Thiopurine withdrawal during sustained clinical remission in inflammatory bowel disease: relapse and recapture rates, with predictive factors in 237 patients

Introduction

Thiopurines have been in clinical use for 50 years and remain the backbone of maintenance strategies for IBD, either as monotherapy or in combination with an anti-tumour necrosis factor agent.¹ Azathioprine (AZA) and its metabolite mercaptopurine (MP) are effective in maintaining clinical remission in patients with Crohn's disease (CD) and ulcerative colitis (UC).^2–8 Around 10–28% of patients report side effects (most commonly nausea) of which 50–80% will discontinue the drug as a result.^9,10 Thiopurines have a narrow therapeutic window and carry a risk of dose-dependent myelosupression^4,9,11,12 and hepatotoxicity.^10,13 A subset of the population who carry two loss of function thiopurine methyltransferase (TPMT) alleles have the greatest risk of myelosupression and serious adverse events.¹⁴ Continuous use of thiopurines has also been linked with malignancies such as lymphoma and non-melanoma skin cancer.^15,16 A large prospective study of 19 486 IBD patients showed incidence rates of non-melanoma skin cancer for current and previous AZA use at 0.66/1000 and 0.38/1000 patient years respectively (age < 50) and a cumulative increase with age.¹⁵ Beaugerie showed an incidence rate of 0.9/1000 patient years for lymphoma in those receiving AZA (n = 19 486), casting doubt on its long-term safety.¹⁶ These long-term risks make clinicians and patients wary about indefinite use of thiopurines despite the risk of relapse on withdrawal.

The relapse rates after stopping thiopurines have been reported in CD at 21–41% at 1 year with a cumulative increase to 61–85% at 5 years.^3,17–21 In UC, one randomised controlled trial and three retrospective studies showed relapse rates of 35–77% at 1 year and 65–75% at 5 years.^9,22–24 However, most of these studies had patients on treatment for a short period of time (Table1) and perhaps therefore overestimate the risk of disease relapse in patients who are in sustained clinical remission.

We aimed to examine relapse rates following thiopurine withdrawal along with predictive factors and the success of recapture in a large group of patients with at least 3 years of continuous thiopurine therapy for CD or UC.

Methods

Results

Across all centres, 264 patients were submitted. 27 were excluded, for reasons detailed in Table S1, leaving 237 patients, 129 with CD and 108 with UC, in the primary analysis (Table2; breakdown by study centre in Table S2). The median duration of thiopurine use prior to drug withdrawal was 6.0 years (IQR 4.4–8.2) for CD and 5.8 years (IQR 4.5–8.5) for UC. The median follow-up post drug withdrawal in those without relapse was 32 months (IQR 24–51) for CD and 36 months (IQR 21–52) for UC. Median CRP was 4.0 mg/L (IQR 2.5–6.0) in CD and 2.5 mg/L (IQR 2.5–4.0) in UC (Table3).

All patients were in sustained clinical remission at the time of thiopurine withdrawal; 35/237 (22 CD; 13 UC) had an additional trigger for drug cessation (Table S3). Thiopurines were tapered prior to withdrawal in 87 patients (37%). Data on length of taper were available in 48 of these patients, with a median duration of 12 weeks (IQR 8–26).

Discussion

In patients with CD, our study shows a moderate-to-severe relapse rate of 24% at 1 year and 39% at 2 years after thiopurine withdrawal. This is similar to published series showing a relapse rate of 21–41% with a cumulative increase in the rate with time (Table1). In addition, our study demonstrates a significant greater relapse rate in CD patients compared to UC. While this study does not address the rate of flare in those continuing therapy, a recent meta-analysis showed that the odds ratio of a flare in those stopping azathioprine versus those continuing was 0.15 [95% Confidence Interval (CI) 0.05–0.44] at 12 months and 0.30 (95% CI 0.08–1.23) at 18 months.²⁵

Previous studies have shown that CRP of 20 mg/L or higher, a neutrophil count of 4.0 × 10⁹/L, a haemoglobin (<12 g/dL), male gender, age ≤31 and duration of remission less than four years were factors predictive of relapse.^17,18 However, duration of AZA/MP use and the definition of remission varied with each study making it difficult to compare their study outcomes. Fraser et al., with n = 222 patients (79 CD, 143 UC), found no correlation between disease flare and clinical or laboratory indices.⁹ Our study shows that CRP is highly predictive of relapse in this cohort, a finding similar to Lemann et al.¹⁹ Tapering of thiopurine prior to withdrawal was also noted here to be associated with relapse, but practice with relation to tapering was quite different between the included centres and it is likely the observed differences in relapse rates relate to other, unmeasured factors rather than tapering itself.

In patients with UC, our study showed a lower relapse rate of 11% at 12 months and 21% at 24 months. This contrasts with a relapse rate after drug withdrawal in other published studies as high as 35–77% at 1 year and 65–75% at 5 years (Table1). We used strict criteria to define sustained remission which included continuous thiopurine use for a minimum ≥3 years and subsequent withdrawal when in sustained clinical remission (absence of symptoms and no corticosteroids for >6 months). This will have impacted the subsequent relapse rates.

