A meta-analysis of liquid biopsy versus tumor histology for detecting EGFR mutations in non-small cell lung cancer

Highlights

• •The most common driving gene for non-small cell lung cancer (NSCLC) is epidermal growth factor (EGFR) mutation. EGFR mutation status is often determined using invasive histopathologic specimens or, less commonly, with more convenient liquid biopsies. However, the consistency of the two detection methods has yet to be definitively established.
• •Our study is a meta-analysis to assess the consistency of liquid biopsy and histopathologic specimens for determining EGFR mutations and other driver gene mutations in NSCLC patients.
• •We found that liquid biopsy had a high sensitivity for detecting EGFR mutations. In particular, our analysis showed that liquid biopsy is suitable for detecting tyrosine kinase inhibitor-sensitive and drug resistance mutations (T790M mutations) in advanced NSCLC.
• •Our findings provide a rationale for using liquid biopsy to monitor disease during tyrosine kinase inhibitor therapy and assess minimal residual disease.

Abstract

Objective

To assess the consistency of liquid biopsy and histologic analysis for detecting epidermal growth factor receptor (EGFR) gene mutations in patients with advanced non-small cell lung cancer (NSCLC).

Methods

The PubMed, Cochrane Library, and CNKI et al. databases were searched to collect studies comparing liquid biopsy and histopathologic specimens. The EGFR mutation status was extracted from the studies, and meta-analysis was carried out using Stata 12.0 software.

Results

We included 22 studies of 3359 NSCLC patients. In the meta-analysis, eight papers with a sample size of size <150 had an OR of 45, indicating that liquid biopsy had high sensitivity for detecting EGFR mutations. In addition, seven papers with a sample size ≥150, with an OR of 70, reported that liquid biopsy was highly susceptible to detecting EGFR mutations. The pooled diagnostic effect size of 6 for literature that included the T790M mutation was smaller than that of 69 for literature that did not include the T790M mutation, and I² >50 %, showing that literature that did not include the T790M mutation was more heterogeneous. The combined diagnostic effect size of 34 in the exon 19 group was smaller than that in the group with no exon 19, with an I²>50 %. There was substantial heterogeneity in both the exon 19 group and the non-exon 19 group. The group with the L858R mutation had a greater diagnostic effect size of 28, lower I², and less heterogeneity than the group without the L858R mutation. The exon 21 group had a larger pooled diagnostic effect size of 66, a smaller I², and less heterogeneity than the group without exon 21.

Conclusion

Liquid biopsy and histologic analysis have high concordance for detecting EGFR mutations in NSCLC. Liquid biopsy can provide an alternative technology for individualized treatment and monitoring of minimal residual disease (MRD) in advanced NSCLC patients with EGFR tyrosine kinase inhibitor-sensitive and drug resistance (T790M) mutations.

Introduction

Non-small cell lung cancer (NSCLC) is the major pathological type of lung cancer, accounting for approximately 90 % of cases [1]. The most common driving gene for NSCLC is epidermal growth factor (EGFR) mutation [2]. A report that analyzed 151 global studies of 33,162 patients with NSCLC found that the EGFR mutation rate at the population level was approximately 29 % [3]. Regionally, the EGFR mutation rate is highest in the Asia-Pacific region, reaching 47 %, and lowest in Oceania at only 12 %. Deletion of exon 19 (accounting for 45 % of NSCLC cases) and mutation of exon 21 L858R (44 %) are the most frequent mutations [4]. The use of tyrosine kinase inhibitors (TKIs) is consistently recommended by multiple guidelines for patients with EGFR exon 19 deletions and the exon 21 L858R mutation. The T790M mutation is associated with acquired drug resistance following treatment with both first- and second-generation TKIs but may be overcome by third-generation TKIs.

Gene mutation screening improves the accuracy of targeted therapy. Detection of circulating tumor DNA (ctDNA) during therapy is also useful for monitoring minimal residual disease (MRD) and for the early detection of drug resistance or disease progression [5]. Currently, gene mutation screening for NSCLC patients is often performed using histopathologic specimens and liquid biopsies. The high demands of acquiring histopathology specimens make some cases complex and difficult to perform, and the patient's physical status, general condition, and body position limit the use of biopsies [6]. Compared with histopathologic specimens, liquid biopsy samples are more convenient to obtain; however, the consistency of the two detection results needs further investigation. Accordingly, this study is a meta-analysis to assess the consistency of tests on liquid biopsy and histopathologic specimens to detect EGFR mutations and other driver gene mutations in NSCLC patients to provide a reference for accurate treatment and MRD surveillance of NSCLC.

