Association of RAS Mutation Location and Oncologic Outcomes After Resection of Colorectal Liver Metastases

Abstract

Background

RAS mutations are prognostic for patients with metastatic colorectal cancer (mCRC). We investigated clinical, pathologic and survival differences based on RAS exon for patients with colorectal liver metastases (CRLM).

Methods

This retrospective, single-center study included patients with R0/R1 resection of CRLM from 1992–2016. Patients with unresected extrahepatic disease or liver-first resection were excluded. Overall survival (OS) and recurrence-free survival (RFS) were assessed and stratified by mutation status and location. Fisher’s exact test, Wilcoxon Rank Sum test, and Log-rank test were used, where appropriate.

Results

938 mCRC patients were identified with median age of 57 (range 19–91). Of the 445 patients with KRAS mutations, 407 (91%) had a mutation in exon 2, 14 (3%) exon 3, and 24 (5%) exon 4. Median OS was 71.4 months (95%CI:66.1–76.5). Patients with KRAS mutations had worse OS compared to KRAS wild-type patients (median 55.5 versus 91.3 months, p<0.001). While there was no significant difference in OS based on the exon mutated (p=0.12), five-year OS was higher for patients with exon 4 mutations (68.8% (95%CI:0.45–0.84)) compared to those with mutations in exon 2 (45.7% (95%CI:0.40–0.51)) or exon 3 (39.1% (95%CI:0.11–0.68)). Patients with NRAS mutant tumors also had worse OS compared to NRAS wild-type patients (median 50.9 versus 73.3 months, p=0.03).

Conclusions

NRAS and KRAS exon 3/4 mutations are present in a minority of mCRC patients. Patients with exon 4 mutant tumors may have a more favorable prognosis, although the difference in oncologic outcomes based on mutated exon appears to be smaller than previously reported.

Introduction

Colorectal cancer (CRC) is the fourth most common cancer in the U.S., with 53,200 deaths predicted in 2020 alone.¹ Metastases to the liver are common, affecting up to 50% of patients.² While the only potentially curative treatment for patients with colorectal liver metastases (CRLM) is complete resection and/or ablation³, other locoregional therapies, such as hepatic arterial infusion, and cytotoxic systemic chemotherapy have achieved promising oncologic outcomes.^4–6 Appropriate selection of patients for these available and emerging therapeutic strategies relies on a fundamental understanding of the biologic drivers of carcinogenesis.

Genetic molecular profiles are increasingly utilized to direct therapeutic interventions and assist with prognostication. Mutations of the KRAS gene are present in 35–45% of CRC patients, who have been shown to have worse overall survival.² KRAS mutated cells, which exhibit increased motility and poor adhesion due to a remodeled actin cytoskeleton, display more aggressive and invasive cell behavior in patients with CRC.^7–9 KRAS mutations have also been associated with resistance to anti-EGFR treatment, higher rates of positive surgical margins, and worse oncologic outcomes.^{7,10,11,12,13–15} While more rare, mutations in the NRAS gene have been similarly associated with poor survival and worse outcomes.^16,17

While traditionally tumor mutations have been treated as binary biomarkers, an increasing body of literature suggests that not all RAS mutants have the same biologic properties and thus may differentially affect survival and cancer recurrence.^16,18,19 Our initial experience evaluating RAS mutation location in a small sample demonstrated that K/NRAS mutations located in exon 3 portend a more aggressive phenotype, whereas KRAS exon 4 mutations were associated with fewer tumor recurrences and improved overall survival.¹⁸ Given the rarity of exon 3 and 4 mutations, larger datasets are critical to confirm the previously characterized behavior of these mutations, prior to widespread adoption of these biomarkers within clinical guidelines.

With improvements in genetic sequencing technology, which have lowered cost and allowed for broader analyses, and more ubiquitous utilization of extended RAS testing in patients with CRC, larger sample size studies, with longer follow-up periods, are now possible to validate early findings. This study aimed to define associations between the mutated exon of KRAS or NRAS and tumor characteristics, recurrence, survival, and pathologic features in patients undergoing resection of CRLM.

