The Utility of ctDNA in Lung Cancer Clinical Research and Practice: A Systematic Review and Meta-Analysis of Clinical Studies

Abstract This systematic review with embedded meta-analysis aimed to evaluate the clinical utility of circulating tumor DNA (ctDNA) in lung cancer. After screening and review of the Embase database search, 111 studies from 2015 to 2020 demonstrated ctDNA’s value in prognostication/monitoring disease progression, mainly in patients with advanced/metastatic disease and non–small cell lung cancer. ctDNA positivity/detection at any time point was associated with shorter progression-free survival and overall survival, whereas ctDNA clearance/decrease during treatment was associated with a lower risk of progression and death. Validating these findings and addressing challenges regarding ctDNA testing integration into clinical practice will require further research.


Introduction
Released by cancer cells through apoptosis, lysis of circulating tumor cells, and active secretion, circulating tumor DNA (ctDNA) constitutes a small fraction of cell-free DNA (cfDNA) in blood (1,2). Unlike tumor tissue sampling, which is limited by invasiveness and associated risks, tissue accessibility, sampling frequency, and cost, ctDNA offers a convenient, minimally invasive method to measure tumor burden and genomic profile. Additionally, because of the short half-life in circulation, ctDNA is considered a "real-time" snapshot of tumor activity (2). These features make ctDNA a promising molecular biomarker in oncology and have triggered interest in ctDNA utility during the past decade. Recent technological advances in the detection and quantification of low-abundance ctDNA have accelerated the progress of integrating ctDNA testing into clinical decision-making in cancer care. Recently published US Food and Drug Administration draft guidance and the European Society for Medical Oncology recommendations on ctDNA utility in cancer further confirm the value and potential of ctDNA while also highlighting the knowledge gaps and data limitations in its clinical utility (3,4). Within the field, there is substantial interest in refining study designs and standardizing assessment methods to improve the level of evidence for the different ctDNA utilities in early-and late-stage disease (3)(4)(5).
ctDNA biology, evaluation techniques, and potential applications have been reviewed in numerous publications (2,(6)(7)(8)(9). As the treatment landscape for lung cancer has increasingly shifted toward the use of targeted therapies and immunotherapies (10)(11)(12), there has been substantial interest in using ctDNA for the detection of specific gene mutations to inform treatment selection. Multiple actionable mutations have been identified in non-small cell lung cancer (NSCLC), including mutations in EGFR, KRAS, and ALK (11). ctDNA-mediated detection of these mutations has the potential to help direct targeted therapy selection or monitor treatment response in situations where biopsies are not feasible.
Previous systematic reviews on ctDNA in lung cancer have primarily focused on diagnostic accuracy, evaluating the concordance between ctDNA results and tissue-based testing (13)(14)(15)(16). Another systematic review in NSCLC focused on the association between ctDNA and response to immune checkpoint inhibitors (ICIs) (17). The literature on ctDNA utility within small cell lung cancer (SCLC) is more limited, with 1 systematic review that examined the role of ctDNA for disease monitoring and genomic profiling (18). Although these analyses furnish valuable insights, they do not provide information on the bigger picture of how ctDNA is being used across lung cancer in this rapidly developing field. This systematic review aimed to comprehensively summarize data on the utility of ctDNA from published clinical studies of lung cancer. The findings detail the current landscape of ctDNA applications in lung cancer, serving as a foundation to inform study design development and result interpretation.

Literature identification and eligibility criteria
A systematic literature review was performed using the Embase database to identify published literature on lung cancer and ctDNA from January 1, 2015, to December 31, 2020. The final search terms were "('lung cancer'/exp OR 'lung cancer') AND ('ctDNA' or 'circulating tumor DNA' or 'cell-free tumor DNA') AND [english]/lim AND [2015-2020]/py". No limitations were placed on lung cancer type or disease stage.
Full-text original clinical research articles were included; review/meta-analysis articles and non-English articles were excluded. Conference abstracts, editorials, commentaries, case reports, and study protocols were excluded because of incompleteness of data or small sample sizes.
Articles were included for studies in which ctDNA was linked to clinical features (eg, tumor stage) or outcomes (eg, progression-free survival). Articles that were limited to the following topics were considered out of scope for this review and therefore excluded: cfDNA, specimens other than blood/plasma/serum, ctDNA analytic validation (eg, descriptive analysis of ctDNA dynamics or correlation with tumor tissue DNA), or assessment of the tumor burden without relating ctDNA to tumor characteristics or clinical outcomes. However, articles that measured the genomic alterations of specific molecular tumorrelated alterations in cfDNA were included. Initial selection based on title and abstract was performed by 2 independent reviewers; any disagreements on inclusion/exclusion were resolved by a third reviewer or consensus-based discussion. One reviewer also identified additional articles using reference lists from systematic review or meta-analysis reports on the same or a similar topic.

