Article Text

Download PDFPDF

Original article
Impact of thiopurines on the natural history and surgical outcome of ulcerative colitis: a cohort study
  1. Carl Eriksson1,
  2. Sara Rundquist1,
  3. Yang Cao2,3,
  4. Scott Montgomery2,4,5,
  5. Jonas Halfvarson1
  1. 1 Department of Gastroenterology, Faculty of Medicine and Health, Örebro University, Örebro, Sweden
  2. 2 Clinical Epidemiology and Biostatistics, School of Medical Sciences, Örebro University, Örebro, Sweden
  3. 3 Unit of Biostatistics, Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden
  4. 4 Clinical Epidemiology Unit, Department of Medicine, Karolinska Institutet, Stockholm, Sweden
  5. 5 Department of Epidemiology and Public Health, University College London, London, UK
  1. Correspondence to Dr Carl Eriksson, Department of Gastroenterology, Faculty of Medicine and HealthÖrebro, University, Örebro SE-70182, Sweden; carl.eriksson{at}regionorebrolan.se

Abstract

Objective Thiopurines are used as maintenance therapy in ulcerative colitis (UC), but whether these drugs influence the natural history of the disease is unknown. We aimed to assess the effect of thiopurines in terms of colectomy, hospital admission, progression in disease extent and anti-tumour necrosis factor (TNF) therapy within 10 years from initiation.

Design Patients diagnosed with UC within the Örebro University Hospital catchment area, during 1963–2010, who initiated thiopurines (n=253) were included. To overcome the risk of confounding by indication, we compared patients who stopped treatment within 12 months because of an adverse reaction (n=76) with patients who continued therapy or discontinued due to other reasons (n=177) and assessed long-term outcomes using Cox regression with adjustment for potential confounding factors.

Results The cumulative probability of colectomy within 10 years was 19.5% in tolerant patients compared with 29.0% in intolerant (adjusted HR 0.49; 95% CI 0.21 to 0.73). The probability of hospital admission was 34.0% in tolerant versus 56.2% in intolerant patients (adjusted HR 0.36; 95% CI 0.23 to 0.56). The risk for progression in disease extent was 20.4% in tolerant patients compared with 48.8% in intolerant (adjusted HR 0.47; 95% CI 0.21 to 1.06). Within 10 years, 16.1% of tolerant and 27.5% of intolerant patients received anti-TNF therapy (adjusted HR 0.49; 95% CI 0.26 to 0.92).

Conclusion Based on the novel approach of comparing patients tolerant and intolerant to thiopurines, we reveal that thiopurines have a profound beneficial impact of the natural history and long-term colectomy rates of UC.

  • chronic ulcerative colitis
  • azathioprine
  • 6-mercaptopurine
  • tnf-alpha
View Full Text

Statistics from Altmetric.com

Significance of this study

What is already known on this subject?

  • Emerging evidence indicate that colectomy rates in UC are decreasing.

  • The decrease in colectomy rates may reflect improved medical treatment.

What are the new findings?

  • Patients tolerating thiopurines are at decreased risk of colectomy, hospital admission, progression in disease extent and of anti-tumour necrosis factor exposure within 10 years of commencement compared with patients discontinuing therapy within the first year due to adverse drug reactions.

How might it impact on clinical practice in the foreseeable future?

  • We provide long-term population-based data that show that thiopurines have a pronounced impact on the natural history of UC.

  • Our novel data strongly support the role of thiopurine drugs as maintenance therapy of UC.

Introduction

The clinical course of UC varies substantially between individuals, ranging from an indolent disease with minimal symptoms to a severe disease that strongly interferes with the patients’ daily life and has a pronounced adverse impact on quality of life.1 2 Most patients with UC achieve long-term control of symptoms by medical therapy and emerging data from population-based cohorts show that colectomy rates have decreased during recent decades, potentially reflecting improved medical management.3–5

