Long-term outcome of using allopurinol co-therapy as a strategy for overcoming thiopurine hepatotoxicity in treating inflammatory bowel disease


Dr A. Ansari, Department of Gastroenterology, Guy’s and St Thomas’ Hospital, 1st Floor College House, London SE1 9EH, UK.
E-mail: azhar.ansari@gstt.nhs.uk


Background  Hepatotoxicity results in the withdrawal of thiopurines drugs, azathioprine (AZA) and mercaptopurine (MP), in up to 10% of patients with inflammatory bowel disease. Our group previously demonstrated that allopurinol with AZA/ciclosporin/steroid ‘triple therapy’ improved renal graft survival.

Aim  To confirm the hypothesis that allopurinol may alleviate thiopurine hepatotoxicity by similar mechanisms as proposed in our renal study.

Methods  Unselected patients with acute thiopurine hepatotoxicity were offered allopurinol co-therapy with low-dose AZA or MP. The starting AZA/MP dose was determined by thiopurine methyltransferase (TPMT) activity (two patients were intermediate TPMT); then this dose was reduced to 25% for allopurinol co-therapy. Response to treatment was assessed by clinical severity indices, endoscopy and blood tests.

Results  Of 11 patients (three Crohn’s disease, eight ulcerative colitis) treated, nine (82%) remain in long-term remission (median 42 months) with normal liver tests. One patient also successfully bypassed flu-like symptoms. Two stopped: one nausea, one abnormal liver function (steatosis on biopsy). Leucopenia occurred in two cases and resolved with minor dose reductions.

Conclusions  Allopurinol co-therapy with low-dose AZA/MP can alleviate thiopurine hepatotoxicity. It appears safe and effective for long-term use, but requires monitoring for myelotoxicity. Assessing the TPMT activity helps tailor the AZA/MP doses.


The thiopurine drugs, azathioprine (AZA), mercaptopurine (MP) and thioguanine were developed by the Nobel laureates Gertrude Elion and George Hitchings as cancer antimetabolites.1 The therapeutic efficacy of the thiopurine drugs has been proven through long-term and widespread use in the treatment of chronic inflammatory, autoimmune disorders and malignancies such as acute lymphoblastic leukaemia. Their relative safety, tolerance and efficacy in inflammatory bowel disease (IBD) is now well established2 and reflected by the frequent use in Crohn’s disease (CD), with reports of over 60% of patients receiving AZA/MP within 4 years of diagnosis.3

However, a proportion of patients beginning AZA or MP (15–40%)4–6 fail to benefit from these drugs because of poor response, intolerance, hypersensitivity or idiosyncratic reactions. A highly cited retrospective study by Present et al.4 found the incidence of hepatotoxicity from thiopurines to be 2.8%. More recently, a large prospective study by Bastida et al.7 reported a greater incidence (>10%). The reason for the higher rate of heptotoxicity is not clear, although more aggressive use of thiopurines (i.e. higher doses) in modern practice could account for a proportion of this effect. This may be inferred by several observations. First, AZA/MP might have been used in the past in relatively conservative, i.e. less effective doses.8 Conversely, supra-normal doses of thiopurines are more frequently used in modern practice to improve efficacy of AZA/MP, increasing the risk of hepatotoxicity (24%).9

Chemically, MP is a synthetic analogue of the natural purine base hypoxanthine and is bioconverted by the purine salvage pathway in a multi-enzyme process to form cytoactive 6-thioguanine nucleotides (6-TGNs). This bioconversion competes with two partly predictable catabolic mechanisms: methylation and oxidation (Figure 1). Deactivation of MP and its intermediate metabolites by thiopurine methyltransferase (TPMT) produces 6-methyl-mercaptopurine byproducts (6-MMPs), some of which have been implicated in side effects.9, 10 Alternatively, oxidation of MP by xanthine oxidase (XOD) produces 6-thiouric acid, but the contribution of this pathway to drug efficacy or side effects has been little studied.

Figure 1.

 The possible influence of metabolites of azathioprine and effect of allopurinol on hepatotoxicity. 6-TGNs, 6-thioguanine nucleotides; TPMT, thiopurine methyltransferase.

Although clinicians and scientists familiar with the work of Trudy Elion and George Hitchings will be aware that they invented not only thiopurine drugs but also allopurinol, few may realize that the latter drug was devised originally to enhance thiopurine response by inhibiting XOD subtraction of MP from the 6-TGN bioactivation pathway.1 At that time, Elion found that allopurinol did not significantly improve the response to MP of childhood acute lymphoblastic leukaemia and the drug subsequently took on its more prominent role as an inhibitor of uric acid production for gout therapy.