Methods

Study Population

A prospectively maintained, single-institution database was queried for patients who underwent resection of CRLM between 1992–2016. Patients who underwent R0 or R1 resection of CRLM with available KRAS sequencing data were included. We included 132 patients from a prior published report from our institution.¹⁸ Patients were excluded if the primary tumor was not resected; if they had unresected extrahepatic disease (EHD); if they had incomplete liver resection; or if they underwent a liver-first surgical approach. Patients, who only underwent KRAS exon 2 testing and had a negative test, were also excluded, given the inability to definitively rule out the presence of an exon 3 or 4 mutation. Patients who had exon 2 mutations by Sanger sequencing were included in the study, since intra-tumoral heterogeneity for KRAS mutational status has been shown to be highly unlikely in this population. Patient characteristics, pathologic features, and survival data were collected from prospectively maintained clinical databases and retrospective clinical chart review. Last follow up was September 2019. Institutional Review Board approval was obtained for this study through Memorial Sloan Kettering Cancer Center.

DNA Sequencing Techniques

KRAS mutational status was determined from one of the following genetic sequencing tests: KRAS exon 2 Sanger Sequencing, Sequenom mass spectrometry-based genotyping, next generation sequencing (NGS) for specific mutation in 45 or 47 genes, MSK-AmpliSeq (98-genes) or MSK-IMPACT.^19,20 NRAS status was determined from either Sequenom testing or NGS. Methods for these sequencing techniques are described elsewhere.^18–23 KRAS exon 2 Sanger sequencing provided data on KRAS codons 12 and 13 mutations.²¹ The MSK Sequenom panel can detect hotspot mutations at select codons in KRAS (codons 12, 13, 22, 61, 117, 146), NRAS (codons 12, 13, 61), BRAF (codon 600), and PIK3CA (codons 345, 420, 542, 545, 546, 1043, 1047).²² MSK-IMPACT, which is the current standard of care at our institution, analyzed 341–468 genes, depending upon assay version with all versions sequencing all coding exons of KRAS and NRAS.²⁰ For patients who underwent multiple sequencing tests during the course of their disease, the first test was preferentially used in the analyses; with the exception of those patients who underwent KRAS exon 2 sequencing first and subsequently underwent analysis with Sequenom or MSK-IMPACT. In these patients, results from the more complete genetic testing were utilized, given availability of additional mutational information.

Statistical Analysis

Clinicopathologic characteristics were compared between mutation and wild-type (WT) KRAS patients and between KRAS exon location using Fisher’s Exact test and Wilcoxon Rank Sum test, where appropriate. Overall survival (OS) was assessed from time of CRLM resection until death. Patients alive at last follow up were censored. Recurrence free survival (RFS) was estimated from the time of CRLM resection until recurrence or death. Patients alive and recurrence free at last follow up were censored. OS and RFS were estimated with Kaplan Meier methods and associations with KRAS and KRAS exon mutated were tested with the log-rank test. As patients were tested using a variety of gene panels, not all mutation statuses were available for all patients. A heatmap was used to assess the distribution of these values within patients.

We combined KRAS and NRAS exons in two ways. First, we looked at any mutation in exon 2, 3, or 4, regardless of KRAS or NRAS. If a patient had incongruous exons between KRAS and NRAS, the highest exon was taken, as these were rarer. Next, we looked at KRAS and NRAS together in a single plot. If a patient had both KRAS and NRAS, the patient was classified as NRAS given the rarity.

As additional sensitivity analyses, we examined the above genes and gene combinations only for those with MSK-IMPACT testing available, excluding patients with only Sequenom or other methods, and we also examined only MSK-IMPACT mutations that were considered oncogenic drivers, based on OncoKB annotations (results not presented). OncoKB is a clinical tool that annotates the oncologic effects and significance of genetic alterations.²⁴ We also examined the relationship between KRAS status and outcomes accounting for disease free interval (DFI). Two-sided p-values were considered statistically significant. No adjustments were made for multiple hypothesis testing. All analyses were performed with SAS 9.4 TS1M6 (The SAS Institute, Cary, NC).

Results

Of the 3370 resections for CRLM performed at MSK during this period, a total of 938 patients had sufficient tumor genomic profiling to be included in this study. KRAS mutations were present in 47% (445/938). In KRAS mutants, there was an equal distribution of females and males; however, a higher proportion of mutant KRAS patients were female compared to WT patients (50.1 vs 38.9%, p<0.001). KRAS mutations were more common in patients with right-sided primary tumors (41.6 KRAS mutant vs 19.9% KRAS WT, p<0.001) and in those with EHD (15.3 KRAS mutant vs 10.3% KRAS WT, p=0.024). Far fewer KRAS mutant patients received anti-EGFR therapy (postoperative 7.2 vs 32.3%, p<0.001). KRAS mutational status was not associated with patient age, race, synchronous metastatic disease at diagnosis, DFI, preoperative CEA, or receipt of preoperative chemotherapy (p=0.16–0.92). (Table 1)

Table 1.

