br Recurrence during follow up
Recurrence during follow-up was assessed using a validated algo-rithm, described in detail elsewhere . In brief, patients who died within 180 days after surgery and patients with metastases at time of diagnosis or within 180 days after surgery were excluded as it was not relevant to investigate recurrence in these patients. Furthermore, pa-tients with other cancers (except non-melanoma skin cancer) before the date of colorectal cancer diagnosis or within 180 days after surgery were excluded as the NPR registered metastases codes mentioned below do not distinguish between which primary cancer the 丝裂霉素 C was referring to. Patients were censored at time of diagnosis of a new pri-mary tumor if it occurred after the 180 days following surgery.
In the registries described above the diagnosis of recurrence were estimated by meeting at least one of the four mentioned criteria: 1) Metastases code registered in the NPR 2) NPR-registered cytostatic therapy codes 180 or more days after the first colorectal cancer surgery and 60 or more days after the last cytostatic therapy code. 3) DPR-registered SNOMED combinations indicating recurrence. 4) An ICD-10 code specific for local colorectal cancer recurrence in the NPR any time Surgical Oncology 28 (2019) 62–66
The study population was followed from 180 days after surgery until death (by any cause), occurrence of a new primary tumor, end of observation (December 31, 2015), or occurrence of relapse of colorectal cancer, whichever came first.
2.2. Drug exposure and statistical methods
Diabetes was defined as medically treated diabetes, because in-formation regarding diet-treated diabetes is not available in the regis-ters used and because use of antidiabetics with very few exceptions indicates presence of diabetes. Study participants were classified into patients with diabetes (patients who were in active antidiabetic treat-ment at the beginning of the radiotherapy) and non-diabetes patients. Patients who were classified as in active treatment had to have re-deemed at least one prescription for antidiabetic medication (Anatomical Therapeutic Chemical code (ATC-code), A10) in the year preceding the first date of neoadjuvant treatment and at least one re-deemed prescription within 180 days after surgery. Patients with dia-betes were matched with non-diabetics 1:2 by propensity score. The propensity score included age at diagnosis, sex, Charlson comorbidity index, BMI, smoking, alcohol consumption, lymph node status, and year of surgery. The matching was done by nearest neighbor matching without replacement and using a caliper of width equal to 0.2 of the standard deviation of the logit. Subgroup analysis were performed in patients receiving metformin in mono-therapy (ATC-code A10BA or A10BD) compared to non-diabetic patients.
The primary outcome of the study was disease-free survival (DFS) and the secondary outcomes were recurrence free survival (RFS) and all-cause mortality. Cox regression model was used. Disease-free sur-vival was defined as time from 180 days after surgery to diagnosis of recurrence, new primary tumor or death regardless of cause. Recurrence-free survival was defined as time from 180 days after sur-gery to diagnosis of recurrence. Patients were censored at date of new primary tumor occurrence or death from any cause. All-cause mortality was defined as time from 180 days after surgery to time of death re-gardless of cause. Results were presented as hazard ratios (HR) with 95% confidence intervals (95% CI). The study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations  and approved by Danish Data Protection Agency (Reg- 052–2017). The statistical ana-lysis was performed using the SAS® Proprietary Software 9.4, SAS In-stitute Inc., Cary, NC USA.
A total of 9799 patients were undergoing rectal cancer surgery with curative intend in the period. Of those, 2379 received neoadjuvant treatment up to one year preceding surgery. In total 172 patients were classified as diagnosed with diabetes, of whom 58 were in monotherapy with metformin and 2056 were not using antidiabetic medication. Patients only prescribed antidiabetic medication before or after the neoadjuvant treatment were excluded (Fig. 1).
The 172 patients diagnosed with medical treated diabetes were matched by propensity score with patients not diagnosed with diabetes. It was possible to match 154 of the users to non-users, and for three patients it was only possible to match with one non-diabetes patient. The 18 patients who were not able to be matched were classified as five metformin users, eight using other antidiabetic medication and five receiving a combination of treatments and 17 of them had a Charlson Comorbidity Index > 2. In total 477 patients were included in the study; 154 with diabetes and 305 not diagnosed with medical treated diabetes, 53 in the diabetes group were treated with metformin in monotherapy. (Table 1a and Table 1b).