; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; MPLC, multiple primary lung cancer; MS/MS, tandem mass spectrometry; NSCLC, non-small cell lung cancer; PC ae, phosphatidylcholine with acyl-alkyl chain; PE, phosphatidylethanolamine; PGF2α, prostaglandin F2 alpha; PLC, primary lung cancer; PMC, pulmonary metastatic carcinoma; Q-TOF, quadrupole time-of-flight; Q-TOF MS, quadrupole time-of-flight mass spectrometry; RP-HPLC-MS/MS, reversed-phase high-performance liquid chromatography-tandem mass spectrometry; SCC, squamous cell carcinoma; SCLC, small cell lung cancer; SM, sphingomyelin; SPLC, single primary lung cancer; SVM, support vector machine; TMT-LC-MS/MS, tandem mass tag-liquid chromatography-tandem mass spectrometry; UHPLC-MS/MS, ultra-high-performance liquid chromatography-tandem mass spectrometry; UPLC, ultra-performance LC; UPLC-MS/MS, ultra-performance liquid chromatography-tandem mass spectrometry; ¹H-NMR, proton nuclear magnetic resonance.

Besides, early diagnosis is crucial in lung cancer; however, it is challenging since there are usually no symptoms in the early stage. A study quantified serum metabolites from computed tomography (CT)-screening diagnosed NSCLC patients and paired healthy controls, using gas chromatography-mass spectrometry (GC-MS) approach. A classifier compromising 9 metabolites proved its efficacy in early diagnosis of asymptomatic lung cancer patients (AUC =0.99) (23). Two other studies focused on lipid metabolism, which may play a crucial role in early-stage lung cancer. Lipids are essential for tumor growth and metastasis, by regulating cell membrane composition and energy metabolism. By comparing the metabolic signatures of early-stage NSCLC patients and healthy individuals, a classifier based on 4 lysophosphatidylcholines detected by LC-MS (16:0, 18:0, 18:1 and 18:2) (24) and a panel of 4 lipids—lysophosphatidylethanolamine, ether PE, cholesteryl ester and sphingomyelin (25)—were developed, both demonstrating strong diagnostic performance (AUC =0.80, 0.808, respectively). Another study proposed a five-metabolite model, comprising two lipids (lysophosphatidylcholine 20:3 and phosphatidylcholine alkyl-acyl C40:6) and three organic acids (β-hydroxybutyric acid, citric acid, and fumaric acid), which accurately distinguished early NSCLC from healthy individuals with an AUC of 0.959 using direct injection MS (DI-MS) and reverse-phase high-performance liquid chromatography coupled with tandem mass spectrometry (RP-HPLC-MS/MS) (26).

Benign lung diseases including benign lung nodule are important differential diagnoses of lung cancer. Using tandem mass tag labeling LC-MS and KEGG pathway analysis, a study involving NSCLC patients and a benign pulmonary disease group (including tuberculosis, non-tuberculosis infection, lung nodules, etc.) identified three metabolites—citrulline, glycodeoxycholic acid and taurochenodeoxycholic acid. A diagnostic model based on these 3 metabolites and 6 protein biomarkers effectively distinguished NSCLC patients from the benign pulmonary disease group (AUC =0.871) (27).

It remains a challenge in clinical practice to determine whether a lung nodule detected on screening CT is benign or malignant. Non-invasive liquid biopsy, utilizing blood metabolites, represents a promising approach for discriminating indeterminate lesions with minimal hazards. In one previous study, GC-MS-based metabolic profiling of non-smoking females, including NSCLC patients, a benign lung tumor group, and healthy controls, identified a three-metabolite panel (cysteine, serine, and 1-monooleoylglycerol) that demonstrated excellent discrimination of NSCLC patients from the other groups with an AUC of 0.956 (28). A subsequent study with larger groups (110 NSCLC patients, 108 benign pulmonary disease groups and 100 healthy individuals) developed a model of 16 metrics, including 14 metabolites, age and sex, achieving an AUC of 0.81 in distinguishing NSCLC from benign lung nodules. Ornithine and palmitoylcarnitine were key metabolites of this model (29). The overexpression of ornithine decarboxylase gene has been proposed as a critical factor in lung cancer progression (31). Fatty acid metabolism, with carnitine playing a pivotal role, is implicated in the proliferation and metastasis of lung cancer (29).

Another study, combining quadrupole time-of-flight mass spectrometry (Q-TOF MS) and machine learning methods, revealed 3 metabolites—octanoylcarnitine, retinol and decanoylcarnitine—which effectively discriminated between pulmonary metastatic cancer and primary lung cancer (AUC =0.96–0.99) (30). Since it is crucial for staging to distinguish multiple primary lung cancer, intrapulmonary metastasis and pulmonary metastatic cancer, which in turn influences treatment decision, metabolites biomarkers may also play a key role in guiding treatment strategies. Notably, a study identified multi-functional metabolite biomarkers for lung cancer diagnosis using LC-MS, developing three distinct models, each consisting of five specific metabolites, for diagnosis, subtype classification, and staging. Palmitic acid, heptadecanoic acid and ornithine were present in all 3 models, highlighting their potential as versatile biomarkers (15). Recently, a meta-analysis regarding metabolic profiles of lung cancer reviewed relevant literature from 2012 to 2022. A total of 31 studies, involving 2,768 NSCLC patients and 9,873 healthy controls, were included, identifying 46 metabolites associated with glucose, lipid, and nucleotide metabolism. These metabolites are expected to serve as promising biomarkers for the early diagnosis, improving diagnostic accuracy for NSCLC patients in clinical practice (32).

