Statistics from Altmetric.com
Thiopurines have been considered the reference treatment for patients with steroid-dependent moderate to severe IBD for many years. This has been based on evidence describing the efficacy and safety of azathioprine and mercaptopurine in Crohn's disease (CD) and UC. In CD, efficacy has been established through controlled trials, meta-analyses, cohort studies and data on mucosal healing, as well as in withdrawal studies. A Cochrane meta-analysis concluded that azathioprine and mercaptopurine effectively maintained remission in CD with an OR between 2 and 3 compared to placebo.1 Mucosal healing was initially studied in patients on azathioprine with longstanding steroid-free remission.2 In this setting, full mucosal healing was achieved in 54% of patients with ileitis and 70% with colitis. The majority of the remaining patients had at least partial healing, and only a small minority experienced no healing at all. This healing potential was confirmed in a controlled study showing the superiority of azathioprine over budesonide.3 When patients achieve sustained steroid-free remission under azathioprine, it is usually stable and longstanding. A controlled withdrawal study from the GETAID indicated sustained remission over more than 5 years in 80% of patients when the treatment was continued, compared to 40% of patients when it was stopped.4
The evidence base in UC is weaker. Nevertheless, a controlled trial comparing azathioprine to mesalazine in steroid-dependent UC clearly showed the superiority of azathioprine with an OR of 4.8 (95% CI 1.6 to 14.5).5 This benefit was confirmed in a meta-analysis.6 Additionally, a withdrawal study, although not controlled, showed around 60% of patients with UC relapse over 5 years after azathioprine withdrawal.7
In CD, these data have also been reinforced by cohort studies suggesting a beneficial effect of thiopurines on long-term outcome, including a decrease in the rate of surgery in adult and paediatric patients.8 ,9 This benefit has, however, not been uniformly described, and a recent, large, population-based study from Denmark did not show any convincing surgery-sparing effect of azathioprine, despite a significant reduction in surgery rates in successive cohorts of patients evolving in parallel with an increased use of thiopurines and anti-tumour necrosis factor (TNF).10 Similarly, it is important to note that other data question the efficacy of thiopurines when used early in adult CD patients,11 ,12 supported by disappointing mucosal healing rates in SONIC for CD13 or in SUCCESS for UC,14 and by safety issues, particularly regarding the risk of lymphomas, skin cancers and maybe other tumours, as well as liver toxicity. Nevertheless, a recent meta-analysis, demonstrated a 40% decrease in the risk of first surgical resection.15
The RAPID trial failed to show superiority of early azathioprine compared to classical step-up therapy in patients with risk factors for the development of disabling disease.11 However, early azathioprine was associated with a significant reduction in new perianal fistula development. Furthermore, the outcome in the classical step-up group has to be interpreted cautiously, bearing in mind that the majority of the patients in this group were also rapidly treated with azathioprine (around 60% within 1 year) and that the remainder never required this drug even after a median follow-up longer than 2 years suggesting in these cases a mild disease course. Similarly, in the AZTEC trial, azathioprine was not superior to placebo at inducing sustained steroid-free remission in early CD.12 Nevertheless, here also, sustained steroid-free remission was achieved with placebo in more than one-third of the patients highlighting a benign disease course in a significant number of them. Furthermore, a posthoc analysis suggested a significantly lower rate of moderate-severe relapse (CDAI>220) with azathioprine compared with placebo (12% vs 30%). Finally, the mucosal healing rate in the SONIC and SUCCESS trials in the azathioprine monotherapy arms was only around 15%, which may be considered insufficient when the main objective of the treatment of IBD has become one of tissue healing.
Therefore, and due to the relatively better efficacy of anti-TNF and perhaps newer biologics, as well as a possibly better safety profile, some are questioning the use of purines in IBD.
However, thiopurines have several advantages including their low cost and the quality and stability of the remission they induce in patients who respond to and tolerate them. In patients who achieve stable clinical remission on thiopurines, annual loss of response is around 5% in CD4 which compares favourably to the rate seen in patients on anti-TNF, ranging between 10% and 15% per year16 and which occurs mainly due to immunisation.17 ,18 Patients achieving such remission may represent at least 20% of those initially started on thiopurines. Thus, in our view it would be a mistake to abandon what is still a valuable drug in IBD. However, we consider that the time has come for improved selection of patients who receive thiopurines, as well as better timing and optimisation of thiopurine use to improve the benefit/risk profile.
