The Role of Circulating Tumor DNA in Advanced Non-Small Cell Lung Cancer Patients Treated With Immune Checkpoint Inhibitors: A Systematic Review and Meta-Analysis

Background The use of circulating tumor DNA (ctDNA) to reflect clinical benefits of advanced non-small cell lung cancer (NSCLC) patients during immune checkpoint inhibitor (ICI) therapy remains controversial. This study aimed to determine the association of pre-treatment and early dynamic changes of ctDNA with clinical outcomes in advanced NSCLC patients treated with ICIs. Methods Electronic databases (PubMed, Embase, Web of Science, and Cochrane) were systematically searched to include relevant studies published in English up to November 2020. The primary outcomes were overall survival (OS) and progression-free survival (PFS) and the secondary outcome was objective response rate (ORR) with RECIST criteria. Results A total of 1017 patients from 10 studies were identified. The baseline ctDNA levels (detected versus not detected) showed no significant association with clinical outcomes regarding OS (hazard ratio [HR], 1.18; 95% confidence interval [CI], 0.93-1.51), PFS (HR, 0.98; 95% CI, 0.80-1.21), and ORR (odds ratio [OR], 0.89; 95% CI, 0.54-1.46). Interestingly, when taken early longitudinal assessment of ctDNA into consideration, the early reduction of the concentration of ctDNA was associated with significant improvements of OS (HR, 0.19; 95% CI, 0.10-0.35), PFS (HR, 0.30; 95% CI, 0.22-0.41) and ORR (OR, 0.07; 95% CI, 0.03-0.18). Further subgroup analyses revealed that the reduction magnitude did not significantly impact the association between ctDNA and clinical outcomes, suggesting that both patients with decreased ctDNA or a ≥50% reduction of ctDNA was associated with improved OS, PFS and ORR. Conclusion Early reduction of ctDNA was associated with improved OS, PFS and ORR in advanced NSCLC patients treated with ICIs. Systematic Review Registration https://www.crd.york.ac.uk/PROSPERO, CRD42021226255.


INTRODUCTION
Over the past decade, immunotherapy, targeting immune checkpoint molecules, programmed-death-1(PD-1)/PD ligand-1 (PD-L1) axis, has turned into one of the most important breakthroughs in cancer treatment, including non-small cell lung cancer (NSCLC) (1). Although treatment of NSCLC with immune checkpoint inhibitors (ICIs) can produce remarkably durable responses, a considerable proportion of patients cannot derive meaningful benefits from ICI therapy owing to drug resistance (2), disease hyper-progression (3), or immunerelated adverse events (irAEs) (4).
Currently, several promising biomarkers for ICI therapy have been widely investigated, such as tumor mutation burden (TMB), PD-L1 expression, germline genotype of HLA-I, the molecular profiling of the tumor microenvironment, mutations in DNA mismatch repair and replication genes (1,5,6). However, most of these biomarkers are far from perfect biomarkers owing to being invasive, not always feasible, and its spatial and temporal heterogeneity (6,7).
In response to the demand for genetic predictive and noninvasive molecular biomarkers in NSCLC, liquid biopsy, including circulating tumor DNA (ctDNA), have been developed. ctDNA, referring to a sub-set of cell-free DNA, is released by tumor cells undergoing apoptosis, necrosis, and in extracellular vesicles (exosomes) secreted from tumor cells and can be found in plasma. ctDNA is highly fragmented and ranges between 100 and 200 base pairs in size and represents genetic material from the primary tumor as well as metastases (8). ctDNA can be quantified using multiple metrics, such as mutant allele fraction or mutant allele concentration (that is, copies per milliliter) (9). The level of ctDNA in plasma has been demonstrated to correlate with tumor size (10,11), disease stage (12), the clinical responses and prognosis of patients receiving anti-tumor treatment (13)(14)(15)(16). The short half-life of ctDNA (17,18), as well as the reduced and ease risk of repeating liquid biopsies relative to tissue biopsies (19) or imaging (20), enables ctDNA to be used for real-time monitoring of tumor burden in response to treatment. Although NSCLC is the cancer type for which plasma ctDNA testing has the most comprehensive and compelling evidence (21), it is still controversial regarding the role of ctDNA in predicting survival and clinical response.
ctDNA assays for the evaluation of cancers that harbor EGFRsensitizing or EGFR-resistance mutations have already entered into clinical practice (22). A recent prospective study using ctDNA to guide matched targeted therapy in lung cancers supported the incorporation of plasma ctDNA into clinical practice (23). In addition to the direction of molecular targeted treatment, ctDNA could also potentially help monitor ICI response as the quantitative level in plasma has been demonstrated to reflect the tumor burden in patients even earlier than clinical detection (17,(24)(25)(26)(27) and might identify response earlier than clinical detection (28,29). However, these studies are mostly retrospective design, lacking high-level medical evidence, and some results were even inconsistent. Therefore, we conducted this meta-analysis to comprehensively investigate the predictive value of ctDNA for advanced NSCLC patients who received ICI therapy.