The thiopurine drugs, azathioprine, 6-mercaptopurine and 6-thioguanine, are often used as second-line therapy in UC.6 A recent Cochrane Review concluded that thiopurines may be effective as maintenance therapy for patients who do not respond or cannot tolerate 5-aminosalicylates and for patients who require repeated courses of steroids.7 However, adverse drug reactions to thiopurines frequently occur, most commonly within 3 months from initiation of treatment.8–11 The impact of thiopurines on the long-term disease course and risk of colectomy is still largely unknown. In a recent population-based study, from Örebro University Hospital, we observed that patients diagnosed with UC in 1991–2005 had a 39% lower relative risk for colectomy within 10 years of diagnosis, compared with patients diagnosed in 1963–1975.5 In parallel with the decrease in colectomies, we found an increased use of thiopurines and anti-tumour necrosis factor (TNF) agents. Increased use of thiopurines in UC has also been reported in several other cohorts.5 12–14 The fact that patients exposed to thiopurines represent a treatment refractory subset of patients challenges the assessment of long-term effectiveness of thiopurines in a real-world setting since the thiopurine-exposed population have a different risk for colectomy and other long-term outcomes compared with the general UC population.15

We aimed to explore if thiopurine treatment influences the natural history of UC. More specifically, we wanted to assess whether the treatment reduced the long-term risk of colectomy, hospital admissions, progression in disease extent and need of anti-TNF therapy in a population-based cohort of patients with UC. In order to avoid differential bias due to confounding by indication, a novel study approach was applied. We only included patients exposed to thiopurines and compared patients who stopped treatment because of adverse drug reactions within 12 months from initiation of treatment with patients who continued treatment or who discontinued the therapy because of other reasons.

Materials and methods

Patients

Information on all patients diagnosed with UC within the catchment area of Örebro University Hospital in 1963–2010 who had received at least one dose of a thiopurine drug (azathioprine, 6-mercaptopurine or 6-tioguanin) at any time from the diagnosis until the end of follow-up on 31 December 2015 was extracted from the Örebro UC cohort. The cohort and the study area have been described in detail elsewhere.5 16 17 In short, the cohort consists of 1007 incident patients diagnosed with UC. Using the personal identification number, unique to every person in Sweden, we followed patients over time. The primary catchment area of Örebro University Hospital is located in central Sweden. In total, the region had 189 603 inhabitants in 2010.18 There are no private gastroenterologists in the area, and all individuals with a suspected or verified IBD are referred to the University Hospital. Overall, 1405 patients (UC, n=899; Crohn’s disease, n=506) from the primary catchment area were registered at the outpatient clinic, Department of Gastroenterology, by 31 December 2010. Flexible colonoscopy was implemented at the unit in 1976 and has since been the preferred way to determine the disease extent.5 The use of thiopurine drugs in UC began during the mid-1980s, while anti-TNF treatment was approved for UC in Sweden in 2005 and intravenous anti-TNF has since then been provided at the infusion unit, integrated within the outpatient clinic.

Data collection

The medical records of all patients in the study population were assessed using a standardised case report form including all surgical procedures, progresses in disease extent, hospital admissions because of UC or serious infections, information on medical therapy and cancer diagnoses. Admissions because of other reasons than UC and serious infections were not recorded. Consequently, admissions due to adverse drug reactions to thiopurines or infusions of anti-TNF agents were not accounted for. The following classification was used for medical treatment: oral 5-aminosalicylates (balsalazide, mesalazine, olsalazine and sulfasalazine), systemic corticosteroids (betamethasone and prednisolone), thiopurines (azathioprine, 6-mercaptopurine and 6-thioguanine), methotrexate and anti-TNF agents (infliximab, adalimumab, golimumab and certolizumab). The reason for discontinuation of thiopurine drugs was classified according to the following criteria: lack of effect or loss of response, because of an adverse drug reaction, pregnancy, scheduled termination by the treating physician (including withdrawal in patients who were in sustained clinical remission) and other reason (including non-compliance with drug therapy). Adverse drug reactions to thiopurines were categorised according to a previously defined classification system11 and were grouped into myelotoxicity (anaemia, leucopenia, neutropenia or thrombocytopenia), GI intolerance (abdominal pain, nausea, vomiting or diarrhoea), hepatotoxicity (all types of hepatotoxic reactions), allergic reactions (arthralgia, myalgia, fever, rash, general malaise) or pancreatitis.

Study design and outcome measurements

All patients were followed from the commencement date of the first thiopurine drug until emigration, death or end of follow-up on 31 December 2015. Patients were divided into two groups based on if they terminated their first thiopurine drug, because of an adverse drug reaction, within 12 months from initiation of treatment (yes/no). Patients who stopped treatment due to an adverse drug reaction within this period were defined as intolerant and patients who continued therapy with thiopurines or stopped treatment due to other reasons were classified as tolerant.