Previous studies identified a total of 39 upregulated and 38 downregulated metabolites exhibiting significant differences between lung cancer patients and healthy controls (Table S2). KEGG pathway analysis revealed key metabolic pathways in lung cancer (Table S3). For upregulated metabolites, the most enriched pathways were arginine biosynthesis and alanine, aspartate and glutamate metabolism (Figure 1A). In contrast, downregulated metabolites highlighted valine, leucine and isoleucine biosynthesis as the top enriched pathway (Figure 1B).

Figure 1.

Biochemical metabolic pathways associated with metabolites exhibiting significant alterations. The metabolites identified as significant in blood samples were compared between lung cancer patients and healthy individuals. Using MetaboAnalyst (https://www.metaboanalyst.ca), the pathways associated with upregulated and downregulated metabolites were identified. These results indicate that both upregulated and downregulated metabolic pathways are significantly associated with lung cancer, suggesting their potential relevance in the prediction and diagnosis of the disease. (A) Upregulated metabolites: pathway analysis revealed significant associations with various biosynthesis and metabolic pathways. Among these, arginine biosynthesis exhibited the most significant association, with a P value of 3.45e−05. Other pathways with notable significance included alanine, aspartate, and glutamate metabolism (P=3.69e−05), glyoxylate and dicarboxylate metabolism (P=6.19e−05), arginine and proline metabolism (P<0.001), phenylalanine, tyrosine, and tryptophan biosynthesis (P=0.001), and the citrate cycle (TCA cycle) (P=0.003). These findings suggest a strong link between these metabolic pathways and lung cancer. (B) Downregulated metabolites: pathway analysis of downregulated metabolites revealed valine, leucine, and isoleucine biosynthesis as the most significantly associated pathway, with a P value of 0.003. Other significantly associated pathways included glycine, serine, and threonine metabolism (P=0.006), biosynthesis of unsaturated fatty acids (P=0.008), neomycin, kanamycin, and gentamicin biosynthesis (P=0.02), galactose metabolism (P=0.04), and phenylalanine, tyrosine, and tryptophan biosynthesis (P=0.046). TCA, tricarboxylic acid.

Blood metabolites as prognostic marker

Blood metabolites have potential to predict prognosis by reflecting treatment responses (Table 2). In order to identify biomarkers capable of predicting response before treatment initiation, blood samples were collected prior to the treatment. A study conducted metabolic profiling of advanced NSCLC patients receiving first-line platinum-based chemotherapy and, using LC-MS, identified four metabolites—caffeine, paraxanthine, stachydrine, and methyl glucopyranoside—that significantly differed between good (>18 months) and poor survival (<12 months) groups, highlighting their potential as biomarkers for predicting chemotherapy benefit (P<0.05) (33). A subsequent study focused on patients with advanced AD who underwent first-line chemotherapy—pemetrexed in combination with either cisplatin or carboplatin. A total of 354 patients were categorized into three groups based on their treatment responses—stable disease, partial response, and progressive disease—with each group exhibiting metabolic variation, and seven metabolites—hypotaurine, uridine, C12:0-carnitine, choline, dimethylglycine (DMG), niacinamide, and C16:0-carnitine—being detected. A panel of these metabolites successfully discriminated between the disease control group (stable disease + partial response) and the progressive disease group, achieving an AUC >0.9 (34).

Table 2. Blood metabolite biomarkers for lung cancer prognosis.

