A Belgian Population-Based Study Reveals Subgroups of Right-sided Colorectal Cancer with a Better Prognosis Compared to Left-sided Cancer

Abstract Background Patients with left-sided colorectal cancer (L-CRC) are known to have a significantly better prognosis than those with right-sided CRC (R-CRC). It has been hypothesized that RAS, BRAF mutations, or deficient mismatch repair status (MMR) might be responsible for the prognostic effect of primary tumor location (PTL). This study aims to evaluate the prognostic effect of PTL in the Belgian population and to determine the role of biomarkers (MMR, BRAF, and RAS status) in this effect. Patients and Methods We performed a retrospective analysis of Belgian Cancer Registry data. First, we studied the prognostic effect of PTL on 5-year relative survival of 91,946 patients diagnosed with CRC (all stages) from 2004-2015. Second, we investigated the interaction between biomarkers and the prognostic effect of PTL in 1818 patients diagnosed with stage IV CRC in 2014-2015. Results L-CRC was associated with a significantly better 5-year relative survival compared to R-CRC in all stages and ages combined (68.4%, 95% CI, 67.7-69.1% vs 65.6%, 95% CI, 64.7-66.4%). Also, when stratified by age, sex, and stage, the prognosis of L-CRC was better compared to R-CRC in most subgroups. Only in stage II and certain subgroups of elderly patients, the opposite was observed. Furthermore, our data showed that none of the biomarkers had a significant interaction with the effect of PTL on survival. Conclusion This population-based study confirms that L-CRC is associated with significantly better relative survival compared to R-CRC, in all stages and ages combined. Furthermore, in stage IV L-CRC is associated with a longer survival than R-CRC, regardless of MMR, RAS, and BRAF status.


Introduction
Colorectal cancer (CRC) is the third most frequently diagnosed cancer worldwide and is ranked second in terms of mortality. 1 During the past years, the impact of primary tumor location in CRC has been intensively studied, demonstrating different disease characteristics in left-and right-sided CRC. 2 The division in the right and left part of the colon is based on its embryological origin. The right part of the colon originates from the midgut, while the left part is derived from the hindgut. The embryological border between both parts of the colon is located at the proximal two-thirds of the transverse colon. 3 However, most researchers use the splenic flexure as the boundary between left-and right-sided tumors. In some studies, tumors of the transverse colon are excluded since classifying them as right-sided is not entirely correct. 4 Furthermore, some authors classify rectal cancers as a third separate group and therefore exclude rectal cancers from leftsided CRC. 5 From an anatomical perspective, blood supplies, innervations, lymphatic drainages, and lumen environments are among the differences between the right and left colon. Bufill et al (1990) were the first to describe CRC based on the anatomical site, in an attempt to sub-classify CRC. 6 Right-sided CRC is associated with female gender, old age, BRAF mutations, and microsatellite instability, while left-sided CRC is associated with male gender, younger patients, and chromosomal instability. Left-sided tumors often cause infiltrating, constricting lesions, causing obstruction, and early symptoms. In contrast, right-sided tumors are mostly exophytic, polypoid lesions, causing anemia, and unfortunately, symptoms only in later stages.
In recent years, various studies have convincingly shown that patients with tumors originating on the left side of the colon have a significantly better prognosis than those with tumors originating on the right side, in all stages. [7][8][9][10][11][12][13][14][15][16] This means that primary tumor location has a prognostic value as it provides information about the overall cancer outcome, independent of treatment received. However, these conclusions are mostly based on retrospective analysis of data from clinical trials and therefore based on data from selected patients, often in small subgroups. Therefore, these findings need to be confirmed in population-based studies, representing an unbiased set of patients.
Furthermore, the underlying mechanisms that cause this difference in survival have not been identified yet. It has been hypothesized that KRAS, NRAS, or BRAF mutations and deficient mismatch repair status (MMR) might be partially responsible for the prognostic effect of primary tumor location.
The aim of this study is to evaluate the prognostic effect of primary tumor location in the Belgian population and to determine the role of biomarker status (MMR, BRAF, and RAS mutational status) in this prognostic effect.