In 1993, we reported that allopurinol co-therapy with thiopurine-based ‘triple therapy’ (AZA, ciclosporin, prednisolone), used for renal transplantation, greatly reduced the incidence of acute kidney graft rejection.11 We proposed that the improvement in graft survival by allopurinol co-therapy might be attributable to the preferential shunting of AZA into the 6-TGN pathway, as well as to the inhibition of reactive oxidative free radicals (ROS) that may cause tissue damage. Clinical applications of the synergism of allopurinol with AZA then lay dormant until a renal study in 1999 examined the effect of this co-therapy in relation to 6-TGN and 6-MMPs.12 More recently, we extended this concept to thiopurine induced hepatotoxicity and demonstrated for the first time in humans that this co-therapy could effectively bypass this problem.13 Sparrow et al.14, 15 confirmed this benefit in a well-constructed study in a subset of patients with thiopurine-induced raised transaminases and 6-MMPs.

Improving the efficacy of thiopurines would appear worthwhile, as these drugs are widely used, effective, relatively inexpensive and safe. Therefore, reducing the rate of hepatotoxicity may salvage a significant number of patients who would otherwise withdraw from this therapy (up to 10%). We have hypothesized that this experimental low-dose co-therapy of AZA/MP plus allopurinol would help thiopurine hepatotoxicity, potentially through a reduction in generation of both reactive oxygen species and 6-MMPs.9, 11

We describe here a retrospective analysis of the outcome of using long-term, reduced dose AZA/MP plus allopurinol co-therapy in a consecutive series of individuals with thiopurine-induced hepatitis. This co-therapy was introduced into our unit in 2001 for those patients with clearly defined acute thiopurine induced hepatotoxicity (greater than double the upper limit of normal of transaminases). Unique points of this report, compared to others, include the fact that our patients were strictly dosed according to TPMT status with consistent further dose adjustment for allopurinol (reduction to 25%). The hepatotoxicity was well defined, not biased towards production of 6-MMPs and this is a relatively long-term analysis.



Consecutive adult IBD patients who had experienced hepatitis after receiving AZA/MP according to unit protocol at two London IBD clinics were offered low-dose AZA/MP plus allopurinol co-therapy (Table 1). The experimental co-therapy was not commenced until the patients’ hepatotoxicity had improved following withdrawal of AZA/MP. Clinical information was obtained from patient records.

Table 1.   Dosing used in the study
TPMTDrug dose (mg/kg/day)
  1. AZA, azathioprine; MP, mercaptopurine; TPMT, thiopurine methyltransferase.

  2. To avoid severe myelotoxicity, the dose of AZA/MP must be lowered appropriately when allopurinol co-therapy is used.

Ideal initiating doses for inflammatory bowel disease
 Wild type21
Hepatotoxicity: dosing for allopurinol 200 mg/day co-therapy*
 Wild type0.50.25

Hepatitis was defined by published guidelines,16 typically alerted by greater than double the upper normal limit of transaminases. All participating patients were screened for viral hepatitis (A, B, C cytomegalovirus and Epstein–Barr virus) and other common causes of liver disease: ceruloplasmin levels, ferritin, alpha-1 antitrypsin. An autoimmune liver profile and a transabdominal ultrasound were performed and a thorough history was obtained regarding other potential hepatotoxins specifically antibiotics and nonsteroidal anti-inflammatory drugs. The 6-TGNs and 6-MMPs metabolites were not measured.

AZA/MP dosing

The study group had received AZA or MP by local guidelines, which were based on the patient’s weight and TPMT phenotypic status: normal (high) activity (red-cell TPMT ≥26 units – pmol/h/mg Hb) and intermediate activity (10–25 units) (Table 1). Blood tests were used to monitor initiation of AZA/MP and included weekly liver function tests (LFTs). AZA/MP was withdrawn after an abnormal LFT was confirmed or if the patient felt too unwell to continue treatment.

Once the LFTs normalized on two consecutive measurements, AZA/MP co-therapy with allopurinol was offered. The thiopurine dose was reduced to 25% of the ‘ideal initiating dose’, with 200 mg/day of allopurinol co-therapy (Table 1).

Following commencement of AZA/MP plus allopurinol co-therapy, LFTs, full blood counts and inflammatory markers were monitored fortnightly for the first 12 weeks and then gradually spaced out to 12-weekly. At each clinic visit, the clinical severity index was ascertained: Harvey–Bradshaw index for CD or Truelove and Witt’s index for ulcerative colitis (UC) (data not presented). An endoscopic assessment with an ileo-colonoscopy was performed on all the patients presented.


