Prognostic Impact of Tumor Status, Nodal Status and Tumor Sidedness in Metastatic Colon Cancer

Background Locally advanced primary tumors have been associated with poor overall survival (OS) in non-metastatic colon cancer. However, their impact on metastatic colon cancer (mCC) is not fully defined. The association between primary tumor location and prognosis in mCC is also evolving. Methods Using National Cancer Data Base, we identified a cohort of 25,377 patients diagnosed with mCC from 2004-2009. Chi-square test was used for descriptive analyses, while all potential prognostic factors were evaluated using Kaplan-Meier survival estimates and Cox proportional hazards regression modeling. Results The five-year OS for the entire study cohort was 12.3%. Factors associated with significant survival impact in multivariate analysis included age, gender, race, comorbidity index, academic level of treating institution, insurance status, income, year of diagnosis, primary tumor site, histologic differentiation, pathologic tumor stage (pT), pathologic nodal stage (pN), and modality of chemotherapy. pT1 lesions demonstrated poor prognosis in stage IV colon cancers, not statistically different when compared to survival outcomes observed in cases with pT4 lesions. Regional nodal involvement demonstrated poor OS in full cohort analysis and subgroup analysis independent of primary tumor location. Both right-sided and transverse colon tumors had similarly worse OS compared to left-sided tumors (right-sided: HR: 1.21, 95% CI: 1.17-1.25; transverse: HR: 1.21, 95% CI: 1.15-1.27). Conclusions T1 lesions arising from right-side or transverse colon portend a poor prognosis in mCC, while regional lymph node involvement by itself is an independent poor prognostic factor. Right-sided tumors are associated with poor outcomes than left-sided tumors, suggesting the role of underlying molecular or biologic variants.


Introduction
Epidemiological reports have always combined colon and rectal cancers for calculation of incidence and mortality rates. According to the most recent cancer statistics data published by the American Cancer Society's Surveillance and Health Services Research Program, colorectal cancer is the third leading cause of cancer both among men and women in the United States [1]. In 2020, an estimated 147,950 new colorectal cancer cases will be diagnosed in the United States, with an estimated 53,200 deaths. It is now the second leading cause of cancer-related deaths in men and women combined. As of January 2019, there will be an estimated 1.5 million people in the U.S. diagnosed with colorectal cancer [2].
A small proportion of metastatic colon cancer patients with limited metastases confined to liver and/or lungs undergo resection and systemic treatment with curative intent, while the rest are treated with palliative treatment modalities, including systemic therapy with or without palliative resection [3,4]. Sometimes, radiofrequency ablation or external beam radiation is also utilized to address oligometastatic disease in the liver. Several small registries and single institution-based studies have looked into the impact of various clinic-pathologic features on disease prognosis in stage IV colon cancer [5][6][7][8]. Few of these studies also evaluated the prognostic impact of primary tumor resection.
The current study was undertaken to assess the prognostic impact of various patient-, tumor-and treatment-specific factors in stage IV colon cancer patients who underwent palliative resection of the primary tumor, utilizing a large population-based cohort extracted from the National Cancer Data Base (NCDB). This study cohort included patients treated with palliative intent only while excluding those who underwent metastasectomy. Our study objective was to identify the prognostic impact of tumor status (T), nodal involvement (N), and location in stage IV colon cancers, managed with palliative intent.

Data source
The NCDB is a joint quality improvement initiative of the American College of Surgeons' Commission on Cancer and the American Cancer Society. It contains 34 million patient records and represents approximately 70% of all newly diagnosed cancer cases in the United States [9]. Data was provided through Participant User File (PUF) by the NCDB after a thorough review of the application. The PUF included patients diagnosed with colon cancer from 2004 to 2014.

