Evaluating Combinations of Biological and Clinicopathologic Factors Linked to Poor Outcomes in Resected Colorectal Liver Metastasis: An External Validation Study

Abstract

Background.

Recent studies have suggested that certain combinations of KRAS or BRAF biomarkers with clinical factors are associated with poor outcomes and may indicate that surgery could be “biologically” futile in otherwise technically resectable colorectal liver metastasis (CRLM). However, these combinations have yet to be validated through external studies.

Patients and Methods.

We conducted a systematic search to identify these studies. The overall survival (OS) of patients with these combinations was evaluated in a cohort of patients treated at 11 tertiary centers. Additionally, the study investigated whether using high-risk KRAS point mutations in these combinations could be associated with particularly poor outcomes.

Results.

The recommendations of four studies were validated in 1661 patients. The first three studies utilized KRAS, and their validation showed the following median and 5-year OS: (1) 30 months and 16.9%, (2) 24.3 months and 21.6%, and (3) 46.8 months and 44.4%, respectively. When analyzing only patients with high-risk KRAS mutations, median and 5-year OS decreased to: (1) 26.2 months and 0%, (2) 22.3 months and 15.1%, and (3) not reached and 44.9%, respectively. The fourth study utilized BRAF, and its validation showed a median OS of 10.4 months, with no survivors beyond 21 months.

Conclusion.

The combinations of biomarkers and clinical factors proposed to render surgery for CRLM futile, as presented in studies 1 (KRAS high-risk mutations) and 4, appear justified. In these studies, there were no long-term survivors, and survival was similar to that of historic cohorts with similar mutational profiles that received systemic therapies alone for unresectable disease.

INTRODUCTION

The only modality that can potentially cure patients with colorectal liver metastases (CRLM) is surgery.¹ However, it has long been observed that some surgically treated patients fare so poorly that these individuals may derive no benefit from surgery. Ideally, matching surgically versus medically treated patients should be performed before drawing definitive conclusions regarding the benefit of surgery. However, this is usually not feasible since the liver tumor burden and distribution of extrahepatic spread among medically treated patients are often too varied to be accounted for by a matched analysis.

Thus, most studies have approached this question by devising clinical risk scores to find subsets of patients who fare worse than what medical treatment alone can offer. For example, the Fong score was devised in 1999 to aid patient selection for surgery by predicting which patients would have lower median and 5-year overall survival (OS) than that conferred by medical therapies that were available in the early 90s.² The authors reported that even the highest risk group had a median survival of 22 months, which was higher than the 12 months conferred by the alternative to surgery (5-fluorouracil monotherapy).³ Several other studies have subsequently failed to find a subset of patients with technically resectable disease who may not benefit from surgery.⁴ Of note, this was likely because systemic therapies were not effective, and thus even a very small benefit from surgery was enough to justify it.

The introduction of more potent chemotherapeutics and biologic agents after 2000 improved the median OS of patients treated with systemic therapies, which created a new standard that surgery must improve median OS by more than 2 years.^5,6 This prompted new studies that utilized biomarkers along with traditional clinicopathologic factors to identify patients who would not benefit from surgery.⁷ However, these studies did not validate the outcomes of high-risk patients in external independent cohorts. As such, whether these findings can be generalized outside the cohorts used to develop them remains unknown. Furthermore, KRAS was used as a binary variable (yes or no mutation), while recent studies have demonstrated that specific hotspot mutations confer poor survival while others do not.⁸ Specifically, mutations associated with decreased median overall survival compared with the overall KRAS group were classified as high risk (G12V, G12A, G12R, G13R, and G13D), and individuals with improved survival compared with the overall KRAS group were classified as low risk (G12C, G12D, G12S, G13C, exon 3, and exon 4). Thus, the aim of the current study was to externally validate previously described combinations of biomarkers and clinical factors proposed to render surgery for CRLM futile and investigate whether the use of high- versus low-risk KRAS point mutations (instead of overall KRAS status) and BRAF mutation can improve patient selection for surgery.

