Quantitative ctDNA Profiling of RAS Mutations as a Prognostic Biomarker in Metastatic Colorectal Cancer

TL;DR

This study shows that measuring RAS mutation levels in blood DNA can predict survival in metastatic colorectal cancer patients.

Contribution

The study identifies a 5% RAS variant allele frequency threshold as a novel prognostic biomarker for metastatic colorectal cancer.

Findings

A RAS VAF threshold of 5% best predicted overall survival in metastatic colorectal cancer patients.

High RAS VAF was linked to worse survival and more aggressive tumor features.

Quantitative ctDNA profiling improves risk stratification and treatment decisions.

Abstract

Circulating tumor DNA (ctDNA) analysis offers a non-invasive approach to molecular profiling. While RAS mutations are well-established predictive biomarkers in metastatic colorectal cancer (mCRC), the prognostic value of their variant allele frequency (VAF) remains unclear. We retrospectively analyzed individual patient data with mCRC who underwent ctDNA testing using the FoundationOne® Liquid CDx assay. The primary objective was to determine the optimal RAS VAF cutoff for overall survival (OS) prognostication. Between November 2020 and July 2024, 282 patients were enrolled. Among 265 eligible patients, 134 (50.6%) were ctRAS mutant, 25 (9.4%) ctBRAF
V600E mutant, and 106 (40.0%) were ctRAS/BRAF wild-type. A RAS VAF threshold of 5% yielded the highest prognostic discrimination for OS (HR = 2.41; 95% CI 1.65–3.55; p < 0.0001; C-index = 0.601). ctRAS-high mutant tumors (VAF ≥ 5%) were associated with synchronous metastatic disease, multiple metastatic sites, higher blood tumor mutational burden, and elevated tumor fraction. ctRAS-low mutant tumors (VAF < 5%) were more frequently metachronous, presented with a single metastatic site, and showed liver involvement. High RAS VAF in ctDNA is a strong and independent prognostic marker for OS in mCRC. Quantitative ctDNA profiling may enhance risk stratification and guide personalized management strategies.

Figures

Full text

1. Introduction

Colorectal cancer (CRC) is the third most common cancer worldwide and the second leading cause of cancer death, with about 1.9 million new cases and more than 900,000 deaths annually, reflecting a substantial and growing population health burden [1,2]. Global burden analyses based on the Global Burden of Disease framework show sustained increases in CRC disability-adjusted life years from 1990 to 2021, with projections indicating further growth by 2040, underscoring rising morbidity and premature mortality across regions [3]. Advances in molecular profiling have significantly improved our understanding of CRC biology, with mutations in the RAS gene family—particularly V-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) and neuroblastoma RAS viral oncogene homolog (NRAS)—emerging as key biomarkers for guiding therapeutic decisions in metastatic CRC (mCRC) [4,5]. These mutations are known to predict resistance to anti-epithelial growth factor receptor (EGFR) monoclonal antibodies such as cetuximab and panitumumab, and are routinely assessed in clinical practice [6,7,8,9].

Beyond their predictive role, the prognostic significance of RAS mutations—especially their variant allele frequency (VAF)—remains insufficiently characterized. Circulating tumor DNA (ctDNA) analysis via liquid biopsy has enabled non-invasive, real-time monitoring of tumor molecular dynamics, offering a more comprehensive view of tumor burden and clonal heterogeneity than traditional tissue biopsies [10,11]. VAF, defined as the proportion of sequencing reads harboring a specific mutation, may reflect tumor burden and heterogeneity.

In this study, we explore the prognostic value of circulating RAS VAF in patients with mCRC, focusing on its association with overall survival (OS). By integrating molecular quantification with clinical outcomes, we aim to determine whether RAS VAF can serve as a biomarker of prognostic relevance and contribute to improved risk stratification in mCRC.

2. Results

2.1. Patient Population

Between November 2020 and July 2024, liquid biopsy was performed in 282 patients with mCRC. Seventeen patients were excluded due to technical failure (n = 8), absence of reportable genomic alterations (n = 6), unavailable VAF for RAS genes (n = 2), or consent withdrawal (n = 1). Among 265 eligible patients, 106 patients (40.0%) were ctRAS/BRAF wild-type, 134 (50.6%) harbored ctRAS mutations, and 25 (9.4%) carried the ctBRAF^V600E^ mutation (Figure 1).