The primary outcome was the cumulative probability of colectomy within 10 years of commencement of a thiopurine drug. Secondary outcomes included cumulative probability of a hospital admission because of UC, a progression in disease extent, cumulative probability of anti-TNF therapy and long-term adverse outcomes of thiopurine therapy (death, lymphoma, non-melanoma skin cancer, solid tumours and infections requiring hospital admissions).

Statistical analysis

Age, disease duration and median time on a thiopurine drug are presented as median and IQR. Differences in demographical and clinical characteristics at baseline were evaluated using Mann-Whitney U test for continuous data with skewed distribution, χ2 tests for categorical data or Fisher’s exact test when appropriate. Life tables, survival plots and the cumulative probability of colectomy, hospital admission because of UC, progression in disease extent, anti-TNF exposure and of long-term adverse outcomes were constructed using Kaplan-Meier analysis, and the log-rank test was used to compare groups. As thiopurines have a slow onset of action,19 all events that occurred within 3 months from commencement were censored and the start of follow-up was delayed with 3 months. Consequently, patients who underwent colectomy or had a first exposure to anti-TNF agents within 3 months were excluded from the relevant analysis, while patients who experienced a potentially recurrent event as a progress from proctitis to left-sided colitis or a hospital admission remained in the analysis with the event censored. Cox proportional hazards regression models were used to evaluate the effect of the independent variables on time to events of interests. Covariates (sex, age, hospital admission at diagnosis, disease extent and previous medication at first thiopurine exposure) were selected on the basis of their known associations with the outcomes. Disease duration at baseline was not entered in the final models as high collinearity with age was observed. Further, as the study period corresponds to major changes in the management of UC, all main analyses were adjusted for the year of thiopurine initiation in 5-year intervals (1981–1985, 1986–1990, 1991–1995, 1996–2000, 2001–2005, 2006–2010, 2011–2015). The proportionality of the hazards assumption was verified for all analyses. In order to preclude bias, sensitivity analyses were conducted (a) with all events occurring during the first 12 months censored and start of follow-up delayed for12 months and (b) with no left-censoring or delay of follow-up. All tests were two-tailed and a P <0.05 was considered statistically significant. Ninety-five per cent confidence intervals (CIs) are provided for all outcome measurements. Statistical analyses were performed using SPSS V.22 (IBM, Armonk, New York, USA).

Results

Of the 1007 patients diagnosed with UC in 1963–2010 within the catchment area of Örebro University Hospital, 253 patients had received at least one dose of a thiopurine drug before the end of follow-up on 31 December 2015 and were included in the study. Azathioprine was used as the first thiopurine drug in 252 patients while one patient received 6-mercaptopurine as the first thiopurine drug. Within 12 months from commencement, 76 (30%) individuals had stopped therapy because of intolerance (myelotoxicity, n=4; gastrointestinal intolerance, n=19; hepatotoxicity, n=17; allergic reactions, n=30; and pancreatitis, n=6) while 177 (70%) patients continued the therapy or stopped treatment because of other reasons. In total, 13 of the tolerant patients stopped therapy during the first 12 months (due to lack of effect or loss of response, n=7; non-compliance, n=6). Although some intolerant patients reinitiated treatment with 6-mercaptopurine or azathioprine at a reduced dose during the 10-year follow-up period and some tolerant patients stopped treatment, the median duration of thiopurine treatment within 10 years from initiation of therapy was 1.6 (IQR 0.9–9.6) months among thiopurine-intolerant patients compared with 86.4 (IQR 31.2–120.0) months among tolerant patients. Clinical and demographical characteristics of the patients with UC, stratified by tolerance, are described in table 1.

Table 1

Clinical and demographical characteristics of patients with UC

Colectomy within 10 years from first thiopurine exposure

In the analysis of cumulative rate of colectomy within 10 years, 11 patients were excluded (thiopurine tolerant, n=6; intolerant, n=5) as they underwent colectomy before or within 3 months from commencement of thiopurines. The risk of colectomy within 10 years was 19.5% (95% CI 12.8 to 26.2) in patients tolerating thiopurines compared with 29.0% (95% CI 17.8 to 40.2) in the patients who stopped treatment because of intolerance (figure 1). In the Cox proportional hazards regression, continued thiopurine use was associated with a reduced risk for colectomy after adjustments for sex, age, disease extent, previous medical therapy, hospital admission at diagnosis and year of thiopurine initiation in 5-year intervals (adjusted HR 0.39; 95% CI 0.21 to 0.73) (table 2). The association remained in the sensitivity analyses, both when all events that occurred within 12 months from commencement were censored (adjusted HR 0.40; 95% CI 0.20 to 0.79) and when no data were censored (adjusted HR 0.35; 95% CI 0.20 to 0.62) (see online supplementary tables S1 and S5).