Materials and methods

Literature inclusion criteria

Studies that met all the following criteria were included: (1) a comparative study of EGFR mutations and other non-EGFR mutations in liquid biopsy and histopathologic specimens from NSCLC patients; (2) detection of gene mutation by Sanger sequencing, amplification refractory mutation system, next-generation sequencing, and other methods; (3) provides data that shows that results obtained by the two detection methods are consistent, or provides the original data so that the consistency can be calculated; (4) the subjects had primary NSCLC regardless of nationality, race, and time of publication; and (5) the language of the study was either Chinese or English.

Literature exclusion criteria

Studies that met one or more of the following criteria were excluded: (1) duplicate published studies or data; (2) the research methods and outcomes were not comprehensive, and the consistency of the two detection methods could not be determined; (3) includes patients with small cell lung cancer or other types of lung cancer; (4) the specific mutation type could not be determined; and (5) case reports, book chapters, conference abstracts, and so on.

Literature retrieval and literature screening

A computer-assisted search was performed to collect published studies of gene mutation in patients with NSCLC by liquid biopsy and histologic gene detection from database inception to May 2023 using the following databases: the Chinese CNKI, Wanfang, Web of Science, PubMed, American Studies, ESPEN, UK NICE guidelines, and Cochrane Library. The Chinese search terms included non-small cell lung cancer, lung cancer, liquid biopsy, circulating tumor cells, pathologic tissue, and mutations. The English search terms include lung neoplasms/non-small cell lung cancer/NSCLC, liquid biopsy/circulating tumor DNA/cell-free DNA/circulating tumor DNA/cell-free DNA, and tissue NEAR/2 biopsy/pathology specimen/tissue/biopsy/pathology specimen. At the same time, the website https://www.connectedpapers.com/ was used to find the references of the included literature.

The relevant papers were retrieved based on the search terms, the titles and abstracts were read for preliminary selection, and papers that were not related to detecting gene mutations in NSCLC were excluded. The full text of the remaining literature was read and screened according to the prespecified inclusion and exclusion criteria described above. Two researchers (JC and WW), both of whom are trained in meta-analysis, completed the literature search and screening process independently. Then, cross-checks were conducted by DC to resolve differences between the papers identified by each researcher, and the remaining papers were included in the meta-analysis. The specific search results are shown in Fig. 1.

Fig. 1.

The process and results of literature screening. After initial screening, 1433 relevant research papers were obtained. After reading the title and abstract, the remaining 1005 articles were screened, and literature that clearly did not meet the inclusion criteria was excluded. This exclusion left 97 articles that were retained for full-text review. After a full-text review, 22 articles were selected for meta-analysis.

Data extraction and quality appraisal

Data extraction

The following data was extracted from the studies: background information such as author, publication date, and geographical region where the study was conducted; the number of cases, pathological type and so on; the methods used to detect mutations including liquid biopsy and tissue biopsy techniques, the target gene involved, and so on; and the detection rate of EGFR mutations by each method and the consistency between the two methods. The weaknesses or limitations of the research were also determined.

Quality appraisal

Based on the QUADAS statement, each study was rated on nine aspects, and "yes", "no", or "unclear" were used to judge each aspect. Within the Quadras framework, a study is classed as high quality if "yes" ≥7 and "unclear" if “yes” <2; as medium quality if "yes" = 5–6 and "unclear" <3; and low quality if "yes" ≤ 4 and "unclear" ≥3. High-quality and medium-quality studies were included in the meta-analysis, and low quality studies were excluded (Table 1).

Table 1.

Indicators and methods used to assess the literature according to the QUADAS statement.