Biomarkers	Sample	Method	Group (N)	Performance	Confounding factors†	Reference
Caffeine(↑), paraxanthine(↑), stachydrine(↓), methyl glucopyranoside (alpha + beta)(↓)	Serum, plasma	LC-MS	Advanced NSCLC =220. Discovery: good survival =55; poor survival =55. Validation: good survival =55; poor survival =55	P<0.05 (*) (good vs. poor)	Age, gender, smoking status (P>0.05), medication (P=0.035*)	(33)
7 metabolites panel: hypotaurine(↑), uridine(↑), C12:0-carnitine(↑), choline(↑), DMG(↑), niacinamide(↓), C16:0-carnitine(↑)	Serum	LC-Q-Orbitrap-MS/MS	LUAD =354 (stage IIIB/IV) after Tx. Discovery: DC =198 (SD =116, PR =82); PD =53. Validation: DC =80 (SD =44, PR =36); PD =23	AUC =0.912, sensitivity =90.8%, specificity =79.5%. AUC =0.909, sensitivity =90.8%, specificity =79.5% (DC vs. PD)	Age, gender, smoking status (P>0.05)	(34)
3-hydroxyanthranilic acid(↑)	Plasma	LC-MS/MS	Healthy =10; NSCLC=19 (Objective response after Tx =10)	AUC =0.83, sensitivity =87.5%, specificity =83.3% (OR vs. not)	Age, gender (adjusted), smoking status	(35)
4 metabolites model: serine(↓), glycine(↓), arginine(↓), quinolinic acid(↑)	Plasma	LC-ESI-MS/MS	Advanced NSCLC =53	C-index =0.775. HR =3.23, 95% CI: 2.04–5.26 (overall survival)	Age, gender, medication (P>0.05)	(36)
N-(3-Indolylacetyl)-L-alanine(↑)	Serum	LC-MS	Advanced NSCLC =250 (stage IIIB–IV): CTx = 50; PD-1 inhibitor + CTx =200. Discovery: R =28, NR =22. Validation: R =75, NR =75	P=0.003**, 0.041* (R vs. NR)	Age, gender, smoking status (P>0.05)	(37)
6 metabolites model: hexadecylthio-PC(↓), mononervonin (15c)(↓), thioetheramide-PC(↓), 7α-hydroxycholesterol(↓), myristoyl-PC(↓), N6-(1-Iminoethyl)-l-lysine(↓)	Serum	UHPLC-Q-TOF MS	Advanced NSCLC =97. Cohort 1 =55 (PD-1 inhibitor monotherapy, 2nd-line)	AUC =0.81, sensitivity =76.9%, specificity =75.9% (long vs. short survival)	Age, gender, smoking status	(38)
3 upregulated metabolites model: agmatine(↓), linoleoyl-carnitine(↓), menthyl salicylate(↓); 3 downregulated metabolites model: curcumin(↑), 2-deoxyribose 1-phosphate(↑), leukotriene B4(↑)			Cohort 2 =42 (PD-1 inhibitor + CTx combination therapy, 1st-line)	AUC =0.948 (up), sensitivity =100%, specificity =81.1%. AUC =0.816 (down), sensitivity =59.1%, specificity =100% (long vs. short survival)	Age, gender, smoking status
Acetate, alanine, glutamate, proline, glycoprotein, phenylalanine, tyrosine, tryptophan (NSCLC > healthy), glucose, taurine, glutamine, glycine, phosphocreatine, threonine (NSCLC < healthy)	Serum	1H-NMR	Healthy =43. NSCLC =38. NSCLC + MWA =39	P<0.05* (healthy vs. NSCLC vs. after MWA)	Age, gender	(39)
GpAEA (LC > healthy), sphingosine (LC < healthy)	Serum	LC-Q-TOF-MS, GC-MS	Healthy =30. LC =30	AUC =0.916, 0.966; sensitivity =76.67%, 96.67%; specificity =93.33%, 90.00% (pre- vs. post-op)	Age, gender (P>0.05)	(40)
SM 42:2(↑), SM 35:1(↑), PC 30:0(↑), PC 30:1(↑), Cer 42:2(↑), SM 38:3(↑)	Serum	UPLC-Q-TOF MS, UPLC-MS/MS	Healthy =51. LC =51. f/u after surgery recurrent (SD + PD) =14. Non-recurrent (PR + CR) =31	P<0.05* (recurrence vs. non-recurrence)	Age, gender (P>0.05)	(41)

^†, this column summarizes whether confounding variables (e.g., age, gender, smoking status, comorbidities) were reported or statistically controlled; *, P<0.05; **, P<0.01. ↑, upregulated in poor prognosis group; ↓, downregulated in poor prognosis group. AUC, area under the curve; Cer, ceramide; CI, confidence interval; C-index, concordance index; CR, complete response; CTx, chemotherapy; DC, disease control; DMG, dimethylglycine; f/u, follow-up; GC-MS, gas chromatography-mass spectrometry; GpAEA, glycerophospho-N-arachidonoyl ethanolamine; HR, hazard ratio; LC, lung cancer; LC-ESI-MS/MS, liquid chromatography-electrospray ionization-tandem mass spectrometry; LC-MS, liquid chromatography-mass spectrometry; LC-MS/MS, liquid chromatography-tandem mass spectrometry; LC-Q-Orbitrap-MS/MS, liquid chromatography coupled with quadrupole Orbitrap tandem mass spectrometry; LC-Q-TOF-MS, liquid chromatography coupled with quadrupole time-of-flight mass spectrometry; LUAD, lung adenocarcinoma; MWA, microwave ablation; NR, non-response; NSCLC, non-small cell lung cancer; OR, objective response; PC, phosphatidylcholine; PD, progressive disease; PR, partial response; R, response; SD, stable disease; SM, sphingomyelin; Tx, treatment; UHPLC-Q-TOF MS, ultra-high-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry; UPLC-MS/MS, ultra-performance liquid chromatography-tandem mass spectrometry; UPLC-Q-TOF MS, ultra-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry; ¹H-NMR, proton nuclear magnetic resonance.

