3.1. Types of VOCs and samples

A total of 268 VOCs were reported across the selected studies. When classified based on chemical structure, alkenes were the most frequently reported accounting for 20.9 % of all VOCs followed by ketones (15.7 %), alkanes (14.6 %), alcohols (12.3 %), carboxylic acids (12.3 %), aldehydes (7.5 %), aromatic compounds (5.6 %), esters (4.5 %), sulfides (1.9 %), and amines (1.5 %). Additionally, heterocyclic compounds, ethers, and alkaloids each comprised 0.7 %, while pyridines, amides, and alkynes each accounted for 0.4 % (Fig. 2).

Among the VOCs identified, the eight most frequently reported compounds were cited in at least four and up to seven different studies. These eight VOCs including phenol, 2-ethyl-1-hexanol, acetic acid, p-cresol, octanoic acid, 2-pentylfuran, guaiacol, 4-tert-butylphenol with their chemical classification, names, CAS numbers, molecular weights, structures, frequencies in which they were reported, and sample matrixes have been compiled in Table 2. The eight VOCs summarized in Table 2 exhibit distinct olfactory properties that are relevant to both sensory evaluation and their biological or environmental origins. Phenol was the most frequently mentioned VOC across the literature. Among these compounds, alcohols were the most prevalent, comprising 63 % followed by carboxylic acids at 25 %, and aromatic compounds at 12 %. All eight VOCs were detected in urine and 2-ethyl-1-hexanol and octanoic acid also identified in cell cultures and urine, while phenol and acetic acid were found in breath, cell cultures, saliva, and urine.

Phenol emits a sharp, medicinal, and smoky odor, often associated with antiseptic or burnt wood characteristics. Its relevance with breast cancer would be estrogenic activity linked to inflammation and microbial metabolism. 2-Ethyl-1-hexanol possesses a slightly sweet, oily scent with plastic-like undertones, commonly perceived as synthetic or industrial. Acetic acid is characterized by a strong, pungent, vinegarlike aroma that is widely recognized and highly volatile.

p-Cresol presents a barnyard or fecal odor, with tar-like notes often linked to animal waste and body odor. Similarly, octanoic acid (caprylic acid) has a rancid, fatty, and sweaty scent profile, typically associated with sebum and human body odor. 2- Pentylfuran contributes a green, beany, and nutty aroma, commonly attributed to lipid oxidation processes and plant-derived volatiles. Guaiacol, a methoxyphenol, is known for its distinctive smoky and woody aroma with subtle vanilla-like sweetness; it is widely used in flavoring compounds such as roasted coffee and smoked meats. Finally, 4-tert-butylphenol is perceived as mildly sweet but exhibits a synthetic, rubber-like odor frequently linked to industrial materials and certain perfumery applications.

In summary, while several of these compounds contribute pleasant or food-related aromas (e.g., guaiacol, 2-pentylfuran), others are associated with unpleasant or biologically derived odors (e.g., phenol, p-cresol, octanoic acid), and some with artificial or environmental exposures (e.g., 2-ethyl-1-hexanol, 4- tert-butylphenol). These odor characteristics provide insight into their potential roles as biomarkers or sensory indicators in environmental or biomedical contexts. These findings would highlight the diversity of VOCs associated with breast cancer and their potential as diagnostic biomarkers.

3.2. Sampling methods

The key parameters were matrix and number of samples, temperature of sample storage, age and BMI of breast cancer patient. An analysis of sample matrixes used in the reviewed studies revealed that urine was the most commonly utilized accounting for 46 % of studies. This was followed by cell samples (33 %), exhaled breath (17 %), and saliva (4 %). Although cell samples were the second most frequently used sample type, their applicability in clinical settings is limited as they do not fully account individual metabolic variability. As a result, their diagnostic accuracy in real-world applications remains uncertain. There were very few studies that explored noninvasive sampling methods such as skin secretions or sweat. It highlights the need for further research in these areas to expand the applicability of VOC-based breast cancer diagnosis.

The number of samples used in each study varied significantly ranged between 2 and 256. The age of sampled patients were also varied between 18 and 85. Most studies did not specify body mass index (BMI) as an inclusion criterion, except one study. It restricted the BMI range between 17 and 28, while another limited participants to those without hypertension or diabetes. Regarding sample storage temperatures, 74 % of studies reported that -80 °C was the most frequently used. Additionally, -20 °C and 4 °C were each used in 10.5 % of studies, while -30 °C was reported in 5 % of cases (Fig. 3).