Aspect	Project	Contents	Answers
1	Included disease spectrum	Represent the examination population in clinical practice	Yes/No/Unclear
2	Object of study	The selection criteria are clear	Yes/No/Unclear
3	Interval time	The reference test and the test to be evaluated are sufficiently short	Yes/No/Unclear
4	Reference test	All samples were accepted	Yes/No/Unclear
5	Refer to the standard test	Independent of the test to be evaluated	Yes/No/Unclear
6	Interpretation of the results of the test to be evaluated	Without knowing the results of the reference test	Yes/No/Unclear
7	Interpretation of results of reference standard tests	Without knowing the test results to be evaluated	Yes/No/Unclear
8	Intermediate test results	Difficult to explain and report	Yes/No/Unclear
9	Case information	All cases were reported	Yes/No/Unclear

Statistical methods

Statistical analysis was performed using Stata 12.0 software. To compare the difference between liquid biopsy and tissue biopsy for detecting EGFR gene mutations, the standardized mean difference and its 95 % confidence interval (CI) were used, and P < 0.05 was considered to indicate statistical significance. To test for between-study heterogeneity, I²=0 % indicates no statistical heterogeneity, I²=50 % indicates moderate heterogeneity, and I²>50 % indicates the greatest amount of heterogeneity. The literature included in this review was heterogeneous and a random-effect model was used. Funnel plots were used to test for publication bias across studies. A sensitivity analysis was conducted to determine if the results were robust. A subgroup analysis based on the sample size of the included studies was conducted.

Results

Basic information of included literature

Twenty-two papers were included, and the basic information extracted from the literature is shown in Table 2.

Table 2.

Basic information obtained from the literature.

Research (year)	Region	NSCLC type	Method for testing liquid specimens	Method for detecting tissue samples	Genes involved
Cho (2022) [7]	Asia	Advanced non-squamous NSCLC	NGS technology	Not specified	EGFR, ERBB2, and KRAS; BRAF V600E; MET; and ALK, ROS1, RET, or NTRK1
Jiang (2011) [8]	China	Advanced non-squamous NSCLC	ME-PCR-based assay	ME-PCR assay and a non-enriched PCR-based assay	EGFR mutations in exons 19 and 21
Sriram (2011) [9]	Australia	Surgically resected stage I-IV NSCLC	ME-PCR and HRM	ME-PCR and HRM	EGFR mutations in exons 19 and 21
Xu (2012) [10]	China	Advanced NSCLC	DHPLC, ME-liquidchip, and Scorpion-ARMS	Scorpion-ARMS	EGFR exon 19 deletion and exon 21 L858R
Kim (2013) [11]	Korea	Stage IIIA, IIIB and IV	Peptide nucleic acid (PNA)-mediated polymerase chain reaction (PCR)	Not mentioned	EGFR exon 19 deletion and exon 21 point mutation
Mok (2015) [12]	Asia	Stage IIIB and IV	Two allele-specific PCR assays	Two allele-specific PCR assays	EGFR mutations
Li (2014) [13]	China	Stage IIIb and IV	ARMS	ARMS	EGFR 19 deletion, L858R and T790M
Bai (2009) [14]	China	Stage IIIB and IV	DHPLC	DHPLC	EGFR mutations in exons 19 and 21
He (2009) [15]	China	Advanced stage	ME-PCR	Direct sequencing	EGFR exon 19 deletions and exon 21 L858R
Huang (2012) [16].	China	Stage I to IV	DHPLC	DHPLC	EGFR mutations in exons 19 and 21
Yam (2012) [17]	Hong Kong	Stage IB to IV	ME-PCR	Not mentioned	EGFR mutations in exons 18–21
Zhang (2013) [18]	China	Stage IIIB and IV	MEL	SurPlex-xTAG70plex	EGFR, KRAS, BRAF, and PI3KCA
Zhao (2013) [19]	China	Stage I-IV	ME-PCR	ME-PCR	EGFR mutations in exons 19 and 21
Weber (2014) [20]	Denmark	Stage IIA-IV	Cobas EGFR blood test	Cobas EGFR tissue test	EGFR mutations in exons 18–21
Duan (2015) [21]	China	Stage IIA-IV	Scorpion-ARMS	Scorpion-ARMS	EGFR mutations
Karlovich (2016) [22]	Global	Stage IIIB and IV	Cobas EGFR plasma test	Cobas EGFR tissue test	EGFR T790M
Ma (2016) [23]	China	Stage IIIA-IV	ARMS	ARMS	EGFR mutations
Wang (2017) [24]	China	Stage IIIB and IV	DHPLC	DHPLC	EGFR mutations in exons 19 and 21
Kobayashi (2018) [25]	Japan	Advanced NSCLC	COBAS ver2	COBAS ver2 and PNA-LNA PCR clamp	T790M
Liu (2018) [26]	China	Advanced NSCLC	ARMS	ARMS	EGFR mutations in exons 18–21
Veldore (2018) [27]	India	Stage IV	NGS	Allele-specific real-time PCR	EGFR mutations in exons 19 and 21
Denis (2019) [28]	France	Stage IIIA-IV	ARMS	ARMS	EGFR mutations