Figure 1

Kaplan-Meier plot demonstrating the cumulative probability of colectomy in patients tolerating thiopurines and of patients who discontinued treatment because of an adverse drug reaction.

Table 2

Risk factors for colectomy within 10 years*

Hospital admission within 10 years from first thiopurine exposure

The cumulative probability of hospital admission because of UC or UC-related conditions within 10 years of first thiopurine exposure was 34.0% (95% CI 26.6 to 41.4) among tolerant patients compared with 56.2% (95% CI 44.2 to 68.2) in patients who discontinued treatment due to adverse reactions (figure 2). In the Cox regression model, thiopurine tolerance was associated with decreased risk for hospital admission after adjustment for potential confounders (adjusted HR 0.36; 95% CI 0.23 to 0.56). In the same model, paediatric patients (HR 1.77; 95% CI 1.08 to 2.90) and patients exposed to anti-TNF agents (HR 2.60; 95% CI 1.38 to 4.90) seemed to be at increased risk for hospital admission in the univariate analysis (table 3). However, after adjustment for potential confounding factors, only the association between use of anti-TNF agents and hospital admission remained significant (adjusted HR 2.88; 95% CI 1.43 to 5.81). Consistently, thiopurine tolerance and anti-TNF exposure were significantly associated with the risk of hospital admission in both sensitivity analyses (see online supplementary tables S2 and S6).

Figure 2

Kaplan-Meier plot of the cumulative probability of hospitalisation in patients tolerating thiopurines and of patients who discontinued treatment because of an adverse drug reaction.

Table 3

Risk factors for hospital admission within 10 years

Progression in disease extent within 10 years from first thiopurine exposure

Patients with ulcerative proctitis (E1) or left-sided colitis (E2) at 3 months after initiation of first thiopurine agent (thiopurine tolerant, n=67; intolerant, n=35) were included in the analysis of progression in disease extent. Among thiopurine-tolerant patients, the risk of disease progression within 10 years was 20.4% (95% CI 10.4 to 30.4) compared with 48.8% (95% CI 27.4 to 70.2) in intolerant patients (figure 3). Thiopurine-tolerant patients were at decreased risk for progression in disease extent in the univariate analysis (HR 0.45; 95% CI 0.21 to 0.97) and the risk estimate was similar in the adjusted model although the association did not remain statistically significant (adjusted HR 0.47; 95% CI 0.21 to 1.06) (table 4). Consistently, similar risk estimates were observed in the sensitivity analyses, both when all events during the first 12 months of therapy were censored (adjusted HR 0.44; 95% CI 0.17 to 1.10) and when no data were censored (adjusted HR 0.43 95% CI 0.20 to 0.89) (see online supplementary tables S3 and S7). In contrast, paediatric patients were at increased risk for progression in disease extent (adjusted HR 4.50; 95% CI 1.55 to 13.08) and the association remained in both sensitivity analyses.

Figure 3

Kaplan-Meier plot demonstrating the cumulative probability of progression in disease extent in patients with ulcerative proctitis or left-sided colitis tolerating thiopurines and of patients who discontinued treatment because of an adverse drug reaction.

Table 4

Risk factors for progression in disease extent within 10 years*

Anti-TNF therapy within 10 years from first thiopurine exposure

Twenty-three patients (thiopurine tolerant, n=17; intolerant, n=6) were excluded from the analyses of thiopurine exposure and anti-TNF treatment since they had received anti-TNF agents before or within 3 months of first thiopurine exposure. At 10 years, 16.1% (95% CI 10.2 to 22.0) of the tolerant and 27.5% (95% CI 16.5 to 38.5) of the intolerant patients had received anti-TNF therapy (figure 4). Cox regression analysis showed that thiopurine tolerance was associated with decreased risk of anti-TNF treatment after adjustments for potential confounders (adjusted HR 0.49; 95% CI 0.26 to 0.92) (table 5). The association between thiopurine tolerance and anti-TNF exposure remained when no data were censored, but was not significant when all events within 12 months from initiation of thiopurines were censored (see online supplementary tables S4 and S8). In contrast, paediatric disease (adjusted HR 3.60; 95% CI 1.76 to 7.35) and year of thiopurine initiation (adjusted HR 2.26; 95% CI 1.60 to 3.19) were significantly associated with the risk for anti-TNF exposure (table 5). The observed associations remained in both sensitivity analyses (see online supplementary tables S4 and S8).