3.3. HS-SPME conditions

Solid-phase microextraction (SPME) is a widely adopted sample preparation technique for the extraction and preconcentration of volatile and semi-volatile organic compounds in complex biological matrixes.⁸⁴ The technique relies on a coated fiber, typically made of fused silica and embedded with polymer, which is exposed to the sample matrix or its headspace. Analytes partition between the sample and the fiber coating based on their physicochemical properties. After the extraction phase, the fiber is transferred directly into the heated injection port of a gas chromatograph, where the analytes are thermally desorbed for subsequent analysis (Fig. 4).⁸⁵ SPME offers several advantages, including high sensitivity, minimal sample preparation, low solvent consumption, and compatibility with automation. It is especially useful for VOC analysis in breath, urine, and cell culture media, where matrix interferences can otherwise complicate direct injection techniques.⁸⁶ The efficiency of SPME depends on various factors such as fiber coating type, extraction temperature and time, agitation, and sample pH or salinity. Because of its simplicity and effectiveness, SPME has become an essential technique in VOC biomarker discovery and has been frequently applied in cancer-related VOC studies, including those targeting breast cancer.⁸⁷

Key parameters included the type of SPME fiber, the temperature and time of saturation, adsorption, and desorption, as well as the chromatographic column and analytical instrument employed. The comparative results of these conditions are summarized in Figs. 5 and 6. The most commonly used SPME fiber was the moderately polar DVB/CAR/PDMS (Divinylbenzene/Carboxen/Polydimethylsiloxane), accounting for 35 % of studies. Carboxen is a porous carbonbased sorbent with a high surface area, widely used in SPME fibers for its strong affinity toward small and volatile analytes, particularly in trace-level of VOC analysis.⁸⁸ The low-polarity CAR/PDMS (Carboxen/Polydimethylsiloxane) was used at 26 %, the non-polar PDMS/DVB (Polydimethylsiloxane/ Divinylbenzene) at 15 %, and the non-polar PDMS (Polydimethylsiloxane) and polar PA (Poly(acrylic acid)) at 6 % each, respectively. Other fiber types including DVB/Car1000/CarX (Divinylbenzene/ Carborane-Silica(1000)/Carborane-Silica(X)), CW/ DVB (Carbowax/Divinylbenzene), and the nonpolar ZIF-7 (Zeolitic imidazolate framework-7) were used in 3 % of the studies (Fig. 5). Car1000/CarX fiber is composed of carborane–silica composites which was designed to enhance the adsorption of a wide range of volatile and semi-volatile organic compounds, particularly under low-polarity and thermally stable conditions.⁸⁹ Most studies utilized commercially available SPME fibers (e.g., DVB/ CAR/PDMS, CAR/PDMS, PDMS/DVB, PDMS, PA). However, given the lack of detailed information of SPME fibers for breath samples, it is reasonable to infer that breath sample was collected from the sampling bag and injected into the analytical instrument directly.

The sample volume used for analysis varied significantly among studies ranging between 5 μL and 1 L. The large variation in sample volumes might be attributed to differences in sample matrixes. Breath samples typically required larger volumes, while cell samples were generally processed in smaller volumes than gas samples. To enhance VOC extraction efficiency, 32 % of the studies added sodium chloride (NaCl), while others incorporated additives such as hydrochloric acid (HCl), protease inhibitors, and doxorubicin. Doxorubicin (DOX) is one of the most widely used and effective chemotherapy drugs, often administered alone or in combination with other agents. It has a broad spectrum of activity and is commonly used to treat both solid tumors and hematologic cancers, including breast cancer, sarcomas, and pediatric malignancies.⁹⁰ In addition to additives, 29 % of the studies adjusted the sample pH to a highly acidic range (pH 1~3). Sample pH was mostly controlled with the HCl solution to optimize VOC extraction. Most urine samples were treated with NaCl as it was, as NaCl was added to enhance VOC extraction efficiency from water-based matrixes.⁹¹ In contrast, pH adjustment was commonly performed to cell samples, as pH plays a critical role in the extraction efficiency and stability of VOCs in cellular matrixes.⁹²