The initiation of programmed cell death protein 1 (PD-1) inhibitor, a widely used immune checkpoint inhibitor (ICI) in lung cancer treatment, remains a complex decision due to high costs and uncertainty regarding its effectiveness for individual patients, making biomarkers that can identify those likely to benefit from treatment highly valuable. 3-hydroxyanthranilic acid was reported to distinguish the objective response group (complete response + partial response) from the non-response group, achieving an AUC =0.83, highlighting its potential as a predictive biomarker for ICI efficacy (35). In another study, NSCLC patients treated with PD-1 inhibitor were stratified based on overall survival following treatment, and through metabolic analysis of pre-treatment plasma samples and Cox proportional hazards analysis, four metabolites—serine, glycine, arginine, and quinolinic acid—were revealed to be correlated with overall survival. The multivariate model incorporating these metabolites effectively predicted overall survival in NSCLC patients receiving PD-1 inhibitor treatment [concordance index =0.775, 95% confidence interval (CI): 2.04–5.26]. The model also exhibited a correlation with immune-related gene expression patterns, further validating its potential in predicting ICI response (36). Recently, N-(3-Indolylacetyl)-L-alanine showed a significant difference between the response (PFS ≥12 months) and non-response (<12 months) groups (P=0.003, 0.041), with patients in the low N-(3-Indolylacetyl)-L-alanine group (below the median) achieving an objective response rate of 43.0%, while those in the high group (above the median) achieved only 24.0%. N-(3-Indolylacetyl)-L-alanine is known to modulate immune cell function through amino acid metabolism (37). A subsequent study using ultra-high-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (UHPLC-Q-TOF MS) compared patients with short and long PFS treated with a PD-1 inhibitor, where cohort 1 received second-line monotherapy and cohort 2 received first-line combination therapy, and the model with the highest performance (AUC =0.948) included three downregulated metabolites—agmatine, linoleoyl-carnitine, and methyl salicylate—which effectively distinguished between the short and long PFS groups (38).

A study focusing on NSCLC patients who received microwave ablation (MWA) treatment revealed differences in metabolic patterns compared to the healthy population. However, after MWA treatment, these differences were reversed, with 8 metabolites increasing and 6 decreasing in NSCLC patients, indicating the effectiveness of the treatment and suggesting that these altered metabolites could be valuable in evaluating treatment response (39).

Another study comparing pre- and post-operative metabolic profiles identified potential biomarkers for recurrence. Glycerophospho-N-arachidonoyl ethanolamine and sphingosine demonstrated high efficacy in distinguishing between pre-operative and post-operative groups, with AUC of 0.916 and 0.966, respectively. These metabolites could serve as indicators for monitoring recurrence, since metabolic alternations following surgical resection may reverse upon recurrence (40). Other research compared recurrent (stable disease + progressive disease) and non-recurrent (complete response + partial response) groups following surgery, revealing significant increases in 6 metabolites—2 phosphatidylcholines, 3 sphingomyelins, and ceramide—in the recurrent group (P<0.05) (41).

Other blood metabolite biomarkers

Several studies have evaluated blood metabolites as predictors of lung cancer risk (42-44). Low serum bilirubin levels were associated with an increased risk of lung cancer and mortality in male smokers, with a P value for interaction of 0.001 in a validation cohort of 425,660 participants. Bilirubin acts as an anti-oxidative and anti-inflammatory molecule, thereby reducing oxidative stress and inflammation in smokers, ultimately lowering lung cancer risk (43). Another study identified cotinine, a nicotine metabolite, as a relevant biomarker for lung cancer risk. In a case-control cohort study involving 5,364 matched participants, cotinine exhibited a dose-response relationship with lung cancer risk in the current smoker group, with odds ratio of 1.39 per 500 nmol/L of increase in cotinine levels (44).

With the advancement of metabolomic analysis techniques, blood metabolites have emerged as promising biomarkers for lung cancer. Numerous studies have demonstrated their diagnostic and prognostic value. However, further research is needed to identify multi-functional biomarkers for both diagnosis and prognosis, as well as to explore the metabolic pathways associated with each metabolite. Nevertheless, some of the studies summarized in this section are limited by small sample sizes, lack of independent validation, or patient heterogeneity, which may affect the robustness and generalizability of their findings. Accordingly, these results should be interpreted with caution, and future research involving large, well-designed, and externally validated cohorts is necessary to establish the clinical utility of blood metabolite biomarkers.

Lung cancer metabolomics in urine

Urine has emerged as a particularly valuable biological sample for detecting lung cancer biomarkers due to its simplicity, non-invasiveness, and potential for repeated sampling. Urine-based diagnostics allow for continuous monitoring of disease progression and therapeutic response, making it an attractive tool for both clinical and research applications. This part reviews recent studies that have investigated urinary metabolomics to identify lung cancer biomarkers, highlighting their diagnostic potential and clinical relevance. Representative studies with larger cohorts or validation are summarized in Table 3.

Table 3. Metabolomics of fluid-based biomarkers for lung cancer beyond blood.

