Fluid biopsy versus conventional biopsy – a revolutionary approach in oral cancer diagnosis: a systematic review and meta-analysis

Introduction

Oral squamous cell carcinoma (OSCC), comprising over 90% of oral cancers, ranks as the sixth most common malignancy globally, with high prevalence in South Asia due to risk factors, such as tobacco, betel quid, and alcohol use [1, 2]. Despite therapeutic advances, the 5-year survival rate remains poor due to late diagnosis and recurrence [3]. Diagnosis traditionally relies on visual examination followed by incisional biopsy and histopathology, the gold standard, but this method is invasive, time-consuming, and limited in capturing tumor heterogeneity [4, 5].
Liquid biopsy (LB), a minimally invasive technique analyzing tumor-derived components in blood or saliva, has emerged as a promising diagnostic tool. It detects circulating tumor DNA (ctDNA), circulating tumor cells (CTCs), microRNAs (miRNAs), and extracellular vesicles (EVs), providing dynamic insights into tumor behavior [6-8]. Its utility has been demonstrated in lung, breast, and colorectal cancers [9]. In OSCC, early studies have reported high sensitivity and specificity using salivary and plasma biomarkers, such as miR-21, miR-210, cell-free DNA (cfDNA) integrity, and EV-associated miRNAs [10-12].
Despite promising results, no systematic review has comprehensively evaluated the diagnostic performance of LB in OSCC. Comparisons across fluid types (saliva vs. plasma), biomarkers, and conventional biopsy are lacking, while the current literature remains fragmented, with varied methodologies and outcome metrics, complicating clinical interpretation.
This review aimed to assess the diagnostic accuracy of LB in OSCC, stratified by biofluid and biomarker type. It also compared LB with standard tissue bio­psy, identified promising molecular targets, and highlighted methodological limitations to guide future research. LB has evolved from a concept in systemic cancers to a viable diagnostic approach in head and neck onco­logy. Its early application in OSCC, highlighted by studies, e.g., Gai et al. [13] and Lin et al. [16], demonstrated cfDNA and EV-miRNAs in saliva and plasma, paving the way for the field of “salivaomics”. Saliva’s proximity to the tumor and ease of collection make it an ideal medium.
Advances in detection methods, including ddPCR, qPCR, and NGS, have improved diagnostic performance. For example, Mattox et al. [7] found ddPCR superior in HPV-positive oropharyngeal cancers. Salivary/plasma EV-miR-210 and miR-375-3p have shown > 90% sensitivity and area under the curves (AUCs) near 0.95 [8, 12]. cfDNA integrity markers (ALU115/ALU60, ALU247/ALU60) were significantly elevated in OSCC patients [12, 13]. However, heterogeneity in biomarker selection, sample handling, and detection techniques still persist. Most studies are small and single-center, limiting generalizability. Standardized diagnostic thresholds and regulatory approval are lacking [14].
LB may soon be used for screening high risk populations, real-time disease monitoring, and early recurrence detection [8]. Combining biomarker classes with AI-driven multi-omics could enhance diagnostic accuracy [15-17], while the progress in point-of-care salivary testing may enable rapid, chairside screening, especially in low-resource settings.
Clinical-grade assay development, standardized pipelines, and multicenter validation would be a key to integrating LB into routine OSCC care. Therefore, while promising, LB requires further validation before becoming a clinical mainstay in oral oncology.

Material and methods

Protocol and registration

This systematic review and meta-analysis followed the PRISMA 2020 guidelines, and was registered in PROSPERO (No. CRD420251002067).

Eligibility criteria

The included studies were original clinical research on LB diagnostics in OSCC, involving biomarkers, such as CTCs, cfDNA, miRNAs, mRNA, or salivary exosomes in saliva or plasma. Only peer-reviewed English-written studies reporting sensitivity, specifi­city, and AUC were considered. Excluded were reviews, case reports, editorials, conference abstracts, non-human studies, and papers lacking full-text or diagnostic data.

Information sources and search strategy

PubMed, Scopus, and Web of Science were searched for studies published between June 2019 and June 2024 using the following terms: (“liquid biopsy” OR “fluid biopsy”) AND (“oral cancer” OR “OSCC”) AND (“biomarkers” OR “cfDNA” OR “miRNA” OR “salivary exosomes”). Reference lists were also screened.

Study selection

After duplicate removal in EndNote, two reviewers independently screened titles and abstracts, and assessed full texts for eligibility. Disagreements were resolved via discussion or a third reviewer. The PRISMA flowchart in Figure 1 outlines the selection process.