2.2. Sampling and Blood Analysis

Liquid biopsies were performed in first-line setting in 149 (56.2%) patients and second- or subsequent line setting in 116 (43.8%) patients. The median turnaround time was 12.0 days (range 7.0 to 23.0).

2.3. Determination of VAF Threshold

The cut-off corresponding to VAF 5% yielded the most significant separation in OS (p < 0.0001), with a HR for death in the high-expression group of 2.41 (95% CI 1.63 to 3.55), indicating an unfavorable prognostic association. Moreover, this cut-off also yielded the highest concordance index (C-index 0.601, 95% CI 0.555–0.648) (Table 1, Figure 2).

To strengthen the definition of the 5% VAF threshold, we performed a receiver operating characteristic (ROC) analysis for OS. The ROC curve yielded an area under the curve (AUC) of 0.681 (95% Boodstrap CI 0.590 to 0.769; p < 0.0001), indicating moderate discriminatory ability. The Youden index identified an optimal cutoff at >4.8% (95% CI > 2.15 to >16.2) with sensitivity and specificity rate s of 78.8 and 57.4 (Figure 3).

ROC curve presents a plot of the true positive rate (Sensitivity) in function of the false positive rate (100-Specificity) for different cut-off points of RAS VAF. Each point on the ROC curve represents a sensitivity/specificity pair corresponding to a particular decision threshold.

Based on these findings, the 5% value of VAF for RAS genes was selected as the optimal cut-off for subsequent analyses. The mean VAFs were 1.5% (SD 1.4) and 25.9% (SD 18.8) in patients with ctRAS-low and ctRAS-high tumors, respectively.

2.4. Patient and Tumor Characteristics

Patient and tumor characteristics at the time of circulating CGP are summarized in Table 2. The study population included both sexes, with a nearly balanced distribution: 49% female and 51% male patients. The overall mean age of patients was 64.4 years (SD 13.5), while ctRAS mutant patients had a mean age of 65.0 years (SD 12.7).

The ctRAS-low (VAF < 5%) and ctRAS-high (VAF ≥ 5%) subgroups exhibited significant clinical and biological divergence. ctRAS-low tumors were more commonly associated with metachronous metastatic presentation (p = 0.046), a solitary metastatic site (p = 0.017), and predominant hepatic involvement (p = 0.001). Conversely, ctRAS-high tumors exhibited significantly elevated blood tumor mutational burden (bTMB; p < 0.0001) and increased TF (p < 0.0001). Metachronous dissemination was more frequently observed in ctRAS-low and ctRAS/BRAF wild-type profiles, whereas synchronous metastatic disease was significantly enriched in ctRAS-high and ctBRAF-mutant cases (p = 0.020).

Patients with ctRAS-mutant tumors exhibited distinct clinical and molecular characteristics across treatment lines, with high-grade histology, prior resection of the primary tumor, presence of multiple metastatic sites, pulmonary involvement, and elevated bTMB occurring more frequently in those receiving second-line or subsequent therapies compared to those treated in the first-line setting (Table A1).

2.5. Prognostic Analysis of VAF

The median follow-up for OS was 35.3 months (95% CI 32.1–58.6). Stratification using a RAS VAF threshold of 5% revealed a significantly increased risk of death, with a HR of 2.41 (95% CI 1.63–3.55; p < 0.0001) (Table 1, Figure 4). No statistically significant difference in OS was observed between the ctRAS/BRAF wild-type and ctRAS-low subgroups (HR = 1.40; 95% CI 0.98–2.00). Moreover, patients with ctRAS-high mutant tumors exhibited survival outcomes comparable to those with ctBRAF^V600E^ mutations (HR = 0.98; 95% CI 0.51–1.91).

The poor prognostic impact of RAS-high on OS was observed in all pre-specified subgroups, confirming consistent prognostic impact of VAF RAS in mCRC (Table A2, Figure 5).

The strongest prognostic value of VAF RAS was observed in patients with metachronous metastatic disease (HR 8.87, 95% CI 3.64 to 21.6), lung involvement (HR 4.56, 95% CI 2.38 to 8.75) and high-grade tumors (HR 4.41, 95% CI 1.01 to 19.25). Of note, there was no statistically significant association between VAF RAS and OS in a small subset of patients with either NRAS (n = 9) or exon 3 (n= 12) tumors.

