Predictive role of circulating tumor DNA based molecular residual disease for long-term outcomes in non-small cell lung cancer patients: a meta-analysis

Abstract

Purpose

To identify the predictive role of circulating tumor DNA (ctDNA)-based molecular residual disease (MRD) for long-term outcomes in non-small cell lung cancer (NSCLC) patients.

Methods

Several databases were searched. The primary outcome was progression-free survival (PFS), and the secondary outcomes included overall survival (OS) and cancer-specific survival (CSS). Hazard ratios (HRs) and 95% confidence intervals (CIs) were combined, and subgroup analyses based on the time point of MRD detection (landmark vs. longitudinal) and treatment (surgery vs. chemoradiotherapy) were further performed.

Results

Ten studies with 1859 cases were included. Pooled results demonstrated that positive ctDNA MRD significantly predicted worse PFS (HR = 11.19, 95% CI: 6.10–20.52, P < 0.001), OS (HR = 6.34, 95% CI: 2.27–17.74, P < 0.001) and CSS (HR = 16.67, 95% CI: 10.00–25.00, P < 0.001). Subgroup analysis by the time points of MRD detection (landmark: HR = 8.93, P < 0.001; longitudinal: HR = 17.52, P < 0.001) and treatment (surgery: HR = 11.63, P < 0.001; chemoradiotherapy: HR = 5.56, P < 0.001) revealed consistent results.

Conclusion

ctDNA-based MRD could serve as a valuable prognostic indicator in NSCLC, and patients with positive ctDNA-based MRD are at significantly greater risk of recurrence and lower survival.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12957-025-03890-3.

Introduction

Surgery remains the primary treatment for most non-small cell lung cancers (NSCLCs) [1]. However, even when the visible tumor is completely removed, there can still be lesions that are not detectable with the naked eye, leading to a greater risk of postoperative recurrence. Therefore, regular postoperative follow-up is crucial for NSCLC patients. Traditional methods for detecting tumor recurrence after surgery include blood tests for tumor markers and routine pulmonary imaging. However, postoperative inflammation can cause changes such as dense or patchy shadows and fibrous streaks, which may obscure the accurate identification of tumor recurrence by imaging. Additionally, tumor markers can be influenced by other factors, reducing their specificity [2]. Imaging tests also have limited sensitivity, as they can detect only disease recurrence that is visible to the naked eye. Therefore, monitoring the risk of recurrence in patients with NSCLC remains a pressing issue that needs to be addressed in clinical practice.

Liquid biopsy is a new, noninvasive testing method that analyzes tumor-derived substances in blood or any other bodily fluids. Compared with tissue biopsy, liquid biopsy offers the advantages of simplicity and safety. It relies primarily on the analysis of circulating tumor cells (CTCs) or circulating tumor DNA (ctDNA) in blood samples to reflect the tumor mutation burden (TMB) and minimal residual disease (MRD) in cancer patients. MRD, also known as measurable residual disease or molecular residual disease, was initially used in hematologic malignancies, referring to residual tumor cells or molecular remnants remaining in the body after treatment, which are potential sources of tumor recurrence [3, 4]. In recent years, liquid biopsy using ctDNA sequencing to detect MRD has become more widely applied in solid tumors, including NSCLC [5–8]. Several studies have explored the relationship between ctDNA-based MRD monitoring and long-term clinical outcomes in patients with NSCLC, but the results have been inconsistent [9–18].

Therefore, this meta-analysis aimed to determine the predictive role of ctDNA-based MRD for long-term outcomes among NSCLC patients on the basis of available evidence contributing to prognosis assessment and early intervention for high-risk patients.

Materials and methods

Our meta-analysis was conducted according to the PRISMA 2020 [19]. This meta-analysis has been registered in PROSPERO with the registration number INPLASY202550050 (DOI: 10.37766/inplasy2025.5.0050; https://inplasy.com/?s=INPLASY202550050).

