Lung cancer is the leading cause of cancer-related mortality worldwide in women and men (1). Non-small cell lung cancer (NSCLC) accounts for 80-85% of all lung cancer cases (2). Numerous oncodriver alterations have been identified in NSCLC, defining the tumor mutation profile and contributing to cancer progression (3). Recent advances in understanding the molecular basis of lung cancer have led to the identification of various genomic signatures, facilitating the reclassification of NSCLC and accelerating the development of targeted therapies (4).

The epidermal growth factor receptor (EGFR) is a member of a subclass of receptor tyrosine kinases. Its signaling pathway is one of the most important mechanisms regulating growth, proliferation, and differentiation in mammalian cells (5). EGFR is frequently over-expressed during the development and progression of NSCLC; mutations in EGFR can lead to its constitutive activation, increased protein expression, and tumor progression. This is why alterations in EGFR have significant therapeutic implications for lung cancer (6). EGFR mutations are common among women, never smokers, and those with adenocarcinomas (7), the frequency of these mutations shows significant differences among various racial groups being higher in Asians (49.1%), followed by Latinos (23%) and Caucasians (12.8%) (8).

The tissue biopsy remains the gold standard for tumor diagnosis and molecular analysis due to its high level of standardization and accuracy of results (9); however, concerns regarding DNA integrity and the inadequate representation of the tumor’s genetic landscape can limit the molecular diagnosis (10). LB is emerging as a valuable alternative due to its many advantages. It is a minimally invasive procedure that assesses a tumor’s molecular profile when tissue samples are not available, it also allows for easy repeatability, provides a comprehensive tumor profile in real-time, and enables early detection of resistance mechanisms (11).

Despite advancements in the field and the evident advantages of LB, there are still significant limitations that need to be addressed. There is currently no standardized methodology for its use, leading to low reproducibility in preclinical studies (12). Additionally, LB may miss critical genetic alterations that occur in the early stages of disease (13). The risk of false negatives is a significant concern; this can occur due to a limited amount of ctDNA in the bloodstream (14).

Different molecular techniques are increasingly being used to detect EGFR mutations. However, the existing literature is inconsistent, making it difficult to determine the most effective method. Additionally, significant heterogeneity has been identified among the different testing methods used in plasma. LB utilizes technologies such as digital polymerase chain reaction (ddPCR), RT-qPCR, amplification refractory mutation system (ARMS) and NGS (15). In cases where there are limited variants of interest, targeted PCR methods can offer advantages over NGS, however, one limitation of all PCR methods compared to sequencing-based techniques is their potential to detect only known mutations, which may hinder the identification of new alterations (16). In contrast, technologies like NGS provide high-throughput sequencing, making it ideal for large-scale analysis, allowing a more comprehensive understanding of the tumor’s molecular profile due to their broader capabilities (17).

The clinical usefulness of LB to monitor disease progression is well known, having been studied across multiple cancers (18). Although the role of baseline ctDNA has not been thoroughly explored, monitoring levels of the initial activating EGFR mutation may facilitate more reliable detection of progression (19). Current studies recognize the significance of EGFR testing during initial diagnosis; however, few studies have shown that the yield of mutant allele burden in ctDNA at baseline was associated with aggressive disease (20). More research involving various molecular techniques and comparisons of data from multiple sources is necessary to define the effectiveness of LB in cancer as this remains an active development area (21).

Some systematic reviews have validated the accuracy of EGFR mutation detection using LB (22). Some of these explored treatment associations with response and survival outcomes (23) while others focused on establishing concordance among the two biopsy methods or testing the suitability of LB as an alternative to tissue biopsy (24). In contrast to these, our goal is to compare the diagnostic performance of EGFR mutation detection in LB to tissue biopsy in treatment-naive patients with NSCLC.

Registration and publication of study protocol. The study protocol of this systematic review and meta-analysis has been registered on the International Prospective Register of Systematic Reviews (PROSPERO), and its registration number is CRD420251027919.

Search strategy. A systematic search in the Medline and LILACS databases identified studies that compare molecular techniques in tumor biopsy and LB to detect and quantify EGFR mutations in treatment -naive patients with NSCLC.

