Introduction

Determining the RAS (KRAS/NRAS) mutational status on a tumor biopsy is mandatory to guide the best treatment recommendation in metastatic colorectal cancer (mCRC). Anti-epidermal growth factor receptor (EGFR) inhibitors in combination with chemotherapy doublet for mCRC treatment is associated with an overall survival that exceeds 36 months in selected populations [1].

Genotyping of RAS mutations is routinely done on tumor tissue to predict drug resistance to anti-EGFR in patients with mCRC [2]. The ESMO Clinical Practice Guideline 2022 has included the possibility that in situations where adequate tissue is not available, RAS mutation status can be analyzed by plasma-derived cell-free DNA (cfDNA) [1].

Several mechanisms of acquired resistance to anti-EGFR have been described in mCRC. The most common are the emergence of RAS, BRAF or EGFR mutations [3,4,5].

The analysis of ctDNA could provide dynamic genetic information about the tumor while the patients are being treated [6,7,8,9,10].

This is an observational and prospective study based on two cohorts of first-line therapy in mCRC. The results of cohort 1 have already been reported [11]. Here, we present the results of cohort 2 by assessing (I) the percentage of patients with RAS, BRAF and EGFR mutations and (II) the concordance rate for RAS mutational status in tumor-tissue and plasma samples (using the Idylla™) at baseline in RAS wild-type mCRC patients starting their standard first-line treatment.

Methods

This was a nationwide, observational, multi-center, prospective study that evaluated the mutational status of RAS, BRAF and EGFR using liquid biopsies (Idylla™ as in vitro diagnostic (IVD) test [Biocartis, Mechelen, Belgium]) [12] in RAS wild-type mCRC patients according to baseline tumor-tissue biopsy per local practice (PERSEIDA study, NCT02792478).

The mutational status of BRAF at baseline in solid biopsy was also assessed.

The participants were following clinical practice, including the baseline tumor-tissue biopsy and the selection of the first-line treatment. The methodology have been previously published [11].

RAS and BRAF and EGFR mutational status were assessed at baseline, at Week 20 and at disease progression in liquid biopsy (Idylla™). The mutant allele fraction (MAF) threshold for KRAS and BRAF was 0.4% and for NRAS was 1.5%.

Patients fulfilling the following inclusion criteria (as reported for cohort 1 [11] were recruited consecutively and followed-up until disease progression: patients ≥ 18 years, with mCRC measurable by RECIST, who start first-line treatment and with a histologically confirmed diagnosis of mCRC and wild-type RAS (according to baseline tumor-tissue biopsy).

Blood samples collection and storage was performed as already described [11]. Additionally, the frozen baseline plasma of 20 patients (subgroup of cohort 2 patients) was shipped to Sysmex Inostics GmbH (Hamburg, Germany) for BEAMing analysis [13] to determine the RAS mutation status concordance between the two techniques. The frozen plasma (at disease progression) of 8 of these patients with progressive disease was prospectively selected when sufficient sample was available and shipped to a central laboratory (Instituto Maimónides de Investigación Biomédica, Córdoba, Spain) for analysis using the next-generation sequencing (NGS) technique AVENIO® (cfDNA expanded panel and Illumina NextSeq 550) [14].

The protocol was approved by an independent ethics committee CEIC Consorcio Hospitalario Provincial de Castellón (Spain), and all patients gave their written informed consent before enrollment. The study has been performed according to the Declaration of Helsinki.

Statistical analysis

The primary endpoint was the percentage of patients with RAS, BRAF and EGRF mutations in liquid biopsies at baseline. The percentage of discordant patients and concordance rate (liquid vs solid biopsies results) was calculated for RAS, along with its 95% confidence interval (CI). Secondary endpoints included the description of RAS, BRAF and EGRF mutations in liquid biopsies at disease progression and at 20 ± 2 weeks. As an exploratory endpoint, the percentage of concordance rate (liquid vs solid biopsies) and Cohen’s kappa coefficient were also calculated for BRAS in the panitumumab subgroup.

The conversion rate for RAS at disease progression was assessed as earlier reported [11]. The conversion rates for BRAF and EGFR were also analyzed.

The mutations analyzed in RAS, BRAF and EGFR using Idylla™ are described in Supplementary Table S1.

The association between cfDNA RAS and BRAF mutational status, respectively, with the overall response rate (ORR) and progression-free survival (PFS) was performed. Moreover, PFS analyses using Kaplan–Meier approach, and a multivariable Cox regression analysis to assess the predictive factors of PFS were also performed as formerly described [11].