The SPME saturation temperature was primarily set at 60 °C in 67 % of studies by considering the boiling points of VOCs which generally range between 50 °C and 100 °C. The saturation time was predominantly 30 min accounting for 80 % of studies. The SPME adsorption conditions varied based on the analytical instruments and columns used in different matrixes. However, adsorption at 60 °C was most common, which reported in 28 % of studies followed by 37 °C (22 %), 50 °C (22 %), 40 °C (17 %), 38 °C (6 %) and 200 °C (5 %). Adsorption time was 30 min in 30 % of studies while 75 min (20 %), 60 min (15 %), 20 min (10 %), 45 min (10 %), 10 min (5 %), 40 min (5 %), and 50 min (5 %) were also reported. A distinct trend was observed in the selection of saturation and adsorption temperatures between cell and urine samples. For cell samples, saturation and adsorption were predominantly performed at temperatures below 50 °C reflecting their thermally labile nature, whereas urine samples, which are more thermally stable, were generally processed at temperatures above 50 °C. For complete desorption of the adsorbed VOCs from the fiber, 90 % of studies conducted desorption at 250 °C while 5 % used 200 °C and 270 °C. The desorption time was 5 min and 6 min in 22 % of studies followed by 2 min (17 %), 10 min (17 %), 12 min (11 %), 0.5 min (6 %), and 3 min (5 %) (Fig. 6).

3.4. Detection and characterization of VOCs

Among the analytical instruments used for VOC analysis, gas chromatography/mass spectrometry (GC/ MS) was the most commonly employed, accounting for 77 % of studies. GC has been widely adopted analytical technique that separates volatile compounds through differential partitioning between a stationary phase and a mobile gas phase to enable high-resolution analysis of complex mixtures.⁹³ Because of its high sensitivity, reproducibility, and capability for VOC identification, the combination of chromatographic separation with mass spectrometric detection allows for accurate qualitative and quantitative analysis of low-abundance VOCs, making it particularly wellsuited for biomarker discovery in cancer research. This technique enables the separation of complex mixtures and the detection of VOCs at trace levels, even in biologically noisy matrixes such as urine or breath.⁹⁴ Moreover, GC/MS offers a robust platform for interlaboratory comparability when standardized protocols are applied.⁹⁵ Gas chromatography/mass spectrometryquadrupole time-of-flight (GC/MS-QTOF) was used in 10 %, selected ion flow tube-chemical ionization/mass spectrometry (SIFT-CI/MS) was used in 7 % of cases, while gas chromatography-flame ionization detector (GC-FID) and secondary electrospray ionization-mass spectrometry (SESI-MS) were applied in 3 % of studies. GC/MS-QTOF offers high mass accuracy and resolution, making it suitable for untargeted VOC profiling and the identification of novel biomarkers.⁹⁶ Compound identification in GCbased methods is typically achieved by comparing acquired spectra against established mass spectral libraries, such as the NIST database.⁹⁷

The stationary phases of the GC columns used in these studies also varied significantly. Polyethylene glycol (PEG), a polar stationary phase, was the most frequently used, accounting for 40 % of studies. Columns with a 5 % phenyl methyl siloxane was used in 20 % of studies and columns with 6 % cyanopropyl/phenyl/ 94 % polydimethylsiloxane was 16 %. Other stationary phases included 6 % cyanopropylphenyl methylpolysiloxane, poly(5 % diphenyl/95 % dimethyl siloxane), nitroterephthalic-acid-modified polyethylene glycol, phenyl arylene polymer, silphenylene polymer (with polarity equivalent to poly(5 % diphenyl/95 % dimethyl siloxane)), and 100 % dimethylpolysiloxane, each of which was used in 4 % of studies (Fig. 7).

