In the current study, a total of 45 patients who attended the University Hospital of Ioannina (Ioannina, Greece) between September 2018 and November 2022 were enrolled. All participants were required to have a new diagnosis of metastatic CRC and an adequate tumor tissue sample available for molecular profiling. In addition, blood samples were collected from patients at multiple time points during the course of treatment. Specifically, blood collection for ct-DNA isolation and analysis was performed at diagnosis (prior to therapy initiation) and at the time of first and second disease progression. Both tumor tissues and plasma samples were subjected to NGS for comprehensive mutational profiles. All patients provided written informed consent prior to enrollment and the study protocol was approved by the Ethics Committee of the University Hospital of Ioannina (approval no. 11/18-04-2018).

Specifically, the inclusion and exclusion criteria were as follows. Eligible participants were adults aged 18 years or older with a new diagnosis of metastatic colorectal adenocarcinoma who had adequate FFPE tumor tissue available for molecular profiling, were able to provide serial blood samples at baseline and during follow-up, and provided written informed consent for participation and molecular testing. Patients were excluded if they had received prior systemic treatment for metastatic disease before baseline sampling, lacked sufficient tumor tissue for NGS analysis, were unable to provide the required blood samples for ctDNA testing, or had a coexisting malignancy or medical condition that could interfere with proper sample collection or reliable NGS analysis.

Peripheral blood samples were collected into Cell-Free DNA BCT tubes (Streck LLC.), particularly designed to preserve cfTNA. Plasma was obtained through two sequential centrifugations at 447 × g for 10 min at 4°C to ensure complete removal of cellular components. cfTNA was then extracted from 2 ml plasma using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Inc.), according to the manufacturer's instructions. To minimize pre-analytical variability, all blood samples were processed within 2 h of venipuncture. Plasma aliquots were stored at −80°C, and repeated freeze-thaw cycles were avoided to preserve cfTNA integrity.

Formalin-fixed, paraffin-embedded (FFPE) tumor biopsy sections were stained with hematoxylin and eosin to ensure that the tumor cell content exceeded 75% whenever feasible. A pathologist identified and marked the representative tumor areas for DNA extraction. Genomic DNA was then isolated from 10 µm-thick unstained FFPE sections using the QIAamp FFPE Tissue Kit (Qiagen, Inc.). DNA concentration and purity were determined using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Inc.). To ensure consistency, DNA quality and yield were further verified by Qubit fluorometric analysis prior library preparation.

For the analysis of FFPE-derived DNA samples, a custom Ion AmpliSeq panel was employed, based on the Ion AmpliSeq Colon and Lung Cancer Research Panel v2 (Thermo Fisher Scientific, Inc.). This panel was modified to include one additional amplicon targeting exon 14 skipping mutations in the MET gene and two additional amplicons covering exons 2 and 3 of the HRAS gene. Fusion RNA transcript analysis was performed using the Ion AmpliSeq™ RNA Fusion Lung Cancer Research Panel (Thermo Fisher Scientific, Inc.). A detailed list of the 23 targeted genes is available upon request. The analyzed genes included AKT1, ALK, BRAF, CTNNB1, DDR2, EGFR, ERBB2, ERBB4, FBXW7, FGFR1, FGFR2, FGFR3, KRAS, MAP2K1, MET, NOTCH, NRAS, PIK3CA, PTEN, SMAD4, STK11, TP53, and HRAS. The DNA concentrations from FFPE samples were measured using the Qubit™ 2.0 Fluorometer and the Qubit™ dsDNA HS Assay Kit (Thermo Fisher Scientific, Inc.). Libraries were prepared from 10 ng of FFPE-derived DNA or 6 µl of ct-DNA using the Ion AmpliSeq Library Kit 2.0 (Thermo Fisher Scientific, Inc.), according to the manufacturer's protocol. Amplicon amplification was achieved with the Ion AmpliSeq™ HiFi Master Mix (Thermo Fisher Scientific, Inc.), and the amplified products were digested with FUPA reagent followed by barcoding with the IonCode™ Barcode Adapters 1–384 Kit (Thermo Fisher Scientific, Inc.). Libraries were purified using the Agencourt AMPure XP PCR purification system (Beckman Coulter, Inc.).