METHOD
The authors declare that all supporting data, study materials and analytic methods within the article and the online supporting information are available to other researchers. This systematic review was performed in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (30). The PRISMA checklist is provided in Supplementary Table S1.

Search Strategy
The systematic search of the scientific literature was performed. The search was conducted up to November 2020 in PubMed, EMBASE, Web of Science, and the Cochrane Library database. The main keywords used for the online search were "Circulating Tumor DNA," "Carcinoma, Non-Small Cell Lung," "Immune Checkpoint Inhibitor." The full online search strategies were demonstrated in Supplementary Text S1. We also manually examined the references of each screened study until no additional articles could be added.

Exclusion and Inclusion Criteria
Studies were selected if they met the following inclusion criteria (1): patients with advanced NSCLC; (2) patients received ICIs alone or ICI-based therapy; (3) ctDNA was analyzed in these groups; (4) information on the clinical response or prognosis of these patients were provided; (5) if multiple studies from the same populations were available, to avoid repetition information, only the one with the largest sample size was included. Reviews, case reports, conference reports, abstracts, phase I studies and non-English publications were excluded. Endnote (Thomson Research Soft; Stanford, Connecticut, the United States) was used to select and screen the literature.

Data Extraction
Data extraction and analyses were performed by 2 independent reviewers (HW and FZ). Any disagreement was discussed with the senior author (XC). Extracted study characteristics included: first author, publication year, country of the study, number of patients who underwent ICIs alone or combined therapy, gender, median/mean age, histological types, stage, smoking history, PD-L1 expression, ECOG PS, follow up duration, drugs, sample, extraction method, detection method, platform, detection time, most mutated genes.

Outcomes
The primary outcomes were overall survival (OS) and progression-free survival (PFS), then the secondary outcome included objective response rate (ORR) with RECIST criteria.

Risk of Bias Assessment
To assesses the quality of non-randomized studies (31), the Newcastle-Ottawa scale was used to assess the risk of bias. The scale assesses risk of bias in the following 3 aspects: selection of the study which include adequate definition and representation of the case, comparability of groups, and ascertainment of exposure and outcome for cases and controls. Studies with score less than 4 were considered as a high risk of bias, those with scores of 4 to 6 were regarded to have an intermediate risk of bias, and scores of 7 or more represented a low risk of bias. The results displayed in Supplementary Table S2

Statistical Analysis
The heterogeneity of different studies was tested by using the Q test. The HRs and ORs with 95% CIs were directly extracted from the research article or calculated using previously published methods, as proposed by Tierney et al. (32). We calculated the I 2 to assess the extent of variability attributable to statistical heterogeneity across studies. I 2 < 50% and P > 0.10 were interpreted as signifying low-level heterogeneity. Across the studies, if no significant heterogeneity was found, the results were combined with the fixed-effects model (Mantel-Haenszel) (33); otherwise, the random-effects model (DerSimonian-Laird) was used (34). Publication bias was assessed by visual inspection of a funnel plot, Begg's and Egger's tests. A sensitivity analysis was performed by serially excluding each study to determine its influence. P values were two-sided and considered significant if less than 0.05. STATA 15.1 software for Mac was used to evaluate the outcomes.

Subgroup Analysis
The following subgroup analyses were performed. Firstly, as already assessed, the patients were stratified with two groups depending on different cutoff value of longitudinal ctDNA dynamic. Then, we also performed subgroup analysis considering that different platforms in studies might provide heterogeneity. The details of groups and outcomes were provided in the results of subgroup analysis.

Study Selection
Of the 427 articles retrieved by the literature search, 97 duplicates were removed, left 330 studies available for screening. 50 studies underwent full-text review, after screening the title and abstract. Of these studies,10 pieces of literature met the inclusion criteria and were chose for the current meta-analysis (7,(35)(36)(37)(38)(39)(40)(41)(42)(43). The detailed flow chart is shown in Figure 1.

Study Characteristics
The basic characteristics of the studies are summarized in Table 1 (1) and (2) to distinguish them.
The Association of Baseline (Pre-ICI Therapy) ctDNA and Clinical Outcomes 4 cohorts from 3 studies (36,37,43) were included for analyzing the association between baseline ctDNA and OS (Figure 2A). Considering that the results were not heterogeneous, we chose fixed effects models. There was no statistical significance between the two groups regarding OS (HR, 1.18; 95% CI, 0.93-1.51; I 2 <0.1%; p=0.181). Data from 3 studies (7, 36, 43) and 4 cohorts were pooled for the analysis of PFS ( Figure 2B), with a total of 625 patients. Considering that the results were homogeneous, we also used fixed-effects models. There was still no evident difference between the groups in terms of PFS (HR,0.98; 95% CI, 0.80-1.21; I 2 = 38.1%; p=0.865) The results of 4 studies (35,36,39,43) and 5 cohorts were pooled to examine the relationship of baseline detected ctDNA and ORR ( Figure 2C), with a sample size of 641 patients. The pooled OR in the detected group with ORR (OR,0.89; 95% CI,0.54-1.46; I 2 = 7.6%; p=0.641) showed no significant difference between the two groups.

