Decoding Brain Metastasis With Liquid Biopsy: Biomarkers and Technologies in EGFR ‐Mutant NSCLC

ABSTRACT

Brain metastasis (BM) is a leading cause of cancer‐related mortality, particularly in patients with lung cancer. Traditional diagnostic methods, such as tissue biopsies, are often impractical, and there are currently no reliable biomarkers for early detection in clinical practice. In this context, liquid biopsy has emerged as a valuable non‐invasive diagnostic tool for cancer patients with BM. This review summarizes the advantages and current limitations (such as the low sensitivity in early‐stage disease and the potential for false positives from non‐cancer sources) of liquid biopsy samples, specifically focusing on peripheral blood and cerebrospinal fluid (CSF). We discuss key biomarkers, including circulating tumor cells (CTCs), circulating tumor DNA (ctDNA), extracellular vesicles (EVs), and tumor‐derived platelets (TEPs), and current technologies for their detection, highlighting their potential clinical applications. Additionally, we explore the role of liquid biopsy in patients with epidermal growth factor receptor‐mutated (EGFR‐mutant) non‐small cell lung cancer (NSCLC), a group that experiences a high incidence of BM. Finally, we address emerging biomarkers and innovative detection methods, offering recommendations for enhancing the application of liquid biopsy in clinical settings. Our review aims to underscore the significance of liquid biopsy in improving the detection and management of BM.

Liquid biopsy, particularly cerebrospinal fluid analysis, plays a key role in managing brain metastasis in EGFR‐mutant NSCLC. This review highlights its superior ability to detect CNS‐specific resistance and guide targeted therapeutic decisions, offering a minimally invasive complement to tissue biopsy.

Abbreviations

AFM

atomic force microscopy

AI

artificial intelligence

BBB

blood‐brain barrier

BM

brain metastasis

B‐Pt‐Au NPs

branched Pt Au nanospheres

CAPP‐Seq

cancer personalized profiling by deep sequencing

circRNA

circular RNA

CM

classification model

CNS

central nervous system

CRP

C‐reactive protein

CSF

cerebrospinal fluid

CT

computed tomography

CTCs

circulating tumor cells

ctDNA

circulating tumor DNA

ddPCR

droplet digital polymerase chain reaction

EGFR‐mutant

epidermal growth factor receptor‐mutated

EMT

epithelial‐to‐mesenchymal transition

EpCAM

epithelial cell adhesion molecule

EVs

extracellular vesicles

HBMECs

human brain microvascular endothelial cells

HDS

high‐depth sequencing

ISET

isolation by size of epithelial tumor cells

LM

leptomeningeal metastasis

lncRNA

long non‐coding RNA

MACS

magnetic‐activated cell sorting

MAF

mutant allele frequency

MBP

myelin basic protein

MET

mesenchymal epithelial transition factor

miRNA

microRNA

MRI

magnetic resonance imaging

NGS

next‐generation sequencing

NSCLC

non‐small cell lung cancer

OS

overall survival

PCI

prophylactic cranial irradiation

PFS

progression‐free survival

SCLC

small cell lung cancer

sEV

small extracellular vesicle

SNPs

single nucleotide polymorphisms

TEPs

tumor‐derived platelets

TGFβ1

transforming growth factor beta 1

TMB

tumor mutational burden

TME

tumor microenvironment

TTI

time to treatment initiation

WES

whole exome sequencing

WGS

whole genome sequencing

1. Background

Advancements in precision medicine have led to earlier diagnoses of lung cancer compared to previous methods. Tissue biopsy is routinely performed during surgery for cancer classification, grading, and staging, thereby guiding subsequent treatment decisions. Despite these advancements, lung cancer remains the leading cause of cancer‐related mortality worldwide, accounting for approximately 50% of all BM cases [1]. 16%–60% of patients with NSCLC develop BM [2], with a significant subset of these BM patients in Asia harboring mutations in the EGFR gene [3]. Unfortunately, the blood–brain barrier (BBB) limits the effectiveness of conventional chemotherapy [4]. The distinct brain microenvironment may foster tumor heterogeneity through genetic and epigenetic mechanisms, thereby adding complexity to therapeutic decisions [5]. Furthermore, difficulties in obtaining tissue samples hinder the precise diagnosis of BM lesions. As a result, there is a growing need for non‐invasive methods to characterize tumors. Liquid biopsy offers a promising alternative for traditional tissue biopsy by detecting real‐time changes of biomarkers specific to tumors. The potential of cross‐molecular infrared fingerprints from serum and plasma has been investigated for lung cancer prognosis and stratification, paving the way for more personalized treatment decisions [6]. CTCs, ctDNA, platelets, and tumor‐derived EVs carry significant genetic information and are released by primary or metastatic tumors into body fluids, including peripheral blood, cerebrospinal fluid, saliva, and urine [7]. Although the BBB prevents the release of these biomarkers into peripheral blood, liquid biopsy provides a complementary diagnostic approach for patients with BM (Figure 1). While numerous reviews have addressed liquid biopsy in NSCLC or BM broadly, a critical synthesis specifically dedicated to EGFR‐mutant NSCLC—the subgroup at highest risk for BM—is lacking, particularly one that reconciles the unique challenges of the CNS compartment. This review aims to fill this gap by providing a focused analysis that bridges advances in liquid biopsy with the specific biological and clinical complexities of CNS metastases in this molecularly defined population.

FIGURE 1.

Molecular mechanisms of brain metastasis and the clinical translation pathway of liquid biopsy. This schematic illustrates the multi‐step process of brain metastasis originating from lung cancer, integrating key molecular pathways and biomarker applications. The left panel depicts the transition from the primary microenvironment to the metastatic brain microenvironment. Lung cancer cells undergo epithelial‐mesenchymal transition (EMT), characterized by upregulation of Snail, Twist, and N‐cadherin, facilitating migration, invasion, and intravasation into circulation as circulating tumor cells (CTCs). The right panel summarizes the clinical translation pathway of liquid biopsy using blood and cerebrospinal fluid (CSF). Key biomarkers such as CTCs, circulating tumor DNA (ctDNA), tumor‐educated platelets (TEPs), EVs, protein and metabolite enable early diagnosis, treatment monitoring, and prognostic assessment in patients with brain metastases. CSF, cerebrospinal fluid; CTC, circulating tumor cell; ctDNA, circulating tumor DNA; EMT, epithelial‐mesenchymal transition; EV, extracellular vesicle; IL6, interleukin‐6; TEP, tumor‐educated platelet; TGF‐β, transforming growth factor‐beta; TNF‐α, tumor necrosis factor‐α.

2. Biomarkers for Liquid Biopsy in NSCLC Brain Metastasis Diagnosis

2.1. Circulating Tumor Cells

Circulating tumor cells (CTCs), which originate from the primary tumor and travel via the bloodstream, carry the potential to establish metastases [8]. This metastatic potential is correlated with tumor stage, as later stages are associated with increased CTC burden and aggressiveness [9]. CTC burden in peripheral blood has shown predictive value for progression‐free survival (PFS) and overall survival (OS) in patients with breast cancer, colorectal cancer, and prostate cancer [10, 11, 12], which supports the potential of CTCs as a valuable, real‐time biomarker for assessing tumor status. Mutations in the Keap1‐Nrf2‐ARE pathway were found via whole genome sequencing (WGS) in most NSCLC patients with BM and in CTCs derived from metastatic cancer patients, and key genes including Keap‐1, Nrf2, and P300 indicate possible strategies for the prevention of CTC dissemination and BM [13].