Database and Study Population
In Belgium, data on patient and tumor characteristics of all newly diagnosed cancers are collected in a national cancer registry; the Belgian Cancer Registry (BCR). In this study, a retrospective analysis of data of the BCR was performed. Data collected and followed by the BCR includes demographic characteristics, primary tumor location, stage at diagnosis, and survival.
The stage at diagnosis was obtained from a compilation of pathological (pTNM) and clinical (cTNM) stages. If both pStage and cStage are available, pStage was used for the combined stage. An exception to this rule was cases with clinical metastases (cM = 1): in this case, the combined stage was IV. If either the pathological or the clinical stage only was available, the combined stage was derived from the available stage. If both pStage and cStage were absent, the combined stage was considered unknown.
In the first analysis, the prognostic effect of primary tumor location (PTL) was studied in all patients diagnosed with CRC (all stages) between January 1, 2004 and December 31, 2015. The last date for follow-up was July 1, 2017. Five-year relative survival was selected as outcome, to correct for competing causes of death. The 5-year relative survival rate was combined with the number of patients at risk in each category (according to age, sex, stage, and primary tumor location) to calculate the number of patients surviving and not-surviving. These numbers were used for further statistical analyses.
In a second analysis, the effect of biomarkers on survival and on the prognostic effect of PTL was investigated in a random sample of approximately 2000 patients diagnosed with de novo stage IV CRC in 2014 and 2015. The data entry was closed in July 2017. The same inclusion and exclusion criteria as the first analysis were applied for the second analysis, but only stage IV CRC was studied. To avoid survival being influenced by a second malignancy, we excluded patients who were diagnosed with a second tumor. However, in situ tumors (stage 0) or basal cell carcinomas were allowed.
We reviewed the histopathological reports of all patients included in the second analysis and collected information about mismatch repair status (deficient mismatch repair status (dMMR) vs. proficient mismatch repair status (pMMR)) and BRAF, KRAS, and NRAS mutational status (mutant vs. wild-type). Biomarker data was obtained according to the protocols of the laboratory of the treating hospital. This implies that which laboratory techniques were used, which exons of KRAS and NRAS were assessed and whether mismatch repair testing was done by immunohistochemistry or molecular testing, was decided by the laboratory of the treating hospital.

Statistical Analyses
Statistical analysis was performed by the statistical software package R version 3.3.2. A P-value of < .05 was considered statistically significant.
For the first analysis, we constructed a logistic regression model using the fraction of deceased patients as the dependent variable and location, stage, age, and gender as the independent variables. Across all logistic regression models, 5-year relative survival was entered as outcome variable.
In the second analysis, Cox proportional hazard models were constructed to determine the effect of KRAS, NRAS, BRAF mutational status, and MMR status on survival. Overall survival was defined as the time from diagnosis of CRC to death from any cause. Patients who were either (1) alive at The Oncologist, 2023, Vol. 28, No. 6 e333 the data entry closure date or (2) were lost to follow-up were censored at the last date that they were known to be alive.