NGS, next-generation sequencing; ME-PCR, mutant-enriched PCR; HRM, high-resolution melt; DHPLC, denaturing high-performance liquid chromatography; ME-liquidchip, mutant-enriched liquidchip; Scorpion-ARMS, scorpion amplification refractory mutation system; ARMS, amplification refractory mutation system; MEL, mutant-enriched liquidchip technology.

Quality assessment of the included literature

Overall, the included studies were of moderate quality, and most did not report intermediate or uninterpretable results, as noted in QUADRAS. (Table 3).

Table 3.

Quality assessment of the included literature.

Research (Year)	Included disease spectrum	Object of study	Interval time	Reference test	Refer to the standard test	Interpretation of the results of the test to be evaluated	Interpretation of the results of reference standard tests	Intermediate test results	Case information
Cho (2022)	Yes	Yes	Not given	Yes	Yes	Not known	Not known	No	Yes
Jiang (2011)	Yes	Yes	Given	Yes	Yes	Not known	No	No	Yes
Sriram (2011)	Yes	Yes	Given	Yes	Yes	No	No	No	Yes
Xu (2012)	Yes	Yes	Given	Yes	Yes	No	No	No	Yes
Kim (2013)	Yes	Yes	Not given	Yes	Yes	No	No	No	Yes
Mok (2015)	Yes	Yes	Given	No	Yes	Yes	Yes	No	Yes
Li (2014)	Yes	Yes	Given	No	Yes	No	Yes	No	Yes
Bai (2009)	Yes	Yes	Given	Yes	Yes	Yes	Yes	No	Yes
He (2009)	Yes	Yes	Not given	Yes	Yes	Yes	Yes	No	Yes
Huang (2012)	Yes	Yes	Given	Yes	Yes	No	Yes	No	Yes
Yam (2012)	Yes	Yes	Given	Yes	Yes	No	Yes	No	Yes
Akca (2013)	Yes	Yes	Not given	No	Yes	No	Yes	No	Yes
Zhang (2013)	Yes	Yes	Given	Yes	Yes	No	Yes	No	Yes
Zhao (2013)	Yes	Yes	Given	Yes	Yes	No	Yes	No	Yes
Weber (2014)	Yes	Yes	Given	Don't know	Yes	No	No	No	Yes
Duan (2015)	Yes	Yes	Given	Don't know	Yes	No	No	No	Yes
Karlovich (2016)	Yes	Yes	Given	Yes	Yes	No	No	No	Yes
Ma M (2016)	Yes	Yes	Given	Yes	Yes	No	No	No	Yes
Wang (2017)	Yes	Yes	Given	Yes	Yes	No	Yes	No	Yes
Kobayashi (2018)	Yes	Yes	Given	Yes	Yes	No	Yes	No	Yes
Liu (2018)	Yes	Yes	Given	Yes	Yes	No	Yes	No	Yes
Veldore (2018)	Yes	Yes	Given	Yes	Yes	No	Yes	No	Yes
Denis (2019)	Yes	Yes	Given	Don't know	Yes	No	Yes	No	Yes