For total cfTNA analysis, the Oncomine™ Pan-Cancer Cell-Free Assay (Thermo Fisher Scientific, Inc.) was utilized. This assay incorporates unique molecular identifiers (UMIs) in the form of random molecular tags to label each molecule prior library amplification, thus enabling the distinction of true variants from random sequencing errors, thereby enhancing accuracy and minimizing false-positive results. The assay targeted hotspot mutations, including single-nucleotide variants (SNVs) and small insertions or deletions, in genes such as AKT1, ALK, AR, ARAF, BRAF, CHEK2, CTNNB1, DDR2, EGFR, ERBB2, ERBB3, ESR1, FGFR1-4, FLT3, GNA11, GNAQ, GNAS, HRAS, IDH1, IDH2, KIT, KRAS, MAP2K1, MAP2K2, MET, MTOR, NRAS, NTRK1, NTRK3, PDGFRA, PIK3CA, RAF1, RET, ROS1, SF3B1, SMAD4, and SMO. Copy number variations were also assessed in CCND1-3, CDK4, CDK6, EGFR, ERBB2, FGFR1-3, MET, and MYC, and RNA fusions involving 12 key driver genes, including ALK, BRAF, ERG, ETV1, FGFR1-3, MET, NTRK1, NTRK3, RET, and ROS1, were also evaluated. The analytical sensitivity of the cfTNA assay was validated at ~0.1–0.5% mutant allele fraction (MAF) for SNVs/indels, ~1% for gene fusions, and ≥1.4-2-fold change for copy number variation (CNV) detection. For FFPE-derived DNA, the Ion AmpliSeq workflow reliably detected mutations with a limit of detection of ~5% MAF, applying a minimum coverage threshold of 500× for confident variant calling. cfTNA concentrations were quantified using the Qubit™ 2.0 Fluorometer and the Qubit™ ssDNA Assay Kit (Thermo Fisher Scientific, Inc.), while libraries were prepared according to the manufacturer's instructions. Briefly, reverse transcription of cfTNA was carried out using the SuperScript™ VILO™ Master Mix. Subsequently, the resulting cDNA was subjected to a two-cycle PCR to attach random sequence tags and A/P1 adaptors to both fragment ends, followed by purification using the Agencourt AMPure XP PCR system (Beckman Coulter, Inc.). An additional 18-cycle PCR was then performed to amplify the tagged fragments and introduce a unique barcode for each sample using the Tag Sequencing Barcode Set (Thermo Fisher Scientific, Inc.). Purification and size selection of the barcoded libraries were carried out with Agencourt AMPure XP beads, and library concentrations were subsequently quantified by real-time PCR using the Ion Library TaqMan^® Quantitation Kit (Thermo Fisher Scientific, Inc.).

For clonal amplification, 100 pmol of DNA or cfTNA libraries were pooled and amplified on Ion Sphere™ Particles (ISPs) via emulsion PCR using the Ion OneTouch™ 2 instrument and the Ion 540™ Kit-OT2, following the manufacturer's guidelines. Quality control was then performed with the Ion Sphere™ Quality Control Kit to ensure that ~10–30% of ISPs were template-positive. Enrichment of template-positive ISPs was carried out on the Ion OneTouch™ ES instrument. The enriched particles were then loaded onto an Ion 540™ Chip and sequenced on the Ion GeneStudio™ S5 Prime System (Thermo Fisher Scientific, Inc.), according to the manufacturer's instructions.

NGS data analysis was performed using Ion Reporter™ software version 5.10.1.0 integrated with Torrent Suite™ version 5.10.1 (Thermo Fisher Scientific, Inc.). The sequencing results were subsequently reviewed and further analyzed using Sequence Pilot (version 4.3.0; JSI Medical Systems GmbH). Coverage analysis was conducted with the Coverage Analysis plug-in (version 5.0.4.0), providing quality metrics for each library within the sequencing run. For FFPE DNA libraries, CNV analysis was carried out using Ion Reporter™ software, which determines CNVs based on relative copy number compared with a reference control sample. CNVs were assigned confidence scores, with values >10 indicating high-confidence amplifications, serving as the threshold for defining copy number gains. For cfTNA samples, mutation detection, RNA fusion identification, and CNV analysis were conducted according to the manufacturer's workflow using Ion Reporter™ Oncomine TagSeq Pan-Cancer Liquid Biopsy v2.1-Single Sample software (Thermo Fisher Scientific, Inc.). Furthermore, the distribution of MAFs in both plasma and tissue samples was also evaluated. All sequencing runs incorporated internal quality controls, including coverage metrics and spike-in control samples, to ensure reproducibility and instrument stability across all experiments.