Study population
Patients diagnosed with stage IV colon adenocarcinoma from 2004 to 2009, who underwent palliative resection of the primary tumor, were identified from the NCDB. Patients undergoing metastasectomy were excluded from the study population, and those cases with missing or unknown data were also excluded from the analysis. Using the American Joint Committee on Cancer's (AJCC) Collaborative Stage Data Collection System (CCS), we excluded patients with appendiceal and overlapping adenocarcinomas. A final cohort of 25,377 patients was identified for analysis. Only cases with pathologic confirmation were included in the final cohort. As a cut-off point, the year 2009 was chosen to ensure a minimum follow-up of five years for all patients included in the analysis.

Outcomes and variables
Five-year overall survival was chosen as the primary endpoint for the study. Censoring was performed at the last recorded follow-up of 60 months. The variables of interest were age, gender, race, Charlson/Deyo comorbidity score, academic level of treating institution, geographic region, rural-urban continuum, insurance status, median family income, year of diagnosis, location of the primary tumor, differentiation of tumor, histologic subtypes of adenocarcinoma, pathologic AJCC tumor status (T), pathologic AJCC nodal status (N), chemotherapy, surgical resection of primary tumor and receipt of palliative external beam radiation.
Age was recoded into ordinal groups of 18-64 years, 65-74 years, and over 75 years. Charlson/Deyo comorbidity index was used to stratify the study population into three different relative risk categories: (i) patients with score zero, (ii) those with a score of one and, (iii) patients with a score of two or more, with zero indicating no comorbidities at baseline. Tumor locations were divided into three categories: left-sided, right-sided, or transverse colon tumors. Median family income was analyzed as four quartiles representing income levels ranging from less than $38,000 to more than $63,000. Tumors were graded into welldifferentiated, moderately differentiated, and undifferentiated tumors. Based on the histology of the tumor, cases were classified into mucinous or non-mucinous adenocarcinomas. A tumor (T), nodal (N), and metastasis (M) staging was based on AJCC, 6th edition.

Statistical analysis
Statistical analysis was performed using the Statistical Analysis Software (SAS, version 9.4; SAS Institute, Cary, NC). Kaplan-Meier survival analysis was utilized to generate adjusted survival curves and compared using the log-rank test. We conducted this part of the analysis using the SAS procedure PROC GPLOT. Cox proportional hazards regression model was used to perform multivariate analysis to assess the influence of the above-mentioned variables by generating hazards ratios (HR) and 95% confidence intervals (CI), with an HR less than 1.0 indicating survival benefit. Multivariate regression analysis was completed using SAS procedure PROC PHREG. Differences were considered statistically significant when the p-value was <0.05.

Patient characteristics
A cohort of 25,377 stage IV colon cancer patients diagnosed from 2004 to 2009, containing information on all the variables of interest, was identified from the NCDB. Patient demographics and clinicopathologic features of the entire study cohort are demonstrated in Table 1. The five-year overall survival rate for this cohort was 12.3%, which is similar to the national average for stage IV colon cancer [10]. Approximately, 50% of the study population was less than 65 years old and Caucasians had majority representation among various ethnic groups (77%). Most of the patients had no baseline comorbidities (73.7%) and almost threefourths of the study population was treated at non-academic centers. With regards to the rural-urban continuum, almost 85% of patients were from metropolitan areas and 95.6% of the entire study cohort had some kind of health insurance. Right-sided tumors constituted almost half of the patient population (49.6%) while mucinous adenocarcinoma represented only 11.5% of all tumors. T1 tumors formed a very small minority (0.6%) of the entire study cohort, which was otherwise predominantly represented by T3 tumors (62%). More than half of them had N2 disease. Approximately 60% of the study population underwent subtotal colectomy. Palliative external beam radiation was administered in only 2.7% of patients and more than half of the study cohort received multi-agent chemotherapy.

Prognostic impact of T and N by tumor location
Subgroup analysis by tumor location was performed to further delineate into the prognostic differences seen among various T and N stages (

sidedness, primary tumor status (T) and nodal status (N)
Variables analyzed as part of multivariate analysis include age, gender, race, Charlson/Deyo comorbidity score, academic level of treating institution, region, location, insurance status, median family income, year of diagnosis, histology, tumor differentiation, chemotherapy, palliative radiation and surgery.