Figure 4

Kaplan-Meier plot demonstrating the cumulative probability of anti-tumour necrosis factor (TNF) therapy in patients tolerating thiopurines and of patients who discontinued treatment because of an adverse drug reaction.

Table 5

Risk factors for anti-tumour necrosis factor (TNF) therapy within 10 years*

Long-term adverse outcomes of thiopurine drugs

Among patients exposed to thiopurines, 79 long-term adverse events were observed (lymphoma, n=3; non-melanoma skin cancer, n=12; solid tumour, n=30; death, n=34) during a median follow-up of 12 (7–17) years. No significant differences were observed between thiopurine-tolerant and thiopurine-intolerant patients (data not shown).

Discussion

To our knowledge, this is the first study to assess the long-term clinical effectiveness of thiopurines in UC by applying a novel approach where patients who stopped treatment because of adverse drug reactions were compared with patients who tolerated the therapy. We observed a profound beneficial effect of thiopurine treatment on the natural history of UC, a reduced risk for colectomy, hospital admission, progression in disease extent and need for anti-TNF therapy, within 10 years from initiation of thiopurines, was observed.

Randomised controlled trials are generally regarded as the highest quality source of data on the effects of medical treatment.20 Early randomised controlled trials showed that thiopurines have a steroid-sparing effect in UC and are effective in maintaining clinical remission.21–24 More recently, a study found that azathioprine is superior to 5-aminosalicylates in inducing clinical and endoscopic remission in steroid-dependent UC.25 However, randomised controlled trials are often limited by insufficient follow-up and poor generalisability because of extensive eligibility criteria for inclusion.26 27 In contrast to randomised controlled trials, observational studies based on real-world cohorts reflect treatment practice in clinical routine and allow longer follow-up periods. However, comparative studies of real-world cohorts often suffer from differential bias because of unequal distribution of characteristics between groups. Thiopurines are recommended in patients who previously failed 5-aminosalicylates or with a steroid-dependent disease.6 This subset of patients are at increased risk of colectomy compared with the general patient with UC, regardless of exposure of thiopurines.15 28 Thus, comparisons of thiopurine-treated patients with patients without thiopurine exposure are easily confounded by indication. This is illustrated by our recently reported colectomy rate in the Örebro UC cohort. The probability of colectomy within 10 years from diagnosis was 22.4% among patients exposed to thiopurines compared with 13.5% in the entire cohort.5 To overcome confounding by indication, we applied a novel approach and included thiopurine-exposed patients only. By comparing patients who were tolerant with those who were intolerant to thiopurines, we identified two groups likely to have similar disease severity and refractoriness to prior treatments, and this assertion is supported by the similarity of clinical characteristics at initiation of therapy for both groups. Notably, patients continuing thiopurines had a 61% lower relative risk for colectomy compared with patients whose therapy ceased due to intolerance. Temporal associations between increasing use of thiopurines and decreasing colectomy rates have been observed in some population-based studies,15 29 30 but no protective effect of thiopurines has been confirmed. However, studies comparing thiopurine-treated patients with non-treated patients have shown an increased risk for colectomy among exposed patients.15 28 Chhaya et al reported a three times greater risk for colectomy for thiopurine users compared with non-users, in a large primary care database of patients diagnosed with UC during 1989–2009 in the UK.15 Consistent with our findings, patients who continued thiopurines >12 months had a 72% lower risk of colectomy compared with patients who stopped therapy within 3 months. However, the results may easily been influenced by bias since data on the reason for termination of thiopurines were not available and early discontinuation of thiopurines may be explained by colectomy.15