Data collection

A standardized form captured study details (author, year, location, design, sample size), biomarker info (biofluid, detection method, biomarker type), diagnostic metrics (sensitivity, specificity, AUC, DOR), and risk of bias (QUADAS-2).

Risk of bias assessment and effect measures

QUADAS-2 was used by two reviewers to evaluate bias across four domains. The results are shown in Table 5.
Primary measures included sensitivity, specificity, AUC, and DOR.

Synthesis methods

In meta-analysis, pooled estimates were calculated using a bivariate random-effects model; summary receiver operating characteristic (sROC) curves and forest plots were generated with RevMan 5.3, STATA 17, and Meta-Disc. In subgroup analysis, saliva versus plasma biomarkers (Table 3) were compared.

Heterogeneity assessment and reporting bias

Cochran’s Q and I² statistics were employed, and an I² of 0% indicated no heterogeneity (Table 6). Deeks’ funnel plot test assessed publication bias, while p > 0.10 signified no significant bias.

Results

Study selection

A total of 925 records were identified through electronic searches of PubMed, Scopus, and Web of Science. After removing 120 duplicates, 805 records were screened by title and abstract. Of these, 585 articles were excluded based on irrelevance or ineligibility. The remaining 28 full-text articles were retrieved for detailed assessment; however, six articles could not be accessed due to unavailability. Following full-text review of the 22 remaining studies, 12 met the inclusion criteria for the systematic review. Of these, 10 studies were eligible for inclusion in the meta-analysis based on availability of complete diagnostic data. The overall study selection process is outlined in the PRISMA 2020 flow diagram (Figure 1).

Study characteristics

The included studies evaluated a diverse range of fluid biopsy biomarkers for the diagnosis of OSCC, including circulating cfDNA, ctDNA, miRNAs, sali­vary exosomes, and HPV DNA. These biomarkers were assessed using various platforms, such as ELISA, PCR-based techniques, and next-generation sequencing. Sample sizes and study designs varied, with most studies being case-control in nature. The detailed summary of the biomarkers investigated and their reported utility is presented in Table 1, whereas Table 2 provides the comprehensive overview of each study’s diagnostic performance, including sensitivity, specificity, and AUC.

Meta-analysis results

The meta-analysis was conducted on 10 studies, which met the inclusion criteria for quantitative synthesis. The pooled analysis using a bivariate random-effects model yielded an AUC of 0.894, with a 95% confidence interval ranging from 0.874% to 0.915%. This indicated high diagnostic accuracy for fluid biopsy biomarkers in detecting OSCC. The sROC curve and forest plots (Figure 2 and Figure 3, respectively) further demonstrated consistent performance across the studies.
The sensitivity and specificity values varied across individual studies, ranging from 74.2% to 92.3% and 63.9% to 100%, respectively. Among the highest-performing biomarkers were plasma-derived EV-miR-210 and salivary miR-375-3p, each with reported AUC exceeding 0.95. The individual study AUCs and corresponding confidence intervals are visualized in Figure 4.

Subgroup analysis: Saliva vs. plasma

The subgroup analysis comparing salivary and plasma biomarkers revealed that both fluid types exhibited strong diagnostic capabilities, although plasma-derived markers demonstrated slightly higher accuracy. For example, EV-miR-210 in plasma showed an AUC of 0.951, and several salivary miRNAs, including miR-133a-3p and miR-375-3p, also achieved high AUCs and diagnostic precision. The comparative results are summarized in Table 3.

Statistical approach for measuring sensitivity and specificity

A bivariate random-effects model was used to pool sensitivity and specificity. Diagnostic accuracy across the studies was visualized using forest plot (Figure 3), while an overall assessment was conducted through a sROC curve (Figure 2).
Findings indicated that extracellular vesicle-associated miRNAs and cfDNA integrity indexes exhibited strong diagnostic performance, with some biomarkers achieving high sensitivity and specificity. Salivary biomarkers, particularly miRNAs, showed promising accuracy for early OSCC detection, while plasma-based biomarkers, such as EV-miR-210, demonstrated the highest diagnostic accuracy, supporting their potential clinical application. Additionally, HPV16 DNA analysis was demonstrated as a valuable tool for identifying HPV-associated oral cancers [7, 12]. Overall, LB techniques provided a reliable and less invasive diagnostic approach, with some biomarkers achieving accuracy comparable with traditional methods (Table 2). The subgroup analysis comparing saliva and plasma as fluid biomarkers is presented in Table 3.
Although fluid biopsy methods (cfDNA, miRNAs, and EVs) provide high sensitivity and specificity, their accuracy is slightly lower than conventional biopsy. However, cfDNA and ctDNA demonstrate diagnostic performance nearing that of tissue biopsy.