In first-line setting, median OS were 31.5 months (95% CI 12.5 to 31.5) and 11.6 months (95% CI 6.4 to 19.3) in ctRAS-low and ctRAS-high, respectively (HR 2.89, 95% CI 1.57 to 5.33; p = 0.0006). In later lines, median OS were 16.7 months (95% CI 9.8 to 31.0) and 7.2 months (95% CI 4.7 to 23.8) in ctRAS-low and ctRAS-high, respectively (HR 2.68, 95% CI 1.57 to 4.58; p = 0.0003) (Table A1).

Multivariate analysis was conducted incorporating variables that demonstrated statistical significance in univariate assessments (Table A2). Independent adverse prognostic factors for OS included elevated (≥10%) tumor fraction (HR 2.58, 95% CI 1.69–3.95; p < 0.0001), administration of second- or later treatment line (HR 2.46, 95% CI 1.61–3.75; p < 0.0001), and increased (≥10) bTMB (HR 1.71, 95% CI 1.09–2.70; p = 0.021).

2.6. Molecular Characteristics and Co-Occuring Alterations

ctRAS-high mutant tumors were associated with higher bTMB (p < 0.0001) and higher TF (p < 0.0001) than ctRAS-low mutant tumors. There was no difference between the 2 subgroups in terms of location or type of mutation. Both ctAPC and ctTP53 mutant high (VAF ≥ 5%) tumors were significantly more frequent in patients with ctRAS-high than ctRAS-low tumors. Significant differences were observed in terms of molecular co-alterations between the 2 groups (p = 0.006), TGFβ and Wnt/β-catenin pathways being more frequently altered in patients with ctRAS-high than ctRAS-low tumors (Table 3).

3. Discussion

This study demonstrates that the VAF of RAS mutations detected via ctDNA is a robust prognostic biomarker for OS in patients with mCRC. Stratifying patients based on the value of 5% VAF for RAS genes provided the most significant prognostic discrimination for OS. This threshold yielded a clear separation between survival curves, with a HR for death of 2.41 indicative of a clinically meaningful increase in mortality risk among patients with higher VAF value (≥5%).

Although the cut-off was selected through exploratory testing rather than a statistically corrected optimization procedure, the approach remains valid within a clinical research framework, particularly when supported by biological plausibility and reproducibility across subgroups. The prognostic significance of RAS VAF was particularly consistent across treatment settings, with ctRAS-high status associated with significantly reduced OS both in first-line (HR 2.89) and later-line therapies (HR 2.68), indicating that its adverse prognostic effect is preserved irrespective of treatment line. Furthermore, in the context of metachronous metastatic disease, a high RAS VAF was associated with particularly poor prognosis, reflected by a markedly elevated HR for death of 8.87. These findings underscore the clinical relevance of quantitative ctDNA metrics beyond mere RAS mutation status. For clinical translation, we selected 5% as the operational threshold, which closely aligns with the statistically derived value while offering a pragmatic and reproducible benchmark across laboratories. While the discriminatory power is modest, the validated threshold provides a clinically feasible tool for risk stratification and highlights the potential of quantitative ctDNA metrics to complement existing prognostic markers in mCRC.

High RAS VAF tumors were associated synchronous presentation, multiple metastatic sites, and elevated tumor burden—as evidenced by higher bTMB and TF. Additionally, ctRAS-high mutant tumors were associated with molecular co-alterations from TGFβ or Wnt/β-catenin pathways. Conversely, ctRAS-low tumors were more frequently metachronous, limited to a single metastatic site, and predominantly involved the liver. Our findings should be interpreted as demonstrating a statistical association between higher ctDNA-based RAS VAF and poorer survival outcomes, rather than implying a biological causal role of RAS mutations in driving tumor aggressiveness. While RAS mutations are well-established oncogenic events in colorectal cancer, the quantitative burden detected in plasma likely reflects tumor load and shedding dynamics rather than intrinsic biological behavior. Thus, VAF ≥ 5% should be considered a prognostic biomarker that correlates with disease state, not a mechanistic determinant of progression. Future studies integrating functional assays and longitudinal sampling will be required to explore whether quantitative mutation burden has any direct biological implications.