All searches were conducted during September 2024 and included articles published up to 2024. The PRISMA-DTA (25) guidelines were followed during data extraction, analysis, and reporting. No limits regarding language or publication period were considered. Appropriate truncations and word combinations were applied and adjusted for each database. The search was done with MeSH and DeCS. The terms considered included: Lung neoplasm, Lung carcinoma, Carcinoma, Non-small Cell Lung, Non-Small Cell Lung Carcinoma, Liquid Biopsy, Cell-free Nucleic Acids, “Exosome, Cells, Circulating Neoplastic, Protein Kinase Inhibitors, TKIs, Next generation sequencing, digital droplet PCR. Reference lists of previous systematic reviews were also reviewed to identify additional studies that might be eligible for inclusion in the analysis.

Selection criteria. The systematic review applied the following inclusion criteria: 1) descriptive studies, cohorts, and clinical trials; 2) comparative studies that evaluate the sensitivity and specificity between tissue biopsy and liquid biopsy in patients with EGFR mutations; and 3) untreated subjects.

The following exclusion criteria were applied: 1) the research methods and outcomes were not comprehensive, and the consistency between the two detection methods could not be established; 2) the study included patients with cancers other than the subtypes of NSCLC.

Data extraction. Two reviewers independently assessed the studies eligible for inclusion and subsequently reached a consensus on which should be included in the analysis. In cases of disagreement, resolution was achieved through discussion or with the involvement of a third reviewer. The extracted data from each article included the author’s name, year of publication, country of publication, study type, reference standard, objective, and methods. The diagnostic performance measures were true positives (TP), false positives (FP), true negatives (TN), false negatives (FN).

Quality evaluation. Data extraction and quality assessment were conducted independently. Quality was evaluated using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool.

Summary measures. The primary outcomes evaluated were sensitivity and specificity. Positive predictive values (PPV) and negative predictive values (NPV) and likelihood ratios (LR) were also estimated.

Data synthesis and analysis. Contingency tables were generated for each study, including true positives and negatives, false positives and negatives cases. Meta-analysis was performed with a random effects model to calculate pooled sensitivity and specificity with their 95% confidence intervals (CIs) of the studies comparing the same technique in tissue biopsy and LB. These values are presented in forest plots. I2 values were used to measure the heterogeneity between studies. I2 values >50% suggested that there was significant heterogeneity. Statistical analysis was performed using Meta-disc software 2.0 (Clinical Biostatistics Unit of the Ramon y Cajal Research Institute, Madrid, Spain).

Literature selection. Figure 1 presents the flowchart for the literature selection process A total of 446 relevant articles were retrieved by searching the literature. After removing 192 duplicate studies, we screened the titles and abstracts of 254 articles. Of them, 209 were irrelevant, 45 were relevant and their full-text articles were assessed for eligibility. After assessment based on the inclusion and exclusion criteria, 33 articles were excluded because they included patients previously treated or under treatment or it was not possible to perform the correct data extraction, leaving a total of 12 eligible studies.

Characteristics of included studies. A total of 12 articles met the inclusion criteria for this analysis providing data from 1,314 patients with NSCLC (Table I) (26-37). These articles reported the diagnostic efficacy of detecting EGFR mutations in plasma, comparing the results to paired tumor tissue samples. The sample sizes of the studies ranged from 39 to 262 subjects. The most common characteristics of patients included: males: 59.5% (n=723); adenocarcinomas: 82.2% (n=929); and clinical stage IV: 77% (n=725).

Among the 12 studies, five performed NGS in plasma and seven applied PCR-based techniques. Six studies were prospective cohorts, and six were retrospective cohorts. Seven studies utilized the same detection method for both tissue and biopsy samples: four performed NGS, and three employed PCR-based methods.

Diagnostic efficacy of NGS and PCR-based methods for detecting EGFR mutations in tissue and plasma. As shown in Table I, all studies provided data for calculating the sensitivity and specificity related to LB’s effectiveness in detecting EGFR mutations. NGS was used to detect EGFR mutations in 280 plasma samples across four eligible publications: Rachiglio et al. (26), Yao et al. (27), Park et al. (28), and Choudhury et al. (29). Choudhury et al. (29) analyzed two cohorts of patients: the first cohort compared NGS in plasma with targeted PCR in tissue, and the second cohort compared NGS in plasma and tumor tissue. Rachiglio et al. (26) reported the highest sensitivity for NGS in plasma at 77.2% while Park et al. (28) obtained the lowest at 67%. Regarding specificity, Yao et al. (27) achieved 100%, whereas Choudhury et al. (29) reported the lowest at 83% in the group that compared NGS in plasma and tumor tissue.

