The Association of KRAS Mutation With Primary Tumor Location and Survival in Patients Undergoing Resection of Colorectal Cancers and Synchronous Liver Metastases

Niccolo Allievi; Paolo Goffredo; Alan F. Utria; Michele Pisano; Elia Poiasina; Alessandro Lucianetti; Paige Zhou; Imran Hassan


Chin Clin Oncol. 2019;8(5) 

In This Article


The present study investigated the associations between KRAS mutational status, primary tumor location and survival in a cohort of 806 patients from a population-based dataset. African-American ethnicity and right-sided lesions were independently associated with m-KRAS. Moreover, m-KRAS and the presence of a right-sided primary lesion were negative prognostic factors in patients with colorectal cancer and synchronous liver metastases undergoing surgical resection of both their primary colorectal cancer and liver metastases.

Given its relatively high incidence, the relative ease of detectability, and the various targeted therapeutic options available,[3,4] m-KRAS is an important prognostic consideration in the management of metastatic colorectal cancer. In the context of multi-disciplinary strategies, the decision-making process is often driven by clinical and radiological features, rather than pathologic and biologic elements. However, the mutational status of KRAS may predict the responsiveness to upfront chemotherapy in presence colorectal cancer and synchronous liver metastases[15] and may also explain intrinsic aggressive tumor biology,[16,17] resulting in increased risk of progression and relapse after resection. m-KRAS is acquired early in the mutational cascade, it remains stable over the course of the disease, and has a high concordance between primary tumor and liver metastases.[18] Several preoperative clinical risk scores for patients presenting with resectable colorectal cancer and synchronous liver metastases were developed[19–22] based on institutional cohorts, although there have been concerns regarding the absence of external validation of these scores as well as the accuracy and applicability of the reported prognostic factors.[23] Therefore, using current knowledge of the mutational profile of colorectal cancer to guide clinical practice could harbor the potential of improving oncologic outcomes. This has led to the creation of the Genetic and Morphological Evaluation (GAME) Score,[24] which is based on morphologic and biologic tumor information and represents the first clinical risk score including the genetic status (KRAS) of the primary tumor. The score was developed using a derivation cohort, including 502 patients, and an external validation cohort (747 patients) and its discriminatory capacity resulted to be superior to other institutionally derived scores. Integrating genetic, biologic and clinical information not only is likely to improve comparison of different cohorts, but may also guide treatment selection and provide relevant prognostic details.[24]

The proportion of patients with m-KRAS in the current study (39.8%) was consistent with previous reports in the literature (14–46%). In an analysis of all Stage IV colorectal cancers within the SEER database, Charlton et al.[25] reported an overall mutation rate of 44% among 6794 patients. Similarly, in their study utilizing the NCDB, Goffredo et al.[10] reported m-KRAS in 42% out of 2,655 patients who underwent surgical treatment of both their primary colorectal cancer and isolated liver metastases. Two meta-analyses of institutional studies by Brudvik et al., that included 14 studies with 1,809 cases, and Tosi et al., which included 11 studies with 1,369 patients, reported an overall m-KRAS incidence of 30.6% and 34.3%, respectively. While there was substantial overlap between these two studies, they differed in that Tosi et al. used a more stringent criteria for inclusion in their analysis and added three newer studies.[6,9] Both these meta-analyses demonstrated that m-KRAS was negatively associated with OS and relapse free survival, irrespective of chemotherapy agent received.

In our cohort, the proportion of patients with a right-sided cancer within the m-KRAS subgroup was high at 54%, as compared to 31% within the wt-KRAS subgroup. In a meta-analysis including 66 studies and 1,437,846 patients, Petrelli et al.[26] reported that right-sided lesions were associated with worse OS. There have been several explanations in the literature for this difference in survival based on location. Firstly, surgical approaches to right-sided and left-sided colon/rectal cancer are different: while the standard of care is well defined for left-sided and rectal lesions, the optimal surgical resection for right-sided lesions remains debated, particularly in regards to the extent of mesocolic excision and lymphadenectomy.[27,28] Secondly, benefits from anti-EGFR agents is less pronounced in right-sided lesions, as reported by two recent trials.[29,30] Finally, right-sided colon cancer is associated with increased incidence of genetic mutations [microsatellite instability (MSI), BRAF and KRAS mutations] which may explain the survival difference in this subgroup of patients.[31]

In the current study, right-sided tumors and African-American ethnicity were significantly associated with the presence of m-KRAS status. Right-sided lesions, but not African-American ethnicity, were associated with lower DSS on multivariable survival analysis. Using the NCDB, Goffredo et al. found comparable results in a U.S. based national cohort of patients,[10] even though the NCDB and SEER databases are different in terms of quality and origin of data. The former collects data from Commission on Cancer-accredited cancer program registries, capturing around 70% of all newly diagnosed cases of cancer at an institutional level; the latter gathers data from population-based cancer registries and covers 34.6% of the U.S. population. Notwithstanding the possible overlapping between the two datasets, the reproducibility of results between these two studies underscores the validity of the association between m-KRAS and survival outcomes. Our findings appear to support the hypothesis that specific tumor biology and mutational status may be related to primary tumor location and could be the main driver for worse outcomes.

The present study has several limitations. Firstly, the retrospective nature of the analysis of the SEER database might be flawed by intrinsic biases of large databases. Secondly, the SEER is a population-based dataset and several potentially significant variables are not collected routinely. Thirdly, the vast range of therapeutic possibilities, including type of surgical resection, various lines of chemotherapy, radiotherapy and their temporal combination, could not be investigated in depth. Finally, the accuracy and extent of surgical treatment as well as other therapeutic maneuvers could not be evaluated.