Subgroup Analysis in Longitudinal Observation of Patients With ICIs Therapy and Platform
Taking into account these studies, the authors selected different degrees of the reduction of ctDNA during treatment as the cutoff to define the positive group. Therefore, we further adopted subgroup analysis to divide the studies that dropped into the positive group and those with different degrees of decline into two subgroups: a decline of >50% as the threshold to define ctDNA response (subgroup 1) vs others; decreased vs others (subgroup 2). 73 patients from two studies (37,39) were included in subgroup 1 with a HR ( Figure 5A     analyze the difference of illumina platform vs other platforms (Supplement Figure S1). The results still maintained coherence in different platforms both in baseline detection and dynamic observation, showing consistent with previous research (44).

Publication Bias
As shown in Figure 6, the funnel plots were almost symmetrical and the test results indicated that no publication bias existed regarding the HRs for OS (

DISCUSSION
To our knowledge, this was the first meta-analysis to comprehensively investigate predictive significance of ctDNA in advanced NSCLC patients treated with ICI therapy. The current study found that early reduction of ctDNA was associated with improved PFS, OS and ORR while there was no significant association between baseline ctDNA and clinical outcomes.
The concept of monitoring tumor burden and therefore biological effects of treatment by analyzing circulating biomarkers has been known for long, but only recently the availability of techniques able to detect ctDNA has opened new perspectives (17,45). In lung cancer, the first experience concern EGFR-mutated disease (46). The use of ctDNA as a quantitative biomarker for assessment of ICI response has been investigated in prior studies (47)(48)(49)(50), most of which have focused on tracking driver mutations in patients using digital PCR or allele-specific PCR. Even previous studies concluded that pretreatment ctDNA level appears to be an independent, inversely prognostic variable across tumor types, characterized by an association with OS and other known prognostic variables, but not with ORR (51). In this meta-analysis, we did not find a significant association between baseline ctDNA and clinical outcomes. Compared with tumor tissue, the abundance of tumor DNA in plasma samples is relatively low, which poses a great challenge to the sensitivity of plasma detection. False-negative results may occur due to insufficient plasma ctDNA content or lacking of inclusion of altered genes in the targeted NGS panel (21). On the other hand, false-positive results may also occur because of germline variants or the presence of somatic mutations in hematopoietic stem cells owing to clonal hematopoiesis, although most plasma ctDNA assays use matched sequencing of white blood cells (52)(53)(54), there still be sequencing errors and artifacts (55). A recent study found that high allele frequency blood TMB was strongly correlated with the ctDNA amount (56). However, high allele frequency blood TMB was a negative prognostic factor rather than a predictive factor (56), this may partially explain that there  was no significant association between baseline ctDNA and clinical outcomes. Interestingly, in this meta-analysis, patients with the reduced dynamic ctDNA obviously obtained more clinical benefits from ICI therapy. The early reduction of ctDNA may reflect an early response of tumors to effective treatment. This was consistent with previous studies, the amount of ctDNA may be an independent factor to predict the efficacy of patients receiving ICIs, and combined with other predictive indicators can better distinguish potential benefit populations (42). Subgroup analysis also indicated that a decline of >50% as a threshold to define ctDNA response and even achievement of undetectable ctDNA may prove to be a stronger predictor of long-term response, compared with just defining ctDNA response as decrease during ICIs therapy, and may identify patients who comprise the "tail" of the survival curve (57). Nevertheless, it still needs prospective studies to confirm these findings. Interestingly, some studies report a transient spike preceding a decline in ctDNA levels in a subset of patients, likely reflecting DNA release as tumor cells are killed. It would be important to avoid misinterpreting such a spike as disease progression (37,58,59).
There still were some limitations in our analysis. Firstly, all data were extracted from retrospective or post-hoc analysis studies. Secondly, the quality of data was heterogeneous as several pieces of important information such as prior therapy were not consistently reported. Finally, a high proportion of patients in these studies had KRAS and TP53 mutations. As TP53 and KRAS mutations showed remarkable clinical benefits from ICIs (60), it remains undermined that whether detecting TP53 and KRAS in ctDNA for predicting the response to ICIs was as effective as total ctDNA.
In summary, this current study suggested that early reduction of ctDNA was associated with improved PFS, OS and ORR in advanced NSCLC patients treated with ICIs. Further large-scale and rigorously designed prospective studies are still warranted to  verify its clinical value. In the future, clinical trials will be conducted in NSCLC patients to determine whether ctDNA could be selected as a significant factor to monitor clinical response to ICI therapy.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

AUTHOR CONTRIBUTIONS
HW and FZ collected the relevant papers and data and drafted the manuscript text. HW, MQ and XL performed statistical analyses. HW, CZ, and LC performed the quality assessment. XC and CCZ gave critical comments and revised the paper. All authors contributed to the article and approved the submitted version.