The primary objective of the current study was to evaluate the concordance between mutation detection in tumor tissue DNA and cell-free plasma DNA. To accomplish this, tumor mutations were identified and compared in both sample types, and Cohen's κ statistic was then performed to assess the level of agreement. More particularly, all statistical analyses were conducted using the R programming language (R Core Team: R: A Language and Environment for Statistical Computing. Vienna, Austria: Foundation for Statistical Computing; http://www.r-project.org/). Continuous variables are expressed as the mean ± standard deviation and median with interquartile range (IQR), while categorical variables are presented as counts (n) and percentages (%). Categorical variables were compared using either the χ² test or Fisher's exact test, as appropriate, whereas paired nominal data, defined as categorical measurements obtained from the same patient in tumor tissue and plasma samples, were analyzed using the McNemar test. The normality of continuous variables was assessed by Kolmogorov-Smirnov test. Based on data distribution, comparisons of continuous data between independent (unpaired) groups were performed using the Mann-Whitney U test for non-normally distributed data. KRAS MAF values did not follow a normal distribution; therefore, comparisons between tissue and plasma samples were conducted using the Mann-Whitney U test. Correlations between continuous variables were evaluated using Pearson's or Spearman's rank correlation coefficients, as appropriate. Sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV), and overall agreement were estimated with 95% confidence intervals (CIs) using binomial distribution. Agreement represents the degree of concordance between the two methods, and Cohen's κ was employed to quantify intermethod reliability. Two-tailed P-values and 95% CIs calculated using Fisher's r-to-z transformation. P<0.05 was considered to indicate a statistically significant difference. For Spearman's rank correlation analyses, two-tailed P-values and 95% confidence intervals were calculated using Fisher's r-to-z transformation, with standard error estimated according to the method of Fieller, Hartley and Pearson.

Among the 45 patients enrolled in this study, 18 were female and 27 were male, all diagnosed with metastatic colorectal adenocarcinoma. The primary tumor was located in the left colon in 32 patients, the right colon in nine patients, and the transverse colon in three patients, while one patient had an unknown primary tumor site. The most common metastatic sites were the liver, lungs, and peritoneum. A total of eight patients had previously received adjuvant chemotherapy with 5-fluorouracil, either alone or in combination with oxaliplatin, following an earlier diagnosis of stage III CRC. The remaining patients presented with de novo metastatic disease at initial diagnosis. MSI testing was performed in a total of 33 patients. Among them, 2 patients showed loss of mismatch repair protein expression, consistent with MSI, while the remaining 31 patients had microsatellite-stable (MSS) tumors. All patients were treated with first-line therapy for metastatic CRC, according to the current clinical guidelines and the treating physician's discretion. More specifically, 9 patients received first-line therapy with mFOLFOX6 combined with an anti-EGFR agent, seven received mFOLFOX6 plus bevacizumab, and one patient was treated with mFOLFOX6 alone. Fourteen patients received XELOX plus bevacizumab, one received capecitabine plus bevacizumab, and one received XELOX alone. In addition, 3 patients were treated with capecitabine monotherapy, 4 received FOLFIRI plus an anti-EGFR agent, 2 received FOLFIRI plus bevacizumab, 2 were treated with FOLFOXIRI plus bevacizumab, and 1 received FOLFIRI alone. The demographic and clinical characteristics of the study population are summarized in Table SI.

Among the 45 tissue samples analyzed by NGS, 24 samples harbored KRAS mutations. The most common variants detected were KRAS c.35G>A (p.G12D) in 6 samples, KRAS c.38G>A (p.G13D) in 6 samples, KRAS c.35G>T (p.G12V) in 3 samples, and KRAS c.34G>A (p.G12S) in 3 samples. Less common variants included KRAS c.64C>A (p.Q22K), KRAS c.436G>A (p.A146T), KRAS c.183A>C (p.Q61H), and KRAS Q61R in 1 patient each, and KRAS c.35G>C (p.G12A) in 2 patients. In addition, 2 patients carried NRAS mutations, and more specifically NRAS c.38G>T (p.G13V) and NRAS c.181C>A (p.Q61K). BRAF V600E mutations were detected in three patients, while 1 patient carried a BRAF G466E mutation. Notably, KRAS, NRAS, and BRAF mutations were mutually exclusive in all tissue samples. During tissue profiling with NGS, among the remaining genes included in the sequencing panel, 27 patients carried TP53 mutations, 6 PIK3CA mutations, 1 patient harbored an APC mutation, 2 SMAD4 mutations, and two FBXW7 mutations. Overall, mutations in genes other than KRAS, NRAS, and BRAF were identified in 35 of the 45 tissue samples analyzed.

Among the 45 baseline plasma samples analyzed by NGS, 20 harbored KRAS mutations, including KRAS c.35G>A (p.G12D) mutation in 5 samples, KRAS c.38G>A (p.G13D) in 6 samples, KRAS c.35G>T (p.G12V) in 2 samples, KRAS c.34G>A (p.G12S) in 3 samples, KRAS c.35G>C (p.G12A) in 1 sample, KRAS c.183A>C (p.Q61H) in 2 samples, while 1 sample carried a KRAS c.436G>A (p.A146T) mutation. In addition, 2 patients carried NRAS mutations, specifically NRAS c.38G>T (p.G13V) and NRAS c.181C>A (p.Q61K). Furthermore, BRAF V600E mutations were identified in 3 samples. As observed in tissue samples, KRAS, NRAS, and BRAF mutations were mutually exclusive in plasma samples. Beyond the aforementioned mutations detected in plasma samples after ct-DNA isolation, additional changes were identified in other genes included in the NGS panel. Specifically, 17 samples carried TP53 mutations, four harbored PIK3CA mutations, two had APC mutations, and PTEN, MET, FBXW7, and AKT mutations were identified in one sample each. Overall, 24 plasma samples tested positive for alterations in genes other than KRAS, NRAS, and BRAF.