Data collection and extraction
One reviewer extracted data from included articles into a predefined extraction Excel table, covering article information (authors, title, publication year), study design (study period, location, research/clinical settings, study population, sample size), tumor characteristics (lung cancer subtype, stage), and ctDNA-related information (platform type, test timing, clinical utility). Extracted data were quality checked by a second reviewer.
The ctDNA utilizations proposed by Wan and colleagues and Narayan and colleagues were adapted (Table 1) (2,19). Studies were classified into 3 categories based on ctDNA testing aims, the timing of ctDNA tests, and endpoints/outcomes in the analyses: (1) diagnosis (early diagnosis, screening), (2) prognostication (detection, profiling, prognostication), and (3) monitoring (monitoring disease progression, treatment response, or genomic evolution using longitudinal ctDNA samples). Because disease history and ctDNA dynamics are a continuous course, categories were not mutually exclusive. Studies that fulfilled more than 1 predefined category were reported in multiple categories.

Meta-analysis
A meta-analysis was conducted to quantitively review and synthesize the association of ctDNA with clinical outcomes for patients with lung cancer. Studies that reported hazard ratios (HRs) with corresponding 95% CIs for the association between ctDNA and progression-free survival (PFS) or overall survival (OS) were included in the meta-analyses. ctDNA detection was assessed as a binary variable (positive vs. negative or decrease/clearance yes vs. no). For longitudinal ctDNA assessments, results were grouped according to 3 broad categories (before, during, or after treatment). The definition of ctDNA decrease/clearance varied across studies in terms of the timing and number of ctDNA assessments. However, because of the small number of studies, all definitions of ctDNA decrease/clearance were grouped together and treated as the same. Separate analyses were performed for the detection of EGFR mutation(s) in ctDNA. Although mutations in other tumor genes were of interest, data were insufficient for meta-analyses.
Pooled estimates were calculated using fixedeffect and random-effect models (20), respectively, depending on the p value (cut point, 0.05) and I 2 for the test of heterogeneity (>50%). Results were presented in forest plots. Stratified analyses were performed by potential effect modifiers, including ctDNA assessment timing (before, during, or after treatment), study type (observational study vs. clinical trial), lung cancer histology subtype (NSCLC vs. other), and disease stage (advanced/metastatic vs. early) when adequate data were available (!3 studies in a category). The appropriateness of the assumptions made for the analyses was assessed by performing the I 2 test of heterogeneity and using metaregression to assess effect modifiers (21,22). Egger tests (when study number was >10) and funnel plots were used to assess publication bias. All statistical tests were 2-sided with a of 0.05. All analyses were performed using R version 4.1.0 with the "meta" and "metafor" packages.

Study selection
A total of 2200 publications were identified in the initial Embase search ( Figure 1). Of the 238 articles selected from the initial screening for full-text review, 101 articles were selected for inclusion. Ten additional articles were identified by reviewing the reference lists for recent selected meta-analysis or review articles (23-25). The final review included 111 articles (Table 2).         (42,. The association between detectable ctDNA before treatment and baseline tumor features (eg, tumor size, type of metastasis) was not consistently reported. Several studies reported higher ctDNA levels with advanced disease stage; this association was not consistently observed, perhaps owing in part to the inclusion of relatively few patients with early-stage disease (30-34, 36,42,49). Similarly, several studies reported ctDNA positivity or a higher concentration of ctDNA associated with specific metastatic sites or the number of metastatic sites, but results were sparse and inconsistent (32,33, [49][50][51][52]86,87).