Hospital admissions accounts for a substantial proportion of healthcare cost of UC.31 Whether an earlier and more aggressive approach of immunosuppressive therapies is associated with reduced rates of hospital admission is yet to be determined.32–34 The hospital admission rate and length of stay for UC remained stable in Canada between 1991 and 2001, while the rates for Crohn’s disease dropped.32 Interestingly, we observed a 64% decreased risk for UC-related admissions in thiopurine-tolerant patients. This finding is contrary to previous data reported from Denmark, where no protective effect of thiopurines was observed in an inception cohort of 300 patients with UC followed for 7 years.35 Randomised controlled studies have shown a decreased risk of hospital admission in anti-TNF-treated patients,36 whereas we observed a 1.9-fold increased risk for hospital admissions in patients who were exposed to anti-TNF at the initiation of thiopurine treatment. Our finding might be influenced by confounding by indication, that is, that anti-TNF-exposed patients represent a selected group of treatment refractory patients.

Some data indicate that treatment with anti-TNF agents and immunomodulators may delay progression in disease behaviour of Crohn’s disease,37 but to our knowledge, preventive effects of medication on extension of disease extent have not been demonstrated in UC. Consistent with a Danish population-based study,38 we have previously reported that 34.5% of patients with ulcerative proctitis and 18.5% of patients with left-sided colitis in the Örebro UC cohort had progressed in disease extent 10 years after diagnosis, but no temporal trend was observed.5 The present study found that patients who tolerate thiopurines seemed to be at decreased risk for progression in disease extent compared with intolerant patients, although the result should be interpreted with caution since only 102 patients were included in the analysis and this was the only outcome that was not statistically significant after adjustment for calendar period of treatment initiation.

The efficacy of anti-TNF agents in the treatment of UC has been widely documented.39–42 Historically, anti-TNF therapies have often been used as third-line treatment, when 5-aminosalicylates (5-ASA) and thiopurines have failed. Similarly to the observed reduced risk of colectomy and of hospital admission, our study indicates that patients tolerating thiopurine treatment are less likely to require anti-TNF therapy within 10 years from initiation of treatment compared with intolerant patients. Patients who were thiopurine intolerant seemed to start anti-TNF therapy soon after termination of the thiopurine drug in the year following initiation since no significant association between thiopurine intolerance and increased risk for anti-TNF therapy was observed, when all events during the first 12 month were censored (online supplementary table 4).

In our cohort, paediatric patients had a more severe disease course with increased risk for progression in disease extent, anti-TNF exposure and hospital admissions. This observation is supported by a recent national population-based study of 27 834 persons diagnosed with IBD in Sweden as well as by a cohort study conducted in the Netherlands.43 44 The underlying biology behind this observation remains largely to be explored, although differences in medical management may contribute since paediatric patients generally are taken care of by paediatricians in contrast to adult patients, who are managed by gastroenterologists.

The main strength of the present study is the study design, where we made use of the fact that 9%–34% of thiopurine-exposed patients stop the treatment because of adverse drug reactions8–11 and compared patients who continued treatment with those who discontinued because of adverse drug reactions, to avoid confounding by indication. Adverse drug reactions to thiopurines most often occur soon after initiation of treatment, that is, within the first 3 months, and rarely beyond 12 months.11 We defined thiopurine intolerance as discontinuation of thiopurines because of an adverse drug reaction within 12 months from first commencement. Correspondingly, thiopurine tolerance was defined as continued therapy for 12 months or termination due to any other reason. In order to prevent potential bias, we did not take into account that some patients may have reinitiated thiopurines at a reduced dose or restarted treatment with another type of thiopurine drug during follow-up. For the same reason, we did not take into account that adverse drug reactions may theoretically have occurred beyond 12 months. However, the observed difference in median duration of thiopurine treatment within 10 years from initiation of the therapy (86.4 (IQR 31.2–120.0) months among tolerant patients and 1.6 (IQR 0.9–9.6) months among intolerant patients) supports the effectiveness of our study design. Theoretically, thiopurine-tolerant patients who had their thiopurine therapy stopped by the prescribing physician (n=6) or who discontinued therapy because of non-compliance (n=13) during the entire follow-up may represent patients with less severe disease. However, after excluding these groups and re-running all main analyses, results entirely consistent with the main analysis were obtained (data not shown). Since thiopurines have a slow onset of action, any outcome event that occurred within 3 months from initiation of treatment was censored and we started the follow-up at 3 months from commencement of treatment in all analyses.19 To provide further evidence of an effect, we performed a sensitivity analyses, where the start of follow-up time was postponed to 12 months from initiation of thiopurines and with no censoring of data.