Patient and Clinical Characteristics based on KRAS mutation status

		All Patients	Mutant	Wild Type	p-value
# Patients		938	445 (47.4)	493 (52.6)
Age at CRLM Resection, years	Median (Range)	57 (19–91)	58 (28–91)	57 (19–88)	0.39
Gender	Male	523 (55.8)	222 (49.9)	301 (61.1)	<.001
	Female	415 (44.2)	223 (50.1)	192 (38.9)
Race	White	796 (84.9)	372 (83.6)	424 (86)	0.16
	Black	49 (5.2)	29 (6.5)	20 (4.1)
	Asian	48 (5.1)	21 (4.7)	27 (5.5)
	Other	6 (0.6)	1 (0.2)	5 (1)
	Unknown	39 (4.2)	22 (4.9)	17 (3.4)
Primary Side	Left	474 (50.5)	189 (42.5)	285 (57.8)	<.001
	Right	283 (30.2)	185 (41.6)	98 (19.9)
	Rectal	173 (18.4)	66 (14.8)	107 (21.7)
	Two-sided	8 (0.9)	5 (1.1)	3 (0.6)
Synchronous Disease	Synchronous	510 (54.4)	250 (56.2)	260 (52.7)	0.29
	Metachronous	428 (45.6)	195 (43.8)	233 (47.3)
Extrahepatic Disease	Yes	119 (12.7)	68 (15.3)	51 (10.3)	0.024
	No	819 (87.3)	377 (84.7)	442 (89.7)
Disease Free Interval (months)	Median (Range)	6.0 (0.0–398.5)	6.0 (0.0–398.5)	5.9 (0.0–157.3)	0.92
# Metastases	≤4	744 (79.3)	365 (82)	379 (76.9)	0.053
	>4	194 (20.7)	80 (18)	114 (23.1)
Preoperative CEA	Median (Range)	7 (1–12325)	7 (1–12325)	7 (1–1749)	0.77
	N Missing	(97)	(52)	(45)
	N =	841	393	448
Preoperative Chemotherapy	Yes	601 (64.1)	282 (63.4)	319 (64.7)	0.68
	No	337 (35.9)	163 (36.6)	174 (35.3)
Anti-EGFR Therapy	Postoperative	191 (20.4)	32 (7.2)	159 (32.3)	<.001
	Preoperative	35 (3.7)	14 (3.1)	21 (4.3)
	None	712 (75.9)	399 (89.7)	313 (63.5)
Resection Status	R0	819 (87.3)	393 (88.3)	426 (86.4)	0.55
	R1	116 (12.4)	52 (11.7)	64 (13)
	Unknown	3 (0.3)	0 (0)	3 (0.6)

Mutation landscape

Figure 1 illustrates the distribution of mutations across our study sample. RAS/RAF mutations, including KRAS, NRAS, and/or BRAF, were present in 54.2% of patients. Of the 445 KRAS mutant patients, 91.5% (407) had an exon 2 mutation; 3.1% (14) in exon 3; and 5.4% (24) in exon 4. The most common exon 2 mutant alleles were G12D (137/445, 30.8%), G12V (98/445, 22%), and G13D (59/445, 13.3%). NRAS mutations were present in 4.2% (34) of the study sample; 52.9% (18) had an exon 2 NRAS mutation and 41.2% (14) had an exon 3 NRAS mutation. One patient (2.9%) had an exon 4 NRAS mutation and another patient (2.9%) had an exon 5 NRAS mutation detected.

Figure 1. Gene Heatmap.

This plot presents a heatmap for each patient along the x-axis with individual genes along the y-axis. When a patient had a mutation present in the given gene, the bar color is blue. If a mutation wasn’t present, the bar color is grey, and if no tests were run, the bar is left blank (i.e. shaded white). The patients are sorted in order of frequency, so all patients with KRAS positive are clustered together, followed by for NRAS, BRAF, TP53 and PIK3CA. As KRAS and NRAS are, on the whole, mutually exclusive, few patients with KRAS mutations have NRAS mutations present. The same pattern can be seen for BRAF with NRAS and KRAS. From this plot, the pattern of mutations present can be visualized.

