br in men and in women Similarly alcohol
659) in men and 18.83% (77/409) in women. Similarly, alcohol drinking status was categorized as drinker vs. nondrinker and drinker was defined as drinking alcohol on at least 12 occasions during the past 12 months prior to interview. Nondrinkers, however, were subjects who reported to have never drank alcohol or drank alcohol mostly six oc-casions per year. Diﬀerences in the type of alcohol (beer, wine, distilled spirit) and the quantity taken daily (moderate or heavy drinker) were not accounted for due to lack of information. Overall, the prevalence of alcohol drinking was 40.21% (265/659) in men and 18.83% (77/409) in women, respectively.
All clinical and pathological information were reviewed retro-spectively and assessed according to the sixth edition of AJCC. The cases with cancer in the proximal two-thirds of the transverse colon, ascending colon, hepatic flexure and cecum were considered RCC while those cases with cancer in the distal third of the transverse colon, splenic flexure, descending colon, and sigmoid colon were considered LCC. Of the 1078 subjects, 234 (21.7%), 241 (22.4%) and 603 (55.9%) were identified to have RCC, LCC and ReC respectively. All patients at stage III-IV underwent curative resection. Adjuvant chemotherapy was carried out either with FOLFOX or another adjuvant including FOLFIRI, based on the China guideline for diagnosis and comprehensive treat-ment of CRC [28,29]. Briefly, FOLFOX regimen is based on at least 6 cycles of monthly PSB 1115 intravenous administration of 5-fluorouracil (400–425 mg/m2/day) and leucovorin (20 mg/m2/day) for 1–5 days.
Informed consent was obtained from each enrolled subject before the recruitment. The institutional review boards (IRB) of the Southeast University (Nanjing China) approved this study under the approval ethical reference code 2017ZDKYSB165.
All patients were followed-up by a trained clinical specialist from Jiangsu Hospital Cancer Center through on personal or family contacts from the time of diagnosis to death or the last follow-up (cutoﬀ in June 2016). The median follow-up duration was 68.62 months (range of 0–112.70 months), during which time 644 (59.74%) patients (165 (70.51%) RCC, 140 (58.09%) LCC and 339 (56.22%) ReC) died (the detailed data has been described elsewhere) . Due to the lack of information in assessing cause-specific mortality, some authorities in-cluding the Food and Drug Administration (FDA) and the European Medicine Agency have advocated overall survival as a more appropriate endpoint for such studies . Overall survival (OS) was defined as the time from initial surgical resection until death due to any cause. Pa-tients alive or lost to follow-up are censored.
2.4. Statistics analysis
Frequency distribution of epidemiology, clinicopathological fea-tures and treatment regimens for RCC, LCC and ReC were summarized using both Chi-square test (χ2 tests) for categorical variables and Kruskal Wallis test for continuous variables. The prognostic eﬀect of each subsite location on overall survival (OS) was first evaluated with the Kaplan-Meier method. Diﬀerences between survival curves for RCC, LCC and ReC were compared with the log-rank test. Secondly, Cox
proportional hazards regression models were used for the unadjusted univariate and adjusted multivariate survival analyses. Proportionality assumptions for the Cox regression models were tested, as well as Schoenfeld residuals plots for each subsite location over time. Variables and pairwise interactions included in the full Cox models are covariates that have been reported previously to have associations with CRC cancer mortality. For the prognostic significance of variables, the Cox regression was performed to obtain the unadjusted and adjusted ha-zards ratios (HRs) at 95% confidence intervals (CIs) in OS for RCC, LCC and ReC patients within each subgroup, controlling for all other vari-ables. Entry and removal probabilities for stepwise-regression were 0.05 and 0.10, respectively. In addition, Lakatos method was used to determine the power of the study. Most of statistical analyses were performed with SPSS software (SPSS Inc., Chicago, IL), using two-sided testing with a significance level at p-value of 0.05. SAS version 9.1.3 (SAS Institute), was used to determine the power of the study.
3.1. Patients’ characteristics by epidemiological and clinicopathological features
With respect to the clinico-pathological features of the cases, the histologic grade was consistent across each subsite location, with only 4.4% of tumors being poorly diﬀerentiated. A greater proportion of cases had a higher level of tumor depth of invasion (T3-4 = 74.1%) with no regional lymph node metastasis (N0 = 57.6%) and no distant metastasis (M0 = 88.9%). When stratified by tumor node metastasis stage, more than 55.4% of all cases were at stage I-II in each of the subsite locations of CRC. In TNM stage III-IV, patients with LCC (23.08%) were slightly frequently higher compared to RCC (20.58%), but no statistical significant (P = 0.693). In addition, patients with RCC exhibited a higher frequency of distant metastasis M1(36.67%) com-pared to only 28.33% and 35.0% for LCC and ReC respectively.