Literature search

We searched the PubMed, Embase, Web of Science and CNKI databases up to August 31, 2024, with the following terms: lung, pulmonary, tumor, cancer, neoplasm, carcinoma, minimal residual disease, molecular residual disease, measurable residual disease, MRD, survival, prognosis and prognostic. Free texts and MeSH terms were applied. The detailed search strategies for each database are presented in Supplementary file 1.

Inclusion criteria

Studies meeting the following criteria were included: (1) primary NSCLC patients; (2) the MRD status was detected and identified after antitumor treatment on the basis of the ctDNA; (3) patients were divided into positive or negative MRD groups, and the clinical outcomes, including progression-free survival (PFS), overall survival (OS) and cancer-specific survival (CSS), were compared; (4) hazard ratios (HRs) with 95% confidence intervals (CIs) for the above endpoints were reported, or Kaplan‒Meier survival curves were generated; and (5) full texts were available.

Exclusion criteria

Studies that met the following criteria were excluded: (1) insufficient, overlapping or duplicated data; and (2) case reports, letters, editorials, reviews, animal trials or meeting abstract studies.

Data extraction

The following information was extracted: first author, country, year, sample size, number of patients with positive MRD, tumor stage, treatment, detection time point of MRD, follow-up period, endpoint, HR and 95% CI.

Methodological quality assessment

All included studies were cohort studies. Therefore, the Newcastle‒Ottawa Scale (NOS) scoring tool was used to assess quality, and studies with an NOS score > 5 were included [20].

Statistical analysis

Heterogeneity between studies was assessed by I² statistics and the Q test. If significant heterogeneity was detected (I² > 50% and/or P < 0.1), a random effects model was applied; otherwise, a fixed effects model was applied. HRs and 95% CIs were combined. Notably, owing to the similarity between PFS and DFS in patients receiving radical surgery, DFS was regarded as PFS during our analysis. Subgroup analyses based on the time point of MRD detection [landmark (within one month postoperative) vs. longitudinal] and treatment (surgery vs. chemoradiotherapy) were conducted. Sensitivity analysis for PFS was conducted to detect the sources of heterogeneity and assess the stability of the overall results. Begg’s funnel plot with Egger’s test was used to detect publication bias, and significant publication bias was defined as P < 0.05 [21, 22]. The above analyses were performed using STATA (v. 15.0) software.

Results

Literature search and selection

As presented in Figs. 1 and 3465 publications were identified, and 397 duplicated records were first removed. A total of 3049 records were excluded after the titles and abstracts were reviewed. Eventually, ten cohort studies were included in our meta-analysis [9–18].

Fig. 1.

Prisma flow diagram of this meta-analysis

Basic characteristics

Among the ten included studies, 1859 participants were enrolled, with sample sizes ranging from 40 to 387. Most studies (8/10) were from China and focused on operated patients. Long-term longitudinal detection and analysis of MRD were applied in most studies. Detailed information is provided in Table 1.

Table 1.

Basic characteristics of included studies

Author	Country	Year	Sample size	Number of MRD+	Tumor stage	Treatment	Detection time of MRD	Follow-up period	Endpoint	NOS
Chaudhuri [9]	USA	2017	40	17	TNM I-III	Mixed (curative)	Within 4 months after treatment completion	35.1 (6.9–56) months	PFS, OS	6
Xia [10]	China	2022	330	26	TNM I-III	Surgery	Within postoperative 1 month	1068 (341–1340) days	DFS	7
Zhang [11]	China	2022	245	21	TNM I-III	Surgery	Postoperative one month and every 3–6 months	19.7 months	DFS	7
Chen K [12]	China	2023	387	19/23	TNM I-III	Surgery	Postoperative 3–7 days/one month and every 3-6months	1071 (987–1137) days	DFS	7
Chen Z [13]	China	2023	114	42	TNM I-IIIA	Surgery	Postoperative 1 week	0–60 months	OS	7
Jung [14]	Republic of Korea	2023	278	16	TNM I-III	Surgery	Postoperative one month and every 3–6 months	62 (2–77)	DFS	6
Pan [15]	China	2023	139	28	TNM IIB-IIIC	CRT	Every 3–6 months after treatment	20.8 months (median)	PFS, CSS	6
Zhang [16]	China	2023	73	13	TNM I-III	Surgery	Postoperative	29 (4–67) months	DFS	6
Fu [17]	China	2024	49	10/14	TNM I-III	Surgery	Postoperative one month and every 3–6 months	9.9 (0.1–54.2)	DFS	6
Li [18]	China	2024	204	36	TNM I-IIIA	Surgery	Postoperative	NR	DFS	7

