Prognostic Impact Of Tumor Location In Colon Cancer - BMC Cancer

Using population-based cancer registries data from the Monitoring of Cancer Incidence in Japan (MCIJ) project, we analyzed colon cancer cases (ICD-10: C18.0–18.7) diagnosed from 2006 until 2008 in 21 population-based cancer registries (Aichi, Chiba, Ehime, Fukui, Fukushima, Gunma, Hiroshima, Ibaraki, Kanagawa, Kumamoto, Miyagi, Nagasaki, Niigata, Osaka, Okayama, Shiga, Shimane, Tochigi, Tottori, Yamagata and Yamanashi) in Japan. Cases were selected according to Japanese standards with regard to (i) proportion of cases reported by death certificate only (DCO%: death certificate only) of less than 25%, (ii) proportion of cases first notified through death certificate (DCN%: death certificate notification) of less than 30%, (iii) mortality to incidence ratio (M/I) of less than 0.67 [24], and (iv) percentage of lost to follow-up of < 5% or adopted linkage to a death certificate database to confirm the vital status of patients. We included those patients diagnosed in 2006–2008 and followed through Dec 31, 2013. Japanese population-based cancer registries start to follow patients at the date of diagnosis and do not register the date of operation or starting treatment. We excluded data from cases that were registered by death certificate only, were secondary multiple cancers, were in situ cases, and those in patients aged > 100 years. We also excluded data from cases that were registered by death certificate notification. The study included colon cancer cases (ICD-10: C18.0–18.7), cecum, C18.0; appendix, C18.1; ascending colon, C18.2; hepatic flexure of the colon, C18.3; transverse colon, C18.4; splenic flexure of the colon, C18.5; descending colon, and C18.6; and sigmoid colon, C18.7. Overlapping lesions of colon (C18.8) and those not otherwise specified (C18.9) were excluded. Colon cancer patients were further categorized into two groups, those with right-sided colon cancer (C18.0–18.4; cecum, appendix vermiformis, ascending colon, hepatic flexure of colon and transversal colon) and left-sided colon cancer (C18.5-C18.7; splenic flexure of colon, descending colon and sigmoid colon). With regard to the extent of disease, patients were categorized into the three disease stages of localized, regional and distant groups. Extent of disease was available in the Japanese population-based cancer registries. The Japanese staging system, extent of disease, was based on the Surveillance, Epidemiology, and End Results (SEER) staging criteria [25]. Extent of disease was unknown for 14.0% of subjects.

Statistical analysis

The frequency of related variables of patients by cancer locations was compared using the two sample t-test for continuous variables and the χ2 test for categorical variables. We calculated 5-year net survival for colon cancer patients diagnosed from 2006 until 2008 by anatomical subsite according to sex, age group (< 40, 40–54, 55–69, ≥70), extent of disease at diagnosis (localized, regional or distant stages). Net survival is regarded as the survival that would be observed in the hypothetical situation that the only possible cause of death was cancer [26]. Net survival is calculated by following two methods: relative survival and cause-specific survival. The population-based cancer registries usually use relative survival to give estimates net survival [27]. We used the recently introduced Pohar Perme estimator [28] of net survival implemented with the program stns in Stata version 14.1. The complete national population life-tables by single year of age, sex and calendar year were used to derive the expected mortality rates. To assess the impact of anatomical location of the colon cancer on survival, the excess mortality model, a multivariate regression approach which adopts the flexible parametric model [29, 30] implemented with the stpm2 function in Stata version 14.1 was used. We applied the excess mortality model to calculate excess hazard ratios (EHRs) and 95% confidential intervals (CIs) with and without adjustment for age, sex and cancer stages to assess the effect of the location of colon cancer. Cases in which the tumor stage was unknown were excluded when the excess mortality model was performed to adjust for tumor stage. The differences in survival rate with location of colon cancer between sex, age groups or tumor stages were statistically tested by including an interaction term into the excess mortality model. A two-sided P-value of < 0.05 was considered statistically significant. All statistical analyses were performed using Stata v. 14.1 (STATA Corporation, College Station, TX).