The use of liquid biopsy offers several advantages, including non-invasive sampling, real-time monitoring, and the ability to capture tumor heterogeneity [12,13,14]. Importantly, our sensitivity analyses confirmed the prognostic value of high RAS VAF across all clinical subgroups, reinforcing its potential utility in routine risk stratification. These findings may inform clinical decision-making, particularly in considering early therapeutic escalation for patients with high ctRAS VAF.

In previous studies, it has been shown that ctDNA metrics provide deeper insights into tumor biology, reflecting disease aggressiveness and burden [15]. Moreover, ctDNA metrics are prognostic for survival. In the RASANC study, ctDNA mutant allele frequency—defined as the highest observed frequency among mutated genes—demonstrated significant prognostic relevance [16]. At the optimal cut-off of 20%, the HR for death was 2.58 (95% CI 2.02 to 3.28). In a translational analysis of the Valentino study, the maximum VAF detected among mutated genes was taken as an estimate of the ctDNA amount in patients with RAS wild-type mCRC [17]. Patients with high VAF (superior to the median of 12.6%) presented with liver metastases and synchronous presentation, and had poorer survival compared to those with low VAF (HR: 1.82, 95% CI 1.20 to 2.76). Together, these studies underscore the potential of integrating quantitative ctDNA analyses into routine practice, both as a baseline prognostic tool and as a dynamic biomarker to guide therapeutic strategies.

RAS mutations are a common oncogenic driver, but their frequency, isoform distribution, and clinical impact vary considerably across different tumor types. In CRC, both KRAS and NRAS mutations occur (KRAS ~40%, NRAS ~3–5%), with diverse codon hotspots (G12, G13, Q61). In pancreatic ductal adenocarcinoma, KRAS mutations dominate (~90–95%), especially at codon 12 (G12D, G12V, G12R), while NRAS and HRAS mutations are essentially absent [18]. RAS mutations are rare in renal cell carcinoma (RCC) [19]. When present, they are usually confined to specific histologic subtypes such as papillary renal neoplasms with reverse polarity, where KRAS mutations occur at high frequency. In the broader RCC spectrum, KRAS mutations affect only about 1% of cases, while NRAS and HRAS mutations are essentially absent. Thus, unlike colorectal or pancreatic cancers, RAS mutations are not a major driver in renal cancer, but when detected, they may define distinct molecularly characterized subgroups.

Limitations of this study include its retrospective design, treatment heterogeneity, and the absence of longitudinal ctDNA testing. First, we acknowledge the limitations inherent to retrospective designs, which inherently restrict the strength of causal inferences and may introduce selection bias. To mitigate selection bias, we included all consecutive patients with mCRC who underwent ctDNA testing at our center during the study period. Exclusion criteria were strictly technical or ethical (e.g., consent withdrawal), and no patients were excluded based on clinical outcomes or treatment response. Second, patient heterogeneity, including differences in prior treatments and disease characteristics, could have influenced survival outcomes and confounded the observed associations. The cohort analyzed in this study was derived from a single-center patient population, which may not fully reflect the heterogeneity of broader mCRC populations. While treatment heterogeneity exists, our primary endpoint—OS—was chosen to minimize confounding from specific therapeutic regimens. Sensitivity analyses across subgroups confirmed the prognostic impact of RAS VAF regardless of treatment line. Additionally, while the FoundationOne^®^ Liquid CDx assay offers standardized quantification, inter-platform variability in VAF measurement warrants further validation across different technologies. VAF values were analyzed as reported by the assay without adjustment for estimated tumor fraction. As tumor fraction correlates with ctDNA burden, non-corrected values may overestimate mutation frequency. This limitation highlights the need for future studies to integrate tumor fraction estimates to refine the prognostic utility of RAS VAF. Finally, the absence of an independent validation cohort limits the generalizability of our findings, underscoring the need for prospective, multicenter studies to confirm the prognostic utility of RAS VAF in ctDNA. Future research should therefore focus on standardized inclusion criteria and independent validation to strengthen the clinical applicability of these results. Beyond ctDNA profiling, nanoscale technologies such as nanomaterials and biosensors are being developed to enhance CRC monitoring, offering complementary sensitivity and theranostic potential [20,21].