PATIENTS AND METHODS

Systematic Search for Studies Proposing Combinations of Biomarkers and Clinical Factors That May Render Surgery for CRLM Futile

Using the search engines PubMed and Embase, relevant literature was identified by using the following search strings in study titles and keywords: prognostic model, risk score, colorectal cancer, liver metastases, liver resection, RAS/BRAF mutation, and survival. References in the retrieved articles and a previously published systematic review by our group and two reviews by other groups were used to identify articles missed in the electronic search.^9–11 Only studies that used overall survival as the endpoint and included patients who underwent surgery for CRLM after 2000 were included. This is because the year 2000 marked a threshold for key developments in systemic chemotherapy, such as the introduction of irinotecan and oxaliplatin as components of cytotoxic combination therapy and the use of biological agents such as bevacizumab, cetuximab, and panitumumab.

External Validation Cohort

All adult patients who underwent curative intent resection of CRLM between 2002 and 2019 and had data on their KRAS or BRAF mutational status at Stanford Hospital, Johns Hopkins Hospital (JHH), Cleveland Clinic (CCF), Charité—University of Berlin, Medical University of Vienna, Medical University of Graz, University of Bergen, CHU Clermont-Ferrand, Université Clermont Auvergne, Hospital Italiano de Buenos Aires, Yokohama City University Graduate School of Medicine, and Graduate School of Life Sciences Kumamoto University were included in this study. We excluded patients who underwent palliative surgery (R2 resection) and those who had two primary malignancies. The institutional review boards of the participating institutions approved the study.

Recorded Variables and Outcomes

Clinicopathologic variables were obtained from the respective databases, which included patient age at the time of hepatectomy, gender, data on the primary tumor [T stage, N stage, and location of the primary (right versus left versus rectal)], disease-free interval (DFI) between primary tumor resection and resection of liver metastases and receipt of pre-hepatectomy chemotherapy, data on the CRLM [size of the largest CRLM, number of CRLMs, laterality of CRLMs (unilobar versus bilobar), presence of concurrent extrahepatic resectable disease, carcinoembryonic antigen (CEA) levels at the most recent time before hepatectomy], operative details [hepatic margin status (R0 versus R1, with R1 defined as the presence of microscopic tumor invasion at the resection margin, and performance of concurrent ablation], receipt of post-hepatectomy chemotherapy, and somatic mutational status for KRAS (KRAS variants) and BRAF. OS was defined as the time between hepatic resection and either death or last follow-up.

Statistical Analysis

All analyses were performed in JMP Pro 16. Continuous variables were reported as median values with interquartile ranges (IQRs), and categorical variables were presented as frequencies (%). Continuous variables were compared using the Kruskal–Wallis test, while categorical variables were compared using Pearson’s chi-squared test or Fisher’s exact test as appropriate. The Kaplan–Meier method was used to assess OS and differences in OS were tested using the log-rank test. A p value <0.05 was considered statistically significant.

RESULTS

Studies Proposing Combinations of Biomarkers and Clinical Factors that May Render Surgery for CRLM Futile

A total of ten eligible studies were identified in the systematic review (Fig. 1). The following studies were excluded because some of the variables used by the authors were not available in our validation dataset. Two studies were excluded because the authors included CA 19–9 levels in the prognostic model; another study was excluded because the authors included prehepatectomy CEA levels in the prognostic model, while a separate study was excluded because the authors used chemotherapy response as one of the criteria for the tumor biology score (TBS) and one study was excluded because the authors used vascular invasion as one of the criteria of their prognostic index (PI).^12–16 In addition, the study by Lang and colleagues was excluded because additional biomarkers beyond RAS and BRAF were utilized.¹⁷

FIG. 1.