MRD: Molecular Residual Disease; NOS: Newcastle-Ottawa Scale; LC: lung cancer; NSCLC: non-small cell lung cancer; CRT: chemoradiotherapy; PFS: progression-free survival; CSS: cancer-specific survival; OS: overall survival; DFS: disease-free survival

Predictive role of ctDNA-based MRD status in long-term outcomes in patients with NSCLC

Nine studies explored the relationship between ctDNA-based MRD status and PFS among NSCLC patients [9–12, 14–18]. The pooled results demonstrated that positive ctDNA-based MRD was related to significantly worse PFS (HR = 11.19, 95% CI: 6.10–20.52, P < 0.001; I² = 77.3%, P < 0.001) (Fig. 2). Subgroup analysis based on the time point of MRD detection (landmark: HR = 8.93, 95% CI: 5.78–13.81, P < 0.001; longitudinal: HR = 17.52, 95% CI: 4.73–64.85, P < 0.001) and treatment strategy (surgery: HR = 11.63, 95% CI: 5.59–24.21, P < 0.001; chemoradiotherapy: HR = 5.56, 95% CI: 3.64–8.49, P < 0.001) revealed similar results. (Table 2)

Fig. 2.

The association of ctDNA based MRD status with PFS in non-small cell lung cancer. MRD: Molecular Residual Disease; PFS: progression-free survival

Table 2.

Results of meta-analysis

Items	Number of studies	Hazard ratio	95% confidence interval	P value	I2	P value
Progression-free survival	9	11.19	6.10-20.52	<0.001	77.3%	<0.001
Time point of MRD detection
Landmark (within postoperative one month)	4	8.93	5.78–13.81	<0.001	0.0%	0.968
Longitudinal	4	17.52	4.73–64.85	<0.001	89.0%	<0.001
Treatment
Surgery	7	11.63	5.59–24.21	<0.001	77.7%	<0.001
Chemoradiotherapy	1	5.56	3.64–8.49	<0.001	-	-
Overall survival	2	6.34	2.27–17.74	<0.001	53.1%	0.144
Cancer-specific survival	1	16.67	10.00–25.00	<0.001	-	-

MRD: Molecular Residual Diseases

In addition, two studies identified the relationship between ctDNA MRD and OS among NSCLC patients [9, 13]. Pooled results indicated that positive ctDNA-based MRD predicted shorter OS (HR = 6.34, 95% CI: 2.27–17.74, P < 0.001; I² = 53.1%, P = 0.144) (Fig. 3). Furthermore, Pan et al. reported that ctDNA-based MRD was also associated with CSS (HR = 16.67, 95% CI: 10–25, P < 0.001) in NSCLC patients receiving chemoradiotherapy [15].

Fig. 3.

The association of ctDNA based MRD status with OS in non-small cell lung cancer. MRD: Molecular Residual Disease; OS: overall survival

Sensitivity analysis and publication bias

The sensitivity analysis for PFS indicated the stability and reliability of our results, and none of the included studies affected the conclusion (Fig. 4).

Fig. 4.

Sensitivity analysis about the association of ctDNA based MRD status with PFS in non-small cell lung cancer. MRD: Molecular Residual Disease; PFS: progression-free survival

Bess’s funnel plot for PFS (Fig. 5) was symmetrical, with P = 0.178 for Egger’s test, indicating nonsignificant publication bias.

Fig. 5.