Agreement between cfDNA biopsies (Idylla™ vs. BEAMing) for RAS at baseline in the subpopulation assessed (n = 20) was assessed by Cohen’s kappa coefficient.

Statistical analyses were performed with the SAS statistical software package (SAS Institute, Inc, Cary, NC).

Results

Baseline characteristics

Between May 2018 and May 2021, 131 patients were screened in 18 Spanish hospitals (117 patients were included, evaluable population). The most frequent first-line treatment was panitumumab plus chemotherapy (n = 103) (panitumumab subpopulation). The main baseline characteristics are shown in Table 1.

Table 1 Baseline demographic and clinical characteristics
Full size table

Liquid biopsy (Idylla™) results

In the evaluable population, 7 (6.0%) patients presented RAS mutations in cfDNA at baseline. Additionally, 6 (5.1%) patients had BRAF mutations, while EGFR mutations were not detected. At baseline, the percentage of RAS mutational status concordance between tissue and liquid biopsies was 94.0% (95% CI 88.1–97.6) with comparable results in the panitumumab subpopulation (Table 2).

Table 2 Liquid biopsy results (Idylla™)
Full size table

At baseline, the percentage of BRAF mutational status concordance between tissue and liquid biopsies was 98.9% (95% CI 94.0–100.0) (Table 2). The Cohen’s Kappa coefficient was 0.739 (95% CI 0.392–1.000), indicative of a substantial agreement.

A logistic regression analysis did not find any variable (liver metastasis, tumor burden, number of affected organs at study entry, primary tumor surgery and serum carcinoembryonic antigen [CEA] levels) associated with discordant cases at baseline (data not shown).

Additionally, the overall percent RAS agreement between the two tests (Idylla™ vs. BEAMing) was 90.0% (Cohen’s kappa = 0.444, indicative of a moderate agreement; n = 18) considering a mutant allele fraction ≥ 0.02% as a threshold.

At 20 weeks, 83 patients had blood sample available, 2 of them (2.4%) presented cfDNA RAS mutations in the evaluable population (both received panitumumab plus chemotherapy) and 3 (3.6%) presented BRAF mutations (2 received panitumumab plus chemotherapy and 1 received bevacizumab plus chemotherapy). EGFR mutations were not detected.

At disease progression, 7/50 (14.0%) patients with blood sample available presented cfDNA RAS mutations in the evaluable population (all 7 received panitumumab plus chemotherapy) and 3 (6.0%) had BRAF mutations (1 received panitumumab plus chemotherapy). EGFR mutations were not detected (Table 2).

Regarding the emerging mutations, 6 patients in the panitumumab subpopulation presented RAS mutations in cfDNA at disease progression that were not present at baseline (Table 3). Accordingly, the RAS conversion rate at disease progression was 13.3% (6/45 patients with baseline RAS wild-type status, both by tumor and cfDNA analysis) and blood sample available at disease progression) in the evaluable population and 15.0% (6/40 patients) in the panitumumab subpopulation. At 20 weeks, 1 patient had emergent RAS mutations (received panitumumab plus chemotherapy) (Table 3). No patients had emergent BRAF or EGFR mutations.

Table 3 Characteristics of patients with emergent RAS mutations at disease progression or at week 20 (liquid biopsy, Idylla™)
Full size table

The characteristics of patients with mutations at any time (per liquid biopsy) are shown in Table 4 (RAS mutations) and in Supplementary Table S2 (BRAF mutations). Among the 6 patients with emerging RAS mutations at disease progression, 4 achieved complete or partial response as the best overall response, while 1 achieved stable disease and 1 progressive disease. In these patients, the most frequent metastatic site was liver (n = 4).

Table 4 Characteristics of patients with RAS mutations at any time as per liquid biopsy (Idylla™)
Full size table

Three patients with RAS mutations in liquid biopsies at baseline were RAS wild-type at disease progression (Table 4, patients #1, #3 and #11).

Association with patient outcomes (panitumumab subpopulation)

The ORR (not confirmed) was 80.6% (95% CI 71.4–87.9) (n = 98 data available), being 85.7% (66/77) and 63.6% (14/22) in patients with left and right-sided primary tumor, respectively. In this subpopulation of panitumumab plus FOLFOX, the ORR (not confirmed) was 83.8% (95% CI 73.8–91.1%).