Comparison of methodological results

Comparison of mutation results from liquid biopsies and tumor tissue specimens

Data from 22 NSCLC patient groups were collected, with the specific data shown in Table 4. Results from tests of tissue specimens were used as the reference standard, and the total effect was combined by Stata software to determine the Pearson correlation coefficient of sensitivity and specificity as Rho=−0.157, P = 0.486, respectively, thus confirming that there was no threshold effect between the studies. It is possible to combine sensitivity and specificity. The pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and exact effect sizes of the likelihood ratio [95 % CI] of the random effects model for liquid biopsy were 0.90[0.75, 0.97], 0.85[0.73, 0.92], 5.9[3.1, 10.9], 0.11[0.04, 0.32], and 52[14, 193], respectively. Using the Q test, we found that the homogeneity of sensitivity and specificity across the data from each publication was low, and the heterogeneity was high (I²>50 %, P < 0.001), as shown in Fig. 2. A forest plot of the sensitivity and specificity of liquid biopsy for NSCLC in the 22 studies is shown in Fig. 2. This figure shows that the effect sizes for the individual studies are very different, and the values of I² are 97.04 % (sensitivity) and 95.46 % (specificity), indicating that there is large heterogeneity between the studies. These results suggest that the accuracy of liquid biopsy for detecting NSCLC mutations is highly variable between studies; this variability could be related to subject characteristics and the detection methods used. Therefore, we used a random effects model to combine the results of the studies to obtain a more conservative estimate of the effect size. Fig. 3 shows the summary receiver operating characteristic (SROC) curve from this meta-analysis. The area under the SROC curve was 0.93(95 % CI: 0.91–0.95), which indicates that liquid biopsy has a high degree of accuracy for detecting NSCLC mutations. The curve is at the top left, away from the diagonal line, indicating that the sensitivity and specificity of the included studies are high, which further supports liquid biopsy as an effective adjunct method for detecting NSCLC mutations.

Table 4.

Comparison of mutation results obtained from liquid biopsy and tumor tissue specimens.

Research	n	tp	fp	fn	tn
Cho (2022)	121	4	21	81	15
Jiang (2011)	58	14	0	4	40
Sriram (2011)	64	3	0	3	58
Xu (2012)	34	4	4	4	23
Kim (2013)	57	8	3	4	42
Mok (2015)	238	72	24	5	137
Li (2014)	141	27	29	3	62
Bai (2009)	230	63	16	14	137
He (2009)	120	80	26	0	14
Huang (2012)	822	188	108	81	445
Yam (2012)	52	14	11	0	17
Akca (2013)	86	54	0	26	6
Zhang (2013)	111	16	29	3	63
Zhao (2013)	196	17	15	0	162
Weber (2014)	94	19	19	0	56
Duan (2015)	174	94	59	0	21
Karlovich (2016)	219	54	36	4	125
Ma (2016)	108	12	45	35	16
Wang (2017)	15	5	0	4	6
Kobayashi (2018)	192	113	6	6	67
Liu (2018)	132	41	4	0	87
Veldore (2018)	126	10	3	0	113
Denis (2019)	121	4	21	81	15

tp, true positive; fp, false positive; fn, false negative; tn, true negative.

Fig. 2.

A forest plot of the mutation diagnosis between liquid biopsy and tumor tissue. Each line in the figure represents a study, and the blocks on the line correspond to the data from one study. The forest plot shows that liquid biopsy has a higher degree of heterogeneity in sensitivity and specificity for gene mutations than tumor tissue samples.

Fig. 3.

SROC curve analysis. The area under the SROC curve (AUC) is 0.93 (95 % CI: 0.91–0.95), indicating that this detection method has high diagnostic accuracy. The point at the top left of the curve represents the data from each study, and the red square represents the comprehensive working characteristic points of liquid biopsy. The coordinates of the SROC are (sensitivity, specificity).

Sensitivity analysis

To establish the robustness of the literature included in this study, we repeated the sensitivity and specificity analysis using a successive removal method. Our results demonstrated that the sensitivity and specificity of each study decreased as the number of excluded articles increased, showing that the overall robustness of the included literature was good.