Begg’s funnel plot about the association of ctDNA based MRD status with PFS in non-small cell lung cancer. MRD: Molecular Residual Disease; PFS: progression-free survival

Discussion

Compared with traditional methods, ctDNA-based MRD monitoring offers greater clinical value in predicting the risk of lung cancer recurrence for several reasons. ctDNA testing can detect extremely small fragments of tumor-derived DNA, which appear in the blood earlier than clinically or radiologically visible recurrences. This high sensitivity allows ctDNA to detect very small residual tumors or early signs of recurrence, even when the patient is asymptomatic or when imaging tests appear normal [23]. The levels of ctDNA can dynamically reflect changes in the tumor burden within the body. By collecting multiple blood samples and monitoring them over time, physicians can track tumor activity in real time, enabling the timely detection of potential recurrences. Unlike traditional tissue biopsies, which provide only localized information, ctDNA offers systemic insights through blood analysis. This means that ctDNA can be used to detect recurrences in any part of the body in lung cancer patients, not only the original tumor site.

Our results indicate that ctDNA-based MRD status can predict postoperative recurrence risk among NSCLC patients. In the study by Xia et al., MRD-positive patients benefited from postoperative adjuvant therapy (P = 0.002). However, adjuvant therapy did not improve PFS in the MRD-negative population (P = 0.283) [10]. Therefore, ctDNA-based MRD status could contribute to the formulation of treatment strategies. Furthermore, Pan et al. explored the clinical application of ctDNA in NSCLC patients receiving chemoradiotherapy and reported novel findings. The concentration of ctDNA significantly decreased during radiotherapy, and patients whose ctDNA-MRD had cleared and remained undetectable until the end of radiotherapy had a median PFS of 22.9 months, which was significantly better than that of those whose MRD had cleared later (10.3 months). Additionally, early MRD clearance did not improve survival in patients who received consolidation immunotherapy (20.3 vs. 22.9 months, P = 0.917) [15]. These results also highlight the role of ctDNA-based MRD in guiding treatment decisions for NSCLC patients.

Several meta-analyses have investigated the clinical role of ctDNA in lung cancer. Lu et al. demonstrated that positive preoperative ctDNA levels predict worse survival in patients with NSCLC [24]. Shen et al. reported that the presence of MRD was associated with a higher recurrence rate and shorter survival [25]. However, our meta-analysis differs greatly from these previous meta-analyses. First, the meta-analysis by Lu et al. focused primarily on preoperative ctDNA levels as a prognostic biomarker. Our analysis specifically investigated ctDNA-based MRD, a dynamic, posttreatment biomarker that reflects minimal residual disease status and is more directly associated with long-term outcomes such as recurrence and survival. Second, our meta-analysis incorporates more recent and MRD-specific studies that were not included in the above meta-analyses, thus providing an updated and more targeted evaluation of ctDNA-MRD. Third, we performed subgroup analyses on the basis of the timing of MRD detection and treatment, offering more granular insight into the prognostic utility of MRD-guided surveillance. Therefore, our study complements and extends prior research by emphasizing the evolving role of ctDNA-based MRD as a longitudinal prognostic tool in NSCLC.

There are several limitations. First, only ten studies were included. Second, most studies are from China, which may affect the generality of the findings. Third, the detection time of MRD varied across the included studies, although we conducted subgroup analysis on the basis of this parameter (landmark vs. longitudinal). More specific differentiation, e.g., postoperative 3 days and 7 days, and analysis should be performed, especially for surgical patients. Fourth, additional subgroup analyses of other important parameters, such as EGFR mutations and ALK translocations, were unavailable.

Conclusion

According to our pooled results, ctDNA-based MRD could serve as a valuable prognostic indicator in NSCLC, and patients with positive ctDNA-based MRD are at increased risk of recurrence and lower survival. However, more studies and detailed investigations are needed to verify the above findings.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Siyuan Che and Dongliang Yu designed this meta-analysis, performed the literature search and selection, collected data, performed the statistical analysis, wrote the paper, and revised the manuscript.

Funding

None.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All procedures performed in studies that involved human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Consent to participate was not applicable for this type of study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.