Study quality interpretation

The methodological quality of the studies included was evaluated using the Newcastle-Ottawa scale (NOS). Five papers, including Bigagli et al. [12] and Sayal et al. [18], received high-quality ratings (8-9/9), indicating strong methodology and reliable findings [7, 8, 11, 19]. Four studies, e.g., Lin et al. [6] and Romani et al. [20], were rated as moderate quality (5-7/9), primarily due to lower comparability, which may introduce bias [2, 13]. Higher quality studies should be prioritized in meta-analyses, while sensitivity analysis may be necessary to assess the impact of lower quality studies on diagnostic accuracy (Table 4).

Risk of bias assessment using QUADAS-2 interpretation

The risk of bias was assessed using the QUADAS-2 tool, which evaluates bias across four domains: patient selection, index test, reference standard, and flow and timing. Most studies were rated as having low risk across all domains. However, two studies, i.e., Lin et al. [6] and Mattox et al. [7], were identified as having a high risk of bias, primarily due to concerns in patient selection and unclear reporting of sample handling and follow-up. The complete risk of bias evaluation is shown in Table 5.

Heterogeneity assessment

Heterogeneity among the included studies was evaluated using the Cochran’s Q test and I² statistic. The results showed no significant heterogeneity, with p-values > 0.10 and I² values of 0% for both sensitivity and speci­ficity (Table 6). This indicated that the included studies were highly consistent in their findings, which strengthens the validity of the pooled estimates. Diagnostic odds ratio The diagnostic odds ratios were calculated based on true positive, false negative, false positive, and true negative values. Several biomarkers demonstrated exceptionally high DOR values, indicating strong discriminatory power, helping in evaluating the diagnostic accuracy of each biomarker.
Biomarkers with high DOR values (> 50), such as plasma EV-miR-210 (Bigagli et al. [12], DOR = 77.468) and miR-133a-3p and miR-375-3p (Crimi et al. [8], DOR = 51.000), showed exceptional diagnostic potential for OSCC detection. Biomarkers with moderate DOR values (range, 12-30), including those in studies by Rapado-González et al. [11] (13.694), Sayal et al. [18] (13.943), and He et al. [19] (12.803), showed reasonable diagnostic performance, but were less effective than the strongest biomarkers. However, Gai et al. [13] (DOR = 25.815) exhibited a strong balance of sensitivity and specificity, making miR-512-3p and miR-412-3p promising candidates for OSCC detection (Figure 5).
Overall, biomarkers with higher DOR values, as shown by Bigagli et al. [12], were more effective for OSCC diagnosis, while those with moderate DOR values may still be valuable, especially when used in combination with other biomarkers (Figure 5). It is essential to interpret DOR alongside AUC, sensitivity, and specificity, to provide a comprehensive assessment of diagnostic accuracy.
The publication bias was assessed using the Deeks’ funnel plot asymmetry test (specific for DTA studies). Since the p-value was greater than 0.10, there was no significant evidence of publication bias in the included studies (Figure 6), which suggest that smaller studies do not disproportionately report more extreme AUC values.

Discussion

This systematic review and meta-analysis highlight the emerging role of LB as a non-invasive, real-time diagnostic tool for OSCC. The data strongly support the diagnostic value of saliva and plasma in detecting molecular biomarkers, such as ctDNA, cfDNA, miRNAs, and EVs. These biomarkers show promise in supplement­ing or potentially replacing traditional tissue biopsies, offering a less invasive, repeatable, and dynamic method for early detection and monitoring of OSCC.

Cell-free DNA and ctDNA

cfDNA and ctDNA have been extensively studied for their potential as tumor-specific, blood- or saliva-based markers. These DNA fragments released into the bloodstream or saliva by apoptotic and necrotic tumor cells, carry valuable genomic information about the tumor. Lin et al. [6] reported significantly elevated cfDNA levels in OSCC patients compared with healthy controls, correlating with tumor size, lymph node metastasis, and late-stage disease. High cfDNA levels also indicated poorer prognosis, supporting its potential as a diagnostic and prognostic biomarker. Similarly, Cheng et al. [21] emphasized the utility of ctDNA in mutation detection, disease monitoring, and early diagnosis, noting its high sensitivity and specificity. Salivary cfDNA has also gained attention for its diagnostic value. Rapado-González et al. [11] showed a sensitivity of 83.3% and specificity of 73.3% for salivary cfDNA using ALU115/ALU60 integrity indices, with an AUC of 0.82, indicating strong discriminative power for OSCC. Also, Sayal et al. [18] reported that both cfDNA and mitochondrial DNA in saliva can serve as effective screening tools in head and neck cancers. Circulating tumor cells CTCs represent another minimally invasive biomarker in LB. As indicators of metastatic potential, their presence in peripheral blood can aid in disease staging and monitoring therapeutic response.
Perumal et al. [22] evaluated the CTC detection in head and neck cancer patients, and proposed its utility in assessing radiotherapy outcomes. Similarly, Horgan et al. [5] emphasized the clinical value of CTCs in capturing tumor heterogeneity and tracking genomic changes over time. When combined with ctDNA analysis, CTC detection enhances real-time cancer surveillance, enabling more precise treatment decisions [14].