Subgroup analysis

Sample sizes were arranged from small to large, and the threshold for large changes in sample size was set at 150. Consistent with the clustering results, there was substantial heterogeneity in sensitivity and specificity in the large sample subgroups, and the combined diagnostic effect size was 70[24,207], which was greater than in the small-sample subgroups, namely 45[7298]. When 200 was used as the cutoff for clustering, there was still a high degree of heterogeneity in sensitivity and specificity in large sample subgroups and the pooled diagnostic effect size of 31[13,74] was smaller than that of 75[3447] for the small-sample subgroups. This finding demonstrated that studies with large sample sizes had greater heterogeneity. The literature that included the T790M mutation had a pooled diagnostic effect size of 6[1,60], which was lower than that for literature without the T790M mutation, namely, 69[29,165], and I² was greater than 50 %. This finding indicated that the literature that did not analyze the T790M mutation was more heterogeneous. The combined diagnostic effect size of 34[19, 61] in the exon 19 group was smaller than that of the group without exon 19, with an effect size of I²>50 %. There was substantial heterogeneity in the exon 19 group and the no exon 19 group. The L858R group had a larger diagnostic effect size of 28[11,74], a smaller I², and less heterogeneity than the no L858R group. The exon 21 group had a larger pooled diagnostic effect size of 66[20,211], a smaller I², and less heterogeneity than the group without exon 21 (Table 5).

Table 5.

Subgroup analysis of the included literature.

Subgroup	Grouping criteria	Number of articles	I2(%)		P	OR [95 %CI]
			Sensitivity	Specificity
Sample size	n<150	15	97.40	94.65	<0.001	45[7298]
	n ≥ 150	7	94.63	97.25	<0.001	70[24,207]
	n<200	18	98.32	95.93	<0.001	75[3447]
	n ≥ 200	4	91.51	80.68	<0.001	31[13,74]
T790M	T790M	7	98.62	97.08	<0.001	6[1,60]
	No T790M	15	88.92	90.11	<0.001	69[29,165]
Exon 19	Exon 19	6	86.73	94.77	<0.001	34[19,61]
	No exon 19	16	94.31	98.68	<0.001	35[19,62]
L858R	L858R	4	69.95	88.42	<0.001	28[11,74]
	No L858R	18	96.88	93.57	<0.001	25[5, 111]
Exon 21	Exon 21	3	0	39.68	<0.001	66[20,211]
	No exon 21	19	97.60	94.81	<0.001	35[7175]

Detection of publication bias

A Deek's funnel plot was used to assess the potential for bias resulting from small studies having large effect sizes in the included literature. We found that there was good symmetry between the studies (P = 0.50), and there was no significant bias in the included literature (Fig. 4).

Fig. 4.

A Deek's funnel plot. This analysis was used to evaluate the possibility of publication bias. The vertical axis represents the number of studies retrieved, while the horizontal axis represents the standard error. Each dot in the figure represents a study. Studies with a large sample size account for a higher proportion of the included studies than smaller studies, and are distributed at the top. Overall, this outcome resulted in a low risk of publication bias with respect to the included studies.

Discussion

Repeated biopsies are used to determine gene mutation and evolutionary status in patients with NSCLC, but they are invasive, require accurate and appropriate sampling, and are prone to abnormal findings due to tumor heterogeneity. Liquid biopsy may negate the need for re-biopsy of tumor tissue during treatment in patients with a positive EGFR mutation [29]. Liquid biopsy uses less invasive blood samples or non-invasive urine specimens, which are more acceptable to the patient. As such, liquid biopsy provides a convenient method of genetic testing for the personalized treatment of patients with NSCLC. Liquid biopsy allows for the non-invasive detection of many gene mutations, thus guiding the direction of clinically targeted therapy, monitoring gene mutation, drug resistance and other changes, overcoming spatial and temporal heterogeneity, and aiding strategies for managing patients with different stages of NSCLC. Screening, detection of MRD to guide adjuvant therapy, early detection of relapse, surveillance of targeted therapy or immunotherapy, and genotyping for drug resistance are additional uses of liquid biopsy [30,31].

One area where liquid biopsy has particular potential is for predicting the effectiveness of TKIs that target EGFRs. Monitoring changes in ctDNA during TKI-targeted therapy could determine early disease progression, identify drug resistance, and be used to adjust treatment strategies promptly, thus reducing the risk of progression and metastasis, particularly the risk of metastasis to the central nervous system metastasis, thereby achieving longer progression-free survival and overall survival [32,33].