Three studies compared results from tissue biopsy with those from LB using RT-qPCR. A total of 334 plasma samples were analyzed. Weber et al. (30) reported the highest sensitivity and specificity for RT-qPCR in plasma at 61% and 96%, respectively, whereas Ulivi et al. (31) obtained the lowest sensitivity at 52%.

ddPCR was used to determine the EGFR mutational status in two eligible studies. Zhu et al. (32) used ddPCR in plasma and compared it with ARMS-PCR on tumor tissue. The results indicated a sensitivity of 76.19% and a specificity of 96.55% for ddPCR in liquid biopsy. Suryavanshi et al. (33) reported that ddPCR achieved a sensitivity of 87.5% in liquid biopsy samples when compared with RT-qPCR performed on tissue samples.

Only one eligible study evaluated the accuracy of ARMS-PCR for detecting EGFR mutations in plasma samples. Duan et al. (34) used ARMS-PCR to identify EGFR mutations in matched tumor tissue and plasma. The sensitivity of the test was found to be 50%, while the specificity was 100%.

Meta-analysis comparing PCR-based methods and NGS. In the meta-analysis, we compared the performance of NGS and RT-qPCR methods used in tissue biopsy and LB, focusing on the sensitivity and specificity (Table I).

NGS was performed in four articles (n=280) achieving an overall sensitivity of 69% (95%CI=62-75%) and specificity of 90% (95%CI=84-94%). The studies with the highest sensitivity were: Rachiglio et al. (26), 77% (95%CI=0.55-0.92), followed by Choudhury et al. (29) with 75% (95%CI=0.43-0.95). The highest specificities were achieved by Yao et al. (27) at 100% (95%CI=0.85-1.00) and Rachiglio et al. (26), at 91% (95%CI=0.71-0.99) (Figure 2).

RT-qPCR was applied in three articles (n=334) achieving an overall sensitivity of 56% (95%CI=46-65%) and specificity of 89% (95%CI=66-97%). Weber et al. (30) achieved a sensitivity of 61% (95%CI=0.41-0.78). Following closely was Paturu et al. (35), who showed a sensitivity of 55% (95%CI=0.40-0.70). Additionally, Weber et al. (30) also achieved a specificity of 96% (95%CI=0.92-0.99), while Ulivi et al. (31) recorded a specificity of 89% (95%CI=0.52-1.00) (Figure 2).

Positive likelihood ratio (PLR) and negative likelihood ratio (NLR). As shown in Table I, PRL and NLR of NGS and RT-qPCR for detecting EGFR mutations were calculated. For NGS, PRL was 6.9 (95%CI=5.91570-7.88430) and NLR was 0.34 (95%CI=0.64430-1.32430). Regarding RT-qPCR the PRL was 5.1 (95%CI=4.11641-608359) and NLR was 0.49 (95%CI=0.49359-1.47359).

Quality assessment of included articles. Methodological quality assessment was performed using the QUADAS-2 tool a total of five domains were evaluated: patient selection, index test, reference standard, flow, and timing. Two review authors independently assessed the quality of the included studies, and the risk of bias (High, Low, or Unclear) was made for each domain. The results are shown in Table II.

Patient selection showed the highest risk of bias (25%), and flow, timing, and index test had the lowest risk of bias (0%). For the applicability of the studies, all the studies were classified as of low concern in the five aspects.

The present article, which included 12 studies involving 1.314 patients, is the first meta-analysis to assess the diagnostic performance of various molecular methods for detecting EGFR mutations in tissue biopsy compared to LB prior to treatment initiation. Our results showed that NGS had a sensitivity of 69% and specificity of 90%, whereas RT q-PCR had a sensitivity of 56% and specificity of 89%. The NGS is generally considered less sensitive than PCR (38). However, our results showed that NGS has better sensitivity and higher LPR compared to RT-qPCR.