These findings suggest that circulating RAS VAF may serve as a useful stratification tool in future trials and could inform risk-adapted treatment strategies. Further validation in independent cohorts and prospective studies is warranted to confirm its prognostic utility and generalizability.

4. Materials and Methods

4.1. Study Design and Patient Population

This retrospective study included adult (≥18 years old) patients diagnosed with mCRC, irrespective of treatment line, who underwent liquid biopsy for ctDNA testing as part of routine molecular profiling. All patients provided written informed consent prior to blood collection, in accordance with ethical standards and institutional guidelines. Patients were treated according to standard-of-care protocols. Clinical data were collected from electronic medical records (DxCare software, version 8.2021.2.8, Dedalus, Antony, France).

4.2. Circulating Comprehensive Genomic Profiling (CGP)

Blood sampling was centralized at Hôpital Franco-Britannique—Fondation Cognacq-Jay (Levallois-Perret, France) and was done using FoundationOne^®^ Liquid CDx cfDNA blood collection tubes (8.5 mL nominal fill volume per tube, 2 tubes per patient). The FoundationOne^®^ Liquid CDx assay was performed using cfDNA isolated from plasma derived from anti-coagulated peripheral whole blood [22]. Extracted cfDNA undergoes whole-genome shotgun library construction and hybridization-based capture of 324 cancer-related genes including coding exons and select introns of 309 genes, as well as only select intronic regions or non-coding regions of 15 genes. VAFs, representing the fraction of sequencing reads in which a variant is observed, were reported as measured by the assay and were not adjusted for estimated tumor fraction. In cases where multiple RAS mutations were detected in a single patient, the mutation with the highest VAF was selected for analysis. The ctDNA fraction (TF), the blood tumor mutational burden (bTMB) and microsatellite instability (MSI) were also evaluated (Appendix A.1) [22,23].

4.3. Molecular Co-Alterations and Pathways

Selected genes were classified into seven prespecified molecular pathways: mitogen activated protein kinase (MAPK), phosphoinositide 3 kinase/protein kinase B/mammalian target of rapamycin (PIK3/AKT/mTOR), DNA damage repair (DDR) system, transforming growth factor beta (TGFβ), Wingless-related integrated site (Wnt), cell cycle, and immune (Appendix A.2). Regarding genomic co-alterations, mutation status was dichotomized as mutated versus non-mutated, irrespective of the VAF level. When multiple co-alterations were present, the variant with the highest VAF was retained.

4.4. Statistical Analysis

To evaluate the prognostic impact of the VAF of RAS genes on survival, we performed a series of manual dichotomizations based on VAF amount. For each dichotomization, patients were stratified into low and high groups, and survival outcomes were compared using Kaplan–Meier estimates and the log-rank test. Harrell’s C-index were also calculated [24]. The confidence interval (CI) of Harrell’s C-index was calculated using the modified τ method [25]. The cut-off yielding the most significant separation between survival curves—defined by a clinically meaningful hazard ratio (HR), the lowest log-rank p-value and the highest C-index—was selected as the optimal threshold. This approach, although exploratory, allowed for transparent assessment of the variable’s discriminative capacity while maintaining interpretability within a clinical context. ROC curve analysis with bootstrap resampling was performed to confirm the optimal threshold for RAS VAF in relation to OS.

Demographic, clinical, and genomic characteristics were obtained using descriptive statistics. Categorical variables were summarized in numbers and percentages, continuous variables were presented as means with standard deviation (SD) and medians with minima and maxima. OS was defined as the time interval between the date of liquid biopsy and the date of death from any cause. Alive patients were censored to the last date known to be alive. Kaplan–Meier method was used for survival, the reverse Kaplan–Meier method for follow-up, and Cox regression for multivariate analyses. A p value < 0.05 was considered statistically significant. All analyses were conducted using MedCalc^®^ Statistical Software version 22.021 (MedCalc Software Ltd., Ostend, Belgium; https://www.medcalc.org; 2023).

5. Conclusions

In conclusion, our study highlights the prognostic significance of RAS VAF in ctDNA, with a ≥5% threshold identifying patients at higher risk of poor survival. Integrating quantitative ctDNA metrics into clinical practice may enhance personalized management strategies in mCRC. Future research should aim to validate the prognostic utility of RAS VAF thresholds in larger, prospective multicenter cohorts to strengthen their clinical applicability.