The study period used here corresponds to a time when major changes in the management of UC occurred and the availability of other treatment options for thiopurine-intolerant patients differed over time. The first anti-TNF agent, infliximab, was approved in UC in 2005, but it is difficult to define a specific year for a widespread use since the anti-TNF treatment has gradually increased.5 12 Therefore, all main analyses of this study were adjusted for the year of thiopurine initiation in 5-year intervals. However, assessment of long-term outcomes requires long follow-up periods. Therefore, the study may be limited by the fact that patients diagnosed after 31 December 2010 were not included and that most of the study period took place in the pre-biologic era.

This study is strengthened by the population-based design with the inclusion of all patients of all ages with UC in a well-defined geographic area as well as by the access to detailed information on medical therapy including reasons for discontinuation of treatment. Further, government-financed healthcare, the absence of private gastroenterologists in the area and the use of the personal identity number, unique to every inhabitant, allowed us to follow patients over a 10-year period.

The main potential limitations of the study are the retrospective non-controlled study design, the limited number of patients included and the fact that our findings are not yet replicated in an independent cohort. Future studies, applying our study design on other cohorts, should be encouraged to confirm the impact of thiopurines on the risk of adverse outcomes of UC. Information on the concentration of thiopurine metabolites, thiopurine methyltransferase (TPMT) genotype and phenotype was not available for all patients. Thus, the extent to which thiopurine-tolerant patients were compliant can only be speculated on, but the treatment was prescribed by their treating gastroenterologist. Adverse events during thiopurine use have been attributed to several different mechanisms, including both individual genetic variation and external factors as concurrent medications and viral infections.45 Polymorphisms in the genes encoding for TPMT and inosine triphosphate pyrophosphatase have been associated with increased risk of adverse drug reactions to thiopurines.46 47 Interestingly, low TPMT activity (<35 pmol/h/mg/Hb) has also been associated with a greater clinical response.48 To our knowledge, none of the genetic variants associated with TPMT deficiency have been associated with a severe disease course of UC or the risk of colectomy. Correspondingly, baseline comparisons of previous medical and surgical management, including past colectomy with subsequent ileorectal anastomosis as well as previous anti-TNF exposure, of thiopurine-tolerant and thiopurine-intolerant patients revealed no significant differences between the two groups (table 1). However, no detailed data on total steroid exposure, number of visits at the outpatient clinic and socioeconomic characteristics of the included patients were available. A steroid-sparing effect of thiopurines was observed in a recent Cochrane systemic review,7 and theoretically, differences in steroid exposure between the groups could represent an explanatory factor for outcomes such as hospital admissions and anti-TNF exposure.

In contrast to the previously reported associations between thiopurine exposure and non-melanoma skin cancer, as well as lymphoma,49 50 we did not observe any differences in long-term adverse outcomes of thiopurines between thiopurine-tolerant and thiopurine-intolerant patients. The absence of association may be explained by the limited number of patients in our cohort.

In conclusion, this population-based cohort study indicates that thiopurine therapy is associated with significant benefits on the natural history of UC, including reduced risk of colectomy, hospital admission, progression in disease extent as well as need of anti-TNF treatment. The results using this unique approach strongly support the role of thiopurine drugs as maintenance therapy for UC.

Acknowledgments

We acknowledge the assistance of Margareta Lundin and Lena Eriksson with data collection.

References

View Abstract

Footnotes

  • Contributors CE, JH and SM planned and conducted the study. CE and SR collected the data. CE and YC performed the data analyses. CE, JH, SM, SR and YC interpreted the data. CE and JH drafted the manuscript. CE, JH, SM and YC revised the manuscript. CE, JH, SM, SR and YC approved the final draft of the manuscript.

  • Funding CE was financially supported by the Swedish government’s agreement on medical training and research (OLL-549221).

  • Competing interests CE: research grant from the Swedish government’s agreement on medical training and research and speaker’s fee from Takeda. JH: research grants from the Swedish Research Council, the Örebro University Hospital Research Foundation, the Swedish Foundation for Strategic Research and speaker’s and/or consultancy fees from Abbvie, Hospira, Janssen, Medivir, MSD, Pfizer, RenapharmaVifor and Takeda.

  • Patient consent Not required.

  • Ethics approval The study was approved by the Uppsala Regional Ethics Committee in 2010 (approval no. 2010/304).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.