Discussion
We evaluated metastatic colon cancer patients to identify factors that have a prognostic impact on overall survival. This is the first study using NCDB data to systematically evaluate the prognostic implications of various patient-, disease-and treatment-specific factors on overall survival in stage IV colon cancer. Previous studies have demonstrated the survival benefit of palliative tumor resection in stage IV colorectal cancers [11][12][13]. Therefore, only patients undergoing primary tumor resection were included in our study cohort.
Multivariate analysis of the full cohort demonstrated patients with pT1 lesions had survival outcomes that were not different from those with pT4 lesions (HR: 0.83; 95% CI: 0.69-1.00). However, earlier studies reported that higher T stage was independently associated with poor survival compared to lower T staging [6,8]. We then performed additional sub-group analysis based on primary tumor location. Multivariate analysis of subgroups based on tumor location showed pT4 and pT1 lesions to have no significant difference in survival outcome both in the right-sided and transverse colon cancers (right-sided: HR: 0.71; 95% CI: 0.54-0.93; transverse: HR: 1.08; 95% CI: 0.4-2.95). Among cancers originating from the left colon, pT1 was associated with better overall survival compared to pT3 and pT4 lesions but performed poorly compared to pT2 lesions ( Table 3). In the full cohort as well as subgroup analysis, pT2 was independently associated with the best survival outcomes. We further looked into the distribution of cases according to nodal status stratified by tumor stage ( Table 4) but could not identify any significant association between pT and pN to explain this intriguing observation. We could hypothesize that T1 lesions may represent a biologically distinct tumor that is more aggressive with unique defects in various molecular and biologic pathways that were not accounted for in this retrospective study.  Nodal involvement was found to be an independent predictor of poor survival. Multivariate survival analysis showed that patients with N1 or N2 disease performed poorly compared to those with no nodal involvement (N0: HR: 0.55, 95% CI: 0.53-0.58). Median survival in patients with N0 disease was 25.5 months as compared to 21.1 months with N1 and 14.4 months in N2 nodal involvement. This difference remained significant independent of primary tumor location, as confirmed in the subgroup analysis ( Table 3). These results are consistent with the findings reported in previous retrospective studies [8,[14][15][16][17][18][19][20]. The proportion of patients with N0 disease in our study was 16%, which is similar to the frequency of 15-20% that was reported in these earlier studies. Since regional lymph node metastasis is considered an early sign of distant metastasis, the presence of 15-20% of node-negative cases in stage IV colon cancers remains quite inexplicable. There is paucity in the literature about various biological mediators related to lymphatic spread in colorectal cancers. Kawada et al. demonstrated the chemokine receptor CXCR-3 positive colon cancers to have poor survival than those without CXCR-3 [21]. It was postulated that patients lacking biologic mediators, like CXCR-3 related to lymphatic spread, may have tumor with less aggressive traits and carry a favorable prognosis.
Location of primary tumor also proved to be an independent factor of overall survival in our study, with right-sided colon cancer showing significantly worse survival compared to left-sided colon cancer (HR: 1.21; 95% CI: 1.17-1.25). Previous studies looking into primary tumor location and outcomes have reported discordant results [22]. However, these studies included a heterogeneous population with different stages receiving non-uniform treatments. According to Surveillance, Epidemiology and End Results (SEER) database analysis by Meguid et al., the prognosis of right-sided tumors was not different from that of leftsided tumors in stage I cancer, while it was shown to be better in stage II cancers and worse in stage III and IV disease [22]. A more recent SEER database analyses by Schrag et al. reported right-sided tumors to have inferior survival in patients with stage III and IV colon cancer after adjusting for various clinical and demographic characteristics [23].