The ORR according to cfDNA RAS, BRAF and RAS/BRAF mutational status at baseline and by primary tumor location is displayed in Table 5. Among patients receiving panitumumab, the ORR did not show statistical differences regardless of the cfDNA RAS and BRAF mutational status at baseline. However, the odds ratio was greater than 1, indicating a trend towards a higher probability of response (ORR) in cfDNA wild-type patients compared to those with RAS or BRAF mutations. This trend was higher in left-sided vs right-sided tumors (RAS and RAS/BRAF mutants).

Table 5 ORR by RAS, BRAF, and RAS/BRAF mutational status in liquid biopsy at baseline (panitumumab subpopulation, total and by primary tumor location)
Full size table

Regarding PFS, the median (95% CI) time was 12.5 (9.9–13.8) months. There were no statistically significant differences in median PFS according to cfDNA RAS, BRAF or RAS/BRAF mutational status at baseline or at any time (Table S3). There were also no differences in median PFS between patients who were always RAS wild-type (as per cfDNA) and patients with RAS mutations at any time. Similar results were observed by RAS status at baseline or at any time in the subgroup of patients with left-sided tumors (Table S4). Additionally, related to the tumor laterality, no differences in median PFS (between patients either cfDNA RAS/BRAF mutations at any time or NGS mutations or baseline microsatellite instability/defective MMR vs patients without these alterations) were found, either in the total population or in the panitumumab subgroup of patients with left-sided tumors (Supplementary Fig S1 and Fig. S2, respectively). The Cox regression analysis did not find any predictive factor of PFS (data not shown). In this subpopulation of panitumumab plus FOLFOX, the median PFS (95% CI) time was 13.3 (10.9–15.0) months.

The multivariable model to predict tumor burden (defined as the sum of the longest tumor diameters [mm]) (which included as predictable factors: liver metastasis, RAS and BRAF mutant/wild-type status at baseline) showed that the presence of RAS or BRAF mutations at baseline was not associated with tumor burden but liver metastasis vs not having liver metastasis was significantly associated with tumor burden (difference of 41.5 mm between them, 95% CI 19.2–63.7; p = 0.0004) (Supplementary Table S5).

NGS results

Subclonal genomic variants potentially associated with anti-EGFR resistance were found in 6/8 patients tested (75%). More details are described in Supplementary Table S6.

Discussion

The PERSEIDA study was a prospective, multicentric and observational study. This prospective approach design has allowed us to get a homogeneous sample with 92% of patients receiving anti-EGFR plus chemotherapy as first-line treatment. Among those, 97% received panitumumab plus chemotherapy, being an informative population for this treatment.

A high concordance for RAS status between tissue and plasma samples (using Idylla™) was observed, being comparable to that previously published in cohort 1 using the BEAMing technique [11]. Moreover, a high percent agreement between both cfDNA tests (Idylla™ and BEAMing techniques was reported (90%, considered as moderate concordance according to Cohen’s kappa). It should be highlighted that Idylla™ is a technique widely used in many centers for the determination of the RAS mutational status in routine practice, so these results are highly informative for clinical practice in patients with mCRC.

Emergent RAS mutations were mostly observed at disease progression (in 6 patients). At 20 weeks, only 1 patient presented emergent RAS mutations. Emergent RAS mutations were located in exon 2 and 3 of both KRAS and NRAS, with a similar frequency between them. Nevertheless, the prevalence of RAS mutations in ctDNA was reported to be higher for KRAS exon 2 compared to the other locations (43% vs 3–4%) [15]. Of note, a trend of emerging Q61 mutations in KRAS which were described as infrequent [16]. As the number of patients with emergent mutations in this study is low, comparison with other studies is difficult.

The highest proportion of patients with RAS mutations at disease progression could explain only in part the appearance of acquired resistance to first-line treatment. Some studies reported a similar rate of emerging RAS mutations during first-line treatment with anti-EGFR plus chemotherapy [4, 17]. Diaz et al. [7] suggested that resistance mutations were highly likely to be present in a clonal subpopulation prior to the initiation of therapy and the time to recurrence was simply the interval required for the subclone to re-populate the lesion. Aligned to this, Takayama et al. [6] also proposed that latent cells from tumors, with undetected KRAS mutations may undergo clonal expansion during the treatment. Other authors reported that RAS mutations can be attributed not only to the selection of pre-existing RAS clones but also to the mutant clone as the result of de novo acquisition of a RAS mutation [10]. By contrast Parseghian et al. [17] suggested that rather than an outgrowth of preexisting clones, a transcriptional mechanism of acquired resistance appears to predominate in patients treated with a combination of an anti-EGFR and cytotoxic chemotherapy at first line. Therefore, we hypothesize in line with prior research that the identification of RAS mutations in ctDNA might have higher clinical relevance in later lines of therapy than in baseline first-line setting, acknowledging the high concordance observed with tissue biopsy and low appearance of emergent RAS mutations during first-line treatment.