Outcomes by RAS mutational status

Median overall survival was 71.4 months (95%CI:66.1–76.5). Median follow-up time in survivors (N=474) was 63.0 months (Range:0.3–300). Patients with KRAS mutations had worse OS compared to KRAS WT patients (median 55.5 vs 91.3 months, p<.001). (Figure 2) While we did not detect a significant difference based on mutant exon (p=0.12), the survival distributions for patients with exon 2 and 4 mutations separated and did not show overlap. Patients with exon 3 mutations overlapped with both exon 2 and 4 mutant patients. Five-year OS was 45.7% (95%CI:0.4–0.51) for exon 2 mutants, 39.1% (95%CI:0.11–0.68) for exon 3 mutants, and 68.8% (95%CI:0.45–0.84) for exon 4 mutants, indicating a potential separation at later time points. Similar to OS, patients with KRAS mutations had worse RFS compared to KRAS WT patients (median 10.8 vs 15.8 months, p<.001). When evaluating RFS by mutant exon, three-year RFS was 21.5% (95%CI:0.18–0.26) for exon 2 mutants, 17.9% (95%CI:0.03–0.43) for exon 3 mutants, and 37.5% (95%CI:0.19–0.56) for exon 4 mutants (p=0.14). More overlap was observed between all three curves for RFS as compared to the curves for OS. (Figure 2)

Figure 2.

Plots and Estimates Stratified by KRAS and KRAS Exon

Patients with NRAS mutations had worse OS compared to NRAS WT patients (median 50.9 vs 73.3 months, p=0.026). (Figure 3) No significant association was found between NRAS and RFS (median 12.2 vs 13.3, p=0.13). When comparing patients with NRAS exon 2 to those with NRAS exon 3 mutations, no significant association was found with OS (median 40.2 vs. 50.9 months, p=0.55) or RFS (median 14.0 vs 9.0 months, p=0.94). When analyzing each exon separately for KRAS or NRAS, no significant statistical association was found with OS (p=0.18) or RFS (p=0.32). When KRAS and NRAS mutations were combined by mutant exon, five-year OS was 45.3% (95%CI:0.4–0.50) for exon 2 mutants, 37.6% (95%CI:0.18–0.57) for exon 3 mutants, and 67.6% (95%CI:0.44–0.83) for exon 4 mutants. While no significant association was found with OS (p=0.11) or RFS (p=0.21), the survival distributions for OS for patients with exon 2 and 4 mutations were separated and did not show overlap, indicating different distributions. (Figure 4)

Figure 3.

Overall Survival and Recurrence Free Survival Plots and Estimates Stratified by NRAS

Figure 4.

Overall Survival and Recurrence Free Survival Plots and Estimates Stratified by KRAS/NRAS Exon Combined

In an analysis controlling for DFI, patients with KRAS mutation had a higher hazard of OS (HR:1.67, 95%CI:1.39–2.00, p<0.001); longer DFI was associated with lower hazard of OS (HR:0.99, 95%CI:0.99–1.00, p=0.037). Similarly, both KRAS mutation status (HR:1.74, 95%CI:1.45–2.09, p<0.001) and DFI (HR:0.99, 95%CI:0.99–1.00, p=0.025) remained significantly associated with RFS.

Clinical and Tumor Characteristics by RAS mutation location

Patient and clinical characteristics were assessed based on RAS mutation location in two ways. First, primary tumor location, synchronous presentation, EHD, preoperative CEA and resection status were compared for patients with KRAS exon 2, 3, and 4 mutations. There was no difference in any of these clinical variables based on the exon mutated (p=0.14–0.93). Next, KRAS and NRAS mutations were grouped by exon and the association with clinical variables was re-assessed. No adjustments were made for multiple testing. Similarly, there was no difference between synchronous presentation, EHD, preoperative CEA, and resection status based on which K/NRAS exon was mutated (p=0.21–0.73). Interestingly, there was a trend toward fewer right-sided primary tumors (29.2% vs 33.3% vs 41.9%), more rectal cancers (20.8% vs 11.1% vs 14.7%), and more two-sided tumors (8.3% vs 3.7% vs 0.7%) in patients with exon 4 mutations compared to those with exon 3 and 2 mutations, respectively; however, this finding did not reach statistical significance (p=0.06).