MicroRNAs

miRNAs are small non-coding RNAs involved in post-transcriptional regulation, and their dysregulation is associated with tumorigenesis. Their stability in body fluids makes them ideal candidates for LB-based diagnostics. Crimi et al. [8] reported that salivary miR-133a-3p and miR-375-3p were significantly downregulated in OSCC patients. ROC analysis revealed an AUC of 0.96, indicating excellent diagnostic performance. Also, Romani et al. [20] identified miR-423-5p through a genome-wide salivary miRNA study as a promising diagnostic marker for OSCC. Bigagli et al. [12] focused on plasma EV-associated miR-210, which demonstrated high diagnostic accuracy (AUC of 0.95), with sensitivity and specificity values of 92.3% and 86.6%, respectively. Beyond diagnosis, miR-210 was also found to correlate with disease-free survival, supporting its prognostic relevance.

Extracellular vesicles and exosomal miRNAs

EVs, particularly exosomes, protect and transport miRNAs, DNA, and proteins, enhancing their stability in circulation. Several studies confirm the potential of EV-associated miRNAs in OSCC detection.
Gai et al. [13] identified salivary EV-miRNAs, such as miR-302b-3p, miR-517b-3p, miR-512-3p, and miR-412-3p as OSCC-specific biomarkers, with ROC analysis confirming their strong discriminatory ability. He et al. [19] further validated miR-24-3p as a salivary exosomal biomarker with reliable screening potential. These findings point to the unique advantage of EV-miRNAs in improving both diagnostic accuracy and patient compliance due to their non-invasive nature.

Clinical utility of salivary biomarkers

Among all LB sources, saliva stands out for its ease of collection, safety, and close interaction with the OSCC tumor microenvironment. Huang et al. [1] emphasized saliva as a rich source for diagnostic biomarkers. Rebaudi et al. [2] proposed a cyto-salivary sampling method combined with sensitive ELISA immunoassays to detect OSCC-specific proteins, such as EGFR, Ki67, p53, PD-L1, HLA-E, and B7-H6. Their findings showed strong correlation with histopathological results, supporting this method as a reliable alternative to tissue biopsy.
Recent studies further support the role of salivary and LB-based biomarkers in oral cancer detection. Chundru et al. [23] demonstrated the diagnostic potential of salivary interleukin-6 levels for distinguishing oral squamous cell carcinoma and potentially malignant disorders. The broader context of genomic reproducibility and LB standardization was discussed by Baykal et al. [24], emphasizing the importance of reliable bioinformatics pipelines. In addition, non-invasive genetic screening approaches for oral cancer prediction were explored by Poell et al. [25].
A scoping review by González-Moles et al. [26] highlighted the ongoing challenges in early oral cancer diagnosis, and emphasized the need for novel diagnostic methods, such as LB. Lousada-Fernandez et al. [27] discussed the promise of LB approaches, particularly focusing on saliva-derived biomarkers in oral cancer. The effects of sample collection and processing on plasma-derived cfDNA, critical for ensuring analytical accuracy, were evaluated by Risberg et al. [28]. Tivey et al. [29] further elaborated on the potential of ctDNA beyond blood, expanding its applicability in cancer diagnostics.
Additionally, an advanced deep learning-based multi-method analysis integrating histopathological images and molecular biomarkers for early OSCC detection was reported by Ahmad et al. [30].

Conclusions

Collectively, the reviewed studies demonstrate that LB, particularly through saliva and plasma, is a robust and reliable method for OSCC detection and monitoring. While miRNAs and cfDNA demonstrate high dia­gnostic accuracy, ctDNA and CTCs offer added value in mutation tracking and disease progression. The integration of EVs further enhances biomarker stability and specificity. However, before LB can become standard practice, additional validation in large-scale, multi­center clinical trials is necessary to standardize protocols, improve sensitivity, and confirm long-term utility in diverse patient populations.

Disclosures

1. Institutional review board statement: Not applicable.
2. Assistance with the article: None.
3. Financial support and sponsorship: None.
4. Conflicts of interest: None.

References