This study included 1530 consecutive patients admitted with colon cancer at Levanger Hospital during 1980–2016. Levanger Hospital is the primary hospital of 10 municipalities in Norway, and the catchment area remained unchanged throughout the study period. The population increased by 18%, from 83,890 inhabitants in 1980, to 99,566 inhabitants in 2016. During this period, the average age of the population also increased. In particular, the number of octogenarian inhabitants increased by 73%, from 2184 individuals in 1980, to 3800 individuals in 2016 [4].
Through the hospital administrative system, we accessed the health records for all patients that were discharged with diagnosis codes of the International Classification of Diseases, 8th revision (ICD-8) from153.1 to 153.9, with ICD-9 codes from 153.0 to 153.9, and with ICD-10 codes from C18.0 to C19. Data on all patients were recorded, crosschecked, and confirmed with data from the Norwegian Cancer Registry, during 1980–2016. From the hospital database, we retrieved data on demographic and logistic variables, comorbidities, treatment, tumour characteristics (including histopathology), complications after treatment, and short- and long-term outcome measures.
We defined colon cancer as any tumour located above 15 cm from the anal verge. Right colon tumours were defined as tumours localized in the caecum, ascending colon, hepatic flexure, or transverse colon. Left colon tumours were defined as tumours localized in the splenic flexure, descending colon, or sigmoid colon. Tumours located within 15 cm from the anal verge were defined as rectal cancer, and we excluded these and cancers localized in the appendix.
We characterized patient comorbidity with the American Society of Anaesthesiology (ASA) score and the Charlson Comorbidity Index (CCI) [12, 13]. We defined anaemia at admission, as advocated by the World Health Organization, as blood haemoglobin levels below 13 g/dL in males and below 12 g/dL in females [14]. We also defined “moderate to severe” anaemia as haemoglobin levels below 11 g/dL in males and 10 g/dL in females. Surgical complications were defined according to the Clavien-Dindo classification of surgical complications, grades I-V [15].
Disease stages were based on the TNM classification, sixth edition [16]. An R0 resection was defined as no detectable residual tumour postoperatively; an R1 resection was defined as a microscopic residual tumour detected in a postoperative histological examination; and an R2 resection was defined as a macroscopic residual tumour detected after surgical treatment [17]. An R0 resection was further classified into two groups: an R0 without tumour perforation and an R0 with tumour perforation. Tumour perforations included both spontaneous (12) and iatrogenic perforations (9).
Patients were categorized into five groups, according to treatment intent: (i) a major resection with curative intent (R0 and R1), (ii) a polypectomy, (iii) a major resection with non-curative intent, (iv) a bypass/stoma, and (v) best supportive care.
Emergency surgery was defined as surgery due to evidence of a large bowel obstruction or large bowel perforation. The laparoscopic colon resection technique was gradually introduced during the last part of the study period. A total of 49 patients underwent laparoscopic surgery. In ten of these patients, the procedure was converted to open surgery.
Staging varied throughout the observation period. Staging was based on complete clinical and histopathological examinations of the resected specimen in 84.9% (1299/1530) of patients; a clinical examination and histopathological examination of a tumour biopsy in 7.8% (120/1530) of patients; a pathological evaluation during an autopsy in 1.4% (21/1530) of patients, and clinical evaluations alone in 5.9% (90/1530) of patients.
Since 1993, the Norwegian national guidelines for treatment of colon cancer advocated that all patients aged 75 years or under with Stage III disease should be evaluated for adjuvant chemotherapy. Later, this recommendation was applied to selected patients with Stage II disease [5].
Follow-ups were initially conducted according to local guidelines. Starting in 1993, they were based on very similar, national guidelines [5]. The follow-up time was calculated as the patient-years at risk, starting from the date of admission. The study endpoints were: local recurrence, metastasis, or death, regardless of cause. The mean follow-up time was 6.05 years (standard deviation [SD] = 6.89, range: 0-38.7 years). The end of follow-up was December 31st, 2018.
Statistical analyses
The Exact Unconditional z-pooled test was used to compare binomial proportions; for example, the percentage of reoperations, relative to the percentage of emergency or elective primary operations. The Cochran Armitage exact trend test was used to test for trends in proportions; for example, the proportions of elective surgeries vs. emergency surgeries in different age groups. The Joncheere-Terpstra test was used to test for the distribution of age, as a dependent variable, across 10-year age groups, as the independent variable. The five-year rates of local recurrence and metastases were estimated with the Kaplan-Meier method.
Logistic regression analyses were performed to assess associations between the 100-day mortality, as the dependent variable, and different explanatory variables. Ordinal logistic regressions were performed to analyse the associations in doubly-ordered r × c tables; for example, the ASA score stratified by age group. The resulting odds ratios (ORs) represent a common OR estimate for any 2 × 2 table that would occur, if the r × c table was collapsed to a 2 × 2 table, based on any cut-off threshold, along the columns and rows. Multinomial logistic regression analyses were performed in singly ordered r × c tables; for example, the type of treatment, stratified by age groups.
Relative survival analysis
Relative survival was defined as a measure of mortality compared to the general population. The observed survival in the group with cancer was divided by the expected survival of a comparable group in the general Norwegian population, matched by age, sex, and the calendar year of investigation. Relative survival was estimated with the Ederer II method and analysed with STATA 16 [18]. Multivariable analyses were performed with a full likelihood approach. Norwegian population survival probabilities were downloaded from the Human Mortality Database, for every year from 1980, calculated for groups stratified by sex and age [19].
Two-sided p-values < 0.05 were considered significant. Means are reported with the range (minimum to maximum) and SD, where relevant. Ninety-five percent confidence intervals (95% CI) are reported, when relevant. Analyses were carried out in Stata 16, IBM SPSS Statistics 25, and StatXact 9.