Biomarkers	Sample	Method	Group (N)	Performance	Confounding factors†	Reference
2 metabolites model: creatine riboside(↑), N-acetylneuraminic(↑)	Urine	LC-MS	Discovery set: NSCLC =469; Healthy =536. Validation set: NSCLC =80; Healthy =78	CREATINE riboside: P<0.00001***; N-acetylneuraminic acid: P<0.00001***; creatine riboside: P<0.00001***; N-acetylneuraminic acid: P<0.00001***	Age, gender, ethnicity (matched), smoking status (adjusted)	(45)
Carnitine(↑), acylcarnitines(↓)	Urine	UHPLC-Q-TOF-MS/MS	NSCLC =20; Healthy =20	AUC =0.958	Age, gender	(46)
4 metabolites model: 1-methylhistidine(↑), xanthosine(↓), 8-hydroxynevirapine(↓), tyrosyl-tyrosine(↓)	Urine	UPLC-Q-TOF/MS	Discovery set: NSCLC =30; Healthy =30. Validation set: NSCLC =38; Healthy =42	1-methylhistidine: AUC =0.95; xanthosine: AUC =0.874; 8-hydroxynevirapine: AUC =0.867; tyrosyl-tyrosine: AUC =0.822	Age (matched), gender, smoking status	(47)
4-Methoxyphenylacetic acid	Urine	1H NMR	NSCLC =28; Healthy =17	AUC =0.85	Age (P=0.022*), gender (matched), smoking status	(48)
PGE-major urinary metabolite (PGE-MUM)(↑)	Urine	Radioimmunoassay	NSCLC =369	Pre-op: P=0.005 (**), HR =3.017 (95% CI: 0.135–24.47); Post-op: P=0.027 (*) (95% CI: 1.284–32.49)	Gender (P=0.006**); smoking status (P=0.002**)	(49)
2 metabolites model: creatine riboside(↑), N-acetylneuraminic(↑)	Urine	UPLC-MS/MS	LC =178; Healthy =351	Creatine riboside: AUC =0.84; N-acetylneuraminic acid: AUC =0.90	Age, gender, ethnicity (P>0.05); smoking status (P<0.0001***)	(50)
Ganglioside GM1 (18:1/12:0)	Sputum	FIE-MS and GC-MS	LC =23; Healthy =33	AUC =0.854	Age, gender, smoking status	(51)
ANN model; 6 metabolites model: phenylacetic acid, L-fucose, caprylic acid, acetic acid, propionic acid, and glycine	Sputum	FIE-MS	LC =23; Healthy =33; SCLC =5; NSCLC =17	AUC =0.99, sensitivity 96%, specificity 94%. AUC =1.00, sensitivity 80%, specificity 100%	Age (P>0.05), gender (P>0.05)	(52)
5 metabolites model: hydroxyphenyllactic acid(↑), phytosphingosine(↑), N-nonanoylglycine(↑), sphinganine(↑), S-carboxymethyl-L-cysteine(↑)	Sputum	LC-MS	LC =76; Healthy =67	Hydroxyphenyllactic acid: P=1.08E−08***; phytosphingosine: P=1.15E−07***; N-nonanoylglycine: P=1.88E−09***; sphinganine: P=1.06E−08***; S-carboxymethyl-L-cysteine: P=8.67E−09***	Age (matched) gender, smoking status	(53)
4 metabolites model: diethanolamine(↑), cytosine(↓), lysine(↓), tyrosine(↑)	Saliva	CE-MS	LC =41; benign lung lesion = 21	Diethanolamine: AUC =0.612; cytosine: AUC =0.586; lysine: AUC =0.620; tyrosine: AUC =0.618	Age, gender	(54)
Glycerol and phosphoric acid	BALF	DI-ESI-QTOF-MS and GC-MS	LC =24; benign lung lesion =31	Glycerol: AUC =0.88, sensitivity 80%, specificity 90%; phosphoric acid: AUC =0.79, sensitivity 80%, specificity 60%	Age, gender	(55)
Phosphoric acid(↓), isocitric acid(↓), inositol(↓), L-Glycine(↓)	Serum, urine, BALF	GC-MS	LC =32; Healthy =29	Phosphoric acid: AUC =0.79; isocitric acid: AUC =0.57; inositol: AUC =0.58; L-Glycine: AUC =0.74	Not available	(56)
Oleic acid (18:1, free fatty acid)(↑), ceramide (d18:1/16:0)(↓)	Pleural effusion	LC-MS/MS	LC =32; tuberculosis =18	Oleic acid: AUC =0.962; ceramide: AUC =0.854	Age, gender (P>0.05)	(57)

^†, this column summarizes whether confounding variables (e.g., age, gender, smoking status, comorbidities) were reported or statistically controlled; *, P<0.05; **, P<0.01; ***, P<0.001. ↑, upregulated in LC; ↓, downregulated in LC. AUC, area under the curve; BALF, bronchoalveolar lavage fluid; CE-MS, capillary electrophoresis-mass spectrometry; CI, confidence interval; DI-ESI-QTOF-MS, direct infusion high resolution mass spectrometry; FIE-MS, flow infusion electrospray ionization mass spectrometry; GC-MS, gas chromatography-mass spectrometry; HR, hazard ratio; LC, lung cancer; LC-MS, liquid chromatography-mass spectrometry; LC-MS/MS, liquid chromatography-tandem mass spectrometry; NSCLC, non-small cell lung cancer; PGE, prostaglandin E; PGE-MUM, prostaglandin E-major urinary metabolite; Post-op, postoperative; Pre-op, preoperative; SCLC, small cell lung cancer; UHPLC-Q-TOF-MS/MS, ultra-high-performance liquid chromatography coupled with quadrupole time-of-flight tandem mass spectrometry; UPLC-MS/MS, ultra-performance liquid chromatography-tandem mass spectrometry; UPLC-Q-TOF/MS, ultra-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry; ¹H-NMR, proton nuclear magnetic resonance.