Some systematic reviews and meta-analyses have analyzed the diagnostic accuracy of LB, with Wang et al. (23) being the only study to report differential results based on the techniques analyzed. The overall LB sensitivity and specificity were 68% and 98%, respectively, slightly higher than ours. In contrast, Franzy et al. (22) analyzed LB using various techniques, such as NGS and PCR-based methods. The results indicated that NGS exhibited slightly higher sensitivity than PCR-based techniques (0.62% vs. 0.56%). This aligns with our findings; although we investigated the same methods, we focused on a different population, specifically the group of treatment-naive patients with NSCLC.

A study by Lawrence et al. (39), reported a tissue and plasma NGS sensitivity of 94.8%, and 52.6%, respectively. The sensitivity of detection in tissue compared to plasma varies across different studies; tissue NGS has shown significantly higher sensitivity across various patient subgroups and disease stages (39); in comparison, on LB, NGS exhibits high specificity, almost 100%, especially for detecting EGFR mutations (40). While RT-qPCR is highly specific, it can only detect a limited number of genetic alterations. Pisapia et al. (41) found that 8.9-9.4% of EGFR mutation went undetected using RT-qPCR but were identified through NGS.

We found that the pooled sensitivity of RT-q-PCR was 56%, which fell within the range of 51.7% to 60.7% reported in other studies (30,31,35). For specificity, our study recorded a pooled specificity of 89% (0.65; 0.97), aligning with data from other authors (30,31,35). Kanaoka et al (42) analyzed causes of discordant results in EGFR detection between two platforms: cobas® EGFR Mutation Test v2, based on RT-qPCR, and the Oncomine Dx Target Test (ODxTT), based on targeted NGS. The main objective was to identify factors that contribute to differences in method sensitivity. The study determined that specific mutations are key factors in certain patient populations and tumor types. Cobas® EGFR Mutation Test v2 could generate false positives for exon 20 insertion mutations in smokers and patients with squamous cell carcinoma. Unfortunately, we could not meta-analyze ddPCR data. Studies included in this systematic review, using ddPCR, have reported sensitivities ranging from 80% to 90%, with specificities exceeding 90% (33).

Recently, LB has gained recognition not only for its usefulness in evaluating advanced NSCLC but also for its potential in early-stage monitoring and screening (43). There is limited data on the effectiveness of LB in early diagnosis, and the results regarding the accuracy of pre-treatment LB are still inconclusive (12). It can lead to false-negative outcomes, primarily due to the small quantity of ctDNA present in the bloodstream (44). The sensitivity of LB when compared to corresponding tissue samples ranges from 70% to 85% in advanced-stage disease, though it diminishes in earlier stages of the disease (45). As a result, it remains unclear which method is the most efficient for molecular profiling in LB during the early stages of diagnosis (46).

The present article has limitations such as the low number of studies that tested tissue biopsy and LB to detected EGFR mutations prior the treatment. The included studies varied across several factors, including the cut-off levels for ctDNA, sample sizes, tumor stages, methodological approaches, pre-analytical conditions, and detection methods. These differences could introduce potential sources of bias and heterogeneity. Despite these limitations, this study provides important insights for selecting the most effective technique for patients who have not undergone treatment in the early stages of diagnosis.

In conclusion, these results contribute to the understanding of the utility of LB in clinical practice for detecting EGFR mutations in treatment-naive patients with NSCLC. We identified that NGS has better diagnostic performance than RT-qPCR in LB compared to tissue biopsy samples tested with the same methodology. In our data, the majority of patients were in advantage stages. Therefore, we recommend more studies that compare the diagnostic performance of the different techniques in treatment-naive patients with NSCLC in different stages mainly in early stages.

The Authors have no conflicts of interest to declare in relation to this study.

(I) Conception and design: M Galeano, R Parra-Medina; (II) Administrative support: M Galeano, R Parra-Medina (III) Provision of study materials or patients: All Authors; (IV) Collection and assembly of data: All authors; (V) Data analysis and interpretation: All Authors; (VI) Manuscript writing: All Authors; (VII) Final approval of manuscript: All Authors.

This research received no external funding.

No artificial intelligence (AI) tools, including large language models or machine learning software, were used in the preparation, analysis, or presentation of this manuscript.