Prognostic Effect of Primary Tumor Location
In the first analysis, the prognostic effect of PTL was studied in patients with CRC diagnosed in 2004-2015. This study population included 91,946 patients: 28,359 (30.8%) with right-sided CRC and 63,587 (69.2%) with left-sided or rectal CRC. Table 1 presents the demographic and clinical characteristics of this study population.
When stratified by stage, left-sided CRC (including rectal cancer) was associated with a significantly better survival in stages I, III, and IV compared to right-sided CRC ( Table  2; Fig. 1). Patients with right-sided CRC in these stages had approximately 1.5 times higher odds compared to patients with left-sided CRC (stage I, odds ratio (OR) 1.49; stage III, OR 1.21; and stage IV, OR 1.58). The odds ratio was here defined as the odds to die of cancer during 5-year follow-up in patients with right-sided CRC versus patients with leftsided CRC. However, in stage II we observed the opposite association; right-sided CRC was associated with a significantly better survival compared to left-sided CRC (OR 0.51). When comparing right-sided colon cancer to left-sided colon cancer without rectal cancer, we observed similar results (Supplementary Tables S1 and S2).
When analyzing the data stratified by stage, age, and sex, left-sided and rectal CRC was again associated with a significantly better 5-year relative survival compared to rightsided CRC in most subgroups. Only in stage II and certain subgroups of patients of 80-years-old or older (stage I >80 -year-old males, stage III >80-year-old males and females, and stage IV >80-year old males) we observed a significantly longer survival in right-sided CRC. A summary of the odds ratios for mortality comparing right-to left-sided (incl. rectal cancer) CRC according to age, stage, and sex are presented in Fig. 2; Supplementary Table S3.

Role of Biomarkers in Prognostic Effect of Primary Tumor Location
Since it has been hypothesized that certain biomarkers might play a role in the prognostic effects of primary tumor location, we studied additional data on biomarkers in stage IV CRC in the second part of the study. Biomarker data of 1035 patients with metastatic CRC (mCRC) diagnosed in 2014 and 1182 diagnosed in 2015 was collected. We excluded 348 patients with a second tumor and 51 patients with tumors located in the appendix or with overlapping or unspecified tumor locations. Supplementary Fig. 1 presents the study population selection process. The final study population consisted of 1818 patients with mCRC; 613 (33.7%) were located on the right side of the colon, 743 (40.9%) were located on the left side of the colon, and 462 (25.4%) rectal cancers. Table 3 provides an overview of the characteristics and biomarker status of the studied patients with mCRC. The median follow-up time in the study population was 17.0 months (until death, alive, or lost to follow-up). As expected, deficient mismatch repair status and KRAS and BRAF mutations were more frequent in right-sided CRC, while NRAS mutations were more frequent in left-sided CRC.
When studying the effect of the covariates (sex and age) on survival separately in a Cox proportional hazards model, only age was significantly associated with survival and therefore included as a covariate in the following analyses (sex, hazard ratio (HR) = 0.98, P = .81; age, HR = 1.03, P = 1.78E-15).
First, we studied the main effect of the biomarkers on survival with primary tumor location and age as covariates in the model. KRAS and NRAS mutational status were not significantly associated with survival (respectively, P = .46 and P = .10). On the other hand, BRAF mutational status did show a significant association with survival (P = 8E-4) ( Table  4). However, it should be noted that there was a very strong association between BRAF mutational status and location, with BRAF mutant-type tumors found more frequently in right-sided CRC (P = 7.1E−9). This strong association could lead to confounding when studying the effect of primary tumor location and the biomarker on survival. Likewise, MMR status had no significant association with survival (P = .20), when location and age were used as covariates in the model. However, it is important to note that there was a strong association between location and MMR status (P = 4.4E-6), with the deficient MMR status being more frequent among patients with right-sided cancer. Due to this multicollinearity, the effect of MMR status was not significant in the model that included location as covariate. Nonetheless, when only age was included as covariate, there was a marginally significant effect of MMR status (P = .03). Fig. 3; Supplementary  Figs. S2-S10 show the Kaplan-Meier survival plots for each biomarker and location. Second, to study the impact of the biomarkers on the association between location and survival, we fitted Cox proportional hazard models with the main effects of each biomarker separately (either KRAS, NRAS, or BRAF mutational status or mismatch repair status) and location, and the interaction between them. The significance of the interaction term then indicates if the effect of the studied biomarker is uniform across all locations. The significance of these interaction terms was tested using a likelihood ratio test. In this analysis, no biomarker showed a significant interaction with location. Therefore, the location of the tumor did not influence the effect of the biomarkers on survival and the biomarker did not influence the effect of primary tumor location on survival (Supplementary Table S4).
Furthermore, primary tumor location remained highly significantly associated with survival in each model, independent of the covariates. Only in the model with BRAF mutational status, primary tumor location was not as strongly associated with survival (but still statistically significant), presumably because of the aforementioned partial confounding with BRAF.
All the aforementioned analyses were performed with primary tumor location coded as a 2-level variable (right versus left combined with rectal). When PTL was analyzed as a 3-level variable (right versus left vs rectal), similar results were obtained.

