The impact of primary tumor sidedness on survival in early‐onset colorectal cancer by stage: A National Veterans Affairs retrospective analysis

Abstract Background The incidence of early‐onset colorectal cancer (EOCRC) is rising. Left‐sided colorectal cancer (LCC) is associated with better survival compared to right‐sided colon cancer (RCC) in metastatic disease. NCCN guidelines recommend the addition of EGFR inhibitors to KRAS/NRAS WT metastatic CRC originating from the left only. Whether laterality impacts survival in locoregional disease and EOCRC is of interest. Methods 65,940 CRC cases from the National VA Cancer Cube Registry (2001–2015) were studied. EOCRC (2096 cases) was defined as CRC diagnosed at <50 years. Using ICD codes, RCC was defined from the cecum to the hepatic flexure (C18.0–C18.3), and LCC from the splenic flexure to the rectum (C18.5–18.7; C19 and C20). Results EOCRC is more likely to originate from the left side (66.65% LCC in EOCRC vs. 58.77% in CRC). Overall, LCC has better 5‐year Overall Survival (OS) than RCC in stages I (61.67% vs. 58.01%) and III (46.1% vs. 42.1%) and better 1‐year OS in stage IV (57.79% vs. 49.49%). Stage II RCC has better 5‐year OS than LCC (53.39% vs. 49.28%). In EOCRC, there is no statistically significant difference between LCC and RCC in stages I‐III. Stage IV EOCRC patients with LCC and RCC have a 1‐year OS of 73.23% and 59.84%, respectively. Conclusion In EOCRC, LCC is associated with better OS than RCC only stage IV. In the overall population, LCC is associated with better OS in all stages except stage II. The better prognosis of stage II RCC might be due to the high incidence of mismatch repair deficient tumors in this subpopulation.


| METHODS
Nationwide data from the National Veterans Affairs Cancer Care Cube Registry (CCCR) were analyzed. No patient charts were accessed. The main data source for the CCCR is the Oncology Domain tables on the Corporate Data Warehouse (CDW) raw server, which is updated every 2 weeks. The Oncology Domain tables are created from the VISTA OncoTrax software package. The registry was accessed on 12 August 2017 and 2 June 2018, and data input after this date is not included in this study. Unique cases of CRC with accession year between 2001 and 2015 were analyzed.
ICD codes C18 to C20 were used to delineate patients with RCC vs. LCC. RCC was defined as cancer from the cecum to the hepatic flexure (ICD C18.1-C18.3), while LCC was defined as cancer from the splenic flexure to the rectum (ICD C18.5-18.7 & C19 & C20) with transverse cancer in between flexures (ICD C18.4). ICD codes C18.8 and C18.9 referring to overlapping and unspecified parts of the colon, respectively, were excluded. The registry defines unique cases as those with the same combination of the following data points: patient social security number, diagnosis date, primary tumor site, sequence number, histology ICD 03 Code, grade differentiation ID, and laterality. Accession year refers to the year in which the patient was first seen at the reporting institution for diagnosis and/or treatment of the primary cancer recorded. The registrar further classifies cases based on abstract status. "Complete" abstract status indicates that all data points have been entered by the Tumor Registrar for that particular case. Only cases with complete abstract status were considered for this study. After application of the above qualifiers (unique cases, complete abstract, ICD code, all stages, and accession year span 2001-2015), 65,940 total cases of CRC were identified.
Demographic data on the CCCR including age at diagnosis, gender, and survival were generated from the VA Health Eligibility Center (HEC) demographic file. Survival in the CCCR was

Lay Summary
Colorectal cancer is increasing in people under the age of 50. We found that most colorectal cancer in young people comes from the left side of the colon. We know that older patients with stage IV left-sided colorectal cancer live longer and respond better to treatment. We show the same is true in young people. We also compared left-and right-sided cancer in stages I, II and III. Left-sided colon cancer has better survival in all stages except stage II. We think this is because stage II colon cancers on the right are mismatch repair deficient (MMRd), which take longer time to spread. defined as <1 year, 1-5 years, 5-10 years, 10-15, and >15 years. Race and ethnicity were derived from the CDW. Demographic characteristics including race and ethnicity were determined based on information provided by patients at initial contact with the VA hospital. Local IRB approval was obtained for the study.
Microsoft Excel (Microsoft Corp) was used for data tabulation and graph formulation. Discreet data points were described using percentages and compared using Chi-squared test with two-sided p-value of <0.05 considered as statistically significant. Graph slopes and Chi-squared tests were calculated using VassarStats.

