Prognostic effect of sidedness in early stage versus advanced colon cancer

Abstract Background and Aims The prognostic effect of sidedness in colorectal cancer has been evaluated in numerous prospective and retrospective cohorts. Most of these have reported overall survival data; there is scant relapse‐free survival data in early stage disease. This study aimed to determine the effect of tumor sidedness in survival in early stage and relapsed colon cancer. Methods Patients with stage I‐III colorectal cancer were identified from the BC Cancer Agency Gastrointestinal Cancer Outcomes Unit. Survival analysis by stage and sidedness was compared with the log‐rank test. Baseline characteristics were controlled by multivariate Cox‐proportional hazard models. In relapsed patients, bevacizumab and EGFR inhibitor (EGFRI) treatments were included and tested for interaction. Results Among 5378 patients with stage I‐III colon cancer, patients with right‐sided stage II tumors experienced better relapse‐free survival compared with those with left‐sided tumors; right‐sidedness was not prognostic for RFS in stage III disease. When survival was considered in patients who relapsed, right‐sided tumors had inferior survival after relapse in both stage II and stage III tumors. At relapse, right‐sided outcomes were inferior regardless of biologic therapy. An interaction test revealed a significant association between sidedness and survival with EGFRIs. Conclusions In this large, population‐based study, right‐sided presentation has a significant prognostic impact: in early stage, right‐sidedness is favorably prognostic among stage II tumors and not prognostic in stage III disease. After relapse, right‐ sidedness is associated with an inferior prognosis, regardless of initial stage of presentation. Colon tumor sidedness is independently prognostic and may be considered in treatment assignment for both early stage and advanced disease.


| BACKGROUND
Multiple studies have demonstrated differences between right and left-sided colon cancer (CCa) in terms of clinical characteristics, pathology, and prognosis. Patients diagnosed with right-sided CCa are more likely to be female, older, and have less acute presentation and more locally advanced tumors. 1,2 Right-sided tumors are characterized by a higher rate of poor differentiation and genetic changes, including microsatellite instability (MSI), BRAF, and hypermethylation (CIMP). 3,4 While the prognostic effect of sidedness has been evaluated in numerous prospective and retrospective cohorts, 1,5,6 most studies have reported overall survival (OS) data only, 1,2,5,7,8  and showing statistical significance, when compared with left-sided advanced disease. 2,7,11 The prognostic value of sidedness is much less conclusive in the early stage I-III setting. In a meta-analysis of 1 437 846 patients, left CCa conferred a favorable pooled prognostic effect on OS [HR 0.82 (95% CI, 0.79-0.84) P < 0.001], and this effect was independent of stage. 12 By contrast, 3 studies of the population-based SEER database showed that right-sidedness confers a favorable prognostic effect in stage II CCa (HRs 0.91, 0.92, and 0.89), and only right-sided stage III tumors had inferior OS (HRs 1.06, 1.12, 1.12). 5,7,8 A limitation of these studies is that they did not include data on Relapse Free Survival (RFS) or more extensive prognostic factors. 2,5,8,13,14 This is illustrated by a SEER propensity score-matched analysis that found no difference in OS between right and left sided stage III tumors. 15 The data points to a dichotomous effect of sidedness, with right-sided tumors faring slightly better in early, stage II presentation, but conferring a substantially inferior prognosis among patients with relapsed or stage IV presentation. Very recent studies have shed more light on this and failed to demonstrate a prognostic effect of sidedness in stage III for DFS. 16,17 The effect of sidedness in stage III disease is unclear and needs to be better described if sidedness is to be used as a prognostic factor.
In this study, our hypothesis was that the prognostic effect of tumor sidedness is different in early-stage versus relapsed colorectal cancer, and the endpoints chosen were Relapse Free Survival (RFS) and Survival after Relapse (SAR). We reviewed demographic and prospectively collected outcome information from patients enrolled in the

| Description of the study population
The BCCA is a province-wide agency that provides comprehensive cancer care, including prevention, screening, diagnosis, and treatment to the residents of British Columbia, Canada. All systemic therapy is centrally reimbursed and documented in the Provincial Chemotherapy Database. The BCCA GICOU prospectively compiles demographic, diagnostic, treatment, and outcome data for all patients referred to the BCCA with gastrointestinal malignancies. Consent is obtained from all patients referred to BCCA for treatment delivery and to prospectively follow outcomes. An active follow-up program including annual letters to primary care providers after discharge results in an estimated loss to follow-up of <5%.
The GICOU database was used to identify all patients diagnosed with stage I, II, or III CCa in 1990CCa in , 1995CCa in , 1996CCa in , and 1999CCa in -2009, and referred to any one of the BCCA centers for treatment of newly diag-

| Definitions of dependent and independent variables
The functional status of patients was characterized with Eastern Cooperative Oncology Group (ECOG) performance status and was ascertained from the patient's records within 1 month of referral.
Staging was based on the AJCC 6th Edition Cancer Staging Manual. 18 Patients were deemed to have undergone a surgical resection if they had a surgery performed with curative intent and without gross (R2) residual disease. Patients were categorized as having received adjuvant chemotherapy if they completed more than 1 cycle of fluoropyrimidine-based chemotherapy following curative resection.
Information about whether patients with relapsed disease were treated with bevacizumab, cetuximab, and/or panitumumab was obtained from the BCCA Provincial Chemotherapy Database.

| Statistical analysis
Baseline characteristics of the patients were summarized by sidedness. Categorical and continuous characteristics were compared, respectively, using Chi-square test and Wilcoxon-rank sum test.
Kaplan Meier survival analysis was performed by stage, and the impact of sidedness compared with the log-rank test. The patients are censored at 10-year of follow-up when RFS were calculated. RFS was measured from date of surgery until local, regional, or distant recurrence, and censored at death or last follow-up or 10-year if followup was longer than 10 years. In the subset of patients who relapsed, SAR was measured from the documented date of local, regional or distant relapse, until death or last follow-up. Cox-proportional hazard models were used to control for baseline clinicopathologic characteristics, including age at diagnosis, sex, grade, lymph node sampling,

| Patient demographics.
A total of 5358 patients with stage I-III resected CCa, diagnosed between 1990 and 2009, were identified, and baseline characteristics described (

| Overall survival after relapse (SAR)
Overall survival analyses were conducted among the patients in the original cohort who developed a local, regional, or distant relapse (Tables 2 and 4

| Limitations
The major limitation of this study, which is one shared with numerous

| CONCLUSIONS
The study shows that right-sidedness confers a favorable prognosis to stage II CCa but is not prognostic of relapse in stage III colon cancer.
Right-sidedness is a strong adverse prognostic factor among patients who relapse and may influence the selection of therapy. While more accurate biomarkers are awaited, sidedness should be considered as a relevant stratification factor in studies of both early and advanced disease, as imbalances between study arms can significantly affect study outcomes. Previously validated prognostic variables remain relevant and should continue to be incorporated in multivariate analysis and reflect the complexity of CCa disease biology.

AUTHOR CONTRIBUTIONS
All of the authors contributed to the conception, data analysis, and writing of this paper.