One of the key advantages of using urine for metabolomic studies is its ease of collection, which can be done frequently and non-invasively, unlike tissue biopsies or blood sampling. Urine contains a wide range of metabolites that reflect various physiological and pathological conditions, including cancer. Recent studies, which will be detailed in the following sections, have focused on analyzing these metabolites to identify potential biomarkers for NSCLC, the most common type of lung cancer. By using advanced techniques such as LC-MS, researchers have been able to quantify and compare the levels of specific metabolites in the urine of lung cancer patients and healthy controls, uncovering significant differences that may serve as diagnostic markers.

Urinary metabolites as diagnostic marker

Previous study using LC-MS to analyze urine samples from 469 lung cancer patients and 536 control subjects identified four key metabolites—creatine riboside (CR), N-acetylneuraminic acid (NANA), cortisol sulfate, and a metabolite labeled 561+. CR (P<0.00001) and NANA (P<0.00001) were significantly more abundant in the urine of lung cancer patients compared to the control group (45). These results suggest that these metabolites may be valuable for diagnosing NSCLC and could potentially be used to monitor disease progression or treatment response. Moreover, the ability to measure these metabolites in urine highlights the potential for developing non-invasive diagnostic tools for lung cancer screening.

In another study, urine samples were collected from 20 lung cancer patients (15 with AD, 5 with SCC) before any medical treatment or surgery, along with samples from 20 healthy volunteers without lung disease for comparison. The researchers found that carnitine and 11 selected acylcarnitines showed significant diagnostic potential, with an AUC of 0.958, after analyzing the samples using ultra-high-performance liquid chromatography coupled with quadrupole time-of-flight tandem mass spectrometry (UHPLC-Q-TOF-MS/MS) (46).

Further research has expanded the scope of urinary metabolomics in lung cancer detection. One study used a case-control design to analyze urine samples from lung cancer patients with ultra-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (UPLC-Q-TOF/MS), identifying 35 potential biomarkers, including 11 overlapping metabolites across different analyses. After investigating common metabolic patterns and finding strong linear correlations with serum biochemical indicators, five key metabolic markers were identified: 1-methylhistidine, xanthosine, 8-hydroxynevirapine, and tyrosyl-tyrosine, with an AUC >0.7 (47). These markers may provide further insights into the biological mechanisms underlying lung cancer and could be integrated with other diagnostic tools to enhance detection accuracy.

This study developed a novel composite method for accurately detecting NSCLC by combining targeted serum microRNA (miRNA) analysis and metabolite analysis, with serum samples for miRNA and urine samples for metabolite analyses collected from NSCLC patients and control groups. Serum miRNA levels were measured using quantitative real-time reverse transcription with an exogenous control, and the relative expression of miR-21 and miR-223 was determined using the delta cycle threshold method. Additionally, the concentrations of six targeted urinary metabolites in patients and healthy controls were measured using ¹H NMR spectroscopy. Receiver operating characteristic (ROC) analysis for miRNA expression demonstrated a sensitivity of 96.4% and a specificity of 88.2% in detecting early-stage NSCLC. When the relative urinary concentration of 4-methoxyphenylacetic acid (4MPLA) was measured, there was a significant difference in concentrations between patients and healthy controls (P=0.008). Furthermore, ROC analysis for 4MPLA also showed high sensitivity (82.1%) and specificity (88.2%), with an AUC of 0.85. The composite process combining miRNA and metabolite expression exhibited nearly 100% sensitivity and specificity, with an AUC of 1, indicating an extremely high predictive value (48). This study suggests that, in the ongoing validation of NSCLC detection, this combined approach may potentially improve early detection, thereby enhancing treatment and survival outcomes for patients.

Urinary metabolites as prognostic marker

Recent studies have also explored the prognostic value of urinary metabolite biomarkers. One study measured the levels of prostaglandin E2 (PGE2), specifically the prostaglandin E-major urinary metabolite (PGE-MUM), in patients who underwent complete surgical resection for NSCLC. Urine samples collected before surgery and 3–6 weeks post-surgery revealed that an increase in post-surgical PGE-MUM levels was significantly associated with worse survival outcomes (hazard ratio 3.017, P=0.005) (49). This suggests that urinary PGE-MUM could serve as a valuable prognostic biomarker, helping to identify patients at higher risk of cancer recurrence following surgery.