CTCs in CSF may be more directly reflective of the CNS tumor burden than those in peripheral blood, as they have direct contact with the tumor microenvironment in the CNS and are less influenced by the BBB. Due to the typically low volume of CSF samples, the enumeration of CTCs in CSF is more technically challenging than in peripheral blood. Nonetheless, studies have demonstrated its satisfactory sensitivity and specificity for diagnosing leptomeningeal metastasis (LM) [14]. These findings support the use of CTC analysis in CSF for BM diagnosis. Furthermore, the quantity of CTCs in CSF can be used to assess treatment response in LM [15], expanding the scope of CTC analysis. While immunoflow cytometry offers similar sensitivity and specificity, the FDA‐approved CellSearch technology allows for longer sample preservation. Additionally, single cells isolated from CTCs can be analyzed for DNA, RNA, and protein, providing valuable genomic information. In a metastatic tropism study, CTCs derived from cancer patients exhibit a characteristic tissue tropism and thus provide novel insights into the mechanisms of BM. This study, which utilized four patient‐derived CTC lines in a murine model (n = 59 mice), identified the SEMAD4 gene as cooperating with the oncoprotein MYC to promote BM, suggesting that disrupting this interaction could offer a therapeutic target. However, the foundational human‐derived component of this research is based on a very limited cohort of only four patients. While the extensive murine modeling provides mechanistic depth, the small patient sample size and a high risk of bias constrain the generalizability of these findings and underscore the necessity for validation in larger‐scale clinical studies [16]. Slow‐cycling cancer cells expressing stemness markers have been identified as critical contributors to BM formation. NDRG1, a clinically established marker of poor prognosis in primary breast cancer, has been shown to regulate a large number of downstream pathways related to cell proliferation, tumor vascularization, differentiation, and invasion. Targeted knockdown of NDRG1 significantly inhibits BM formation in vivo [17]. The association between cathepsin S and BM has also been uncovered and depletion of cathepsin S in tumor and stromal cells reduces BM incidence. Importantly, tumor cell‐derived cathepsin S seems to help in the transmigration and extravasation across the BBB, which indicates a potential therapeutic target for BM [18].

2.2. Circulating Tumor DNA

Circulating tumor DNA (ctDNA), which is primarily derived from apoptotic and necrotic tumor cells, is frequently detected at elevated concentrations in cancer patients harboring genomic alterations. Numerous clinical investigations have systematically examined the correlation between ctDNA and BM [19, 20, 21]. Furthermore, quantitative analysis revealed that both mutant allele frequency (MAF) and tumor mutational burden (TMB) in ctDNA exhibited significant positive correlations with BM tumor volume [22]. The observation represents a promising step towards quantitative disease monitoring. While these initial findings are hypothesis‐generating, the small sample sizes and a high risk of bias limit the statistical robustness of the conclusions. Therefore, the clinical translatability of these specific correlations is yet to be established, and their ultimate utility for guiding patient management will depend on confirmation in larger, well‐designed clinical studies. Notably, the first ctDNA test for EGFR mutations in NSCLC patients was approved by the FDA in 2016, marking a significant advancement in non‐invasive cancer diagnostics.

For patients with BM, CSF ctDNA has demonstrated higher diagnostic value than plasma ctDNA. The high background signal from normal DNA in plasma hinders its clinical application [23]. This is particularly advantageous in LM, where it surpasses conventional CSF cytology. Advanced genomic profiling techniques, including whole exome sequencing (WES) and next‐generation sequencing (NGS) have been employed to detect ctDNA [24]. The application of these sensitive methods allows for a more robust assessment of CSF ctDNA detection reliability. Beyond diagnosis, CSF ctDNA can uncover molecular progression and emergent resistance mechanisms that may not be apparent on MRI, particularly in leptomeningeal disease, due to its direct contact with intracranial lesions. For instance, unique targetable mutations (e.g., EGFR exon 19 deletions) have been identified in CSF from patients with BM that were absent in matched plasma or prior extracranial biopsies, directly influencing therapy decisions [25]. Specifically, in the context of osimertinib resistance monitoring, liquid biopsy (especially CSF ctDNA) holds a pivotal place for the early identification of resistance mechanisms such as C797S mutation, T790M loss, and bypass pathway activation, thereby informing subsequent therapeutic strategies [26]. Additionally, ctDNA enables dynamic monitoring of treatment response [27], though comprehensive multicenter clinical trials are required to establish its clinical utility. Transforming growth factor beta 1 (TGFβ1) has been demonstrated to play a pivotal role in facilitating epithelial‐to‐mesenchymal transition (EMT), a critical biological process driving tumor cell invasion and metastatic dissemination. Significant associations have been reported between single nucleotide polymorphisms (SNPs) in TGFβ1 and radiation pneumonitis in lung cancer patients undergoing radiotherapy or concurrent chemoradiotherapy. Subsequent genetic analysis revealed that NSCLC patients with the TGFβ1 rs1982073 (T+29C) variant genotypes exhibited an increased risk of BM. The identification of this genetic variant may serve as a predictive biomarker for BM development and could potentially guide the implementation of prophylactic cranial irradiation (PCI) in high‐risk patients [28]. A retrospective single‐center study identified ALK rearrangements and MDM2 amplifications in ctDNA from metastatic brain tumors based on NGS (n = 7) [29]. In a separate clinical case, ALK rearrangement was identified in a BM patient, resulting in the expression of anaplastic lymphoma kinase that activates downstream signaling pathways involved in cell survival, angiogenesis and proliferation [30]. It is hypothesized that next‐generation ALK inhibitors, with their improved BBB penetration, will translate into improved clinical outcomes for BM patients. However, the translational journey from a compelling single‐case observation to clinically applicable knowledge requires demonstration of its recurrence in a significant patient subset. In a clinical study involving 29 BM patients, SERPINI1 was identified as the most frequently mutated gene. This finding leads us to hypothesize that SERPINI1 mutations could play a role in metastatic cancer cell adaptation during hematogenous dissemination and subsequent niche formation [31].

2.3. Extracellular Vesicles

Extracellular vesicles (EVs) are nanoscale, lipid bilayer‐enclosed particles released by both neoplastic and normal cells, containing diverse molecular cargo including DNA, RNA, proteins, and metabolites. EVs are systematically classified into three main subtypes—exosomes, microvesicles, and apoptotic bodies—based on their biogenesis pathways and physical characteristics [32]. These nanoscale vesicles serve as essential mediators of intercellular communication and play pivotal roles in cellular signal transduction pathways [33]. Exosomes and microvesicles, derived from viable cells through distinct biogenesis mechanisms, have emerged as particularly promising candidates for liquid biopsy applications [34]. EVs represent a promising diagnostic tool, having demonstrated both high sensitivity and specificity as tumor biomarkers [35]. Comparative analyses have revealed that small EV (sEV)‐derived protein profiles demonstrate superior accuracy in tumor type classification compared to conventional serum protein profiles [36].

In various malignancies, EV‐associated PD‐L1 secretion has been mechanistically linked to acquired resistance against immune checkpoint inhibitor therapies. Elevated PD‐L1 expression has been shown to significantly enhance tumor cell metastatic potential through immune evasion mechanisms. Furthermore, circulating EV‐associated PD‐L1 has been demonstrated to effectively suppress CD8⁺T lymphocyte activity, highlighting the critical role of EVs in facilitating tumor immune evasion. In lung cancer patients with BM, proteomic analysis revealed significantly elevated protein concentrations in circulating small EVs (sEVs) compared to non‐metastatic controls. Tumor‐derived EVs facilitate BM progression by enhancing BBB permeability and reprogramming the brain metastatic niche microenvironment, thereby promoting multiple stages of BM development [36]. Comprehensive proteomic profiling identified elevated CEMIP expression, associated with cell migration and hyaluronan binding, in exosomes from brain metastatic cells. Genetic depletion of CEMIP significantly inhibits BM formation by impairing tumor cell invasion capabilities and disrupting vascular association in the brain microenvironment. Therapeutic targeting of exosomal CEMIP has emerged as a promising strategy for both BM prevention and early detection, with potential clinical applications currently under investigation [37]. High plasma EV integrin β3 level has been demonstrated to mediate a brain‐tropic metastasis pattern. In a cohort of 75 lung cancer patients with BM undergoing whole brain radiotherapy, comprehensive analysis of circulating EV integrins revealed a significant correlation between elevated integrin β3 levels and both reduced OS and poorer intracranial disease control [38]. In small cell lung cancer (SCLC) models, co‐culture with human brain microvascular endothelial cells (HBMECs) induced an increase of S100A16 expression, suggesting potential cross‐talk mechanisms between tumor cells and the brain vasculature. Functional inhibition studies demonstrated that S100A16 upregulation significantly enhanced the pro‐survival effects of HBMEC‐derived exosomes on recipient SCLC cells, establishing S100A16 as a promising therapeutic target for BM intervention [39].