Prisma flowchart of study selection

Validation of Studies Proposing Combinations of Biomarkers and Clinical Factors that May Render Surgery for CRLM Futile

A total of 1661 patients were included in the validation cohort with a median (interquartile range; IQR) follow up of 32.8 (17.1–56.1) months. The demographic and clinicopathologic characteristics of the patient cohort are summarized in Table 1. The 1-, 3-, and 5-year OS were 92.0%, 64.2%, and 45.3%, respectively. The association of KRAS wild type, low-risk KRAS mutation, high-risk KRAS mutation, and BRAF mutation with OS are illustrated in Fig. 2. The demographic and clinicopathologic characteristics of the patient cohort stratified by KRAS (wild-type versus low- and high-risk KRAS mutations) and BRAF mutational status are summarized in eTable 1. In eTable 2 the characteristics of the patients who had median OS less than 2 years, which suggest that these patients may have not benefited from surgery versus chemotherapy alone, are presented.

TABLE 1.

Baseline patient characteristics

Variables	All cohort (n = 1661)
	No.	%
Patient characteristics
Age, years, median (IQR)	62	(53–69)
Male gender	1032	62.1
Primary CRC characteristics
Primary CRC site (right/left/rectum)	407/607/552	26.0/38.8/35.2
Primary CRC T stage (T3 or T4)	1373	86.4
Primary CRC nodal metastases, positive	1034	63.6
Preoperative factors
Disease-free interval (< 12 months)	1002	63.9
Chemotherapy for liver disease, yes	988	59.6
No. of cycles > 7	263	17.3
Preoperative CEA, median (IQR)	7.5	(3.3–25.2)
CEA > 20	399	28.8
CEA > 200	81	5.8
Extrahepatic disease at the time of surgery	189	11.4
Tumor factors
No. of CRLM, median (IQR)	2	(1–3)
Size of largest CRLM, cm, median (IQR)	2.9	(1.8–4.5)
Tumor burden score, median (IQR)	3.6	(2.5–5.7)
Genetic information
KRAS mutation, total	573	34.3
Low risk	316	19.0
High risk	253	15.2
BRAF mutation	127	7.6
Operative factors
Concurrent ablation, yes	197	13.1
Extent of resection, major*	358	22.4
Resection margin, R1	300	18.4
Postoperative factors
Postoperative chemotherapy, yes	811	51.6

CEA carcinoembryonic antigen, CRC colorectal cancer, CRLM colorectal liver metastases, IQR interquartile range, R1 resection margin < 1 mm in pathology specimen; Postoperative chemotherapy, postoperative chemotherapy received within 3 months of resection

^*Major resection was defined as resection of three or more liver segments

FIG. 2.

Overall survival after CRLM resection of patients with wild-type tumors, with tumors that harbor low-risk KRAS point mutations, with tumors that harbor high-risk KRAS point mutations, and with tumors that harbor BRAF mutations

First Study

The study by Passot et al. from MD Anderson Cancer Center (MDACC) proposed that patients with KRAS mutation, primary tumor lymph node metastasis, largest colorectal liver metastasis > 3 cm, and > 7 cycles of preoperative chemotherapy should not be offered surgery on the basis of very poor outcomes.¹⁸

In this validation, the high-risk group had a median and 5-year OS of 30.0 months and 16.9%, respectively (Fig. 3A, eTable 3). When we stratified the high-risk group to patients with high- versus low-risk KRAS mutations (instead of any KRAS mutation), median OS was worse among patients with high-risk KRAS mutations (26.2 versus 31.8 months; p = 0.02) (Fig. 3E, eTable 3). Of note, no patient with a high-risk KRAS mutation survived beyond 30 months. Outcomes were even worse among patients with a BRAF mutation who fulfilled the Passot clinical criteria (node-positive primary tumor, largest colorectal liver metastasis > 3 cm, and > 7 cycles of preoperative chemotherapy); median OS was only 16.0 months and all patients died within 36 months (Fig. 3E, eTable 3).

FIG. 3.

Overall survival after CRLM resection stratified by the recommendations by Passot et al.

Second Study

The study by Margonis et al from Johns Hopkins Hospital (JHH) suggested that the Genetic And Morphological Evaluation (GAME) score may aid in treatment selection.¹⁹ The high-risk group according to the GAME score included patients with a score of four to six. The score was calculated as follows: KRAS mutated tumors: one point; carcinoembryonic antigen (CEA) ≥ 20 ng/mL: one point; primary tumor lymph node metastasis: one point; 3 ≤ TBS < 9: one point, TBS ≥ 9: two points; extrahepatic disease: two points.