Another study involving 529 participants, including 178 lung cancer patients and 351 healthy individuals, reported that CR and NANA levels were positively correlated with tumor size. Unlike previous studies, this study utilized UPLC-MS/MS for metabolite measurement and additionally evaluated correlations with lung cancer metrics, smoking status, gender, and race. The results indicated that CR and NANA levels were higher in smokers compared to non-smokers; however, the association with lung cancer risk persisted even after adjusting for smoking status and quantity. The findings of this study suggest that CR and NANA are associated with lung cancer risk (AUC =0.84, 0.90), indicating a stronger association in Caucasian Americans (50).

Overall, the growing body of evidence underscores the significant potential of urinary metabolomics in lung cancer detection and prognosis. The non-invasive nature of urine collection, combined with the sensitivity of advanced metabolomic profiling techniques, makes it a promising tool for clinical applications. As research continues to identify and validate additional biomarkers, urine-based diagnostics could play a pivotal role in improving early detection, guiding treatment decisions, and monitoring disease progression in lung cancer patients.

Fluid-based lung cancer metabolomics beyond blood and urine

In addition to blood and urine, several other biofluids, such as sputum, saliva, BALF, and pleural effusion, have shown potential in the discovery of metabolite biomarkers for lung cancer. These fluids, closely linked to the respiratory system, provide a more localized perspective on metabolic changes occurring in lung tissue and its microenvironment. Recent studies have explored their unique metabolite profiles, revealing biomarkers that can aid in the early detection and diagnosis of lung cancer. This section reviews the current literature on metabolite biomarker research utilizing these alternative fluid samples, highlighting their diagnostic and clinical significance (Table 3).

Sputum and saliva are easily accessible biofluids that contain both cellular and non-cellular components reflective of the respiratory tract’s condition. Sputum, derived directly from the lower airways, contains exfoliated epithelial cells, mucus, and other secretions, making it a valuable sample for studying lung cancer-associated metabolic changes. In a pioneering study from 2016, the first analysis of sputum metabolites was conducted using spontaneous sputum collected from 34 patients with suspected lung cancer and 33 healthy controls, with 23 patients later diagnosed with lung cancer, including 16 with NSCLC and 6 with SCLC. Using flow infusion electrospray ion mass spectrometry and GC-MS, the study identified discriminatory metabolites through principal component analysis, analysis of variance (ANOVA), and Random Forest, revealing metabolites such as ganglioside GM1 (monosialotetrahexosylganglioside) with AUC values greater than 0.8 that could aid in non-invasive diagnostic approaches and screening programs for at-risk populations (51). Another study selected key metabolites for distinguishing between lung cancer classes using flow infusion electrospray ionization mass spectrometry and evaluated their predictive power through artificial neural networks. The model demonstrated exceptional classification performance, achieving an AUC of 0.99, with sensitivity and specificity rates of 96% and 94% for cancer detection, and identified six metabolites (phenylacetic acid, L-fucose, caprylic acid, acetic acid, propionic acid, and glycine) that accurately discriminated SCLC from NSCLC, with a sensitivity of 80% and specificity of 100% (52). In a paper published in 2021, neutral desorption extractive electrospray ionization mass spectrometry identified 19 altered sputum metabolites in 143 patients with lung AD (76 cases) compared to controls (67 cases), with five key biomarkers (hydroxyphenyllactic acid, phytosphingosine, N-nonanoylglycine, sphinganine, and S-carboxymethyl-L-cysteine) demonstrating significant diagnostic potential (AUC >0.75). Metabolic pathway analysis revealed that sphingolipid metabolism, fatty acid metabolism, carnitine synthesis, and the Warburg effect were notably related to lung cancer (53). A study analyzed salivary metabolites from 41 lung cancer patients and 21 benign lung lesion patients using capillary electrophoresis mass spectrometry, revealing significant differences in metabolite profiles. The multiple logistic regression model identified diethanolamine, cytosine, lysine, and tyrosine as potential non-invasive biomarkers for distinguishing lung cancer from benign lesion, with an AUC of 0.729, further supporting the use of saliva-derived metabolites in clinical diagnostics (54).

BALF is a sample obtained through bronchoscopy, providing a more invasive but highly informative collection of biofluid from the lower respiratory tract. BALF offers direct access to the microenvironment of the lungs, including alveoli and bronchioles, making it a rich source of information on local metabolic alterations in lung cancer. The metabolomic profiling of BALF has been pivotal in identifying key metabolites involved in inflammation, immune responses, and cancer metabolism. Using two complementary metabolomic techniques—direct infusion high-resolution mass spectrometry and GC-MS—42 altered metabolites were identified in BALF samples from 24 lung cancer patients. Metabolic pathway analysis revealed significant alterations in glutamate and glutamine metabolism, while glycerol and phosphoric acid demonstrated potential as sensitive and specific biomarkers for lung cancer diagnosis and prognosis (55). In another analysis of BALF samples, 16 altered metabolites were identified in lung cancer patients, with partial least squares discriminant analysis (PLS-DA) score plots indicating strong statistical performance (R²Y=0.992, Q²=0.763). Among the altered metabolites, only glycerol and phosphoric acid demonstrated AUC values greater than 0.75, and an Euler diagram highlighted that six metabolites were consistently altered across serum, urine, and BALF samples, with four of them showing concordant directional changes in all three (56).