Discussion
In the first analysis of this study, we observed a significantly better survival in left-sided CRC compared to right-sided CRC (when all stages, ages, and sexes were combined) in the Belgian population. This corresponds to the results of other population-based studies comparing survival in left-and right-sided CRC (Supplementary Table S5). In most of these population-based studies, right-sided CRC was associated with significantly poorer survival with hazard ratios ranging from 1.01 to 1.3 (in all stages combined). 4,5,[17][18][19][20][21][22][23][24][25][26][27] However, some studies reported no significant difference or even better survival in patients with right-sided CRC. [28][29][30] This difference in results might be due to different definitions of right-and left-sided CRC, different study populations, or different statistical analyses. A significant finding of the current study was that after stratification by stage, age, and sex, we noticed a better prognosis for left-sided CRC in almost every subgroup. Nonetheless, we observed the opposite association in 2 subgroups of patients: (i) patients of 80 years and older and (ii) patients with stage II disease.
In patients of 80 years and older a better survival was observed in right-sided CRC compared to left-sided CRC. This phenomenon was also observed in other studies; however, the mechanism behind this observation remains unclear. 17 Our observation in stage II confirms the findings of other population-based studies that also observed better survival in right-sided compared to left-sided CRC. 5,17,21,[27][28][29][30][31] It has been suggested that better survival in right-sided stage II CRC is caused by a higher frequency of tumors with dMMR status or microsatellite instability (MSI) in this group. Tumors with MSI are associated with less metastasis and an overall better prognosis. 5 However, the prognostic impact of MSI depends on stage. In stage II, tumors with MSI have a superior prognosis compared with microsatellite stable tumors. While in stage III the prognostic effect of MSI depends on the risk group. In the low-risk group (T1-3 and N1), MSI patients had a longer OS compared to the MSS group, whereas in the high-risk group (T4 or N2) similar survival was observed in the MSI and MSS group. 32,33 In stage IV, a negative prognostic effect has been reported in univariate analysis, but it did not remain significant after multivariable adjustment. As a result, the prognostic effect of MSI in metastatic settings remains unclear. 33,34 As indicated by the complicated relation among stage, MMR status, and prognosis, the relationship between PTL  The Oncologist, 2023, Vol. 28,No. 6 and prognosis in CRC is very complex and the identification of the factors that are responsible for the prognostic value of PTL is hampered by the many factors involved and the potential interaction between them. Therefore, in a second analysis in a group of stage IV CRC, we studied the role of biomarkers in the prognostic effect of PTL.
Consistent with other studies, right-sided CRC in our population harbored more frequently deficient MMR status and KRAS and BRAF mutations compared to left-sided tumors, while NRAS mutations were more frequent in leftsided CRC. 25,[35][36][37][38][39][40] Even though BRAF mutations are known to be associated with poor outcomes and are more frequent in right-sided CRC, our results indicate that right-sided CRC is still associated with significantly poorer survival regardless of BRAF mutational status. However, since there was a strong association between BRAF mutational status and PTL, we hypothesize that the higher frequency of BRAF mutations is at least partly responsible for the prognostic effect of PTL, but additional factors must be involved.
The other molecular markers, including MMR status and KRAS and NRAS mutational status, were not significantly associated with survival in the current study population. This might be due to the fact that only a small number of patients with tumors with dMMR status (n = 39) were included and therefore, the study was underpowered to detect a prognostic effect or interaction. Also, only a small number of BRAF MT (n = 27) patients was included, but even though the study may have been underpowered for this analysis a significant association could be detected. Previous studies also concluded that right-sided CRC was associated with poor survival, independent of the known prognostic factors (including KRAS, BRAF, and MMR status). 