2.4. Tumor‐Educated Platelets

In addition to CTCs, ctDNA, and EVs, blood platelets represent a rich and clinically accessible source of tumor‐derived biomarkers in liquid biopsy applications [40]. Accumulating experimental evidence has established the multifaceted role of platelets throughout all stages of tumorigenesis, from initiation to metastasis. Comprehensive transcriptomic and proteomic analyses have revealed significant alterations in platelet profiles in NSCLC patients [41], indicating their potential diagnostic and prognostic utility. Tumor cells actively transfer tumor‐associated biomolecules to platelets, inducing significant modifications in their RNA and protein profiles, thereby transforming them into TEPs. The RNA profiling of TEPs has demonstrated significant potential as a biomarker platform for tumor detection and localization, particularly in lung cancer, owing to the substantial pulmonary production of platelets [42]. Studies have established that specific alterations in platelet mRNA and circular RNA (circRNA) profiles can serve as promising predictive biomarkers for NSCLC, with combined analysis offering enhanced diagnostic accuracy for early‐stage detection [42, 43]. A molecular signature comprising five distinct features—including two circRNAs (circSLC8A1 and circCHD9) and three mRNAs (PSMB9, RUNX1, and LILRB1)—was identified for combinatorial analysis within an integrated mRNA‐circRNA profiling panel. This early‐stage predictive model demonstrated high diagnostic performance, with sensitivity and specificity values of 85% and 86%, respectively. Notably, platelets function as primary biological responders to tumor presence, enhancing diagnostic accuracy through early molecular alterations. Additionally, the RNA profiles of platelets are dynamically modulated by both primary and metastatic tumors, establishing a direct correlation between plasma platelet RNA and BM development. In many cases, platelets can even identify primary tumor origins, and their RNA profiles can accurately indicate the oncogene status of various tumors [44, 45, 46], thereby complementing the comprehensive understanding of tumor biology.

2.5. Proteins and Metabolites

Dissociated proteins and metabolites detected in serum or CSF are of diagnostic value and can serve as biomarkers for BM. Interleukin‐6 (IL‐6) activates the JAK2/STAT3 signaling pathway, a critical mediator of BM pathogenesis, and its serum levels are significantly elevated in NSCLC patients [47]. Serum S100B levels, a calcium‐binding protein predominantly secreted by astrocytes in the form of homodimers, are significantly elevated in NSCLC patients with BM compared to non‐BM counterparts [48]. Myelin basic protein (MBP), a major structural component of central nervous system (CNS) myelin sheaths, shows significant elevation following BBB disruption [49]. Consequently, serum MBP levels can quantitatively indicate the degree of BBB compromise caused by brain metastatic invasion. A clinical investigation has demonstrated that serum C‐reactive protein (CRP) levels exhibit strong predictive value for differentiating metastatic brain tumors from glioblastoma and forecasting short‐term prognosis in patients with solitary brain lesions. However, the study was limited by the relatively small cohort size, necessitating larger‐scale studies to fully elucidate the underlying molecular mechanisms [50].

Beyond protein biomarkers, metabolic reprogramming in brain metastatic cells generates distinct metabolite profiles that reflect their adaptation to the unique brain microenvironment. The dynamic equilibrium of cytokine and chemokine signaling networks plays a crucial regulatory role in tumor progression and metastatic dissemination [51]. Furthermore, differences in the volume of CSF‐derived metabolites have been observed between patients with and without BM. In a cohort study, eight biomarkers corresponding to observed metabolic differences were identified in adenocarcinoma patients, with six metabolites—inositol phosphate, succinyladenosine, hypoxanthine, creatinine, valyl‐methionine, and homocysteine—showing downregulation in non‐BM patients [52]. Nevertheless, these preliminary findings require validation through larger, multicenter clinical trials before potential translation into clinical practice (Table 1) [53, 54, 55, 56, 57, 58].

TABLE 1.

Summary and comparison of clinical studies on liquid biopsy biomarkers for brain metastasis in EGFR‐mutant NSCLC.

Biomarker	Methods	Statistical endpoints	Clinical significance	Molecular determinants	Limitations	References
ctDNA/blood (n = 205)	PCR	1. DMFS: HR = 1.601 (1.042–2.459), p = 0.032 for rs1800469 CT/TT; HR = 1.589 (1.009–2.502), p = 0.046 for rs1982073 CT/CC 2. Brain MFS: HR = 2.567 (1.155–5.702), p = 0.021 for rs1982073 CT/CC	Predict distant metastasis‐free survival among patients with NSCLC	TGFβ1	1. Single‐center, retrospective design; 2. Small sample size; 3. Limited to radiotherapy cohort	[28]
ctDNA/CSF (n = 12)	ddPCR	1. Sensitivity: CSF ctDNA had significantly higher sensitivity for detecting CNS somatic mutations than plasma ctDNA (p < 0.05, Mann–Whitney test) 2. MAF: In CNS‐restricted disease, MAFs in CSF ctDNA were significantly higher than in plasma	Explore CSF as a complement for liquid biopsy in characterization, diagnosis, prognosis and clinical managing of EGFR‐mut NSCLC BM	MAF	1. Small, heterogeneous cohort; 2. Preliminary diagnostic validation	[53]
protein/serum (n = 459)	Quantitative TMT‐based proteomics	1. Diagnosis (Combined Model): AUC = 0.951 (training), AUC = 0.845 (validation) 2. Prognosis: High tissue CTSF expression associated with shorter PFS (6.7 vs. 10.9 mos, p = 0.047) and was an independent prognostic factor (HR = 2.052, p = 0.04)	Identify CTSF and FBLN1 as novel diagnostic biomarkers for NSCLC BM	CTSF/FBLN1	1. Single‐center, retrospective design; 2. Potential clinical variability; 3. Lack of validation in large‐scale multicenter cohorts	[54]
protein/serum (n = 120)	ELISA	1. Diagnosis: IL6 expression was significantly higher in brain‐metastatic (A549‐F3) vs. parental cells (p < 0.05) 2. Prognosis: High serum IL6 was associated with a significantly increased risk of brain metastasis (35% vs. 17%, p = 0.012) and shorter overall survival (p = 0.0006)	Provide a promising new approach for inhibiting BM in EGFR‐mut NSCLC patients	IL6/JAK2/STAT3 signaling	1. Limited clinical sample size; 2. Retrospective design	[47]
protein/serum (n = 30)	ELISA	1. Diagnosis: Serum S100B levels were significantly higher in NSCLC patients with BM vs. without BM (0.048 ± 0.0029 vs. 0.015 ± 0.0160 μg/L, p < 0.01)	Indicate the correlation between elevated serum S100B levels and BM in NSCLC	S100B	1. Small clinical sample size; 2. Retrospective and single‐center study design; 3. Lack of in vivo validation	[48]
protein/serum (n = 68)	ELISA	1. Diagnosis: Serum NfL was significantly higher in patients with BM vs. without BM (median 35 vs. 16 pg/mL, p = 0.001). AUC for distinguishing groups was 0.77 (95% CI: 0.66–0.89)	Imply NfL as a potential biomarker for EGFR‐mut NSCLC BM	NfL	1. Retrospective design; 2. Limited sample size; 3. Lack of prospective, scheduled neuroimaging	[55]
CTC/blood (n = 48)	CellSearch/mRT‐PCR	1. Detection Rate: CTC positivity rate was 15.2% with CellSearch vs. 29.2% with multi‐marker mRT‐PCR 2. Association with Disease State: CTC detection by mRT‐PCR was highest in M1 chemo‐naïve patients (55.6%, p = 0.015 vs. M0)	Increase the sensitivity of CTC detection in EGFR‐mut NSCLC patients	EpCAM	1. Limited sample size and single‐center retrospective design; 2. Lack of survival analysis	[56]
CTC/blood (n = 8)	MACS	1. Prognosis: High RAC1 expression was significantly associated with worse overall survival (median 1081 vs. 1798 days, p = 0.0015)	Imply CTC as precursor of BM in lung adenocarcinoma	RAC1	1. Retrospective design; 2. Limited sample size; 3. Insufficient statistical power	[57]
CTC/blood (n = 18)	Parsortix	1. Detection Rate: CTCs detected in 50% (9/18) of NSCLC BM patients using a size‐based (Parsortix) system	Suggest a potential role of CD74 and CD44 in survival and trafficking of CTCs in EGFR‐mut NSCLC BM patients	CD74/CD44	1. Small patient cohort; 2. Limited matched CTC, BM tissue, and CSF samples	[58]
CTC/ctDNA/blood (n = 5)	WGS/ddPCR	1. Diagnosis: Keap1‐Nrf2‐ARE pathway mutations in 80% (4/5) of BM patients, enriched in metastases	Suggest the potential role of Keap1‐Nrf2‐ARE pathway and provide therapy strategies for NSCLC	NRF2	1. Very small discovery cohort; 2. No clinical outcome correlation	[13]

Note: A systematic overview of key clinical studies evaluating circulating tumor DNA (ctDNA), proteins, and circulating tumor cells (CTCs) as liquid biopsy biomarkers in NSCLC patients with brain metastasis (BM). Information is categorized by study design, level of evidence (according to the Oxford Centre for Evidence‐Based Medicine, OCEBM), biomarker type and source, detection method, primary statistical endpoints, clinical significance, relevant molecular determinants, and study limitations.

Abbreviations: AUC, area under the curve; BM, brain metastasis; CNS, central nervous system; CSF, cerebrospinal fluid; CTC, circulating tumor cell; ctDNA, circulating tumor DNA; DMFS, distant metastasis‐free survival; HR, hazard ratio; MAF, mutant allele frequency; NSCLC, non‐small cell lung cancer; PFS, progression‐free survival.

3. Methods for Liquid Biopsy in NSCLC Brain Metastasis Diagnosis

3.1. Methods for CTC Detection

The low abundance of CTCs in peripheral blood limits their direct detection. Enrichment techniques, including positive enrichment, negative enrichment, and immunoaffinity‐based methods, are employed to isolate and concentrate CTCs, thereby establishing them as one of the most popular options in liquid biopsy. In positive enrichment, the CellSearch system utilizes ferrofluid nanoparticles to separate epithelial cell adhesion molecule (EpCAM)‐expressing cells from other blood cells, while magnetic‐activated cell sorting (MACS) captures cells tagged with magnetic nanoparticles conjugated to specific antibodies. Negative enrichment methods, represented by the EasySep system and the Quadrupole Magnetic Separator, remove non‐CTC components, thereby selectively retaining CTCs in the sample. Immunoaffinity‐based methods target antigens specifically expressed on the surface of CTCs through the application of antibodies. However, the heterogeneity of CTC populations often leads to the loss of certain subpopulations, presenting a major challenge for this approach [59]. A novel electrochemical cytosensor has been developed employing branched Pt Au nanospheres (B‐Pt‐Au NPs) as tags and MnO₂‐GO‐Au nanosheet‐modified electrodes for CTC detection, which facilitates the analysis of extremely rare CTCs in peripheral blood [60]. A proof‐of‐concept study demonstrated a microfluidic platform integrated with atomic force microscopy (AFM). This platform sorts CTCs from liquid biopsy samples and allows for automated measurement of their mechanical properties [61]. Subsequent investigations should focus on applying this AFM‐based methodology to a wider array of CTCs and other cancer‐associated cells found in liquid biopsies. Such studies are poised to yield significant advancements in tumor monitoring and our understanding of metastatic mechanisms.

3.2. Methods for ctDNA Detection

For the detection of ctDNA, advanced sequencing technologies, including droplet digital polymerase chain reaction (ddPCR), cancer personalized profiling by deep sequencing (CAPP‐Seq), WES and WGS have been employed. ddPCR exhibits high sensitivity, thereby enabling the detection of rare mutations and calculation of copy number variants. CAPP‐Seq identifies major mutations, including insertions, deletions, and rearrangements, allowing the assessment of tumor‐specific alterations at the individual patient level. WES identifies potential oncogenes and tumor suppressor genes throughout the entire exome, providing a comprehensive, genome‐wide analysis. WGS provides a comprehensive evaluation of the entire tumor genome, facilitating the identification of characteristic and deleterious alterations in ctDNA and enabling the detection of a broad spectrum of potential tumor mutations [62]. An interpretable binary classification model (CM) was developed to distinguish between lung cancer patients and healthy subjects by analyzing circulating ctDNA fragmentation patterns [63]. This model provides high interpretability and traceability, thereby facilitating clinical decision‐making. The study cohort comprised 148 healthy controls and 138 patients with lung cancer. The model demonstrated stage‐dependent sensitivity, with values of 66.7%–85.7% for stage I, 77.8%–100% for stage II, and 70.0%–80.0% for stage III disease. The specificity ranged from 79.3% to 90.0% across all stages. However, the retrospective nature introduces potential confounding, which is a key limitation. Consequently, this model should be validated in a prospective trial.

3.3. Methods for EVs Detection

EVs carry intercellular information molecules such as DNA, long non‐coding RNA (lncRNA), microRNA (miRNA), and proteins. Among EVs, exosomes derived from living cells have gained considerable attention in recent years. The extraction and purification of exosomes constitute critical steps for reliable experimental research. Common separation methods rely on distinct physical or biochemical properties of the target. These include techniques based on density, size, and surface component affinity, often implemented through processes such as ultracentrifugation, ultrafiltration, or affinity capture. However, high costs (e.g., for instrumentation or specialized reagents) and suboptimal efficiency (e.g., in terms of throughput or processing time) hinder their translation into routine clinical practice [64] (Figure 2).

FIGURE 2.

Clinical values and detection technologies of key liquid biopsy biomarkers in brain metastasis. This figure summarizes the unique clinical applications and corresponding detection methodologies for four major classes of liquid biopsy biomarkers—circulating tumor DNA (ctDNA), extracellular vesicles (EVs), circulating tumor cells (CTCs), proteins and metabolites—in the context of brain metastases. ctDNA enables dynamic monitoring of tumor burden and genomic evolution. EVs facilitate intercellular communication between tumor cells and the brain microenvironment. Protein and metabolite provide mechanistic insights into driver mutations and may serve as biomarkers of neuronal damage. CTCs allow for prognostic stratification and functional characterization of metastatic potential. The right section outlines representative detection technologies and workflow for each biomarker type: ctDNA: Digital droplet PCR (ddPCR) and whole‐genome sequencing (WGS); Protein and metabolite: Liquid chromatography–tandem mass spectrometry (LC–MS/MS) and enzyme‐linked immunosorbent assay (ELISA). EV: Ultracentrifugation (UC) and ultrafiltration (UF). CTCs: CellSearch System and EasySep System. CTC, circulating tumor cell; ctDNA, circulating tumor DNA; ddPCR, digital droplet PCR; ELISA, enzyme‐linked immunosorbent assay; EV, extracellular vesicle; LC–MS/MS, liquid chromatography–tandem mass spectrometry; UC, ultracentrifugation; UF, ultrafiltration; WGS, whole‐genome sequencing.

3.4. Pre‐Analytical Variables for EGFR Mutation Detection

The choice of blood collection tube and the time‐to‐processing are critically interdependent and fundamentally influence the quality of ctDNA. Andersson et al. [65] demonstrated that plasma processed immediately (< 1 h) from standard K₂EDTA tubes yielded high‐quality ctDNA. However, a processing delay of 168 h (7 days) at room temperature led to a 28‐fold increase in measured DNA concentration, primarily attributable to genomic DNA contamination from leukocyte lysis. This substantial dilution of tumor‐derived fragments by wild‐type genomic DNA can severely obscure low‐frequency EGFR variants, thereby increasing the risk of false‐negative results. In contrast, cell‐stabilizing tubes effectively mitigated this effect. For instance, Streck tubes maintained stable ctDNA yields with minimal genomic DNA contamination over 168 h. Moreover, the author recommended a combined approach utilizing both a long/short amplicon qPCR ratio and fragment size analysis for comprehensive pre‐analytical quality control to achieve optimal analytical sensitivity [65].

Beyond blood, CSF often yields higher tumor DNA fraction than plasma in cases of LM. However, a standardized protocol for CSF ctDNA isolation has been lacking, contributing to inter‐study variability. A 2024 method optimization study directly addressed these technical challenges by systematically comparing isolation methods and parameters using artificial CSF spiked with reference ctDNA. The study demonstrated that magnetic bead‐based extraction methods provide a significantly higher and more consistent yield than column‐based methods for CSF. Furthermore, protocol modifications such as using a 75 μL bead volume (optimized for 2 mL CSF) and implementing a post‐extraction vacuum concentration step maximized the final ctDNA concentration without compromising DNA integrity, as confirmed by droplet digital PCR [66]. This work provides much‐needed empirical evidence for optimizing CSF‐ctDNA isolation, underscoring that extraction chemistry, reagent volumes, and post‐processing steps must be carefully calibrated for the unique matrix of CSF.

4. Role of Liquid Biopsy in Deciphering Intrapatient Heterogeneity of CNS Metastasis

4.1. Heterogeneity Across CNS Lesions

Investigating the genomic heterogeneity among multiple CNS metastatic lesions within a single patient is critical for determining whether a single biopsy can adequately guide therapy. Notably, research into this spatial heterogeneity has revealed a paradigm of relative genomic homogeneity across intracranial sites.

The seminal study by Brastianos et al. directly addressed this question by performing whole‐exome sequencing on multiple, anatomically distinct brain metastases from the same patients [67]. Their analysis involved 86 matched patient trios comprising brain metastases, primary tumors, and normal tissue. Contrary to the expectation under a model of branched evolution, the researchers found that different brain metastases shared nearly all potentially actionable driver alterations, indicating that they were more closely related to each other than to the primary tumor or extracranial metastases. Specifically, 97% (29 of 30) of the potentially clinically informative driver alterations were shared. This finding suggests that the metastatic seeding to the brain likely occurred from a common, late ancestral subclone that already harbored these key drivers, and that subsequent intracranial spread maintains this core genomic profile.

In the context of EGFR‐mutant NSCLC, this relative homogeneity among CNS lesions carries profound practical implications. It implies that the resistance mechanisms driving CNS progression are likely consistent across different brain metastases. Consequently, molecular profiling of a single, accessible CNS lesion—or, alternatively and more comprehensively, via a CSF liquid biopsy—can reliably identify the targetable alterations responsible for intracranial therapeutic escape. This understanding transforms the clinical approach, shifting from one that requires multiple high‐risk biopsies to one in which a single, minimally invasive CSF sample can effectively guide targeted therapy for multifocal CNS disease.

4.2. Divergence From Extracranial Disease

The primary rationale for employing compartment‐specific liquid biopsy, particularly of CSF, in managing CNS metastases is the well‐documented genomic discordance between intracranial and extracranial disease. The divergence stems from distinct branched evolutionary trajectories. The founding clone of a CNS metastasis diverges early and undergoes independent evolution under the unique selective pressures of the CNS sanctuary site and systemic therapies. Consequently, plasma‐derived ctDNA, which primarily reflects the systemic disease burden, often fails to capture the distinct genetic drivers and resistance mechanisms active within the CNS.

The landmark study by Brastianos et al. provided definitive, quantitative evidence for this compartmentalized evolution [67]. In their cohort of 86 matched patient trios, 53% (46/86) of brain metastases harbored at least one potentially clinically actionable alteration that was not detected in the sampled primary tumor. Furthermore, their analysis of eight cases with additional extracranial metastatic samples demonstrated that regional lymph nodes and distant extracranial metastases were genetically distinct from brain metastases and thus served as unreliable surrogates for the oncogenic alterations present in the intracranial compartment.

This critical divergence is particularly consequential and well‐documented in EGFR‐mutant NSCLC with LM. A study focused on 80 NSCLC patients with LM by Miao et al. revealed that while the overwhelming majority (93.8%) harbored druggable driver mutations, the CSF ctDNA profile was starkly different from that of paired extracranial samples [68]. Among 30 matched pairs, CSF achieved a 100% (30/30) driver gene detection rate, vastly outperforming plasma (44%, 7/16). Notably, four patients whose extracranial samples tested negative for driver mutations were found to harbor EGFR mutations exclusively in CSF and subsequently benefited from TKI therapy. These findings underscore the complementary—and often superior—diagnostic value of CSF over plasma for guiding therapy in patients with CNS progression, as it provides direct access to the genetically divergent CNS disease compartment.

5. Liquid Biopsy in Brain Metastasis Diagnosis for EGFR‐Mutant NSCLC Patients

EGFR mutations, which are particularly prevalent among Asian patients with NSCLC, are associated with a higher incidence of BM [69]. A clinical report analyzed a cohort of 148 tissue samples, encompassing 55 surgical specimens from histologically proven primary NSCLC tumors, 17 matched lymph node metastases, and 76 BM [70]. Within this cohort, EGFR expression was detected in 52.3% of primary tumors and 62.7% of BM. Comparative analysis further revealed that the frequency of strong EGFR expression was significantly higher in BM (29.4%) than in primary tumors (18.2%), suggesting a selective upregulation of EGFR in the metastatic niche. EGFR‐mutant NSCLC patients with BM have significantly shorter median OS and reduced quality of life compared to those without BM [71].

5.1. Heterogeneity in CNS Metastasis Propensity Among EGFR Mutant Subtypes

The EGFR gene consists of 28 exons, with exon 19 deletions (19Del) and exon 21 L858R mutations being the most common [72]. It encodes a transmembrane protein with 1186 amino acids, featuring a peptide growth factor ligand‐binding extracellular domain and a cytoplasmic tyrosine kinase domain. This protein plays a crucial role in regulating cellular proliferation, differentiation, and survival [73]. EGFR mutation status is associated with an increased incidence of BM in NSCLC patients and significantly affects prognosis [74]. Clinical guidelines recommend EGFR mutation testing for patients with advanced lung adenocarcinoma [75]. The differential propensity for CNS metastasis conferred by the two major sensitizing mutations—19Del and L858R—remains an unresolved issue in managing EGFR‐mutant NSCLC. The current literature presents seemingly contradictory evidence, necessitating careful contextualization. One retrospective study of 669 early‐stage patients reported that the incidence of BM was 17.1% in patients with 19Del, compared to 13.6% in those with L858R, 13.3% with other mutations, and 6.1% with wild‐type EGFR; however, the difference between the 19Del and L858R groups was not statistically significant (p = 0.565) [76]. Conversely, a study of 813 patients with advanced NSCLC demonstrated that those harboring L858R mutations had a significantly higher risk of developing symptomatic CNS metastases compared to patients with 19Del mutations (hazard ratio [HR] = 3.34, p = 0.001) [77]. These discrepancies may be explained by differences in patient populations (early‐ vs. advanced‐stage), treatment eras, and statistical power. Future analyses utilizing longitudinal paired plasma and CSF samples will be essential to determine whether the mutational subtype influences the kinetics of ctDNA shedding, the development of CNS‐specific resistance, and ultimately, the clinical trajectory of CNS metastases.

5.2. EGFR ‐Mediated Pre‐Metastatic Niche

EGFR signaling not only drives autonomous tumor cell behaviors but also actively reprograms the CNS microenvironment to facilitate metastatic colonization. EGFR‐mutant lung cancer cells can secrete factors that activate astrocytes, inducing a neuroinflammatory state. Tumor‐associated microglia/macrophages can be polarized towards a pro‐tumor (M2‐like) phenotype, releasing immunosuppressive cytokines (e.g., IL‐10, TGF‐β) that further dampen anti‐tumor immunity and support growth [78]. Moreover, sustained EGFR and VEGF signaling from tumor cells stimulates pathological angiogenesis, forming leaky, immature vessels (the blood‐tumor barrier, BTB) that further support growth while paradoxically hindering drug delivery [79].

5.3. Central Nervous System Pharmacokinetics of Key EGFR ‐TKIs

Third‐generation EGFR TKIs, designed for enhanced BBB penetration, have become cornerstone therapies for NSCLC with CNS involvement. However, substantial inter‐agent differences exist in their CNS distribution and exposure, directly influencing intracranial disease control and the interpretation of ctDNA dynamics between CSF and plasma. Osimertinib, the first‐in‐class third‐generation TKI, exhibits improved CNS penetration relative to earlier agents. Nevertheless, its CSF concentration remains significantly lower than plasma levels, which may contribute to isolated intracranial progression during standard‐dose (80 mg) therapy. Consistent with this pharmacokinetic rationale, dose escalation to 160 mg can recapture intracranial responses, indicating the feasibility of overcoming a CNS exposure threshold [80]. Similarly, furmonertinib demonstrates substantial intracranial efficacy. A study reported intracranial disease control rates exceeding 70% with furmonertinib 160 mg as salvage therapy in patients with brain or leptomeningeal metastases progressing on prior third‐generation TKIs [81]. Clinically, aumolertinib demonstrated superior progression‐free survival versus gefitinib in the AENEAS trial, and it was ranked highest for intracranial PFS in a network meta‐analysis, collectively affirming its potent CNS efficacy [82]. The distinct CNS pharmacokinetic profiles of these TKIs therefore carry important implications for applying liquid biopsy in BM management.

While Osimertinib demonstrates effective BBB penetration and broad in vivo distribution, a subset of malignant cells can survive within the brain microvascular tumor microenvironment (TME), eventually proliferating as Osimertinib‐resistant lesions [83]. Therefore, understanding and monitoring the mutation status of EGFR may help advanced‐stage patients. Liquid biopsy allows real‐time monitoring of EGFR mutation status, which is critical for managing BM patients. The FDA has approved two plasma‐based diagnostic tests for detecting EGFR resistance mutations, enabling treatment response monitoring and disease progression prediction, including BM development. Target gene modification, alternative pathway activation, and histological or phenotypic transformation are the three most common TKI resistance mechanisms [84]. Specifically, the detection of T790M, mesenchymal epithelial transition factor (MET) amplification, and EMT enables the early identification of resistance mechanisms. Liquid biopsy offers a comparative advantage over tissue biopsy for such monitoring [85].

5.4. CTC Detection in BM for EGFR ‐Mutant NSCLC

Numerous CTC detection systems have been developed for NSCLC patients [86]. High EGFR protein expression in NSCLC patients with BM is a favorable factor for the efficiency of CTC enrichment using magnetic cell separation. The low CTC detection rate in NSCLC may be attributed to EMT, which allows cancer cells to enter the bloodstream, although surface antigen targeting has improved CTC isolation [70]. Although NSCLC patients with BM more seldom harbor CTCs, they are still predictive for overall survival and might identify oligo‐metastatic NSCLC patients who might benefit from a more intense therapy [87]. The survival data of 87 NSCLC patients were analyzed, revealing a significantly shorter OS in CTC‐positive BM patients, using both CTC cut‐off values of ≥ 2 and ≥ 5 CTCs per 7.5 mL of blood (p = 0.027 and p = 0.008, respectively, log‐rank test). Notably, all oligo‐brain metastatic patients in this cohort who tested positive for CTCs (using a ≥ 1 CTC cut‐off) succumbed to the disease within 1.3–8.0 months of follow‐up, demonstrating a mean overall survival of 4.6 months. In contrast, CTC‐negative oligo‐brain metastatic patients exhibited a significantly longer mean overall survival of 10.4 months (range: 0.7–47.7). The ISET (Isolation by Size of Tumor cells) technique, which utilizes size‐based filtration to isolate CTCs, has been characterized in terms of its diagnostic performance for lung cancer. However, rigorous assessment revealed that the ISET method is suboptimal for broad cancer screening, demonstrating a markedly lower CTC detection sensitivity of only 26% in lung cancer [88]. A novel size‐based microfluidic platform integrated with a semi‐automated cell recognition system, was developed for the detection and comprehensive molecular characterization of CTCs in patients with BM. In the cohort of 18 NSCLC patients, CTCs were detected in 50% (9 out of 18) of cases. Subsequent investigations revealed that CD44 and CD74 co‐expression on metastasis‐initiating CTCs plays a crucial role in BM development in NSCLC. The dynamic expression patterns of CD44 and CD74 during CTC circulation may offer promising liquid biopsy‐based strategies for BM management in NSCLC patients [58]. However, these findings warrant validation in larger, more diverse patient cohorts, including individuals without BM, coupled with more detailed molecular profiling of the detected CTCs.

5.5. ctDNA Detection in BM for EGFR ‐Mutant NSCLC

A manuscript has described the genomic profile of EGFR‐mutant NSCLC with second‐site EGFR mutation and resistance to Osimertinib [89]. 10% of EGFR‐mutant NSCLC with a second‐site EGFR mutation had more than one mutation among disparate tumor clones, highlighting the utility of liquid biopsy in capturing the composite genomic profile shed from multiple metastatic sites. However, the resistance mechanisms may be underestimated due to inadequate ctDNA in blood and the histologic differentiation cannot be inferred. In a multicenter clinical trial, detection of EGFR mutations in plasma ctDNA has been used as a selection criterion for gefitinib treatment in patients with advanced lung adenocarcinoma. Among 426 screened patients, 188 who tested positive for EGFR mutations in ctDNA were enrolled and received gefitinib. The study demonstrated that patients who achieved clearance of EGFR mutations in ctDNA by week 8 of gefitinib treatment (147 out of 167 evaluable patients, 88%) experienced a markedly longer median PFS than those with persistent mutations (11.0 months, 95% CI 9.43–12.85 vs. 2.1 months, 95% CI 1.81–3.65; hazard ratio [HR] = 0.14, 95% CI data 0.08–0.23; p < 0.0001) [90]. WES and NGS enable the detection of low‐frequency mutations in ctDNA, providing a comprehensive representation of tumor genomes from both primary and metastatic lesions [62]. These analyses offer real‐time insights into resistance mechanisms and responses to EGFR TKIs. In a longitudinal study of 21 EGFR‐mutant NSCLC patients, plasma ctDNA analysis identified the EGFR exon 20 p.T790M resistance mutation ahead of radiological progression confirmed by computed tomography (CT) imaging. Specifically, among TKI‐treated patients (Cohort 1, n = 9), the emergence of the T790M mutation in ctDNA identified a subgroup with a significantly more rapid disease progression compared to those without this resistance mechanism (p = 0.04) [91]. Despite the limited sample size, the strength and timing of this signal validate the role of liquid biopsy for dynamic risk stratification and support its integration into clinical protocols for timely therapeutic intervention. Importantly, for EGFR‐mutant NSCLC patients with BM, CSF ctDNA has shown superior accuracy compared to plasma ctDNA in assessing the molecular status of intracranial lesions, highlighting the critical role of CSF in liquid biopsy. A prospective observational study by Ho et al. [92], which enrolled 136 EGFR‐mutation positive NSCLC patients, demonstrated a 78% overall agreement between plasma and tissue biopsy for EGFR mutation detection. This study provided quantitative evidence that a molecular response (defined as the clearance of EGFR mutations from plasma) was highly predictive of clinical response, occurring in 93% (94/101) of baseline plasma‐positive patients. However, the limitations of plasma‐based monitoring were highlighted by three cases of isolated CNS progression where molecular progression was not detected in plasma, likely due to the restricted passage of ctDNA across the BBB. This observation underscores a key constraint of plasma ctDNA analysis in neuro‐oncology. Consequently, for patients with BM, CSF constitutes a more direct and thus potentially more reliable biosource for genotyping, particularly to monitor intracranial disease status. It is imperative to acknowledge, however, that the generalizability of findings from CSF‐based studies is often constrained by their typically small cohort sizes (frequently n < 20), a common limitation in this challenging field of research [92]. A randomized phase III trial (n = 319) demonstrated that a liquid biopsy‐first diagnostic strategy, which identified actionable alterations via ctDNA in 81% of patients, significantly accelerated the time to contributory genomic results (17.9 vs. 25.6 days, p < 0.001). Although the mean time to treatment initiation (TTI) was not reduced in the overall population, pre‐specified analyses revealed significant TTI reductions in key subgroups, most notably in patients with targetable alterations (21.0 vs. 37.4 days, p = 0.004) [93].

5.6. Refining Management Through Liquid Biopsy Monitoring

The management of EGFR‐mutant advanced NSCLC has been transformed by the integration of third‐generation TKIs, such as osimertinib, with longitudinal liquid biopsy monitoring. This paradigm enables dynamic assessment of treatment response and the early detection of resistance. Case series by Sparavelli et al. [94] provides specific insights into the resistance landscape of advanced EGFR‐mutant NSCLC treated with first‐line osimertinib, with a particular focus on histological transformation. While ctDNA analysis proficiently identifies on‐target (e.g., EGFR C797S) and off‐target (e.g., MET amplification) resistance mechanisms, its utility in diagnosing histological transformation is limited, as this phenotypic shift may not be discernible from the mutational profile alone. Notably, this transformation is strongly associated with specific molecular contexts, particularly the co‐occurrence of TP53/RB1 alterations and the original EGFR mutation subtype, with exon 19 deletions conferring a significantly higher risk compared to the L858R mutation. Therefore, although plasma‐based genotyping is indispensable for guiding subsequent targeted strategies upon the emergence of most resistance mechanisms, tissue re‐biopsy remains the gold standard for confirming histological transformation, ensuring the application of appropriate subsequent therapies.

Early detection of BM in EGFR‐mutant NSCLC patients is key to improving survival rates and extending PFS. Liquid biopsy offers a non‐invasive, real‐time, and repeatable diagnostic option, increasing patient compliance and acceptance. This approach complements traditional surgery and radiotherapy by addressing some of their limitations and could potentially transform patient management and therapeutic strategies in the future. However, further clinical trials and evidence are necessary to validate its widespread application.

6. Integrating Liquid Biopsy Into Clinical Decision‐Making: A Stepwise Approach

The integration of liquid biopsy into the management of EGFR‐mutant NSCLC with BM requires a structured approach to maximize its clinical utility. Based on current evidence and expert consensus, we propose a practical algorithm to guide clinicians in selecting the appropriate liquid biopsy modality and interpreting its results within the context of EGFR‐TKI therapy sequencing.

Figure 3 outlines a proposed algorithm for utilizing liquid biopsy in the setting of suspected CNS progression during EGFR‐TKI therapy. The initial critical step is to determine whether the progression is isolated to the CNS or concurrent with systemic disease. This distinction guides the choice of the optimal liquid biopsy compartment.

FIGURE 3.

Proposed clinical decision algorithm integrating liquid biopsy for the management of suspected central nervous system (CNS) progression in EGFR‐mutant non‐small cell lung cancer (NSCLC). This algorithm addresses key clinical questions regarding the use of liquid biopsy in routine practice. It outlines a stepwise approach beginning with suspicion of CNS progression during EGFR‐TKI therapy. The initial decision node separates isolated CNS from concurrent systemic progression. For isolated CNS progression, cerebrospinal fluid (CSF)‐derived ctDNA is prioritized when lumbar puncture is clinically feasible, offering potential for earlier detection of resistant clones compared to imaging alone. If CSF is inaccessible or negative, plasma ctDNA serves as an alternative. The algorithm further illustrates how molecular profiles (e.g., detection of C797S, MET amplification, or sensitive‐only mutations) obtained from liquid biopsy can directly inform subsequent treatment strategies, including clinical trial enrollment or targeted therapy selection. CNS, central nervous system; CSF, cerebrospinal fluid; CTC, circulating tumor cell; ctDNA, circulating tumor DNA; EGFR, epidermal growth factor receptor; EV, extracellular vesicle; TKI, tyrosine kinase inhibitor.

For patients with concurrent systemic progression, plasma ctDNA is the recommended first‐line liquid biopsy modality due to its convenience and proven utility in capturing the comprehensive landscape of resistance mutations driving extracranial disease. However, a negative plasma result does not exclude CNS‐specific resistance, necessitating further CNS‐directed evaluation.

When isolated CNS progression or LM is suspected, CSF ctDNA analysis becomes the procedure of choice if a lumbar puncture is clinically safe and feasible. Its superior sensitivity for detecting CNS‐derived tumor DNA can reveal actionable mutations absent in plasma. In cases where CSF sampling is contraindicated or unsuccessful, plasma ctDNA remains a viable, though less sensitive, alternative. Emerging modalities like EVs (for analyzing cargo such as RNA and proteins) and CTCs (for functional and phenotypic analysis) may provide complementary information, particularly when ctDNA yields are low or to investigate non‐genetic resistance pathways, but their use is currently best reserved for research or refractory cases within clinical trials.

The molecular profile obtained must then be integrated with the patient's treatment history. For example, the detection of C797S in CSF following osimertinib progression would prioritize referral for clinical trials involving fourth‐generation EGFR TKIs (e.g., BDTX‐1535, BLU‐945). In contrast, identification of MET amplification might support combination therapy with a MET inhibitor. This algorithm emphasizes a sequential, compartment‐aware approach to liquid biopsy, ensuring that the most informative sample guides therapeutic decisions against CNS metastases.

7. Perspectives

Despite advances in systemic therapy, brain metastases are increasingly prevalent and remain a major cause of mortality in patients with EGFR‐mutant NSCLC, as these therapies prolong survival but often fail to fully prevent intracranial spread. Late detection, limited drug penetration across the BBB, and drug resistance contribute to the progression of BM and the high mortality of these patients. Early detection and timely intervention are essential to improving quality of life and PFS. Biomarkers such as CTC, ctDNA, EVs and TEPs provide real‐time insights into the molecular characteristics of tumors and offer a non‐invasive diagnostic alternative. As such, liquid biopsy serves as an important complement to tissue biopsy.

This review has systematically evaluated the evolving role of liquid biopsy in the management of BM for EGFR‐mutant NSCLC patients. By focusing on this high‐risk subtype, we have highlighted several distinctions from general NSCLC reviews: the critical importance of CSF over plasma for capturing CNS‐specific genomics, the influence of BBB integrity and TKI pharmacokinetics on biomarker detectability, and the unique patterns of intra‐ and extracranial tumor evolution. To critically evaluate the strength and validity of the evidence presented herein, we systematically assessed the methodological quality of the included liquid biopsy studies using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist. As summarized in Table 2, the evaluated studies encompass a wide range of designs and sample sizes (from n = 1 case report to n = 319 randomized trial), reflecting the evolving yet heterogeneous nature of this field. Many studies to date are limited by retrospective design, small sample sizes, and methodological heterogeneity. Future efforts should prioritize prospective, standardized protocols to strengthen the evidence base.

TABLE 2.

Evidence quality assessment (based on JBI checklist) of liquid biopsy studies in EGFR‐mutant NSCLC with brain metastasis.

Study design	Sample size	Biomarker	Biospecimen	Risk of bias assessment (based on JBI checklist domains)	Major limitations	References
Prospective, multicenter diagnostic/prognostic cohort	N = 267	CTCs	Plasma	Low	1. External validity limited; 2. Comparative analysis bias	[10]
Prospective, single‐center, prognostic cohort study	N = 75	CTCs	Plasma	Moderate	1. Statistical Bias; 2. Cohort Heterogeneity	[11]
Retrospective analysis of a prospective, multicenter cohort	N = 97	CTCs	Plasma	Low to moderate	1. Statistical Bias; 2. Retrospective Biomarker Analysis	[12]
Experimental preclinical study	N = 4	CTCs	Plasma	Moderate	1. Limited patient cohort size; 2. Use of immunodeficient mouse model	[16]
Prospective observational cohort	N = 30	ctDNA	CSF	Moderate	1. Small sample size; 2. Lack of standardized clinical endpoints	[19]
Prospective observational study	N = 29	ctDNA	Plasma	Moderate to high	1. Very low BM detection rate; 2. Lack of external validation cohort	[20]
Retrospective observational cohort study	N = 90	ctDNA	Plasma	Moderate to high	1. Retrospective sample collection; 2. Potential selection bias	[21]
Prospective observational cohort study	N = 20	ctDNA	Plasma/CSF	Moderate	1. Small cohort size; 2. Lack of external validation cohort	[22]
Retrospective observational cohort study	n = 5	ctDNA	Plasma/CSF	Moderate	1. Very small sample size; 2. No healthy controls	[23]
Prospective observational cohort study	n = 41	ctDNA	Plasma	Low to moderate	1. Single‐center design; 2. Short follow‐up	[24]
Prospective observational cohort study	n = 14	ctDNA	Plasma/CSF	Moderate	1. Very small sample size; 2. Limited gene panel	[25]
Retrospective exploratory cohort study	n = 146	ctDNA	Plasma	Moderate	1. Lack of specificity analysis; 2. Limited biomarker scope	[26]
Retrospective cohort study	n = 84	ctDNA	Plasma	Moderate	1. Lack of external validation cohort; 2. Limited generalizability	[27]
Retrospective cohort study	n = 205	ctDNA	Plasma	Moderate	1. Single‐center, retrospective design; 2. Lack of functional validation in patient samples	[28]
Retrospective cohort study	n = 28	ctDNA	Plasma	Moderate	1. Single‐center, retrospective design; 2. No external validation cohort	[29]
Case report	n = 1	ctDNA	Plasma	High	1. Single patient case; 2. Short follow‐up	[30]
Retrospective cohort study	n = 29	ctDNA	Plasma	Moderate	1. Small sample size; 2. No survival or treatment response data	[31]
Prospective observational cohort study	n = 42	EV	Plasma	Moderate	1. No functional validation; 2. Potential selection bias	[36]
Prospective observational cohort study	n = 75	EV	Plasma	Low to moderate	1. No external validation cohort; 2. Potential variability in EV isolation	[38]
Retrospective cohort study	n = 60	Platelet	Plasma	Low to moderate	1. Small sample size; 2. Lack of external validation cohort	[42]
Retrospective cohort study	n = 12	Platelet	Plasma	Low to moderate	1. Very small cohort; 2. Small validation cohort	[43]
Retrospective cohort study	n = 29	CRP	Plasma	Moderate	1. Very small sample size; 2. No long‐term follow‐up data	[50]
Retrospective cohort study	n = 286	ctDNA	Plasma	Moderate	1. Limited generalizability; 2. No long‐term follow‐up or prognostic validation	[63]
Retrospective cohort study	n = 180	ctDNA	CSF	Moderate to high	1. No external validation cohort; 2. Retrospective nature	[68]
Prospective, single‐arm, multicenter phase 2 trial	n = 188	ctDNA	Plasma	Moderate	1. No control arm	[90]
Prospective longitudinal cohort study	n = 21	ctDNA	Plasma	Low to moderate	1. Small sample size; 2. Limited biomarker panel	[91]
Prospective observational cohort study	n = 136	ctDNA	Plasma	Moderate	1. Single‐center, single‐region design; 2. Selection bias	[92]
Multicenter, randomized, phase III trial	n = 319	ctDNA	Plasma	Low to moderate	1. Open‐label design; 2. External validity	[93]

Note: The study design, sample size, biomarker type, and biospecimen for key publications included in this review. A critical appraisal of methodological quality is presented using a Risk of Bias Assessment based on the Joanna Briggs Institute (JBI) Critical Appraisal Checklist domains, with an overall judgment of Low, Moderate, or High risk of bias. Major limitations pertaining to internal/external validity, sample size, study design, and analytical methods are detailed for each study. This structured assessment supports the interpretation of evidence strength and generalizability across the reviewed literature.

Abbreviations: BM, brain metastasis; CSF, cerebrospinal fluid; CTC, circulating tumor cell; ctDNA, circulating tumor DNA; EV, extracellular vesicle; JBI, Joanna Briggs Institute; NSCLC, non‐small cell lung cancer.

While technological advancements in detecting CTCs, ctDNA, EVs, and TEPs are promising, their translation into routine practice is hampered by the standardization and sensitivity challenges outlined herein. The diagnostic sensitivity and specificity of liquid biopsy assays often remain inferior to those of conventional tissue biopsy, which can compromise diagnostic accuracy. In the early stages of disease, biomarkers such as ctDNA can be present at very low concentrations in plasma, partly due to the BBB in the context of brain tumors. Furthermore, some biomarkers may be detectable in non‐cancerous conditions, elevating the risk of false‐positive findings. The lack of standardized protocols for sample collection, processing, and analysis contributes to considerable variability in results across different studies and laboratories. Consequently, establishing rigorously validated, widely applicable, and clinically approved standard operating procedures (SOPs) is a critical prerequisite for the clinical validation and routine implementation of liquid biopsy biomarkers. Additionally, the regulatory landscape for novel liquid biopsy biomarkers remains in a state of evolution. The establishment of clear guidelines for analytical validation, demonstration of clinical utility, and reimbursement pathways is therefore essential to facilitate widespread clinical adoption. Finally, the substantial cost associated with liquid biopsy profiling presents a significant economic barrier, particularly for cancer patients and healthcare systems. The development of more cost‐effective and scalable molecular profiling technologies is therefore imperative to ensure equitable access and sustainable integration into clinical practice [95].

Looking ahead, several promising strategies are poised to address the current gaps in managing BM for EGFR‐mutant NSCLC. High‐depth sequencing (HDS) and ddPCR have revolutionized the detection of ctDNA and CTC in liquid biopsy, yet their utility in early BM detection remains limited. The integration of multi‐omics approaches, including genomics, transcriptomics, and epigenomics in liquid biopsy is crucial for deciphering the complex molecular mechanisms of BM development and resistance, potentially uncovering novel predictive signatures. Novel biomarkers, such as EVs are particularly promising for BM, as they can traverse the BBB via transcellular transport mechanisms and modulate the premetastatic niche, offering a window into the intracranial microenvironment that is otherwise difficult to access [96]. Exosomes exhibit distinct advantages over other nanoparticulate drug delivery systems primarily attributed to their inherently low immunogenicity and biocompatibility [97]. Furthermore, serial liquid biopsy enables the real‐time monitoring of tumor dynamics and facilitates early detection of BM. To translate these complex data into clinical decision‐making, artificial intelligence (AI) algorithms and big data analytics are being leveraged to develop sophisticated models. These models can integrate longitudinal liquid biopsy data with clinical and radiographic information (e.g., from CT and MRI) to generate a more comprehensive diagnosis and improve the prediction of BM risk and prognosis in EGFR‐mutant NSCLC patients. Further clinical trials specifically in the EGFR‐mutant NSCLC population and standardized guidelines for these novel applications are required to facilitate the widespread adoption of liquid biopsy in healthcare institutions of all levels.

Author Contributions

Yuxiang Sun: conceptualization, literature review, writing – original draft. Qinglin Wang: conceptualization, literature review. Pengcheng Zhu: conceptualization, literature review. Zhitong Li: conceptualization, literature review. Tongyan Liu: writing – review and editing. Rong Yin: writing – review and editing. All authors read and approved the final manuscript.

Funding

The work was supported by National Science Foundation of China (82472724, 82372709), Yishan Research Project of Jiangsu Cancer Hospital (YSZD202408), Jiangsu Province Capability Improvement Project through Science, Technology and Education, Jiangsu Provincial Medical Innovation Center (CXZX202224‐01), National Science and Technology Major Project (2023ZD0501904).

Ethics Statement

The authors have nothing to report.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.