Pathologic features by RAS mutation location

Pathologic features were evaluated based on RAS mutation status. When evaluating pathologic findings, no significant association was seen between tumor size and KRAS mutation status (median 2.7 vs 2.6 cm, p=0.28) or the KRAS exon mutated (median exon 2:2.5 cm, exon 3:3.3 cm, exon 4:3.4 cm, p=0.10). There was also no significant association with KRAS exon mutated and the number of metastases (>4 metastases exon 2:18.9%, exon 3:14.3%, exon 4:4.2%, p=0.18). Similarly, no significant association was seen between tumor size and NRAS mutation status (median 2.5 vs 2.7 cm, p=0.56) or NRAS exon mutated (median exon 2:3.2 cm, exon 3:2.4 cm, p=0.25).

Discussion

Somatic mutations in the K/NRAS genes are a negative prognostic biomarker for patients with metastatic CRC (mCRC). These mutations have been evaluated for their potential to prognosticate recurrence and survival, and accordingly serve as a proxy for tumor behavior. We sought to characterize survival and pathologic features in RAS-mutant patients to determine how mutation location within the RAS oncogene impacts outcomes. Our previously published experience reported worse OS for patients with K/NRAS mutations, but notably improved oncologic outcomes in patients with exon 4 mutations, compared to those with exon 2 or 3 mutations. This initial experience included 58 patients with exon 2 mutations, 7 in exon 3, and 6 in exon 4. In our current study, with a larger overall sample size, we did not find a statistically significant difference in OS based on RAS exon mutated. However, the survival distributions for patients with exon 2 and exon 4 mutations were separated and did not overlap. This may suggest that an effect exists, but the effect size is weaker than suggested in our earlier series. We also did not find any difference in pathologic characteristics, although tumor location approached significance for K/NRAS. Taken together, these findings suggest a unique and less aggressive phenotype for KRAS exon 4 mutant tumors, although with a weak to moderate effect size.

The incidence of KRAS mutations in our study was 47%, which is consistent with prior publications from our institution.^18,25 While this is higher than some prior studies of hepatic metastasectomy, many of these studies were limited by their analysis of only exon 2 mutations, as the assays used were not designed to detect mutations in exon 3 or 4.²⁶ As our series includes patients from before 2000, our rate of KRAS mutations may be higher given testing selection in the early cohort (1992–2000). Consistent with prior studies, the majority of patients in our study had exon 2 mutations, with only a small percentage having tumors with exon 3 or 4 mutations. Similarly, NRAS mutations were identified in only a minority of patients (4.2%) with an incidence consistent with previously published data.¹⁶ The small number of KRAS exon 3 and 4 and NRAS patients in this study of 938 patients further highlights the rarity with which these patients present for surgical management. When assessing clinicopathologic characteristics for these patients, those with mutant KRAS were found to have more aggressive phenotypes, with a higher rate of right-sided tumors and EHD at time of liver resection.^27,28 Notably, in patients with exon 4 mutations, we observed a smaller number of right-sided tumors, suggesting a more favorable phenotype for these patients. The lack of statistical significance may be a result of the limited number of exon 3 and 4 mutated patients included in this study.

RAS mutations have been extensively studied in CRC to help clinicians prognosticate recurrence and survival and predict response to anti-EGFR therapy. Early studies evaluating KRAS in patients with CRC concluded that survival was improved for patients WT for KRAS compared to those with mutant KRAS.²⁹ Subsequently, the impact of KRAS mutation on survival and recurrence after hepatic metastasectomy was evaluated, and KRAS mutated patients were found to have significantly worse survival compared to non-mutated patients.³⁰ These data were reproduced in a second study evaluating 202 patients undergoing surgery for CRLM; in this cohort, KRAS mutations were observed in 29% of patients and associated with both worse OS and RFS.²⁶ One potential explanation for these findings may be the difference in metastatic patterns as a function of KRAS mutation status, with patients with KRAS mutant tumors found to have a significantly higher incidence of bone, brain and lung metastases.³¹ The presence of an NRAS mutation has also been previously associated with significantly shorter survival.^32–34 Similarly, in our series, patients with KRAS and NRAS mutations had worse OS compared to those with RAS WT tumors. While KRAS mutated patients were also found to have worse RFS, no significant association was found between NRAS and RFS. This lack of association may be due to the rarity of NRAS mutations in mCRC and thus a lack of statistical power.