When considering the use of an immunosuppressant and/or biologic therapy in the setting of steroid dependence or intolerance, it is important to assess the severity of the underlying disease and the potential consequences of a period of time without full disease control, such as intestinal tissue remodelling and loss of function leading to surgical resections. Therefore, patients with the most severe disease, characterised by early disease onset, extensive and deep intestinal lesions or complex fistulising perianal disease, should preferably be rapidly treated using combination therapy with an immunosuppressant and a biological to maximise the chance of rapid and total control of the disease. Likewise, patients treated with a purine analogue who do not achieve steroid-free remission and a reassuring degree of intestinal healing within 12–24 weeks should not continue on this insufficient regimen. Knowledge accumulated over recent decades on purine metabolism allows clinicians to rapidly recognise patients at risk of poor response or intolerance, and to adapt the treatment accordingly, thus limiting the risk of insufficient disease control and progressive tissue damage.
Traditionally, in clinical practice and in drug trials, thiopurine dosing has been based on body weight with doses of up to 2.5 mg/kg of azathioprine and 1.5 mg/kg of mercaptopurine being recommended.19 However, this may not represent optimal use. Thiopurine metabolism is dependent on the activity of a number of enzymes under the influence of genetic variation. The therapeutically active product is the thioguanine nucleotides (TGN). However, thiopurines are also metabolised to methylated metabolites (MeMP) which are associated with hepatotoxicity and have little or no therapeutic effect. It is possible to measure thiopurine metabolites using high-pressure liquid chromatography,20 and to use the results to optimise thiopurine therapy.
A number of studies have suggested that TGN concentrations correlate with thiopurine efficacy, including a recent meta-analysis describing an OR of remission of 3.15 (95% CI 2.41 to 4.11) in patients with a TGN concentration >230–260 pmol/8×108 erythrocytes.21 Studies investigating the effects of optimising thiopurine dose based on TGN concentration have shown beneficial effect,22–25 while others have shown that TGN-guided management is superior to unguided management and is inexpensive, particularly if used in non-responders, being estimated at approximately £40 per treatment change.22 ,26
However, only two randomised trials have compared dose optimisation of thiopurines with non-guided management, with neither showing a benefit of using TGNs. The first has been criticised for methodological issues,27 while the second was terminated early due to recruitment difficulties. Nevertheless, in the latter underpowered trial there was a signal that TGN-guided management was superior, 40% of subjects in the guided arm being in clinical remission at week 14 compared with only 16% in the weight-based arm.28
Additionally, neither of these studies addressed the problem of hypermethylation of thiopurines which occurs in 15–20% of patients29 and results in shunting away from the production of TGNs towards the non-therapeutic methylated metabolites. Hypermethylator profiles are associated with poor response to thiopurine therapy.24 ,30 ,31 However, by using reduced doses of thiopurines in conjunction with allopurinol, a xanthine oxidase inhibitor that was originally developed with the aim of optimising thiopurine metabolism, it is possible to correct this shunt in almost all patients resulting in improved disease activity in many.30–32 In addition to correcting hypermethylation, allopurinol cotherapy can be used for patients with side effects other than hepatotoxicity, such as nausea, myalgia and fatigue.33 Thus, it has been estimated that the use of combination therapy may increase the proportion of patients who can tolerate thiopurines and in whom it is effective by at least 12%.30 Nevertheless, this combination has only been assessed in observational studies including small number of cases, and the patients should be carefully informed about the risk of toxicity when using it and should also be proposed alternative therapeutic options.