There are concerns as to whether liquid biopsy can replace tissue biopsy to guide accurate molecular-targeted therapy and monitor MRD. This meta-analysis included 22 studies comparing EGFR mutations in 3359 NSCLC patients with liquid biopsy specimens and histologic specimens. Our findings demonstrated that liquid biopsy had a high sensitivity for detecting EGFR mutations irrespective of sample size. In this study, we found that the heterogeneity of the literature excluding the T790M mutation was greater than that of the literature that included the T790M mutation; the combined diagnostic effect value was lower, and significant heterogeneity was present, in both exon19 groups compared with the no exon19 group; the combined diagnostic effect value was higher, and heterogeneity was lower, in the L858R group than in the no L858R group; and the combined diagnostic effect value was higher in the exon21 group than in the no exon21 group, although heterogeneity was lower in the exon21 group than in the no exon21 group. These results suggest that liquid biopsy can be used as an accurate and reliable method for detecting the mutation status of EGFR, T790M, L858R, exon 19, or exon 21 in NSCLC. However, because of the low ctDNA content in the peripheral circulation and the limited sensitivity of the detection technique, some driver gene mutations have a low detection rate. Moreover, the sample size of some genes is small, meaning results related to these genes cannot be used as a clinical basis for determining mutation status [34]. Thus, given the limitations of liquid biopsies for driver genes, histologic biopsies must still be the first choice in some cases.

The main advantages of this study are that it includes a large number of research papers, has a large sample size, and a reliable conclusion. Simultaneously, the process of meta-analysis using established principles such as QUADAS ensured academic rigor. However, because the quality of the included studies was inconsistent, there was some degree of heterogeneity, which may affect the precision of our findings. Also, our meta-analysis was limited by sample size and EGFR mutations, so fewer of the included studies were related to other driver gene mutations, which is a limitation of the study. There was a significant difference in outcomes based on sample size, with smaller studies showing different odds ratios from larger studies. This variability could introduce bias into the meta-analysis results. There is still improvement to be made in the design of the research program and in the statistical methodology. Future prospective studies with higher quality and larger sample sizes are required to verify our findings. Meanwhile, the driver mutations we focused on are the most common and clinically relevant with respect to NSCLC. This narrow focus might limit the applicability of the findings to broader clinical scenarios. Future research should include a wider range of EGFR mutations, and potentially other relevant genetic alterations, in NSCLC. Expanding the genetic scope would provide a more comprehensive assessment of the capabilities of liquid biopsy for early diagnosis, prognosis, and surveillance of MRD. Continued advances in liquid biopsy technology mean that further studies to compare the sensitivity and specificity of different platforms to detect mutations in multiple driver genes are needed. The results of such studies will provide a more accurate means of detecting mutations for comprehensive and individualized molecular diagnostics, treatment decisions, and MRD monitoring of patients with NSCLC. In addition, liquid biopsy, which has the advantages of safety and convenience, still has a way to go before it can be translated into clinical practice, and aspects such as cost-effectiveness, patient outcomes, and practical implementation need to be evaluated further. It is worth emphasizing that most of the data used in this study came from research subjects with locally advanced NSCLCs. Therefore, the consistency between liquid biopsy and histological detection in early resectable NSCLC patients needs to be confirmed by large-scale clinical research. The value of liquid biopsy combined with PET-CT examination for early detection of NSCLC may also be further elucidated in the future.

In summary, this study found that there was a high concordance between liquid biopsy and histological detection of genes for EGFR TKI-sensitive and drug resistance mutations (T790M mutations) in advanced NSCLC. The clinical value of liquid biopsy for detecting additional mutations in NSCLC remains to be further assessed. For this reason, liquid biopsy can be used as a convenient and effective method for detecting EGFR mutations in advanced NSCLC patients as an alternative to tumor biopsy methods.

Funding

This work was supported by grants from Science and Technology Development Project of Jilin Province (YDZJ202201ZYTS092 to WZ) and Natural Science Foundation of Jilin Province (20210401174YY to WZ and 20210101339JC to WW).

CRediT authorship contribution statement

Jing Cai: Writing – original draft, Methodology, Investigation. Wanning Wang: Writing – original draft, Validation, Investigation, Data curation. Wenlong Zhang: Writing – review & editing, Supervision, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.