Impact of Thiopurine Discontinuation at Anti-Tumour Necrosis Factor Initiation in Inflammatory Bowel Disease Treatment

A Nationwide Danish Cohort Study

Sandra Bohn Thomsen; Ryan C. Ungaro; Kristine H. Allin; Rahma Elmahdi; Gry Poulsen; Mikael Andersson; Jean-Frederic Colombel; Tine Jess


Aliment Pharmacol Ther. 2022;55(9):1128-1138. 

In This Article

Materials and Methods

The study cohort was drawn from four Danish health registries, the Danish National Patient Registry (DNPR), the Registry of Medicinal Product Statistics, the Danish Cancer Registry, and the Civil Registration System. These were linked by the unique Danish personal identification number. The DNPR contains data on all inpatient contacts in Denmark since 1977 and outpatient contacts since 1995.[23] Information on IBD diagnoses, intestinal surgery, anti-TNF therapy (anti-TNF unspecified, adalimumab or infliximab), thiopurine therapy (azathioprine or 6-mercaptopurine) and hospital admissions was retrieved from the DNPR. Additional information on anti-TNF, thiopurine and corticosteroid prescriptions were collected from the Registry of Medicinal Product Statistics, which contains data on all prescriptions filled at Danish pharmacies since 1995.[24] Data on cancer were collected from the Danish Cancer Registry which contains information on all cancer diagnoses occurring in Denmark.[25] Death and emigration status were collected from the Civil Registration System.[26] Data on cancer diagnoses were available through December 31st, 2017. The remaining data were available through December 31st, 2018.

Study Population

Patients with a least one International Classification of Disease (ICD) diagnosis of CD or UC were included. Patients with CD were identified with ICD-8 codes 563.00–563.09 and ICD-10 code K50. Patients with UC were identified with ICD-8 codes 563.19, 569.04 and ICD-10 code K51. IBD diagnoses included primary, secondary or underlying codes. One ICD code was considered sufficient as patients were also treated with two IBD specific medications (99.9% of patients had two IBD diagnosis codes). If patients had codes for both CD and UC the most recent code prior to anti-TNF start was used. Anti-TNF therapies were identified using the Danish SKS procedure codes B0HJ18A, B0HJ18A1 and B0HJ18A3 and Anatomical Therapeutic Chemical (ATC) codes L04AB02, and L04AB04 (Table S1). Thiopurine therapy was identified using SKS procedure code BWHB83 (treatment retrieved from outpatient clinic) and the ATC codes L04AX01, L01BB02 (treatment retrieved from pharmacy).

Patients were included if they initiated anti-TNF between January 1st, 2003 and October 2nd, 2018 and were using thiopurine at the time of anti-TNF initiation. Patients who were under 18 years of age at anti-TNF initiation, who were prescribed anti-TNF before first IBD diagnosis code, who were previously prescribed anti-TNF and UC patients with a history of colectomy were excluded. Patients were considered thiopurine continuers if they filled a thiopurine prescription within 90 days after anti-TNF initiation and discontinuers if they did not. The 90-days cut off was chosen as Danish patients will usually fill a thiopurine prescription within this timeframe. Follow-up started 90 days post-anti-TNF initiation to limit potential lead time bias (Figure 1). Patients who had an IBD-related surgery, emigrated from Denmark, were lost-to-follow-up, or died within the 90-day window following anti-TNF initiation were excluded.

Figure 1.

Illustration of thiopurine and anti-TNF exposure in thiopurine continuers and thiopurine discontinuers


We defined the primary outcome as a composite of any new oral corticosteroid prescription, IBD-related hospitalization, IBD-related surgery, or death. Corticosteroid use was defined as any new prescription more than 90 days after anti-TNF initiation. IBD-related hospitalizations were identified as any hospital admission with a primary diagnosis of IBD or a secondary diagnosis of IBD combined with a primary diagnosis of any IBD-related symptom including abdominal pain, diarrhoea, nausea, vomiting, rectal bleeding, stenosis, fistula, abscess or intestinal obstruction (Table S1). IBD-related surgery was defined as any colectomy, intestinal resection or other major surgical intestinal procedure, identified using the Nordic Medico-Statistical Committee Classification for Surgical Procedures (1996 and onwards). Our secondary outcome was any new diagnosis of cancer, ICD-10 C00-C97 (Table S1). Due to the common occurrence of non-melanoma skin cancer (NMSC) (defined by ICD-10 code C44) a separate sub-analysis was undertaken for this cancer type and for all cancers excluding NMSC.[27]