Following first-line treatment and at first disease progression, blood samples were again collected from 33 patients and processed for ct-DNA isolation and mutation analysis. Among them, 10 patients harbored KRAS mutations, including KRAS c.35G>A (p.G12D) in four patients, KRAS c.35G>T (p.G12V) in 2 patients, KRAS c.38G>A (p.G13D) in 2 patients, KRAS Q61R in one patient, while 1 patient carried co-occurring KRAS c.38G>A (p.G13D) and KRAS c.34G>A (p.G12S) mutations. NRAS mutations were detected in 2 samples, specifically NRAS c.38G>T (p.G13V) and NRAS c.183A>T (p.Q61H). In addition, 3 samples harbored a BRAF V600E mutation. No samples showed concurrent KRAS, NRAS, and BRAF mutations. At first disease progression, additional mutations were also identified in other genes. Particularly, APC mutations were found in four samples, TP53 mutations in 13 samples, and SMAD4 mutations in two samples. Furthermore, copy number amplifications were detected in EGFR (one sample) and FGFR1 (one sample). Overall, among the 33 available samples at disease progression, 18 carried mutations in genes other than KRAS, NRAS, and BRAF.

At second disease progression, 2 blood samples were collected. Both samples harbored KRAS c.38G>A (p.G13D) mutations, while remaining wild-type for NRAS and BRAF. Among them, one sample carried an APC mutation, and the other harbored a TP53 mutation.

The results of the statistical analysis demonstrated a high level of concordance between each mutation tested. More specifically, for KRAS mutations, the overall agreement rate between tissue and plasma samples was 87%, with a PPV of 95% and a NPV of 80.8% [Cohen's κ=0.74 (0.55–0.93); Se=79.2%; Sp=95.5%]. The combined analysis of KRAS and NRAS mutations also showed an agreement rate of 87%, with PPV=95.5% and NPV=79.2% [Cohen's κ=0.74 (0.55-.93); Se=80.8%; Sp=95%). For BRAF mutations, concordance was even higher, reaching 97.8%, with PPV=100% and NPV=97.8% [Cohen's κ=0.85 (0.55–1.00); Se=75%; Sp=100%]. For other, less frequent mutations, a positive agreement of 67.4% was also observed, with PPV=40.9% and NPV=67.4% [Cohen's κ=0.33 (0.09–0.57); Se=62.9%; Sp=81.8%). The McNemar test verified these findings, showing no statistically significant differences in the detection of KRAS, NRAS, and BRAF mutations between tissue and plasma samples (P>0.05; Table SII).

Regarding KRAS, the median MAF in tissue samples was significantly higher compared with plasma samples (Mann-Whitney test, P=0.022; Fig. 1). However, Spearman's correlation analysis showed no significant association between tissue and plasma KRAS MAF values (Spearman's ρ=0.25; P=0.313). In several cases, MAF values were higher in plasma compared with tissues, and vice versa. The above data are summarized in Table SIII.

The assessment of changes in mutational status during disease progression is of great importance. At progression, one patient who was initially KRAS/NRAS wild-type developed a novel KRAS/NRAS mutation, while seven patients who harbored KRAS/NRAS mutations at baseline lost these mutations at progression. Similar findings were observed in plasma samples. More particularly, two patients acquired new KRAS/NRAS mutations at progression, while six patients lost their baseline KRAS/NRAS mutations during progression. By contrast, the BRAF mutational status remained largely stable throughout the disease course. All patients who were BRAF wild-type at baseline maintained their wild-type status during progression, while only one patient with a BRAF mutation at baseline reverted to wild-type status during progression. However, no changes in BRAF mutational status were detected in liquid biopsy samples between baseline and progression. Among other genes included in the sequencing panel, six patients developed new mutations at progression compared with baseline tissue samples, while 12 patients lost previously detected mutations. Consistently, in baseline plasma samples, 10 patients developed new mutations, and eight patients lost prior mutations during progression. The detailed mutational status for KRAS, BRAF, and other genes is summarized in Table SIV. Overall, 62.5% of patients with KRAS or NRAS mutations at baseline retained these mutations during progression, whereas 37.5% lost them. Conversely, 11.8% of patients who were wild-type KRAS/NRAS at baseline developed new mutations at progression, while 88.2% remained wild-type (Fig. 2).