Less attention has been paid to the upper limit of TGN concentration, higher levels being more likely to result in myelosuppression and, at least in theory, in other immunosuppression-related side effects such as infection and lymphoma. However, dose reduction in patients with levels >550 pmol/8×108 erythrocytes, appears not to increase the chance of losing response.22
Therefore, the argument for measuring thiopurine metabolites in patients not responding to adequate doses of azathioprine after 3–6 months of therapy is now clear. Four profiles can be identified, each directing different management. Patients with very low or undetectable metabolites are likely to be non-adherent and should have this addressed; patients with subtherapeutic TGNs without a hypermethylator profile are underdosed, and should have their dose increased; patients with subtherapuetic TGNs and a hypermethylator profile could be considered after full patient information for a treatment with low-dose thiopurine and allopurinol (with continued metabolite monitoring); and patients with therapeutic TGNs should be treated with alternative therapy.34 Indeed, the first three of these profiles can be identified once steady state is reached at 4 weeks allowing intervention before waiting to see if treatment is effective at 3–6 months. While the case for dose adjustment based on TGN/MeMP measurement in all patients has not been made, there is an increasing volume of evidence supporting this strategy. Balanced against this is the theoretical risk that, in those patients in whom treatment is increased based on TGNs, the potential for immunosuppression-related side effects is greater. More controlled data are needed on the benefit/risk ratio of higher doses of thiopurines which may be required in some patients to achieve therapeutic levels.
The other area in which knowledge has recently accumulated is the long-term safety of purine analogues. Here also, better selection of patients and specific measures regarding patient follow-up and the duration of treatment may improve the benefit/risk profile.
Safety problems that may restrict use of azathioprine are those that threaten life or lead to irreversible damage. Myelotoxicity was quickly identified as a potential cause of drug-induced death.35 ,36 Appropriate management of the risk has been progressively refined, with or without determination of TMPT status, and the real remaining concern in this field is suboptimal implementation of guidelines.37 Irreversible liver damage can also be prevented through adequate monitoring of liver tests.38 More recently, the risk of nodular regenerative hyperplasia was identified, particularly in men exposed to thiopurines after extensive small bowel resection.39 Patients who develop a progressive drop in platelet count, liver test abnormalities or any sign of portal hypertension should have liver investigations, such as MRI, elastography and liver biopsy if necessary.
It has not been firmly demonstrated that azathioprine increases the risk of opportunistic bacterial or fungal infections, in contrast to anti-TNF agents.40 However, fatal forms of varicella41 and Epstein-Barr virus (EBV) primary infections associated with haemophagocytic lymphohistiocytosis42 have been reported in patients exposed to thiopurines. Varicella can be prevented by vaccinating non-immunised patients before initiating treatment with azathioprine.43 For EBV, the risk can be attenuated on a case-by-case basis by using anti-TNF agents and/or methotrexate in young female and male patients who are seronegative for EBV. In young males, this strategy may also prevent fatal postmononucleosis lymphoproliferative disorders which are exclusively observed in males.44 ,45 The absolute risk of this latter complication is significant (3/1000 patient-years in the Cancers et Surrisque Associés aux Maladies Inflammatoires Intestinales En France (CESAME) cohort) when considering the subgroup of patients at risk (ie, males seronegative for EBV).46
The overall excess risk of malignancies attributable to the use of thiopurines was quantified in transplant recipients in the 1970s.47 However, IBD-specific data only became available in 2013, with an overall 41%48 to 68%49 excess risk of malignancy in patients exposed to thiopurines being described, after adjustment for major confounders. While this figure may be alarming, there are distinct risks that can be addressed to avoid precluding prolonged use of thiopurines. Thiopurine use is associated with an excess risk of non-melanoma skin cancers,50 ,51 although these cancers are generally not life-threatening. The risk possibly persists after drug withdrawal,51 leading to a recommendation of lifelong sun protection and dermatological surveillance in IBD patients exposed to thiopurines. It is likely that this recommendation will be generalised to all IBD patients from diagnosis, because an increased risk of basal cell carcinoma and melanoma has been found to be associated with IBD itself in some cohorts.52 ,53 Similarly, yearly screening for uterine cervix cancer has been recommended in all women with IBD,54 although it is still unclear whether thiopurines increase the risk of cervical dysplasia.