Study and patient characteristics
Most studies reported shorter OS or PFS associated with ctDNA positivity and/or higher concentrations of ctDNA, regardless of lung cancer stage (Supplementary Tables 4 and 5) . Among studies that evaluated overall ctDNA positivity (not specific genetic mutations in ctDNA) before treatment, only 2 studies did not find an association between overall ctDNA positivity/detection and survival outcomes (44,71). Most of the studies that evaluated mutations in a single predefined mutated oncogene (eg, T790M resistance mutation in the EGFR gene, KRAS) did not find an association between pretreatment ctDNA and prognosis (34,46,54,59,63,67,68). One study found patients with a low pretreatment T790M/EGFR ratio had significantly less tumor shrinkage (47). Three of the studies that reported no association between detectable pretreatment ctDNA and subsequent outcomes were clinical trials (54,67,68). Four studies focused on the association between the methylation of specific genes (KMT2C, SOX17, SHOX2) and prognosis and reported improved survival outcomes with lower levels of methylation (30,35,42,45). Ten studies analyzed the association between pretreatment ctDNA and the overall response rate, disease control rate, or tumor response; however, the results were not consistent (38,46,48,60,70,75,77,79,84,85).
The remaining 22 studies included only patients with advanced/metastatic lung cancer. Most studies consistently reported an association between ctDNA positivity during/after treatment and reduced survival (PFS, OS, or relapse-free survival [RFS]) or treatment response. All 8 studies that assessed overall ctDNA positivity during/after treatment in patients with advanced/metastatic disease found worse outcomes for patients with detectable ctDNA.
Most studies (12/14, 86%) that focused on specific genes (EGFR, BRAF, KRAS, methylated SHOX2) in ctDNA assessed during/after treatment observed an association between the detection of variants in these genes and worse survival outcomes. One study of 20 patients with metastatic disease treated with second-or third-line EGFR-tyrosine kinase inhibitor (TKI)-targeted therapy found 50% of non-responders had EGFR detected in ctDNA after treatment compared with none with EGFR detected among the ongoing treatment responders (89). An additional study of 120 patients with advanced disease treated with a first-or second-generation TKI found that detection of the EGFR T790M resistance mutation during treatment was associated with progression at extra-thoracic metastatic sites (p ¼ 0.008) and bone disease (p ¼ 0.003) (91). However, 2 studies found no association between the detection of the T790M resistance mutation in EGFR and patient outcomes (93,97).

ctDNA assessed at nonuniform or unclear time points
Ten additional studies included patients with cross-sectional ctDNA collected at nonuniform or unclear time points across the treatment spectrum (Supplementary Table 7) (106)(107)(108)(109)(110)(111)(112)(113)(114)(115). As with the cross-sectional studies noted above, a consistent association between ctDNA detection or higher levels of ctDNA was associated with worse survival outcomes. A study of 128 patients with stage I-IV NSCLC with ALK assessed before treatment with a TKI, or at the radiographic follow-up evaluation reported a shorter OS for those with !1 ALK mutations than for those with no ctDNA detected; OS and PFS were worse for those with !2 ALK mutations than for those with a single ALK mutation (111).
Association between ctDNA positivity at any time point and clinical outcomes in patients with advanced disease (meta-analysis) To evaluate the association between ctDNA positivity at certain time points and clinical outcomes, we included studies that reported an HR and 95% CI for the association between ctDNA positivity and PFS and OS in a meta-analysis. Among the 8 estimates from 6 studies (n ¼ 622) included, patients with lung cancer who were positive for ctDNA had a significantly shorter PFS than those with no ctDNA detected (HR, 2.34; 95% CI, 1.89-2.89; Figure 2(A)). A stronger association was observed for the 5 observational studies (HR, 2.99; 95% CI, 2.27-3.94) than for the 3 clinical trials (HR, 1.67; 95% CI, 1.21-2.31; Supplementary Table 8). Similar results were observed for 10 studies (n ¼ 743) specifically reporting the detection of the EGFR alterations in ctDNA (HR, 2.19; 95% CI, 1.78-2.68; Figure  2(B)) and among 4 studies (n ¼ 337) evaluating the EGFR T790M resistance mutation in ctDNA (HR, 2.55; 95% CI, 1.67-3.90; Supplementary  Figure 1). However, a stronger association between EGFR and PFS was observed for the studies assessing the detection of EGFR during treatment (HR, 4.29; 95% CI, 2.77-6.67) than for the studies that assessed EGFR before treatment (HR, 1.82; 95% CI, 1.37-3.29; Supplementary Table 8). No significant evidence of publication bias was found in these analyses (p > 0.05; Supplementary Figure 2). Subgroup analyses were performed by study design, ctDNA test timing, lung cancer subtype, and disease stage and are summarized in Supplementary Table 8; results were similar to overall positivity results for both ctDNA overall and EGFR mutations, with positivity associated with shorter PFS and OS. Among 9 estimates from 8 studies (n ¼ 723) with data available for ctDNA positivity and OS, patients with positive ctDNA had a significantly shorter OS than those with no detected ctDNA (HR, 2.33; 95% CI, 1.91-2.85; Figure 2(C)). A stronger association was observed for the 6 observational studies (HR, 2.72; 95% CI, 2.12-3.50) than for the 3 clinical trials (HR, 1.81; 95% CI, 1.31-2.50; Supplementary Table 8). No differences were observed for ctDNA timing (before vs. during treatment) or for analyses limited to patients with advanced/metastatic disease or those with the NSCLC subtype. Similar results were observed for 9 studies (n ¼ 841) specifically reporting circulating mutations in EGFR (HR, 1.96; 95% CI, 1.58-2.43; Figure 2(D)). Stratified analyses showed few significant differences by study design (Supplementary Table 3). A stronger association between EGFR positivity and shorter OS was observed for EGFR-mutation detection during treatment (HR, 5.34; 95% CI, 3.05-9.34) than for detection before treatment (HR, 1.64; 95% CI, 1.30-2.07; Supplementary Table 8). However, the small number of studies within each stratum precludes drawing strong conclusions. The funnel plot for the EGFR positivity analysis shows asymmetry, which may indicate the presence of publication bias (p ¼ 0.0311); little evidence of publication bias was observed in the other OS analyses.
Longitudinal ctDNA for monitoring disease progression or treatment response Longitudinal ctDNA assessment was used in evaluating ctDNA dynamics during treatment, evaluating prognosis, determining lead time compared with radiographic progression, and assessing changes to the genomic profile of tumors.