Pleural effusion, an accumulation of excess fluid between the layers of the pleura surrounding the lungs, is often associated with advanced lung cancer. This fluid can contain malignant cells, immune cells, and a variety of biomolecules, including metabolites that are indicative of tumor activity and the body’s response to cancer. The metabolomics of pleural effusion has uncovered specific metabolites linked to cancer aggressiveness and metastasis, making it a valuable source for identifying biomarkers that could assist in determining the stage of disease and guiding treatment strategies. In a study analyzing metabolomes of 32 malignant pleural effusions from lung cancer patients and 18 benign effusions from pulmonary tuberculosis patients using LC-MS/MS, free fatty acid (FFA) 18:1 (oleic acid) was identified as a potential biomarker with an AUC of 0.96. Notably, the ratio of FFA 18:1 to ceramide (d18:1/16:0) further improved diagnostic performance to an AUC of 0.99, with sensitivity of 93.8% and specificity of 100%, highlighting its clinical utility for diagnosing malignant pleural effusions (57). In another study, distinct metabolic profiles revealed 25 downregulated ether lipids, including plasmalogen phosphatidylcholine (plasmalogen PC) (40:3p) with an AUC of 0.953, and upregulated oxidized polyunsaturated fatty acids in malignant pleural effusions from lung cancer patients (n=46) compared to benign pleural effusions from tuberculosis patients (n=32), indicating high oxidative stress (58).

These findings underscore the significant potential of fluid-based metabolomics beyond traditional blood and urine samples in lung cancer research. By leveraging the unique metabolic insights gained from sputum, saliva, BALF, and pleural effusion, researchers can enhance early detection and improve diagnostic accuracy, ultimately paving the way for more personalized and effective treatment strategies for lung cancer patients. The current evidence base is still in its early stages, with many studies being exploratory and involving relatively small cohorts. Standardization of methodologies and further validation in larger, independent populations will be crucial to fully realize the clinical applicability of these novel biomarker sources.

Collectively, this review presents promising metabolite biomarkers with sufficient accuracy, categorized by sample type and clinical purpose. However, many of the studies have focused primarily on diagnosis; therefore, further research on biomarkers for prognosis and treatment response will be necessary.

Conclusions

This review underscores the growing importance of metabolomics in the diagnosis, prognosis, and management of lung cancer using body-fluid-based biomarkers (Figure 2). Traditional diagnostic methods, such as imaging and biopsies, remain invasive and costly, motivating the search for non-invasive alternatives. Metabolomics, by analyzing metabolites in blood, urine, sputum, and pleural fluid, offers critical insights into the metabolic reprogramming that occurs in cancer cells. Studies have identified specific biomarkers—such as lactate, methanol, and proline—that can accurately differentiate between lung cancer patients and healthy individuals. These biomarkers not only aid in early detection but also facilitate the staging of lung cancer, subtype classification, and disease monitoring. Additionally, machine learning models have enhanced the precision of metabolomics by uncovering complex metabolic patterns that predict therapeutic responses and highlight treatment-resistant cases.

Figure 2.

Significant biomarkers and associated pathways in lung cancer patients from various samples. This figure illustrates the differential metabolites identified in lung cancer patients compared to healthy individuals across different biological samples. In blood samples, metabolites such as C16:0-carnitine, tryptophan, and tyrosine were significantly upregulated in lung cancer patients, while choline, glucose, glutamine, serine, and threonine were downregulated. In urine samples, creatine riboside and N-acetylneuraminic acid showed higher levels in lung cancer patients. Pleural effusion samples revealed lower levels of oleic acid and ceramide in lung cancer patients, while bronchoalveolar lavage fluid samples showed increased levels of tyrosine and decreased levels of glycine, glycerol, and phosphoric acid. Pathway analysis linked upregulated metabolites to arginine biosynthesis, while downregulated metabolites were associated with the Valine, leucine, and isoleucine biosynthesis pathways. These findings suggest potential biomarkers and metabolic pathways relevant to lung cancer. ↑ indicates upregulation in lung cancer; ↓ indicates downregulation in lung cancer. The figure is created via BioRender with credit. PC, Phosphatidylcholine.

Integrating metabolomics into clinical practice holds significant potential for addressing key challenges, including tumoral heterogeneity and drug resistance. Beyond diagnosis, metabolomics paves the way for personalized treatment by targeting metabolic vulnerabilities such as lipid and amino acid metabolism. As research expands, it is expected to play an increasingly central role in guiding treatment strategies, enhancing early detection, and improving patient outcomes.

Looking ahead, future research should focus on leveraging metabolomics to predict responses to ICIs, given the variability in patient outcomes with immunotherapy. Additionally, further investigation into the metabolic changes associated with cancer cachexia—a severe complication in advanced cancer—could improve prognosis and identify novel therapeutic targets. By deepening our understanding of these areas, metabolomics-based research can help address current clinical challenges, leading to better cancer management, optimized treatment responses, and improved quality of life for lung cancer patients.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.