In the UC cohort, our study shows that a raised white cell count is highly predictive of a relapse after drug withdrawal. Hawthorne et al. performed a small RCT trial and found younger age to be the statistically significant predictive factor for relapse.²⁴ Cassinoti et al. performed a multicentre observational study of 127 UC patients and found that relapse during treatment with AZA, withdrawal of AZA due to drug toxicity and disease extent to be predictive of disease relapse at drug withdrawal. Patients in this study had concomitant aminosalicylates, masking the true effects of AZA.²³ On the contrary, a large single centre study with 143 UC patients did not show any factors predictive of relapse.⁹

The definition of clinical remission is important when evaluating drug withdrawal studies. Studies have used various clinical disease activity indices and laboratory markers to define clinical remission. Two randomised controlled trials used the Crohn's Disease Activity Index (CDAI),^19,21 while others used the Harvey Bradshaw Index (HBI).^9,17,20 O'Donoghue et al. and Lobel et al. utilised the Physician Global Assessment (PGA) score to define remission and disease flare in CD.^3,22 We used the PGA clinical index and corticosteroid use (in the last 6 months) to define remission.

Recapture data have only been reported by Treton et al. in CD patients where 22 of the 23 patients were successfully retreated with AZA.¹⁸ Although a small cohort, our study is the only study to show retreatment success in both disease groups. However, it should be noted that 25/29 patients with CD and moderate-to-severe relapse within 12 months (20% of the overall CD cohort) required systemic steroids, anti-TNF or hospital admission and five of these patients required resectional surgery. Further large studies are needed to ascertain re-treatment success as this would have an impact on our decisions to withdraw thiopurines.

With high cumulative relapse rates after thiopurine withdrawal in sustained remission, devising a set of key relapse indicators that encompass clinical, endoscopic and laboratory markers would be beneficial. Our study highlights the importance of risk stratification in patients before considering drug withdrawal. The knowledge of these predictive factors may be translated onto the anti-TNF group of patients; however, multicentre trials are required to validate this. The STORI study, looking at infliximab withdrawal in CD remission showed that the presence of no more than two risk factors (a combination of clinical and biological markers) carried a 15% risk of relapse at 1 year.²⁶ Similarly, risk stratification in patients on long-term AZA/MP treatment who are risk of disease flare post drug withdrawal can be adopted in clinical practice. Future studies could use pre-defined risk groups to assess relapse rates post drug withdrawal.

In addition to risk factors for disease relapse, adverse events with long-term use must be taken into account when considering drug withdrawal. The small but definite association of non-melanoma skin cancer and lymphoma with long-term thiopurine use has been reported.^15,16 The risk of non-melanoma cancer is greater when treating older patients with IBD.¹⁵ In addition, the comorbidity rate is significantly higher in the elderly group (age >65 years) with IBD.^27,28 With an ageing population worldwide, the number of older patients with IBD is also expected to increase.²⁹ Therefore, treatment strategies with thiopurines would need further evaluation and a careful consideration.

The key strengths of our study are threefold. Our study is one of the largest to date looking at AZA/MP withdrawal. While many studies used varied parameters to define remission, we used strict clinical parameters with at least 3 years of continuous thiopurine use prior to drug withdrawal.

Patients within this study were selected for withdrawal by their physicians on the basis of their assessment, and so the withdrawal rates may not be generalisable to all patients in clinical remission. For example, physicians may have been less likely to withdraw patients with perianal disease or rectal disease. There were also limited data available on faecal calprotectin; it is likely that as an accurate marker of endoscopic disease activity³⁰ it would prove highly useful in predicting relapse in this context, as has been seen for infliximab withdrawal in the STORI study.²⁶ The study may have been underpowered to fully assess the predictive power of all of the factors assessed.

Thiopurines remain an integral part of disease management in IBD patients with evidence of its role in sustaining long-term remission. However, bearing in mind the side effects and risks of malignancy with long-term immunosuppression, it is crucial to identify a sub-cohort who are at highest risk of disease flare. Our study and data from the STORI trial suggest that patients in clinical and biochemical remission have a low risk of relapse. Of those who relapse after drug withdrawal, reintroduction of thiopurines allows recapture in the majority of IBD patients, particularly in UC. However, in a select group of patients (CD cohort), long-term thiopurines may be in their best interest, especially if the consequences of disease flare have an impact on morbidity and subsequent remission rates.

Authorship

Guarantor of the article: Dr Charlie Lees.

Author contributions: NAK and CWL conceived the study. PI, JM, MP, TA, JRFC, IA, JS, AJL, MS, JOL, CWL co-ordinated data collection at their respective sites. BW, CJG, RM, SR, RD, NH, RF, SM, SMS, CAL, HAH, DG collected the data. NAK aggregated the data and performed analysis. RK wrote the initial draft of the manuscript. NAK and CWL co-ordinated revision of the manuscript as guided by all of the authors who approved the final version of the manuscript.

Supporting Information

Additional Supporting Information may be found in the online version of this article: