Harnessing minimal residual disease for precision medicine in locally advanced gastric cancer

Abstract

Gastric cancer remains a major cause of cancer-related mortality, and recurrence after radical-intent therapy is a central obstacle in locally advanced gastric cancer (LAGC). Minimal residual disease (MRD) refers to occult residual tumor burden that cannot be detected by conventional imaging or routine laboratory tests but may later drive recurrence. Circulating tumor DNA (ctDNA), a tumor-derived fraction of circulating cell-free DNA that is mainly released into the bloodstream during tumor-cell apoptosis and necrosis, has emerged as a promising non-invasive biomarker for MRD assessment. In LAGC, ctDNA-MRD detection may support recurrence-risk stratification, postoperative surveillance, molecular response assessment after neoadjuvant therapy, and exploratory evaluation of treatment efficacy, including immunotherapy. However, current evidence should be interpreted cautiously. Most available studies are observational, include modest cohorts, use heterogeneous assay platforms and sampling schedules, and apply non-uniform thresholds for MRD positivity. Therefore, the best-supported role of ctDNA-MRD in LAGC is currently prognostic rather than definitively predictive or treatment-directing. This narrative review summarizes the biological basis of ctDNA, major polymerase chain reaction- and next-generation sequencing-based detection strategies, representative clinical evidence in LAGC, and the main barriers to clinical implementation. CtDNA-MRD assessment has substantial potential to refine precision management in LAGC, but routine treatment modification based solely on ctDNA status requires prospective, interventional validation and standardized analytical workflows.

Key Words: Locally advanced gastric cancer; Minimal residual disease; Circulating tumor DNA; Recurrence prediction; Treatment guidance; Neoadjuvant therapy

Core Tip: Locally advanced gastric cancer remains difficult to cure because occult residual disease can drive recurrence after radical or multimodal therapy. Minimal residual disease (MRD) assessment using circulating tumor DNA (ctDNA) offers a non-invasive strategy for molecular surveillance, recurrence-risk stratification, and dynamic evaluation of treatment response. This minireview highlights current ctDNA-MRD technologies, introduces tumor-informed and tumor-agnostic approaches, and clarifies that ctDNA-MRD is promising but still investigational for treatment guidance. Standardized assays, validated thresholds, and prospective interventional trials are needed before routine precision management of locally advanced gastric cancer.

INTRODUCTION

Gastric cancer (GC), as one of the more common malignant tumors worldwide, is diagnosed with nearly one million new cases every year and is also the fifth leading cause of cancer-related deaths[1]. The incidence rate of GC in China is relatively high worldwide[2]. The onset of this disease is relatively insidious, and there are usually no obvious clinical symptoms. Many patients are diagnosed in the late stage, and even show distant metastasis. Locally advanced GC (LAGC), is usually caused by the tumor invading the gastric muscle layer or even deeper tissues, often accompanied by local lymph node metastasis. Early diagnosis of this disease is actually quite difficult, and the prognosis is not good. There have been many new changes in the treatment methods for LAGC in clinical practice. It is no longer just traditional surgery or systemic chemotherapy. Often, it is combined with radiotherapy, targeted therapy, and immunotherapy. Nevertheless, long-term outcomes remain unsatisfactory, and the 5-year survival rate of patients with advanced disease remains limited[3,4]. More accurate tools for early recurrence detection and treatment-response monitoring are therefore needed to improve the prognosis of patients with LAGC[5].

Minimal residual disease (MRD) refers to the residual cancer cells and tumor-related markers that cannot be found through conventional imaging or laboratory testing methods after radical treatment. In malignant tumors, MRD is a crucial prognostic risk factor. If MRD is detected as positive, the probability of tumor recurrence and distant metastasis will increase[6]. The emerging diagnostic technique of liquid biopsy has received a lot of attention in the field of tumor treatment. It primarily involves the detection of circulating tumor DNA (ctDNA), circulating tumor cells, cell-free DNA (cfDNA), and exosomes[7,8]. ctDNA is actually a special cfDNA released by apoptotic or necrotic tumor cells, and has now become a potential biomarker for evaluating MRD. ctDNA has shown promising prospects in postoperative risk assessment, and its potential value in guiding adjuvant therapy decisions for LAGC is currently being validated through multiple clinical trials. However, it has not yet been used in routine clinical practice and cannot be used to guide adjuvant therapy[9,10].

Several prospective and retrospective studies have reported that postoperative or longitudinal ctDNA positivity is associated with recurrence risk in GC. For example, Yang et al[11] found that postoperative ctDNA positivity was associated with later recurrence and preceded radiological recurrence in some patients. However, these findings should be interpreted primarily as evidence for prognostic risk stratification, not as proof that ctDNA can independently guide adjuvant therapy decisions in routine practice. Most available studies remain observational, include modest sample sizes, and differ in assay platforms, sampling windows, and thresholds for MRD positivity. Therefore, this minireview distinguishes the relatively well-supported prognostic value of ctDNA-MRD from its predictive or actionable utility, which remains unproven in LAGC.

OVERVIEW OF CTDNA-MRD DETECTION TECHNOLOGY

Biological characteristics of ctDNA

The half-life of ctDNA itself is relatively short, and it also carries tumor-specific genetic and epigenetic changes, with significant differences between individuals[12]. Its main sources are primary tumors, metastatic tumors, and circulating tumor cells. There are many genetic changes that can be detected, such as mutations, insertions, deletions, chromosomal rearrangements, copy number variations, and methylation modifications, making it a valuable tumor-specific biomarker[13]. Tumor-derived ctDNA exhibits a canonical size profile of approximately 160-200 base pairs, corresponding to the length of DNA wrapped around a single nucleosome - the core characteristic of plasma cfDNA in liquid biopsies[14]. The quantity and nature of ctDNA are actually influenced by many factors. For example, the burden of the tumor itself, which refers to the total volume and quantity of tumor tissue, as well as the biological state of tumor cells, such as proliferation, apoptosis, and invasive activity. In addition, the DNA clearance and degradation mechanisms in the body can also have an impact[15]. The ctDNA fragments carry genetic variations unique to primary GC, so the appearance, development, recurrence, and treatment efficacy of tumors can be monitored through blood samples. Previous studies have shown that ctDNA testing is more sensitive than traditional tissue biopsy and tumor marker detection in identifying tumor-specific gene variations[16]. The expression of ctDNA can actually be influenced by physiological and pathological mechanisms related to tumors, such as inflammatory reactions and tissue damage[17]. The half-life of ctDNA is approximately two hours[18], which can be used to track tumor status in real-time in vivo.

ctDNA testing is a non-invasive liquid biopsy method with potential applications in early tumor screening, prognosis assessment, recurrence detection, treatment monitoring, and drug-resistance analysis[19]. Compared with repeated tissue biopsy, blood-based testing is more convenient and less invasive for patients[20]. Nevertheless, clinical interpretation in LAGC remains challenging because ctDNA abundance is affected by tumor shedding, tumor location, disease burden, sample handling, assay depth, and bioinformatic filtering[21-23]. Therefore, ctDNA results should be interpreted together with pathological, radiological, and clinical information rather than as a stand-alone determinant of treatment.

Application of ctDNA-MRD detection technology in LAGC

Liquid biopsy of ctDNA is the main method for detecting MRD in solid tumor patients. ctDNA, with its dynamic changes in peripheral blood that can be real-time detected, is helpful for tracking tumor evolution and treatment effects[24]. The optimal collection and processing of ctDNA depend on key pre-analytical factors, including blood volume, plasma separation time, and the use of cell-stabilizing tubes (e.g., Streck tubes) containing specific stabilizing agents, which can inhibit the lysis of peripheral blood cells and prevent the release of cellular DNA to avoid contamination of ctDNA samples[25].

At present, ctDNA detection technologies are mainly divided into polymerase chain reaction (PCR)-based techniques and next-generation sequencing (NGS)-based techniques. As the overarching high-throughput sequencing platform, NGS encompasses targeted NGS, whole-exome sequencing and whole-genome sequencing. PCR-based methods are relatively simple to operate and cost-effective, but they have certain limitations in the comprehensiveness (the range of detectable genetic alterations) and sensitivity of detection[26]. NGS technology can more comprehensively and deeply detect genetic changes, identifying somatic mutations along with copy number variations and structural rearrangements[27], but it has limitations such as high detection costs and complex data analysis. Whole-genome sequencing detection of ctDNA in peripheral blood can detect more mutations without template screening, but it also has drawbacks such as high background errors, too many tracked mutations, and low cost-effectiveness[28]. In addition, some emerging technologies based on DNA methylation or other epigenetic features reflecting the chromatin state of tumor cells have been used to detect ctDNA[29]. In recent years, PCR-based and NGS-based methods have been the most common in ctDNA analysis technologies and have been widely studied[25]. A comparative summary of the key characteristics distinguishing PCR-based and NGS-based ctDNA analysis technologies is presented in Table 1. In LAGC, ctDNA-MRD detection technology is mainly divided into two categories: Tumor-informed analysis methods based on tumor tissue information and tumor-agnostic (or tumor-naive) analysis methods that do not rely on tumor tissue information[30]. Representative technologies or platforms of the tumor-informed method include Signatera^TM, Safe-Seqs, and brPROPHET^TM; tumor-agnostic methods include AVENIO^TM and Guardant Reveal^TM.

Table 1 Key characteristics of polymerase chain reaction-based and next-generation sequencing-based circulating tumor DNA analysis technologies.

	ddPCR-based ctDNA-MRD assay	NGS-based ctDNA-MRD assay
Core technical principle	Microdroplet-partitioned single-molecule PCR amplification, absolute quantification of ctDNA mutations based on Poisson distribution[25,26]	Target region capture + high-throughput sequencing, combined with UMI for ctDNA mutation quantification[25,27]
LOD	0.01%-0.1% VAF[26]	0.1%-0.5% VAF (standard depth); < 0.01% VAF (ultra-deep sequencing)[27]
Detectable variant types	Only pre-designed known SNV and insertions/deletions (indel)[25,26]	Simultaneous detection of SNV, indel, CNV, and gene fusions[25,27]
Detection turnaround time	1-3 working days	5-7 working days
Single sample detection cost	Low (approximately 100-300 USD per sample)	Medium-high (approximately 500-1500 USD per sample)
Core advantages	Simple operation, low cost, short detection cycle, high absolute quantification accuracy[25,26]	Covers multi-gene and multi-variant types, tracks tumor heterogeneity, adapts to personalized panel design[25,27]
Core limitations	Only detects known mutations, cannot cover tumor heterogeneity, low throughput[25,26]	Long detection cycle, high cost, complex data analysis, lower sensitivity than ddPCR at standard sequencing depth[25,27]

ddPCR: Droplet digital polymerase chain reaction; ctDNA: Circulating tumor DNA; MRD: Minimal residual disease; NGS: Next-generation sequencing; PCR: Polymerase chain reaction; UMI: Unique molecular identifier; LOD: Limit of detection; VAF: Variant allele frequency; SNV: Single nucleotide variants; Indel: Insertion and deletion; CNV: Copy number variations.

Given that most studies in the field adopt ctDNA mutation-based MRD detection technologies or platforms, this article only introduces and compares ctDNA-MRD detection methods based on mutation analysis[31]. The tumor-informed method involves comprehensive genomic sequencing of tumor tissue to determine patient-specific tumor-specific genetic mutations/alterations (i.e., clonal alterations unique to the patient’s tumor) followed by customized personalized NGS gene detection panels (or multiplex PCR methods) to perform subsequent qualitative and quantitative analysis of a certain number of sites in ctDNA[32]. This strategy has the advantage of high sensitivity and accuracy in detecting ctDNA[33]. By using highly selective small gene detection panels, usually containing tens of tumor-specific single nucleotide variants or insertions/deletions, this method can perform deeper sequencing analysis with a smaller volume of blood samples. It can also identify inherent genetic variants of patients, thereby effectively reducing the interference of clonal hematopoiesis on detection results.

The tumor-agnostic method does not require prior acquisition of tumor tissue sequencing data but instead uses a fixed gene panel preset to be closely related to tumor occurrence and development to screen for gene variants potentially derived from tumors in blood samples for ctDNA detection[34]. Its advantages include the ability to track newly emerging mutations, which is helpful for analyzing drug-resistant mutations. Additionally, it has a shorter turnaround time compared to the tumor-informed method. However, due to cost factors, this strategy faces certain challenges in achieving ultra-deep sequencing, which may result in relatively low detection sensitivity[35]. Moreover, this method cannot effectively avoid the interference of clonal hematopoiesis on MRD detection.

Although tumor-informed methods usually provide higher analytical sensitivity and better control for clonal hematopoiesis, they require adequate tumor tissue, higher cost, and longer turnaround time. Tumor-agnostic methods are more accessible when tumor tissue is unavailable and may detect newly emerging resistance alterations, but their sensitivity for low-level MRD may be lower. Thus, tumor-informed assays may be preferable for postoperative MRD surveillance when tissue and resources are available, whereas tumor-agnostic assays may be useful for complementary resistance monitoring. Importantly, no ctDNA-MRD platform has yet become a universally accepted standard for LAGC, and cross-platform comparability remains limited. Figure 1 illustrates the schematic workflow and clinical applications of ctDNA-MRD detection in LAGC.

Figure 1 Detection procedure and clinical application of circulating tumor DNA-based minimal residual disease detection in locally advanced gastric cancer. ctDNA: Circulating tumor DNA; NGS: Next-generation sequencing; MRD: Minimal residual disease; LAGC: Locally advanced gastric cancer.

CLINICAL APPLICATION OF CTDNA-MRD IN LAGC

Application of ctDNA-MRD detection in treatment decision of LAGC

Dynamic changes in ctDNA may provide early molecular information about treatment response in GC[36]. However, evidence that ctDNA-guided treatment changes improve survival in LAGC is still insufficient. At present, ctDNA should be viewed as a complementary biomarker that may support risk stratification, surveillance planning, and clinical trial enrollment rather than as a routine stand-alone basis for changing therapy. For example, Slagter et al[37] confirmed that conventional serum markers such as carcinoembryonic antigen and carbohydrate antigen 19-9 have prognostic value in resectable LAGC. Combining ctDNA with serological markers may improve risk assessment, but the optimal integrated model and clinical decision threshold remain undefined.

Human epidermal growth factor receptor 2 (HER2)-positive LAGC provides one example of how molecular information may influence treatment selection, because HER2-targeted therapy such as trastuzumab can benefit selected patients[38]. ctDNA analysis may help track dynamic HER2 copy-number changes and emerging resistance during treatment[39]. Nevertheless, this use is conceptually different from postoperative MRD detection: HER2 monitoring may inform molecular evolution under therapy, whereas ctDNA-MRD aims to identify residual disease and recurrence risk. These applications should not be conflated, and both require clinical validation in the specific setting in which they are used.

Application of ctDNA-MRD detection in recurrence prediction and prognosis assessment of LAGC

ctDNA has been explored as a biomarker for MRD detection after treatment of gastrointestinal malignancies. In LAGC, postoperative or longitudinal ctDNA positivity is generally associated with a higher recurrence risk. Table 2 summarizes representative studies while also highlighting assay type, sampling time points, reported effect measures where available, and key limitations[11,40-43].

Table 2 Key studies on circulating tumor DNA assessment of minimal residual disease for recurrence prediction and prognosis assessment in locally advanced or resectable gastric cancer.

Ref.	Population and design	Assay and sampling	Key result/effect size	Critical interpretation/limitations
Yang et al[11]	46 stage I-III GC patients after curative-intent resection; prospective cohort	Targeted deep sequencing; preoperative, postoperative, and serial follow-up samples	Postoperative ctDNA positivity was associated with high recurrence risk; molecular relapse preceded radiological recurrence by a median of about 6 months	Proof-of-concept evidence for prognosis, but small cohort and limited power; not sufficient to establish ctDNA-guided therapy
Leal et al[40]	Resectable GC patients from the CRITICS phase III trial; translational analysis	Ultrasensitive targeted sequencing with leukocyte filtration; after preoperative treatment and postoperatively; follow-up NR in this minireview	ctDNA detection at key perioperative time points predicted recurrence	Analytically rigorous workflow, but timing and assay differ from other studies; clinical actionability remains untested
Lan et al[41]	428 GC patients; large-scale clinical study	Dynamic postoperative ctDNA monitoring; detailed platform not uniformly reported in this minireview	Persistent postoperative ctDNA elevation was more closely associated with recurrence than CEA; preoperative ctDNA was not clearly correlated with recurrence	Highlights that baseline ctDNA and postoperative MRD are not interchangeable; thresholds still need standardization
Min et al[42]	GC patients included in multiple studies; meta-analysis	Mixed ctDNA assays, sampling time points, and follow-up schedules	ctDNA was associated with prognosis; post-chemotherapy ctDNA level correlated with DFS	Broader evidence base, but pooled interpretation is limited by heterogeneity in platforms and thresholds
Wu et al[43]	LAGC patients; clinical detection study	ctDNA mutation-burden analysis before and after treatment; follow-up NR in this minireview	Higher ctDNA mutational burden correlated with shorter OS; reduced mutational frequency after treatment was associated with better PFS/OS	Suggests value of quantitative burden, but reproducibility and cutoff definitions require validation

GC: Gastric cancer; ctDNA: Circulating tumor DNA; NR: Not reported or not consistently extractable from the cited report; CEA: Carcinoembryonic antigen; MRD: Minimal residual disease; DFS: Disease-free survival; LAGC: Locally advanced gastric cancer; OS: Overall survival; PFS: Progression-free survival.

In a landmark prospective cohort study, Yang et al[11] investigated the role of MRD detection via ctDNA analysis in predicting recurrence risk among GC patients after curative resection. The study included 46 patients diagnosed with stage I-III GC and revealed several key findings: (1) ctDNA was identified in 45% of plasma samples collected before treatment; (2) The extent of the primary tumor (T stage) was independently correlated with positive preoperative ctDNA; (3) All patients in whom ctDNA was detectable right after surgery eventually suffered from recurrence; (4) Postoperative ctDNA positivity at any time during follow-up correlated with poorer disease-free survival (DFS) and overall survival (OS); and (5) Appearing a median of 6 months before radiological evidence of recurrence.

Leal et al[40] compared and analyzed the white blood cells and cfDNA of patients, and based on this, developed a method to distinguish ctDNA from cfDNA, which also pushed the field forward. In the phase III randomized controlled trial CRITICS for the evaluation of perioperative treatment of resectable GC, their study using ultrasensitive targeted sequencing found that ctDNA was detected in cfDNA after white blood cell filtration to remove germline mutations, which could predict tumor recurrence whether it was nine weeks before or after preoperative treatment. Lan et al[41] conducted a large-scale study on 428 GC patients, further confirming the association between postoperative ctDNA monitoring and clinical outcomes. Compared with carcinoembryonic antigen, if the postoperative ctDNA concentration continues to increase, the relationship with tumor recurrence is more closely related, especially in the first 12 months after surgery. Another point to mention is that preoperative ctDNA levels are actually not related to recurrence, which also shows that postoperative tracking of ctDNA is particularly useful in guiding monitoring plans[44].

The existing meta-analysis evidence suggests that ctDNA has the potential to become a prognostic biomarker in GC[42]. After completing three courses of systemic chemotherapy, the measured ctDNA levels can be used to reliably predict DFS. If the ctDNA level of patients is relatively low, their DFS tends to be significantly longer[45]. Consistently, reductions in ctDNA levels post-treatment correlate with better treatment response and progression-free survival (PFS), with lower ctDNA levels predicting superior prognosis[46]. When ctDNA is used for genetic testing, mutation characteristics related to treatment efficacy, disease progression, and recurrence can be identified. In GC patients, if the mutation load of ctDNA is high, the OS period is shorter than that of patients with low load[43]. If the mutation frequency decreases after treatment, both PFS and OS will improve[47]. There are also epigenetic changes such as the methylation of RAS association domain family 1A and protocadherin 10 in ctDNA. Compared with cases without methylation, the prognosis is slightly worse[48].

However, the evidence should be interpreted cautiously. Yang et al[11] provided an important prospective proof of concept but included only 46 patients. The CRITICS translational analysis by Leal et al[40] used an ultrasensitive workflow with leukocyte filtering, whereas other studies used different panels, thresholds, and sampling schedules. Lan et al[41] further showed that preoperative ctDNA was not clearly associated with recurrence, indicating that baseline ctDNA detection and postoperative MRD assessment answer different clinical questions. Neutral or less favorable observations are therefore important for balance. CtDNA is not detectable in every patient, particularly in low-shedding tumors or in disease patterns such as isolated peritoneal spread. A negative ctDNA result should not be interpreted as definitive absence of residual disease. Instead, ctDNA should be integrated with pathological stage, lymph-node status, radiological findings, serum markers, treatment response, and patient-specific clinical risk.

Overall, current data support ctDNA-MRD mainly as a prognostic tool for recurrence risk stratification and earlier molecular detection of relapse. The predictive utility of ctDNA - that is, whether a ctDNA result can identify which treatment will improve outcome - and its actionable utility for changing adjuvant therapy remain investigational in GC.

Application of ctDNA-MRD detection in evaluating the effect of neoadjuvant chemotherapy for LAGC

Neoadjuvant chemotherapy (NACT), refers to systemic chemotherapy used before surgery. Its main purpose is to reduce tumor volume, make surgery smoother, and increase the likelihood of cancer lesion resection[49]. Clinical evidence shows that NACT can significantly increase the probability of curative resection in patients with LAGC and improve their long-term prognosis[50]. A prospective study was conducted by Zhang et al[51], who selected 79 patients with stage II to III GC. These patients underwent two cycles of NACT before undergoing D2 lymph node dissection for gastrectomy, which is the standard lymph node dissection procedure for GC. The aim was to control all tumors in the region. We will collect plasma samples at baseline, after completion of NACT, and after surgery, as well as tissue samples obtained before treatment and during surgery.

After NACT combined with surgical treatment, the overall detection rate of ctDNA gradually decreased, indicating a reduction in tumor burden. Patients with detectable ctDNA at baseline, post-chemotherapy, or post-surgery exhibited significantly worse OS compared to those with undetectable ctDNA. Taking the situation after NACT as an example, patients with ctDNA negative conversion have a three-year OS rate of approximately 73%, while if ctDNA is still positive, this proportion is only 34%. Moreover, patients who maintain a negative ctDNA test before and after receiving NACT have the best overall prognosis. This also indicates that continuous monitoring of ctDNA status has good prognostic reference value in the evaluation of treatment efficacy for LAGC.

A follow-up meta-analysis conducted on observational studies focused on a population of non-metastatic cancer patients receiving NACT, ultimately confirming the association between dynamic changes in ctDNA and tumor recurrence[52]. The correlation between ctDNA and prognosis is particularly evident at the two critical time points before and after treatment. Before treatment, the value of ctDNA can reflect the initial tumor burden, and after surgery, it can indicate whether there are still small residual lesions in the body that have not been detected. This time pattern can also be seen, ctDNA can be used as a very sensitive indicator to monitor the effectiveness of early treatment and detect subclinical residual lesions, which are particularly important for knowing the risk of recurrence in advance.

When multiple related studies are considered together, ctDNA-MRD detection may have clinical value in LAGC patients receiving NACT by helping clinicians: (1) Monitor molecular response during treatment; (2) Stratify recurrence risk after surgery; and (3) Identify patients who may be considered for closer surveillance or clinical trial enrollment. However, routine escalation or de-escalation of postoperative therapy based solely on ctDNA status is not yet supported by prospective interventional evidence.

Dynamic monitoring of ctDNA-MRD in LAGC

In the field of cancer diagnosis and treatment, the detection of ctDNA-MRD holds great promise for assessing the tumor status and the effectiveness of treatment[53]. In the preoperative stage, the detection of ctDNA-MRD in patients with LAGC is closely related to the T stage[11]. However, a study by Kim et al[44] drew a different conclusion, finding no statistically significant correlation between preoperative ctDNA positivity and GC T stage (tumor invasion depth) in their cohort of LAGC patients. All enrolled patients had T3 or T4 stage disease, yet no clear disparity in preoperative ctDNA detection rates was observed across different T subgroups. These discrepancies among studies indicate that tumor burden reflected solely by T stage cannot fully determine the shedding level of ctDNA into peripheral blood, and that intrinsic tumor biological characteristics are critical factors governing ctDNA release.

Surgery and perioperative therapy can influence ctDNA detectability. In a study of 79 patients with stage II/III LAGC, serial sampling at baseline, after NACT, and after surgery showed a gradual decrease in ctDNA detection after treatment, with the postoperative detection rate falling to 47.4%[51]. This pattern is consistent with a reduction in tumor burden, but postoperative cfDNA release, blood-draw timing, and assay sensitivity can also influence results.

Longitudinal monitoring may provide additional information beyond a single time point because conversion from negative to positive ctDNA, persistent positivity, or failure of ctDNA clearance can reflect different biological scenarios. Nevertheless, the optimal frequency of testing and the clinical action triggered by each pattern remain unresolved[11].

Qi et al[54] reported a LAGC case in which ctDNA-MRD positivity emerged with rising carbohydrate antigen 19-9 while imaging did not yet show abnormal lesions, and later ctDNA dynamics were consistent with disease progression. Such case-level evidence illustrates the potential sensitivity of molecular monitoring, but it should be regarded as hypothesis-generating rather than sufficient proof of improved outcomes from earlier intervention.

Application and exploration of ctDNA-MRD in immunotherapy of LAGC

Currently, immunotherapy combined with chemotherapy is the most commonly used treatment regimen for LAGC[55]. Therefore, identifying molecular biomarkers that can predict the efficacy of this treatment is crucial for optimizing immunotherapy for GC patients[56].

In a clinical trial conducted by Jiang et al[57], 30 patients with LAGC were enrolled and received first-line treatment with programmed cell death protein 1 inhibitors combined with chemotherapy. Peripheral blood samples were collected at baseline and after two cycles of treatment, and baseline tissue samples were also collected. The plasma and tumor tissue were analyzed using NGS gene detection panels with 47 and 737 genes, respectively, to assess genomic alterations. The study defined ctDNA response as the clearance of maximal variant allele frequency (maxVAF) compared with baseline levels. The results showed that patients with ctDNA response had a significantly higher objective response rate than non-responders (P = 0.0073), and their PFS (15.6 months vs 6.0 months, P = 0.003) and OS (not reached vs 9.0 months, P = 0.011) were also significantly longer than those of non-responders. In another study by Jin et al[58], 46 patients with LAGC who received programmed cell death protein 1 antibody immunotherapy underwent 425-gene NGS testing. The results indicated that patients with undetectable ctDNA after treatment had a median PFS of 7.4 months, compared to 4.9 months for patients with detectable ctDNA (P = 0.025).

In conclusion, dynamic variation in ctDNA may be a potential biomarker for immunotherapy response and prognosis in patients with LAGC. However, the available evidence remains based on relatively small cohorts and non-standardized assays. ctDNA response should therefore be considered an exploratory marker that may help design future trials, rather than a validated criterion for selecting or discontinuing immunotherapy in routine care.

DISCUSSION

Study heterogeneity and critical appraisal

Several sources of heterogeneity limit direct comparison across ctDNA-MRD studies in GC. Patient populations vary from stage I-III resectable GC to stage II/III LAGC. Treatment paradigms also differ, including surgery alone, perioperative chemotherapy, neoadjuvant chemoradiotherapy, adjuvant chemotherapy, targeted therapy, and immunotherapy[11,40,51]. Assay methods include droplet digital PCR, targeted NGS, broad-panel NGS, tumor-informed personalized assays, tumor-agnostic panels, mutational burden approaches, and methylation-based approaches[21-23]. Sampling time points range from baseline and post-neoadjuvant therapy to early postoperative and serial surveillance periods. Finally, thresholds for MRD positivity are inconsistent, including single-variant detection, variant allele frequency cutoffs, mutation burden, methylation signal, or assay-specific algorithms. These differences make it difficult to generalize results across studies or define a universal clinical cutoff.

Clinical application framework and interpretation of ctDNA results

From a practical perspective, ctDNA testing in LAGC may be considered at four clinically relevant time points. First, baseline testing before neoadjuvant therapy can establish whether a tumor is ctDNA-shedding and may provide a molecular reference, but baseline positivity mainly reflects tumor burden and is not equivalent to MRD. Second, testing after neoadjuvant therapy may help evaluate molecular response. Third, early postoperative testing, after immediate surgical cfDNA noise has decreased, may identify patients at high recurrence risk. Fourth, serial surveillance during and after adjuvant therapy may detect molecular relapse before radiological recurrence. However, these time points should be interpreted as a proposed framework for research and multidisciplinary discussion, not as a mandatory standard of care[11,40,51,54].

Potential clinical responses to ctDNA results should also be conservative. A positive postoperative or longitudinal ctDNA result may justify closer imaging surveillance, repeat molecular testing, multidisciplinary review, or enrollment in a ctDNA-guided clinical trial. It may also support discussion of adjuvant treatment intensification in a trial setting. Conversely, a negative ctDNA result should not be used alone to omit standard adjuvant therapy or reduce surveillance intensity, because false-negative results may occur in low-shedding tumors, low-volume disease, peritoneal-only relapse, inadequate sampling, or assays with insufficient sensitivity.

Limitations of current evidence

The main limitations of current evidence include small sample sizes, single-center cohorts, short or heterogeneous follow-up, platform-dependent sensitivity, non-standardized pre-analytical processing, inconsistent MRD definitions, uncertain cost-effectiveness, and limited accessibility of high-depth personalized assays. Another major limitation is the absence of prospective randomized evidence showing that ctDNA-guided management improves survival, quality of life, or cost-effectiveness compared with standard clinicopathological risk assessment and surveillance. These limitations should be explicitly considered when translating ctDNA-MRD findings into clinical practice.

CONCLUSION

GC remains a lethal malignancy in which late diagnosis and postoperative recurrence continue to limit long-term survival. Conventional imaging and serum tumor markers have limited sensitivity for detecting occult residual disease after radical-intent therapy or perioperative treatment. ctDNA-MRD assessment therefore offers a biologically plausible and clinically attractive approach for earlier recurrence-risk stratification and molecular monitoring.

However, ctDNA testing still faces important technical and clinical limitations. The short half-life and low abundance of ctDNA require strict control of blood collection, transport, plasma processing, storage, cfDNA extraction, sequencing, bioinformatic filtering, and result interpretation[59,60]. Across gastrointestinal cancer studies, MRD-positivity criteria remain inconsistent[40,61-66]. Sensitivity can be limited during or after chemotherapy and in tumors with low ctDNA shedding[67,68]. MRD testing also depends on sample availability, enrichment strategies, assay depth, turnaround time, cost, and access to validated platforms[69].

From a clinical perspective, most available studies remain observational or retrospective, and large multicenter prospective trials are still needed to determine whether ctDNA-guided management improves outcomes[70]. Repeated sampling may also increase patient burden and health-care costs. Standardized analytical workflows, clinically meaningful thresholds, reporting formats, and decision algorithms must be established before routine implementation.

Precision oncology is constantly advancing, and the commonly used clinical testing methods used to evaluate disease recurrence and tumor heterogeneity are a bit outdated. ctDNA testing is a non-invasive diagnostic method with good sensitivity and specificity, and is gradually becoming an important tool to guide clinical practice. It can not only serve as a biomarker for MRD, help detect recurrence risks earlier and adjust treatment plans, but also provide some ideas for targeted therapy and immunotherapy research[71]. However, it must be emphasized that ctDNA-MRD assessment is still a promising but investigational technology for LAGC. At this stage, it can be preferentially used for recurrence risk stratification and prognosis evaluation, while its value in guiding clinical treatment decisions has not been fully verified. Routine large-scale clinical application is not yet mature. Perhaps in future research progress, multi omics methods such as whole genome sequencing and methylation analysis will be combined with artificial intelligence to improve the accuracy of detection.

Through technical optimization, unified operation standards and verification of large-scale multicenter intervention trials, ctDNA-MRD detection can become a key auxiliary means for the whole-course management of LAGC and bring new hope for improving the long-term prognosis of GC patients.

ACKNOWLEDGEMENTS

We would like to express our sincere gratitude to Professor Li HT for his invaluable guidance and insightful suggestions throughout this research. We also extend our thanks to Department of General Surgery, The 940^th Hospital of Joint Logistics Support Force of Chinese PLA, for providing the experimental platforms that made this study possible.