| Demographic data and age distribution
Of the 65,940 CRC cases diagnosed at the VA between 2001 and 2015, 19,969 (30.28%) cases originated from the RCC, while 38,754 (58.77%) originated from the LCC. Expressed as a left to right (L:R) ratio of 1.94, LCC is twice as common as RCC in the VA population. Transverse colon is rare with only 6.36% (4,191) cases. Only 4.59% of CRC cases (3,026) were excluded as they were documented as originating from overlapping or unspecified parts of the colon. The demographic data of the VA Cancer Care Cube are described in Table 1.
As expected in a military setting, the VA CRC population is predominantly male (97.56%). Among the 1,602 women veterans, RCC constituted 36.33% of cases compared to only 30.13% in men (p < 0.0001). The described population is predominantly White with 68.80% (45,366) identifying as such. The second most common group is Blacks with 17.15% (11,229) of the population. When afflicted with CRC, Asians are almost three times more likely to have LC than RC as opposed to two times overall (Asian L:R 2.94, Overall L:R 1.94, p = 0.0113). Conversely, Blacks are more likely to have CRC from the right side of the colon and less from the left side (Blacks L:R 1.76, Overall L:R: 1.94, p < 0.0001). Whites, American Indians/Alaskan Natives, and Native Hawaiians/ Pacific islanders have similar rates to the general population; as do the unknowns and patients who declined to answer, indicating no self-selected bias. Overall, only 5.06% of the population identified as Hispanic or Latino. Interestingly, Hispanics are more likely to be afflicted with left-sided colon cancer with 61.05% of all CRC (L:R ratio of 2.15; p = 0.0108).

| Age distribution
Incidence of EOCRC, defined as CRC before the age of 50, constitutes 3.18% of CRC database with most EOCRC occurring in the fourth decade. The laterality distribution in

| Stage distribution
Stage distribution of CRC at diagnosis is shown in Table 3 and Figure 2. CRC is most likely to be stage I (25.49%) at diagnosis, followed by stages II (19.89%), III (18.26%), and IV (15.94%). Stage 0 (carcinoma in situ) is the least frequent stage with 9.20% of all CRC. LCC is more likely to present in early stages, specifically stage 0 (L:R 2.24) and stage I (L:R 2.37) compared to the average L:R of 1.94. LCC remains more common than RCC at late stages but as the stage progresses so does the proportion of RCC, with L:R ratios of 1.54 at stage II, 1.7 at stage III, and 1.75 at stage IV.

| Overall survival (OS)
In EOCRC (Table 4B), RCC carries a significantly worse survival than LCC in the metastatic setting (1-year OS-RCC: 59.84% vs. LCC: 73.23%; p = 0.0086). In the non-metastatic setting, the difference in 5-year Overall Survival observed with CRC is not statistically significant in any stage. More detailed visual representations of overall survival of CRC and EOCRC are presented in Figure 3A and B, respectively. In EOCRC, there is no statistically significant difference between LCC and RCC in stages 0-III. Stage IV LCC in patients <50 has a 1-year OS of 73.23% (vs. 59.84% for RCC).
In CRC (

| Epidemiology
This paper describes the first study of PTS in Veterans Affairs hospitals. Despite the overwhelming White male slant of the Veterans population, it reinforces 21-23 that RCC is more likely to arise in women, Blacks, and the elderly. We also demonstrate that RCC presents at a more advanced stage, often with worse performance status than LCC. Despite the increase in diagnosis of CRC on routine screening colonoscopies, most CRC still presents symptomatically. 24 Symptoms of CRC differ by tumor location. Symptoms more characteristic of RCC are anemia and vague abdominal pain, while LCC usually presents with hematochezia, change in bowel habits, and is more likely to cause obstruction. This difference is thought to be due in part to the larger luminal diameter of the cecum and consistency of the bowel contents, as the tumors need to grow large enough to cause obstructive symptoms.
Traditionally, this anatomical discrepancy was thought to explain the shortened survival associated with RCC. The predominance of RCC in specific epidemiologic subpopulations (Blacks, women, and elderly) belies that explanation.