ctDNA dynamics
In 8 studies, ctDNA detection was assessed for changes at ! 2-time points during the treatment period without evaluating the associations with clinical outcomes (Supplemental Table 9) (51,58,65,(116)(117)(118)(119)(120). These studies described ctDNA changes in a variety of ways, which makes it challenging to compare results. In general, these studies show a decrease in ctDNA positivity during or after treatment compared with before treatment, or a higher proportion of patients with detectable ctDNA in those with progressive disease (PD) than in those with stable disease (SD). For example, among 20 patients with stage IV NSCLC who received either a TKI or chemotherapy, 65% had detectable ctDNA before treatment, which subsequently decreased for patients with SD (35%) and increased for those with PD (80%) (117). Another study of 41 patients with stage I-IV NSCLC reported 8.9% of patients with ctDNA positivity before surgery, which decreased to 0.28% after surgery (116).

ctDNA for monitoring tumor progression
The association between ctDNA collected at multiple time points in relation to lung cancer outcomes was evaluated in 7 studies of patients with nonmetastatic lung cancer or for all disease stages combined, and in 30 studies of patients with advanced/metastatic disease. In each of the 7 studies evaluating the association between ctDNA monitoring and clinical outcomes in patients with the nonmetastatic disease (or all stages combined), more favorable survival (longer PFS, OS, or RFS) was observed for patients with ctDNA clearance, ctDNA decreases, or no detectable ctDNA observed across the treatment and/or follow-up periods (Supplementary Table 10) (26, 43,45,105,[121][122][123]. Each study evaluated multigene ctDNA using a next-generation sequencing (NGS) platform; no studies focused on specific genes such as EGFR.

Association between ctDNA clearance/decrease and clinical outcomes (meta-analysis)
For the meta-analysis of longitudinal ctDNA, we included studies that reported an HR and 95% CI for the association between ctDNA clearance/decrease during treatment and PFS or OS. The change in ctDNA was defined in a variety of ways, including any decrease (67,95) or >50% decrease (71) in ctDNA concentration from before to after treatment; above vs. below median change in detected ctDNA quantity from baseline and at follow-up during treatment (68); increased ctDNA or no change vs. complete clearance in ctDNA (45,46,132); and ctDNA disappearance at 4 weeks (126), 8 weeks (56), or the start of the third therapy cycle (129) during systemic treatment. ctDNA clearance/decrease was associated with a lower risk of progression in 8 studies (Figure 3(A); HR, 0.24; 95% CI, 0.19-0.31) and a lower risk of death in 8 studies (Figure 3(B); HR, 0.40; 95% CI, 0.27-0.60). However, there was evidence of publication bias (p ¼ 0.0005) based on the funnel plot analysis for the OS analysis.
Subgroup analyses of ctDNA clearance/change and PFS did not show differences when limited to observational studies or advanced/metastatic disease (Supplementary Table 12). For the OS analysis, the association between improved survival outcomes and ctDNA clearance/change was stronger in observational studies (HR, 0.21; 95% CI, 0.13-0.36) than in clinical trials (HR, 0.62; 95% CI, 0.49-0.78; Supplementary Table 12). There was no difference in the OS analysis when limiting the analysis to studies of patients with advanced/metastatic lung cancer.

Lead time of ctDNA-detected progression compared with radiographic progression
The lead time or concordance for detection of ctDNA progression compared with radiographic progression after treatment was reported in 12 studies; 5 studies focused specifically on circulating EGFR and/or EGFR T790M, with the remaining 7 studies evaluating the detection ctDNA (Supplementary  43,45,54,65,71,86,99,124,127,132). Most studies (8/12; 67%) were conducted in patients with advanced/metastatic disease. The average lead time varied widely between studies (17 days to 12.6 months) in both patients with early-stage and advanced/metastatic disease. The variation may reflect differences in frequency or timing (eg, weekly, monthly, pre-/post-treatment only), ctDNA assessments (overall positivity or specific alterations), radiographic imaging (eg, every 2 months, every 5-10 weeks), and the study populations (eg, treatment type, early stage vs. metastatic), making it difficult to compare results across studies. However, each of these studies demonstrated that ctDNA may be a useful tool for detecting disease progression earlier than standard radiographic methods.
Assessing tumor evolution and resistance by changes in ctDNA genomic profile Clonal evolution, or the acquisition of new mutations in response to selective pressures, is increasingly recognized to explain tumor heterogeneity and therapeutic resistance (134). Longitudinal profiling of ctDNA in serial monitoring provides an attractive, cost-effective, noninvasive monitoring technique to describe tumor genomic clonal evolution in response to treatment in a real-time manner, and consequently contribute to understanding the mechanisms of acquired drug resistance. Several studies have evaluated the prognostic effect of newly acquired mutations during TKI treatment, with EGFR T790M most frequently evaluated. Additional studies evaluated small samples of patients during treatment for several different ctDNA-based genetic changes. These ctDNA-based genetic changes included the acquisition of new mutations during TKI treatment (122,135) or chemotherapy (36), increases in gene copy numbers or losses of tumor suppressor genes during ICI therapy (136), and the frequency of C > G and C > A substitutions during chemotherapy (77). In this review, 1 clinical study examined genetic evolution with systemic approaches (eg, evolutionary trees). In the TRACERx study, Abbosh and colleagues detected subclonal single-nucleotide variants in early-stage NSCLC and mapped them back to multiregion exome sequencing (M-Seq)-derived tumor phylogenetic trees (137). A limited number of additional studies described changes in the detection of specific genes in ctDNA, which may indicate genomic evolution in response to treatment. For example, a study of 43 patients with advanced NSCLC evaluated the copy number gains in specific genes (eg, MET, EGFR) from before treatment with third-generation EGFR-TKI, rociletinib, to the time of PD; patients with shorter PFS were more likely to have copy number gains (47).

Discussion
In this systematic review, we summarized the data on ctDNA clinical utility in lung cancer, focusing on 3 key areas: diagnosis, prognostication, and monitoring. Among these, prognostication was by far the most studied utility for ctDNA in lung cancer. In meta-analyses, both overall ctDNA positivity and the presence of EGFR mutations within ctDNA were associated with worse PFS and OS outcomes. The second most common ctDNA utility in lung cancer was disease monitoring. The association between ctDNA dynamics and clinical outcomes was evaluated by a meta-analysis, in which ctDNA clearance/decrease was associated with a lower risk of progression and death. In terms of diagnosis or other applications within ctDNA monitoring, such as detecting residual disease or disease relapse, assessment of treatment efficacy, or identification of treatment resistance mechanisms, the data in the literature are limited and/or mixed. As such, although these results are promising, the existing data are insufficient to formulate definitive conclusions or recommendations for use, and additional research is needed. The least studied utility was diagnosis (3 of 111 studies), and many key issues need be addressed before ctDNA can be applied for cancer screening and early detection.
Despite the consistent association between ctDNA and prognosis in patients with lung cancer, several important limitations need to be addressed in future studies. A critical limitation is the small sample size in most studies, with almost 70% of studies having ctDNA data for fewer than 100 patients. Consistent with this, many studies were not statistically powered to detect associations between ctDNA results and outcomes in subgroups where patients were stratified by factors with the potential to contribute to heterogeneity (eg, lung cancer subtype or stage). Further, the small sample sizes also limited the independent prognostic value of the results because it was not possible to estimate and remove potential confounding variables (eg, stage). Sample sizes were especially small in studies measuring ctDNA dynamics longitudinally for monitoring disease progression and/or treatment response. For studies with few patients (ie, <20) with ctDNA encompassing hundreds of cancerrelated or other genes measured at 2 or more time points, it is challenging to determine the true association between ctDNA changes and prognosis. Moreover, most of the identified studies simply estimated the association of ctDNA with clinical outcomes at multiple time points individually or ctDNA change over 2 time points. Consequently, they cannot address important questions, such as quantitatively characterizing ctDNA dynamics throughout the disease course or identifying the optimal timing and frequency of blood sampling for the intended purpose of the study and whether any adjustments need to be made depending on cancer subtype.
Most of the studies evaluated were conducted in patients with advanced/metastatic NSCLC. There were limited prognostication data in patients with early-stage lung cancer or other histological subtypes. Because only 2 of the included studies focused exclusively on patients with SCLC, the utility of ctDNA for predicting prognosis in patients with this aggressive lung cancer is less clear (36,84). Besides prognostication, 3 studies used ctDNA for the detection of molecular residual disease (26, 43,137). These studies suggest that ctDNA analysis, particularly using highly sensitive tumor-informed assays, can identify recurrence earlier than radiographic imaging. If confirmed, ctDNA for molecular detection has the potential to inform and improve patient identification and stratification for treatment selection. A more comprehensive review covering topics beyond our study, such as technical considerations and major barriers for use in clinical practice, on ctDNA molecular residual disease detection in patients with earlystage NSCLC, was published recently (138).
Several limitations should be considered when interpreting the findings of this systematic literature review. First was the heterogeneity across studies. Within each of the 3 clinical applications or articles that addressed similar research questions, studies vary significantly by study design, lung cancer histological type, stage, tumor burden, location, treatment, the timing of ctDNA testing, features of ctDNA assays, and in the preanalytic or analytic parameters of the ctDNA analysis. For example, in studies that assessed a single gene mutation, some studies estimated the mutation status using digital PCR assays (24), whereas others used NGS (139). In addition, the timing of ctDNA sampling varied, with some studies testing for EGFR mutations before (45,51) or during treatment (66,90,126). The differences across studies reflect the complexity of ctDNA research but also highlight the importance of standardizing guidelines when ctDNA becomes part of standard clinical practice. In meta-analyses, when the overall prognostication associations with PFS and OS were estimated, we did not detect heterogeneity by stage, histological type, or ctDNA measurement timing. However, this result may be due to sparse data in certain subgroups and the underpowered analyses. Moreover, there might be other factors that were not evaluated but that also contributed to the associated heterogeneity. Another limitation of the embedded meta-analyses was that not all prognostic studies reported HRs. Of the 85 studies on prognostication, only 26 studies qualified for meta-analyses. Many studies, particularly those with small sample sizes, simply reported the median survival time for patients with detected vs. those with no detected ctDNA alterations. The lack of HR data in these studies may have biased the pooled HRs.
Our study covered a wide range of ctDNA clinical utility and synthesized the data largely based on the availability of published studies. Therefore, some interesting topics were not reviewed in detail. For example, as an actionable alteration in advanced NSCLC, ctDNA EGFR mutations have been extensively studied and reviewed previously (140)(141)(142). However, this review focused on the association of ctDNA EGFR mutations, at both a single time point or longitudinally with clinical outcomes. Beyond that, several studies also demonstrated the value of EGFR or EGFR T790M mutation in monitoring EGFR-TKI treatment and treatment resistance (122,135). However, because data were usually presented descriptively or on a patientby-patient basis, with large interstudy differences, we did not summarize these findings, despite their clinical relevance.
Although ctDNA testing has been increasingly used and studied in recent years, ctDNA is unlikely to replace tumor tissue DNA testing for lung cancer screening and diagnosis or for the detection of disease progression in the shortterm. However, ctDNA will play an important role in lung cancer as oncology advances precision medicine. This systematic review of 111 studies with embedded meta-analyses demonstrated that ctDNA is a sensitive biomarker, reflecting tumor burden and dynamics, that also has prognostic value. These factors make ctDNA a promising tool for predicting clinical outcomes and tracking treatment responses. Large-scale, prospective clinical trials are needed to further validate these findings. Future studies will evaluate the potential for improved patient outcomes or cost savings of ctDNA applications compared with standard clinical approaches to assess treatment response, detect treatment resistance, individualize therapy, and predict outcomes.