To further understand the association between RAS and oncologic outcomes, recent studies have evaluated the impact of mutation location on tumor biology, recurrence, and survival. One study, which compared outcomes in patients with KRAS exon 2, 3, and 4 mutations, observed that mutations in exon 2 and 3 were associated with worse clinical outcomes, when compared to those in exon 4.¹⁹ A purported hypothesis for this finding was lower intrinsic RAS activity in patients with exon 4 mutations. These results were reproduced by our group in a study which evaluated 165 patients who underwent curative-intent R0 resection.¹⁸ In that study, patients with exon 4 mutations had significantly better RFS and improved OS, while patients with exon 3 mutations had significantly worse outcomes. Patients with exon 3 mutations were also found to have a greater number of small tumors and a more diffuse recurrence pattern compared to other mutation locations and WT patients. Given the rarity of exon 3 and 4 mutations, these findings were limited by small sample size and extremely wide confidence intervals. To validate these trends, we sought to re-evaluate the association between mutation location and outcomes in a larger cohort of patients with resectable CRLM. An additional benefit of our analyses was longer follow-up, as we included patients who had been analyzed with multiple unique genetic platforms, conducted over the past two decades. In our analyses, the association between mutated exon and patient characteristics, outcomes and pathologic features did not reach statistical significance and the survival distribution for patients with exon 3 mutations overlapped with those who had exon 2 and 4 mutations. However, the curves for patients with exon 2 and 4 mutations were separated, suggesting that patients with exon 4 mutations may have a survival advantage. While this finding may be driven in part by a small number of long-term survivors, it is important to note as a potential signal for less aggressive tumor biology.

The discrepancy in these findings highlights the complex and still poorly understood relationship between mutations and tumor biology. As KRAS exon 3 and 4 and NRAS mutations are relatively rare, large numbers of patients must be evaluated to appropriately characterize how these mutations translate into tumor behavior. Patients with K/NRAS exon 3 mutations have been previously described as an aggressive subset of mCRC.^16,18 Despite this, a recent study from Cercek et al. found that NRAS mutant patients with liver-limited disease, who were candidates for resection, were able to achieve long periods of disease control.¹⁶ Our patient population represents a very select, extensively pre-treated group of CRC patients with a resectable liver burden and limited and resectable EHD. By selecting patients with liver-predominant disease, amenable to resection, our cohort may have been biased, with over-representation of patients with the most favorable tumor biology. In this highly-selected subset of mCRC patients, it is possible that the incremental differences in the phenotypes of each RAS mutation location are offset. As most patients in this study received varying durations and intensity of preoperative chemotherapy and/or anti-EGFR inhibitors, adjustment for these factors was not feasible in our analyses.

This study had some inherent limitations. First, the data were obtained from a single-institution, which may not be reflective of the general population. There was intrinsic referral bias for patients in this study, as they needed to be candidates for resection of CRLM or curative intent locoregional therapies. Additionally, sequencing may have been more likely to have been offered to patients with worse tumor biology, so our sample may not be representative of all resected patients at our institution. Second, several different sequencing assays were used to detect KRAS mutations, as the technology improved throughout the period of the study. 16 Patients tested with Sanger sequencing may have had exon 3 and 4 mutations which were undetected. We elected to include these patients, given the added benefit of a prolonged follow-up period in these patients. As an additional sensitivity analysis, we evaluated only patients with MSK-IMPACT testing results, which led to similar results (data not shown). Patients were often sequenced multiple times by different techniques including Sanger sequencing, Sequenom, and MSK-IMPACT, with repeat testing performed using more comprehensive platforms due to cancer progression. To address this, we sought to use the first most complete genetic information obtained. Finally, despite the large patient sample, absolute numbers of patients with K/NRAS exon 3 and 4 mutations remained small.

Understanding the clinical significance of RAS mutations is important for prognostication, chemotherapy susceptibility and improved treatment selection in patients with mCRC. In this large-scale confirmatory study, we have demonstrated the association between RAS mutations and oncologic outcomes in patients with CRLM. KRAS exon 3 and 4 and NRAS mutations were found in only a minority of patients with resectable CRLM, and the impact of mutation location on outcomes remains heterogenous and incompletely understood. KRAS exon 4 mutant tumors may have a unique, potentially more favorable phenotype. Such a difference in oncologic outcomes for patients with exon 4 mutations appears to be smaller than previously suggested in patients with resectable CRLM.

Synopsis.

NRAS and KRAS exon 3 and 4 mutations are present in a minority of patients with metastatic colorectal cancer. Patients with exon 4 mutations may have more favorable oncologic outcomes, after resection of colorectal liver metastases.