3.5. Challenges and limitations of VOC-based breast cancer diagnosis

Despite the potential of VOCs as biomarkers for breast cancer, several challenges must be addressed. The lack of standardization in sampling and analytical methods, and data interpretation hinders reproducibility and comparability among studies. A review of 30 studies reveals substantial variability in sample storage temperature, saturation, and adsorption temperatures and times during analysis depending on the matrix of sample. These inconsistencies highlight the need to establish standardized analytical methods to each sample matrix. Additionally, external factors such as diet, lifestyle, medication use, and environmental exposures can significantly influence VOC profiles, making it difficult to distinguish disease-specific biomarkers.^98,99 Notably, the majority of studies were conducted in limited countries such as China, Portugal, and the United States (Fig. 8). To better compare VOC patterns across different populations and support the development of standardized biomarkers, further research is necessary. Another major limitation is the lack of large-scale clinical validation, as many identified VOC biomarkers have not yet been rigorously tested for diagnostic reliability and specificity. Although GC/MS was the most commonly used instruments in the studies reviewed, its clinical implementation faces significant challenges. High-resolution techniques such as GC/MS and SIFT-CI/MS require costly and specialized equipment and expertise, which limits their accessibility and practicality. To overcome these limitations, future research should prioritize the establishment of standardized methodologies, the expansion of clinical studies, and the integration of advanced data analysis approaches to enhance the reliability and applicability of VOC-based breast cancer diagnosis.

4.1. Diagnostic accuracy and standardization of analytical methods

While numerous VOCs have been suggested as potential biomarkers for breast cancer, their diagnostic performance remains uncertain because of the limited clinical validation and inconsistent methodologies. Many studies differ in critical parameters such as sample type (e.g., breath, urine, or cell cultures), SPME fiber selection, adsorption temperatures and times, and chromatographic systems. These variations contribute to inconsistent results and limit crossstudy comparability. To improve diagnostic reliability, future research should aim to validate candidate VOCs in large, diverse populations and establish standardized methods for sampling, sample preparation, and analysis. It includes defining consistent storage conditions, extraction procedures, and implementing rigorous quality control measures such as internal standards, blank samples, and instrument calibration. Establishing methodological consensus will not only enhance reproducibility but also support multi-center collaborations and facilitate the clinical translation of VOC-based diagnostics.

4.2. Integration of machine learning for VOC pattern analysis

Given the complex and multivariate nature of VOC datasets, conventional statistical tools often fall short in capturing subtle patterns associated with disease states. Machine learning (ML), a subset of artificial intelligence that enables systems to learn patterns from data and make predictions without being explicitly programmed, offers a powerful approach to uncover nonlinear relationships and classify VOC profiles with high accuracy.¹⁰⁰ Algorithms such as support vector machines (SVM), random forest (RF), and artificial neural networks (ANN) have already demonstrated promising results in classifying cancer samples based on VOC signatures.¹⁰¹ SVMs are effective for high-dimensional data and operate by finding the optimal hyperplane that separates classes. RF is an ensemble method that constructs multiple decision trees to improve classification performance and reduce overfitting. ANNs mimic the structure of the human brain and are particularly powerful for modeling complex, nonlinear relationships. To utilize the full potential of ML, we should focus on collecting large, high-quality datasets that include diverse sample types and standardized measurements. This will allow for effective training and validation of predictive models. Additionally, explainable AI techniques should be applied to provide insights into the decision-making process of these models, so that it increases their interpretability and clinical trustworthiness. Combining ML with other omics data, such as genomics or metabolomics, could further enhance the specificity and sensitivity of breast cancer detection. Ultimately, the integration of machine learning into VOC research holds the potential to revolutionize early cancer diagnostics by enabling real-time, non-invasive screening tools.

4.3. Expanding research on breast cancerspecific VOCs and their metabolic pathways

Despite the growing interest in VOC-based cancer diagnostics, studies specifically focused on breast cancer VOCs remains limited compared to other cancer types such as lung or colorectal cancer. Moreover, the metabolic pathways responsible for the production and alteration of VOCs in breast cancer are still poorly understood. Elucidating these pathways is essential for validating VOCs as biologically meaningful biomarkers rather than incidental findings. Future investigations should integrate metabolomic profiling and pathway analysis to trace the biochemical origins of key VOCs associated with breast cancer. This will require interdisciplinary collaboration among analytical chemists, biochemists, and clinicians to bridge the gap between chemical signatures and biological mechanisms. Furthermore, in vitro and in vivo models should be used to simulate VOC production under various conditions such as tumor progression, treatment response, and hormonal influences. A deeper understanding of the metabolic context of VOCs will aid in identifying biomarkers with higher disease specificity and reduce the likelihood of false positives.