In the Cancer and Leukemia Group B/South West Oncology Group (CALGB/SWOG) 80405 analysis by Venook et al., tumor sidedness was shown to have both prognostic and predictive value in metastatic colon cancer [24]. In this analysis, patients with left-sided Kirsten rat sarcoma viral oncogene homolog (KRAS) wild-type disease were found to have better overall survival and progression-free survival than those with KRAS wildtype right-sided cancer. In addition, survival outcomes favored treatment with bevacizumab than cetuximab in patients with right-sided primary tumors. A retrospective analysis of rat sarcoma viral oncogene (RAS) wild-type metastatic colorectal cancer patients from Cetuximab Combined With Irinotecan in First-Line Therapy for Metastatic Colorectal Cancer (CRYSTAL) and FOLFIRI Plus Cetuximab Versus FOLFIRI Plus Bevacizumab as First-Line Treatment of KRAS Wild-Type Metastatic Colorectal Cancer (FIRE-3) reported left-sided cancers to have a better prognosis in terms of survival and response rates, irrespective of treatment [25].
Right-sided colon cancers tend to be diagnosed much later than left-sided colon cancers. This clinical observation highlights the tendency for right-sided colon cancers do not produce symptoms until they are relatively advanced with larger tumor size, invasion to adjacent organs, the involvement of regional lymph nodes, and multiple organ metastases. This difference in presentation time-line has been argued to result in a "lead-time" bias for left-sided tumors that may make survival appear longer. All of these prior studies that looked into left-sided vs. right-sided tumors alone when assessing for the difference in survival. However, it is known that during embryonic development, the right side of the colon and proximal two-thirds of the transverse colon develops from the midgut while the left side of the colon and distal one-third of the transverse colon originate from the hindgut. Our study analyzed the site of the primary tumor as originating from either left, right, or transverse colon. By doing this, we were able to show the close similarity between the survival outcomes of right-sided and transverse colon tumors (right-sided Right-sided colon cancers are characterized by a high frequency of microsatellite instability (MSI) and BRAF mutation [26][27][28][29]. BRAF mutation has been shown to carry a poor prognosis. MSI has been shown to impart favorable prognosis in curatively resected stage II colon cancers but is associated with poorer survival in patients with stage IV colorectal cancers. Guinney et al. looked into various gene expression arrays of colorectal cancers and came up with four well-defined consensus molecular subtypes (CMS), which reflect the way colorectal cancer behaves biologically [30]. CMS1 and CMS3 are associated with poor prognosis, and these two subtypes were more likely to be present in the right side of the colon.
There are some potential limitations to our current study. First, information on MSI, BRAF mutational status, or any other molecular correlates was lacking. Although tumor sidedness could be considered a surrogate for these high-risk biological subtypes, this could not be validated in the current study cohort. Second, this is a non-randomized, retrospective analysis that could carry an inherent selection bias. However, we tried to minimize heterogeneity by including only stage IV colon cancer patients who underwent palliative resection and excluding cases undergoing curative metastasectomy. Finally, NCDB does not collect information about disease-specific mortality, and the overall survival outcomes discussed in the current study might have been influenced by secondary competing mortality.
Despite these limitations, the current analysis is the first and largest population-based analysis of metastatic colon cancer patients, utilizing the NCDB.

Conclusions
Using a large national database, we demonstrated pT1 lesions to carry a poor prognosis in stage IV colon cancers, not statistically different when compared to survival outcomes observed in cases with pT4 lesions. This observation was validated in the right-sided and transverse colon only subgroups, highlighting the impact of various underlying distinct biological subtypes. Regional lymph node involvement was associated with poorer overall survival in metastatic colon cancer patients undergoing palliative resection of the primary tumor. Most importantly, right-sided tumors demonstrated worse prognosis than left-sided tumors. This further strengthens the idea of tumor-sidedness being a surrogate marker for underlying high-risk biological or molecular subtypes, which needs additional prospective validation in clinical trials.