38,39 The main strength of this study is the fact that it is based on high-quality information from a large population-based database. We included almost 92,000 patients in total, providing high power to the statistical analyses. Furthermore, we included all Belgian patients to create a heterogeneous group, covering all institutions and regions. Since this was an Odds ratio for mortality Figure 2. Odds ratios comparing right-to left-sided CRC according to age, stage, and sex. The odds ratio was defined as the odds to die of colorectal cancer during the 5 years follow-up in the non-reference category (right-sided CRC) versus the reference category (left-sided CRC). Five-year relative mortality in stage I, patients with left-sided colorectal cancer aged 0-59 year was used as reference survival (in each sex separately). In most subgroups, left-sided CRC is associated with a significantly better 5-year relative survival compared to right-sided CRC. Only in the subgroups of stage I >80year-old males, stage II >70-year-old males and females, stage III >80-year-old females, and stage IV >80-year-old males, a significantly better survival in right-sided CRC was observed. Dots, females; triangles, males. Abbreviations: RC, right colon; LCR, left colon and rectum.
The Oncologist, 2023, Vol. 28, No. 6 e337 unselected population, this study was thus based on unbiased data. This is in contrast to other large cohort studies, which are mostly based on data from clinical trials. Second, we provided a long-term follow-up (January 2004-July 2017). However, one of the limitations is the retrospective and observational character of the study. We considered numerous variables, which could at least in part explain the differences in survival between left-and right-sided CRC. However, other prognostic factors could not be examined (eg, comorbidity, disease bulk, location of metastasis, etc.) since those variables are not registered in the BCR. Second, we did not collect information on therapy. This would have been of particular interest since it has been proven that primary tumor location also has a  Since biomarker testing was mostly relevant for patients considered for anti-EGFR therapy (KRAS, NRAS) or younger patients (MMR testing for Lynch syndrome or BRAF testing for prognostication), it is possible that more young and fit patients with mCRC were tested. Furthermore, in 2014 and 2015, biomarker testing was not yet part of the standard work-up of every patient in every hospital. Therefore, it might be possible that biomarker status was tested more frequently in hospitals with specialized oncology clinics compared to smaller hospitals. Furthermore, the lack of standard techniques to test the biomarkers hampers interpretation. Based on the strengths and limitations of our study, we can conclude that we adjusted for the most important patient-and cancer-related prognostic factors and that this study provides valuable data. To our knowledge, this is one of the largest population-based studies that include biomarker information.

Conclusion
In summary, this Belgian population-based study on CRC confirms in an unselected population that relative survival is significantly higher in patients with left-sided CRC compared to right-sided CRC, in all stages and ages combined. However, in stage II and certain subgroups of patients over 80 years of age (stage I >80-year-old males, stage III >80-year-old males and females, and stage IV >80-year-old males) right-sided CRC shows a significantly better prognosis. Therefore, we can conclude that the prognostic effect of tumor location depends on age, sex, and stage in Belgian patients with CRC.
When studying the role of biomarkers in the prognostic value of PTL in 1818 Belgian patients with stage IV CRC, we found that none of the biomarkers (NRAS, KRAS, and BRAF mutational status and MMR) showed a significant interaction with location. Indicating that the effect of primary tumor location on survival was not influenced by these biomarkers in this population. Furthermore, it appears that in mCRC leftsided tumors are associated with a better survival compared to right-sided tumors, regardless of RAS and BRAF mutational status or mismatch repair status. However, the number of patients with deficient MMR or BRAF mutations in this population was too small to draw definitive conclusions.
Based on this study, the precise mechanism causing the difference in survival between left-and right-sided CRC remains to be elucidated. Further research should focus on determining these underlying complex molecular mechanisms.

Funding
None to declare.