In the validation cohort, the high-risk group had a median and 5-year OS of 24.3 months and 21.6%, respectively (Fig. 4A, eTable 4). Patients with high- versus low-risk KRAS mutations (instead of any KRAS mutation) had comparable median survival (22.3 versus 24.9 months; p = 0.31) (Fig. 4D, eTable 4). However, outcomes did worsen among patients with a BRAF mutation who fulfilled the GAME clinical criteria; median OS was only 6.2 months (compared with 22.3 months for the KRAS high-risk group; p < 0.01) and all patients died within 9.6 months (Fig. 4D, eTable 4).

FIG. 4.

Overall survival after CRLM resection stratified by the recommendations by Margonis et al.

Third Study

The study by Brudvik et al. from MDACC suggested that modified clinical risk score (m-CRS) can be used to preoperatively assess the potential benefits of surgery and help patients make personalized decisions based on the risks and benefits of any proposed intervention.²⁰ m-CRS is calculated as follows—KRAS mutation: one point; positive primary tumor lymph node status: one point; diameter of the largest liver metastasis > 50 mm: one point.

In the validation cohort, the median OS of the high-risk group was 46.8 months, and 5-year OS was 44.4% (Fig. 5A, eTable 5). When we stratified the high-risk group to patients with high- versus low-risk KRAS mutations (instead of any KRAS mutation), 5-year OS did not decrease (44.9% versus 50.5%; p = 0.97) while median OS was not reached (Fig. 5D, eTable 5). Outcomes did worsen (though without reaching statistical significance), when we evaluated patients with a BRAF mutation who fulfilled the m-CRS criteria; median OS was 20.9 months for this patient group months (compared with 46.8 months for the low-risk group; p = 0.26) but there were patients with a BRAF mutation who survived for at least 5 years (Fig. 5D, eTable 5).

FIG. 5.

Overall survival after CRLM resection stratified by the recommendations by Brudvik et al.

Fourth Study

The study by Gagniere et al. from Memorial Sloan Kettering Cancer Center (MSKCC) suggested that patients with BRAF-mutated CRLM and at least two of the following factors (node-positive primary tumor, CEA level > 200 mg/L, and clinical risk score of four) may not benefit from surgery.²¹ Due to the small number of patients with three risk factors and a somatic mutation (only six patients), we combined those with two or three risk factors into a single category for analysis.

In this validation cohort, median and 5-year OS among patients with BRAF-mutated CRLM and one factor (primary tumor lymph node metastasis, CEA level > 200 ng/mL and clinical risk score ≥ 4) was 26.8 months and 20.2%, respectively (Fig. 6C, eTable 6). The median and 5-year OS for individuals with two or three factors was significantly worse at 10.4 months and 0% and all patients died within 21.0 months (Fig. 6D, eTable 6).

FIG. 6.

Overall survival after CRLM resection stratified by the recommendations by Gagniere et al.

DISCUSSION

To the best of our knowledge, this is the first study to externally validate combinations of biomarkers and clinical factors that have been proposed to identify patients for whom surgery for CRLM has questionable value. The search for such combinations started several years ago but has intensified as systematic therapies have considerably improved OS. In turn, some investigators have argued that the survival threshold achieved by surgery now has to surpass 2 years to justify invasive intervention rather than chemotherapy alone.²² Thus, it is important to evaluate these combinations of biomarkers and clinical factors that have been proposed to identify patients for whom surgery for CRLM might be futile to see if they can be generalized outside the centers that produced them.²³

In 2017, Passot et al. from MDACC was the first to publish a combination of biomarkers and clinical factors to identify patients who may not benefit from surgery in their article entitled “Is hepatectomy justified for patients with RAS mutant colorectal liver metastases?”. Specifically, the authors reported that the combination of RAS mutation, node-positive primary tumor, largest liver metastasis > 3 cm, and > 7 cycles of preoperative chemotherapy was associated with a median OS of only 22 months, and no patient was alive 4 years after surgery. These patients were 6% of the entire cohort (n = 14). The researchers argued that these outcomes were comparable with those obtained by modern systemic therapies alone and questioned the benefit of hepatectomy in this population.^18,24 Importantly, the authors acknowledged the need for external validation of their recommendations by stating that their suggestions did not “represent a validated recommendation.” In the current study, we validated their recommendation and noted that the 18 (1.4%) patients with the afore-mentioned combination had a slightly higher median OS at 30 months, while 5-year OS was 16.9%. Although these outcomes seem to justify surgery in this group, it is possible that there is a subset of patients that fares worse. A previous study by our own group demonstrated that this subset may be patients with high-risk KRAS point mutations. Indeed, when we restricted analysis to only those patients, median OS decreased to 26 months and all patients died within 2.5 years from surgery. Outcomes were even worse among patients with BRAF mutations who fulfilled the Passot clinical criteria; median OS was only 16 months, and all patients died within 3 years.

In comparison, the updated analysis of the TRIplet plus BEvacizumab (TRIBE) randomized controlled trial (RCT) reported a median OS of 25.6 months and 5-year survivors for the group of patients with unresectable RAS-mutated CRLM.²⁵ Of note, these superior outcomes were achieved despite the fact that medically treated patients had several unfavorable characteristics. For example, around 80% had concurrent extrahepatic disease, and 20% did not undergo resection of the primary CRC. The updated analysis of the TRIBE trial also reported on the OS of patients with unresectable BRAF-mutated CRLM. Specifically, this group had a median OS of 13.4 months and there were survivors up to 42 months of follow up. Thus, it is plausible that the combination of the clinical factors proposed by Passot et al. and high-risk KRAS point mutations or BRAF mutations form a group of patients for whom surgery has questionable value.

The next study suggesting a combination of biomarkers and clinical factors to identify patients who may not benefit from surgery was conducted by Margonis et al from JHH in 2018. The researchers devised a hybrid clinical and genetic risk score (GAME score) and reported that the high-risk group (score of four to six) had a median OS of 30 months and a 5-year OS of 11%.¹⁹ These high-risk patients comprised 11.4% of the population of the original study. The researchers acknowledged the limitation of the small sample size and suggested that future studies were needed to investigate whether the GAME score can be used to aid patient selection for surgery. In the current study, we noted that the 191 (16.3%) of patients with this combination of risk factors had a median OS of 24.3 months, while 5-year OS was 21.6%. Of note, when we restricted the analysis to patients with high-risk KRAS point mutations, median OS was 22.3 months, and 5-year OS was 15.1%. Outcomes were even worse among patients with BRAF mutations who fulfilled the GAME clinical criteria; median OS was only 6.2 months, and all patients died within 10 months. Similar to the Passot study, and in light of the TRIBE RCT, the value of surgery may be limited among patients with BRAF mutations who fulfilled the GAME clinical criteria.

Around the same time, Brudvik and colleagues from MDACC devised a similar hybrid clinical and genetic risk score (m-CRS) and reported that the combination of RAS mutation, primary tumor lymph node metastasis, and largest liver metastasis > 5 cm was associated with a median OS of only 15 months, although some patients were still alive 5 years after surgery (numbers at risk were not provided by the authors).²⁰ These patients were only 2.5% of the entire cohort (n = 14). Similar to the GAME score study, Brudvik et al. externally validated their findings in an international multicenter cohort. Interestingly, median OS was much better on external validation at 30 months, and there were also 5-year survivors. The researchers acknowledged that their score was not “designed to make decisions on eligibility for resection” and that even in its high-risk group, longterm survival was feasible. In this study, we validated their recommendation and found that the 41 (2.8%) patients with the described combination of risk factors had a high median OS of 46.8 months while 5-year OS was 44.4%. Of note, when we restricted the analysis to patients with high-risk KRAS point mutations, 5-year OS was still high at 50.5%. Although median OS decreased to 20.9 months when we evaluated patients with BRAF mutations who fulfilled the m-CRS clinical criteria, this was still higher than what the TRIBE trial reported for patients with unresectable BRAF-mutated disease. Thus, m-CRS should not be used to identify patients who will derive questionable benefit from surgery.

The last study we validated was by Gagniere et al. from MSKCC in 2019. Unlike the other three studies, the researchers used BRAF instead of KRAS, as the former is believed to be a more potent biomarker. As the title of the study (Is Hepatectomy Justified for BRAF Mutant Colorectal Liver Metastases?) indicated, the authors investigated whether the combination of BRAF and clinical factors can identify a subset of patients with unfavorable outcomes.²¹ BRAF mutation combined with at least two of the factors was associated with worse OS (node-positive primary tumor, CEA level > 200 μg/L, and clinical risk score ≥ 4), as well as a low median OS at 13 months. Although concluding that “the presence of a BRAF mutation does not preclude long-term survival,” the authors did not present Kaplan Meier plots for the high-risk group; thus, it is possible that the existence of 5-year survivors applies only to the average BRAF patient and not one with the high-risk combination. In this study, median OS for patients with BRAF-mutated CRLM and one of the three factors (node-positive primary tumor, clinical risk score equal or greater than 4 and CEA level > 200mg/L) was 26.8 months, and 5-year OS was 20.2%. The median and 5-year OS for individuals with two or three factors was only 10.4 months and 0%. Thus, the latter may represent a group for whom surgery is futile.

This validation study is limited by its retrospective design. Furthermore, the reported outcomes for best medical treatment apply to the average patient with unresectable disease and thus may overestimate outcomes for high-risk patients. In addition, the number of patients in the high-risk groups for studies 1, 3, and 4 was small. However, this is not a limitation of this study cohort as it is somewhat expected that only a few patients with resectable CRLM may not benefit from surgery. Lastly, we were unable to validate the combinations of biomarkers and clinical factors proposed in some otherwise eligible studies due to the lack of relevant data in our multi-institutional dataset. Future studies should address these gaps.

It is well known that despite their prognostic value, biomarkers such as KRAS, BRAF or even the triple RAS/TP53/SMAD4 mutation are not associated with that poor outcome that could be used alone to select patients for surgery.^26,27 A few studies have investigated whether a combination of these biomarkers with clinical factors could be used to find groups with particularly poor outcomes and in turn potentially recommend no surgery in these patients. The current study validated these recommendations for the first time and found that the use of overall KRAS mutation status in combination with the clinical factors proposed in studies 1, 2 and 3 to indicate that surgery is futile does not appear justified. However, patients with high-risk KRAS mutations who fulfilled the Passot clinical criteria (study 1) may represent a group for whom surgery is futile, as there were no 5-year survivors and overall survival was similar to that of historic cohorts with similar mutational profiles that received systemic therapies alone for unresectable disease. Of note, before these recommendations can be used to guide precision surgery in CRLM, prospective validation is needed. Additionally, the prognostic value of high-risk KRAS point mutations must be further investigated in studies involving patients with unresectable disease. Similarly, surgery may be futile for patients with BRAF mutations who fulfilled the GAME clinical criteria (study 2) as well as those with a BRAF mutation who fulfilled the Gagniere criteria (study 4), as there were no 5-year survivors and overall survival was even lower than that of historic cohorts with similar mutational profiles that received systemic therapies alone for unresectable disease.

Collectively, our validation study suggests that the following combinations of biomarkers and clinical factors could potentially be used to identify patients who may not benefit from surgery: 1) patients with high risk KRAS or BRAF mutations, primary tumor lymph node metastasis, a colorectal liver metastasis > 3 cm, and > 7 cycles of preoperative chemotherapy. 2) Patients with BRAF-mutated tumor and a GAME score of four to six (CEA ≥ 20 ng/mL: one point; primary tumor lymph node metastasis: one point; 3 ≤ TBS < 9: one point, TBS ≥ 9: two points; extrahepatic disease: two points). 3) Patients with BRAF-mutated CRLM and at least two of the following factors: node-positive primary tumor, CEA level > 200 ng/mL, and clinical risk score of four.

Supplementary Material

SUPPLEMENTARY INFORMATION The online version contains supplementary material available at https://doi.org/10.1245/s10434-024-16319-0.