Regarding haematogical maligancies, thiopurines promote acute myeloid leukaemia and severe myelodysplatic syndromes, although the absolute risk of such cancers is very low.55 Hepatosplenic T-cell lymphomas are non EBV-related. They are generally fatal and are mostly observed in males below the age of 35 years, who have received the combination of thiopurines and anti-TNF agents for 2 years.56 Again, the absolute risk of such lymphomas is very low and could be further reduced, where possible, by limiting the duration of combination therapy to 2 years in young males.
In adult patients exposed to thiopurines, drug-induced non-Hodgkin EBV-related lymphomas occur due to loss of control of EBV-infected B-lymphocyte proliferation by the immune system.44 These lymphomas are responsible for the twofold to fivefold increase in the risk of lymphomas attributable to thiopurines.45 ,57 ,58 Over a 10-year period of treatment, it can be estimated from the CESAME data that the cumulative risk of thiopurine-induced lymphoma is approximately 0.4% below the age of 50 years, 2% between ages of 50 and 65 years, and 4% after 65 years of age. Importantly, it has been shown that the excess risk of lymphoma attributable to thiopurines is fully reversible as soon as the treatment is stopped.45 ,57
In terms of benefit to risk ratio, it must be remembered that there is no overall excess mortality in IBD patients exposed to thiopurines.59 ,60 Additionally, IBD complications due to uncontrolled disease (severe infections originating from intestinal lesions or malnutrition, inflammation-related cancers and thrombo-embolic events) are associated with excess mortality in CD, and in UC with childhood onset.61 In this context, it has been shown that the benefits outweigh the risk of lymphoma in the first 5 years of thiopurine use in patients requiring immunosuppressive therapy in a step-up approach.62 This conclusion also applies to patients with late-onset disease over the age of 65 years.62 In this particular population, the natural history of IBD is significantly more benign than in other age classes.63 However, in older patients with clinically uncontrolled disease despite the use of 5-aminosalysilic acid (5-ASA), the use of steroids is associated with the highest rates of morbidity and mortality,40 ,64 and the use of anti-TNF is also associated with increased mortality partly attributable to an excess of opportunistic infections by intracellular agents.65 Therefore, choice of immunosuppressive agent (corticosteroids, thiopurines, methotrexate, anti-TNF) in late-onset IBD must be made case by case in a step-up approach, taking into account the general health status and comorbidities of the patient.
In patients with sustained (>5 years) controlled IBD on thiopurines, it is established that drug withdrawal is associated with an ongoing risk of clinical recurrence.4 ,7 Unfortunately, in Crohn's and UC, there is a lack of reliable factors to predict recurrence. However, the possibility to detect preclinical recurrence using non-invasive tools (MRI, CRP, calprotectin) allowing timely resumption of treatment before clinical recurrence occurs perhaps gives reassurance when considering attempts at a drug holiday, having taken into account patient age, psychology and preferences.66
Having considered the known benefits, as well as the recently described limitations in the efficacy and safety of purine analogues, we defend wise use of these drugs. They are most appropriate for the management of non-severe steroid-dependent IBD, and should be tailored to the individual patient based on specific clinical and demographic criteria as well as biomarkers (figure 1). Purine analogues should also be used in the setting of a treat-to-target approach, considering a change in therapy if the target (steroid-free remission) has not been reached within 3–6 months. At a time where drug costs have surpassed hospitalisations and surgeries as the main source of direct medical expense in IBD, and while budget constraints are increasing in Western societies, we think it would be an error to abandon the use of a cheap drug with a well-described risk/benefit profile that can safely maintain at least one-fifth of our patients in a state of stable long-term steroid-free clinical remission.
Contributors EL, PI and LB have all actively contributed in the writing of this manuscript and and reviewed and agreed on the final manuscript.
Competing interests EL has received fees for: research grant: AstraZeneca, Schering-Plough, Abbott. Speaker fees: Abbott, Abbvie, AstraZeneca, Ferring, Schering-Plough, MSD, Chiesi, Menarini, Nycomed, Falk, UCB. Advisory board: Abbott, Abbvie, Ferring, UCB, MSD, Millenium, Mitsubishi Pharma, Takeda. Consultant: Abbvie.
Provenance and peer review Commissioned; externally peer reviewed.
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.