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


In this nationwide population-based cohort study of 2,630 patients with IBD, we found that patients who discontinued thiopurine therapy when initiating anti-TNF had an increased risk of a composite adverse outcome, including new corticosteroid treatment, IBD-related hospitalization, IBD-related surgery or death. When we analysed outcomes separately, thiopurine discontinuers specifically had an increased risk of new corticosteroid use and IBD-related hospitalization. Patients who continued thiopurine therapy did not have an increased overall risk of cancer, although, their risk of NMSC was increased. Our study provides important data on a common clinical situation in which physicians and patients consider the risks and benefits of continuing thiopurine treatment at anti-TNF initiation.

Previous evidence on the impact of thiopurine discontinuation in IBD mainly focus on patients in remission. While patients who achieved remission on thiopurine monotherapy had an increased risk of relapse after thiopurine discontinuation, thiopurine withdrawal after combination treatment with anti-TNF has not been found to increase the risk of a relapse.[18,29] Randomized studies on treatment naïve patients have found that combination therapy with thiopurine and anti-TNF increases infliximab trough levels (serum infliximab levels before infusion) and decreases infliximab antibody formation.[10–14,30,31] Nonetheless, clinical evidence did not find improved efficacy of continuing thiopurine, except that thiopurine reduced the risk of infusion reactions.[20,32,33] However, these results were derived from subgroup analyses of anti-TNF versus placebo studies (baseline thiopurine exposure vs never thiopurine exposure). Our study addresses a different question related to therapy escalation in a real-world setting. Observational and retrospective studies that evaluated thiopurine exposed patients at anti-TNF escalation have conflicting findings.[34–37] A Canadian population-based study evaluated a cohort including both thiopurine experienced and inexperienced patients. Although their objective was to evaluate the impact of thiopurine exposure at initiation of anti-TNF, they found a reduced risk of adverse outcome (hospitalization, intestinal resection, new corticosteroid use or change of anti-TNF agent) in anti-TNF and thiopurine co-exposed patients with CD but not in patients with UC.[36] They attribute their findings to a long follow-up, which may also be the case in our study. Our findings may also be explained by thiopurine persistence, as our sensitivity analyses suggested that re-starting thiopurine one year after anti-TNF initiation reduced the risk of our primary outcome whereas stopping thiopurine within a year increased the risk. This suggests that continuing thiopurine for at least one year may be advantageous. Similarly, an observational study on patients with IBD on infliximab and immunomodulator maintenance therapy found that patients had a lower risk of relapse in later time periods on combination treatment than with only infliximab.[34] Nevertheless, it is complex to compare IBD-cohorts across studies as cohort classifications and outcomes differ. Moreover, local and national guidelines and treatment practices may differ. An important finding in our study was that both patients with UC and CD had benefit from continuing thiopurine, which is interesting as previous evidence on efficacy in UC is weaker.[10,11,38] This presents important evidence that thiopurine continuation should be considered in UC as well as in CD at the time of escalating to anti-TNF.

Hospitalization and new corticosteroid use were specifically improved in thiopurine continuers in our study. A reduced hospitalization risk has previously been demonstrated in combination therapy users.[36] Although we classified hospitalizations to reflect disease flares, they may also reflect admissions for infections or hypersensitivity reactions. Either way, a reduction is important as hospitalizations are a major adverse event for patients as well as an economic burden for health care systems.[39] A decrease in the use of corticosteroids is likewise of tremendous importance as corticosteroids increase the risk of infection and mortality.[40]

Patients who continued thiopurine did not have an increased risk of overall cancer, which is an important finding as cancer has been one of the main concerns regarding thiopurine and anti-TNF combination therapy, in particular the risk of lymphoma.[17,41] We did not have sufficient power to analyse lymphoma separately, and a longer follow-up may be needed to demonstrate differing risks for this low-incidence cancer type. The risk of NMSC was increased in thiopurine continuers, as has been previously demonstrated, and patients using thiopurine should continue to be instructed to use thorough sun protection and have annual dermatology visits.[42]

There are several strengths of this study. First, the study includes a large, unselected nationwide cohort. Second, we had access to data from 2003 to 2018 enabling long-term follow-up. Third, the Danish registries allow for a complete follow-up due to complete data on emigration and death, and patients are not lost to follow-up when moving within Denmark. Moreover, Danish IBD diagnoses have previously been validated and found close to complete (94%) and highly valid (90%-97%).[43] Finally, the Danish registries allow for an accurate outcome ascertainment as registration is used for reimbursement from the state, which results in a low risk of misclassification of the outcomes.

The limitations of registry-based research should also be considered. While we have high-quality data on both anti-TNF and thiopurine prescriptions, there is a risk of exposure misclassification of thiopurine continuers and discontinuers which can potentially diminish differences between the groups. We addressed this by using a strict exposure definition and only included patients who filled a thiopurine prescription within 90 days of anti-TNF initiation. Moreover, we undertook sensitivity analyses of thiopurine exposure one year after anti-TNF initiation. As differences in thiopurine exposure duration prior to anti-TNF initiation may impact treatment efficacy, we also performed sub-analyses of the duration of prior thiopurine exposure and did not find that duration of pre-anti-TNF thiopurine treatment modified the effect of combination therapy (thiopurine and anti-TNF). To account for anti-TNF discontinuation in the follow-up period, we undertook a per-protocol analysis where patients were censored at anti-TNF discontinuation. Although the total number of outcomes decreased, we still found an increased risk of the composite outcome in thiopurine discontinuers. The observational design may cause risk of bias due to lack of control for unknown or unmeasured confounders, particularly by treatment indication. The reasons for thiopurine discontinuation were unknown and may include adverse effects, patient wishes, previous malignancy, and differing clinical practices. However, by using a design where all patients had been exposed to thiopurine and initiated first treatment of anti-TNF we aimed to minimize confounding by indication. We were unable to include subgroup analyses of anti-TNF subtype, as the code "anti-TNF unspecified" was used and in addition we detected subtype misclassification to some extent (Table S6). We did not have data on smoking habits nor disease severity measures such as disease localization, extent, or endoscopic scores. However, we did address disease severity using pre-anti-TNF IBD-related hospitalization, IBD-related surgery, and corticosteroid use as proxies for disease severity and found that the two treatment groups did not differ using these metrics of disease severity. Although the use of methotrexate is rare in Denmark, patients who discontinued thiopurine may have been exposed to methotrexate instead, which may have led to fewer outcomes in this group. We cannot exclude the possibility that certain gastroenterologists more often continued thiopurine, while also having a lower threshold for hospitalization and prednisone prescription, nonetheless, the similar previous healthcare utilization in the two groups likely mitigates potential confounding by this factor. Lastly, a longer follow-up may be needed to demonstrate differing risks for major surgery, death and cancer subtypes.

In conclusion, continuing thiopurine therapy at the time of anti-TNF initiation decreased the risk of adverse outcomes, including new corticosteroid use, IBD-related hospitalization, IBD-related surgery, and death in this nationwide cohort study of 2,630 patients with IBD. Particularly, it reduced hospitalization and new corticosteroid use, key goals in the management of IBD. Clinically this study can aid patients and physicians when considering whether to continue thiopurine at anti-TNF initiation. Data from one of the recently established prospective IBD databases may be used to further guide treatment decisions in this common clinical scenario.