Covariates included age at IBD-diagnosis (<17 years, 17–39 years, ≥40 years), sex, timeframe of anti-TNF initiation (2003-2006, 2007–2010, 2011–2014, 2015–2018), disease duration (<1 year, 1–5 years, >5 years), IBD-subtype (CD or UC) and duration of thiopurine use prior to anti-TNF initiation (0-29 days, 30–90 days, 91–180 days or >180 days) (Table S2). We assessed health care utilization prior to anti-TNF initiation as a surrogate measure of disease severity. We analysed the number of IBD-related hospitalizations (0, 1–2, >2), new oral corticosteroid prescriptions (0, 1–2, >2), IBD-related major surgery (yes/no) or IBD-related minor surgery (yes/no) in the 5 years prior to inclusion in the study. Minor surgeries comprised incision of an abscess or fistula and endoscopic intestinal dilatation (Table S2).

Statistical Analyses

Patients were followed from 90 days post-anti-TNF initiation until emigration, death or end of follow-up (December 31st, 2018), whichever came first. We used intention-to-treat analysis and patients were not censored and did not change categories if treatment changed after the 90-day exposure window. We added a per protocol analysis for anti-TNF exposure, where patients are censored at the end of anti-TNF treatment. For analysis of cancer incidence, the follow-up period ended December 31st, 2017. We used SAS V.9.4 (SAS Institute, Cary, NC) for statistical analyses. Categorical data were summarized using proportions and groups were compared using Pearson's chi-square tests. We calculated incidence rates (IR) and hazard ratios (HR) using Cox regression models with 95% confidence intervals (CI) and used Kaplan-Meier models with log-rank tests to compare outcomes. The analyses were adjusted for the potential confounders including sex, diagnosis age, IBD-subtype, disease duration, timeframe of anti-TNF initiation, pre-anti-TNF thiopurine duration and past disease severity (hospitalizations, surgery and new corticosteroid use as defined above). If few events were identified for an outcome, the adjustments were restricted to three potential confounders (sex, diagnosis age and IBD-subtype) due to power considerations. Stratified sub-analyses for the primary outcome were performed on sex, diagnosis age, timeframe of anti-TNF initiation, IBD-subtype, and duration of pre-anti-TNF thiopurine use. To examine the impact of post-anti-TNF thiopurine adherence, we undertook sensitivity analyses of thiopurine exposure one-year post-anti-TNF initiation. Patients were considered thiopurine exposed one-year post-anti-TNF initiation if they used thiopurine within 30 days of the one-year date. Patients were divided into four groups according to thiopurine use at the one-year date. First, patients who initially continued thiopurine were divided into two groups; a group who continued thiopurine one-year post-anti-TNF initiation and a group who stopped thiopurine one-year post-anti-TNF initiation. Next, initial discontinuers were divided into discontinuers who restarted thiopurine one-year post-anti-TNF, and discontinuers who remained discontinuers. Analyses were restricted to patients who had not yet experienced surgery, death or emigrated after one year. The results were considered statistically significant if the two-sided chi-square test P-value was below 0.05.


Study permission was obtained from the Danish Data Protection Agency. Ethical approval is not required for registry-based research in Denmark. The study adheres to The Strengthening the Reporting of Observational Studies in Epidemiology statement.[28]

Patient and Public Involvement

Patients and the public were not involved in the design or execution of the study. The results will be disseminated through patient organizations and presentations at national and international conferences.

Role of the Funding Source

The funding source was not involved in study design; in the collection, analysis or interpretation of data; in the writing of the report; or in the decision to submit the paper for publication.