| Molecular genetics
More recently, molecular studies revealed that the pathophysiology of the malignancy and not just the anatomy of the digestive system carries prognostic and predictive implications. Tumors in each colonic segments (RC and LC) with distinct embryological origins (midgut and hindgut) express different molecular genetics 25 29 of CRC revealed that the consensus molecular subtypes (CMS) type 1 (CMS1) tumors were more frequently rightsided, while CMS2 tumors were more likely to be left-sided.

| Overall survival
We show metastatic LCC has significantly better survival than RCC, as previously demonstrated in the CALGB80405 30 and FIRE-3 19 studies. We did not evaluate the impact of EGFR-and VEGF-directed therapies in our analysis. The impact of PTS in the non-metastatic setting is less well-studied. To our knowledge, this is the largest retrospective review of PTS not performed within the confines of the SEER database and the third largest overall.
Given that the treatment of early stage disease is primarily surgical, the difference in survival is expected to be less pronounced. Nonetheless, LCC is associated with better OS in all stages except stage II, where, interestingly, RCC performed slightly better. Previous studies using the SEER database yielded mixed results for RCC vs. LCC in stage II. Studies by Yang 31 (2000Yang 31 ( -2012 and Weiss 32 (1992Weiss 32 ( -2003 found a survival benefit favoring RCC in stage II, while Warschkow 33 (2004Warschkow 33 ( -2012 and Meguid 34 (1988Meguid 34 ( -2003 found LCC to have a longer survival. Prior analyses of molecular markers by stage and side have shown that rightsided stage II adenocarcinoma is enriched in MSI-high 35

| EOCRC
We report that EOCRC accounts for 3.18% of all cases of CRC in the VA population. An analysis of the National Cancer Data Base (NCDB) during the same timeframe (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) reports 11% of CRC to be diagnosed before the age of 50. 40 The lower fraction of EOCRC observed in our study can be accounted for by the higher incidence of malignancies in veterans than the general population. 41 On average, veterans are more likely to be older, smoke, drink alcohol, and to have been exposed to Agent Orange. While CRC is almost twice as likely to originate from the left colon in the overall population, EOCRC is almost three times as likely to arise from the left side (L:R ratio 2.84). This ratio is highest for cases diagnosed in patients in their 30 s (L:R of 3.44). The predilection of CRC to arise from the distal colon is consistent with prior observations. 42,43 One potential explanation of this tendency is the notable difference in transit time in each colonic segment by age. 44,45 A relatively longer rectosigmoid transit in adolescents can lead to prolonged exposure to potential carcinogens, such as highfat food and processed meat. 46,47 A consensus 48 is emerging that EOCRC is a pathologically, epidemiologically, anatomically, and biologically different disease than late-onset CRC. We sought to examine the impact of laterality in EOCRC in this study. In stage IV, LCC exhibits a longer 1-year OS than RCC in younger patients, similar to the overall population. While the difference in 5-year overall survival-favoring LCC in stages 0, I, and III and RCC in stage II-is not statistically significant, there remains a trend toward better 5-year OS in stage II RCC and stage III in EOCRC. The smaller number of patients with EOCRC could account for this difference. Due to their young age, EOCRC patients are more likely to receive more intensive chemotherapy than their elder counterparts. 49 Unfortunately, this practice has not resulted in a commensurate increase in overall survival. Identification of specific prognostic and predictive markers is of paramount importance to guide therapy. Recent molecular analyses by sides in EOCRC 12,50 have sought to shed light on these questions.
Our study is limited by the inherent biases of its retrospective design. We acknowledge the lack of availability of data regarding molecular profiling, follow-up, and treatment modalities as limitations. The data presented here may show a need for a greater incorporation of PTS in the design of prospective randomized clinical trials. The rise in EOCRC dictates a greater involvement of these affected in clinical trials.