By contrast to the emergent RAS mutations, there were three patients with RAS mutations at baseline (according to liquid biopsy) that were RAS wild-type at progression. The explanation is unknown, however, the disappearance of RAS mutant clones in plasma has been described before, supporting a negative selection of RAS mutations during the clonal evolution of mCRC. Nevertheless, the extent of conversion to RAS wild-type disease at the time of progression has not been clarified yet [18]. Another explanation could be the sensitivity of the liquid biopsy as ctDNA often represents only a small fraction of total cfDNA [19,20,21]. Other factors as assay type, technical and biological background can also affect RAS mutational detection.

The NGS results found subclonal genomic variants potentially associated with anti-EGFR resistance in 6 out of 8 patients (75%) at disease progression. The NGS analysis suggests that resistance is driven by a complex process of genomics alterations, and not just by RAS mutations. The ORR results showed no statistically significant differences according to baseline cfDNA RAS mutational status. Despite these, there was a trend towards a higher probability of response in wild-type compared to RAS and/or BRAF mutational status. This trend was higher in the left-sided tumors compared to right-sided one (RAS and RAS/BRAF mutants). These better results in left-sided tumors were previously reported, suggesting the importance of the primary tumor location [22,23,24]. Comparable results were reported when considering the RAS and BRAF mutational status at any time. It should be noted that the presence of baseline mutations in cfDNA had no impact on ORR to chemotherapy plus panitumumab. These results make us consider what the threshold of sensitivity would be, for example what fraction of RAS allelic variants in cfDNA, would be clinically relevant to select or not a treatment that includes anti-EGFRs.

Related to the PFS, there were also no significant differences in PFS according to baseline cfDNA RAS mutational status or between patients that were always RAS wild-type (as per cfDNA) and patients with RAS mutant any time point, although a numerically higher trend was observed in RAS wild-type patients. The comparison of these results with cohort 1 of the same study that used BEAMing analysis showed that, despite no significant association between ORR or PFS and cfDNA RAS mutational status was observed, patients with left-sided tumor presented a median PFS significantly longer among cfDNA RAS wild-type patients than those presenting cfDNA RAS mutations at any time [11]. The discrepancies between the two cohorts could be done to the different sensitivity between Idylla™ and BEAMing analysis [25]. Even when we selected the patient population most theoretically susceptible of being benefited from anti-EGFR treatment (patients presenting either cfDNA RAS/BRAF mutations at any time or NGS mutations or baseline microsatellite instability/defective MMR), we did not find statistically significant differences in PFS, probably due to a low number of patients available and the low Idylla™ sensitivity. In addition, patients who gained new RAS/BRAF mutations showed a similar prognosis as those who maintained RAS/BRAF mutations, and shorter PFS and OS than those who remained RAS/BRAF wild-type [26].

In our multicentric hospital setting, FOLFOX plus panitumumab was the most common first-option treatment for patients with RAS wild-type mCRC, with an ORR of 84% and a median PFS time of 13 months. These data are according to previous published results in this population [27, 28].

This study has some limitations. The study presents a large series of patients with native RAS tumors who were treated with chemotherapy and anti-EGFR, specifically panitumumab, as a first-line. However, due to the low frequency of RAS mutations in cfDNA, it is challenging to obtain clinically and statistically significant differences. Moreover, this study was not initially designed to determine the association between RAS status and outcomes and the predictive factors study that were only explorative endpoints.

Some of the strengths are the homogeneity of the prospective studied population, with all the patients being tumor RAS wild-type at baseline, starting their first-line treatment (mostly panitumumab plus chemotherapy). This study compared two highly sensitive techniques for cfDNA (Idylla™ and BEAMing techniques). Finally, it assessed not only RAS mutational status but also BRAF and EGFR mutational status over time.

In conclusion, the concordance rate between liquid and solid biopsies at baseline was very high, as previously reported. The emergence of RAS mutations during disease progression is noteworthy but may only partially explain acquired resistance to anti-EGFRs, as more complex mechanisms are involved. Therefore, the dynamics of the genomic landscape in ctDNA may provide relevant information for the management of mCRC patients both in first-line and later lines.