Thiopurine therapy

Pre-allopurinol.  In a unit with an audited hepatotoxicity rate of 4%, 11 IBD patients (three with CD, eight with UC) who developed hepatitis on thiopurines (10 on AZA, one on MP) between 2001 and 2007 (Table 2) were started on the co-therapy. None of the patients had previous surgery or fibrotic complications of CD. There were five females and six males in the group (age range 31–78 years, median 45 years, mean 50 years), and AZA/MP was prescribed as a steroid-sparing agent. There was no dose escalation which could have otherwise biased patients to get hepatotoxicity (Table 1). The reference median TPMT activity quoted in our laboratory is 32 units. The study group had a range of TPMT activity of 18–40 units, with a median of 33 units mean 32.6). There were two intermediate TPMT patients (both genotype *3A/*1) who received 1 mg/kg/day of AZA. One of these, case 11, developed intense gastric intolerance and was switched to MP (0.5 mg/kg/day), which was better tolerated, but the patient subsequently developed hepatotoxicity. The remaining nine patients had normal TPMT activity and had received a median AZA dose of 1.9 mg/kg/day (range 1.7–2.2).

Table 2.   Patient, disease and treatment demographics
Year started (LOT in months)2004 (44)2003 (5)2003 (62)2001 (71)2004 (<1)
LFT patternMixedMixedMixedTransMixed
Comments Steatosis on liver biopsy 8 weeks after stopping AZA and allopurinolFlu-like symptoms on full dose AZAStopped 2007, pregnancy 
HT onset6 Weeks8 weeks3 days6 weeks6 weeks
SE on co-therapyLP, dose reducedHTNilLP, dose reducedGI intolerance
Clinical outcomeTolerated SFRStoppedTolerated SFRTolerated SFRStopped
Endoscopic outcome 2008ICCER ICCERICCER 
  1. CD, Crohn’s disease; UC, ulcerative colitis; GI, gastric intolerance; Hetero, heterozygote; Chol, cholestatic; Mixed, mixed liver enzymes (cholestatic and transaminitis); SFR, steroid-free remission; Trans, transaminitis; LOT, length of therapy time; Wt, wild type; HT, hepatitis; SE, side effects; LP, leucopenia; CyA, ciclosporin; ICCER, Ileo-colonic complete endoscopic remission; TPMT, thiopurine methyltransferase; LFT, liver function test; AZA, azathioprine.

Year started (LOT in months)2003 (59)2005 (38)2007 (14)2005 (32)2006 (20)2004 (42)
LFT patternMixedMixedMixedCholMixedMixed
CommentsComplete regression of large inflamed caecal massNilNilNilNausea with AZA; HT with MPNausea with AZA; HT with MP; 3 months of CyA at start of co-therapy
HT onset.3 Weeks6 weeks2 weeks3 weeks2 weeks5 weeks
SE on co-therapyNilNilNilNilLP, dose reducedNil
Clinical outcomeTolerated SFRTolerated SFRTolerated SFRTolerated SFRTolerated SFRTolerated SFR

All the patients in the group were also on long-term 5-aminosalicylic acid therapy and eight were also on steroids at the time of developing hepatotoxicity. No other hepatotoxic agents were identified and the liver disease aetiology screens including transabdominal ultrasounds were all normal.

Thiopurine-induced hepatotoxicity occurred between 3 days and 8 weeks (median 5 weeks, mean 4.3 weeks). One patient exhibited malaise soon after starting AZA and developed a transaminitis within days. Other patients had nonspecific symptoms and hepatotoxicity was discovered on routine monitoring. All the episodes of hepatotoxicity resolved within 2 weeks of withdrawing AZA, without any clinically detectable consequence.

Thiopurine plus allopurinol therapy.  All the hepatotoxic patients were concurrently commenced on allopurinol (200 mg/day) plus low-dose AZA or MP as per departmental protocol. The desired AZA or MP doses for normal TPMT patients were 0.5 or 0.25 mg/kg/day respectively. This was further adjusted down for the two intermediate TPMT patients, whose doses were additionally reduced by half (Table 1). The median duration of observation of the AZA/MP plus allopurinol therapy in the nine patients who tolerated this combination was 42 months (range 14–71 months, mean 42.5 months).

Two patients had to stop the combination therapy because of recurrent side effects. Patient 2 (Table 2) suffered from ischaemic heart disease and experienced recurrent hepatitis after 8 weeks. A liver biopsy 8 weeks after stopping this co-therapy confirmed early mild fibrosis with steatosis without any evidence of nodular regenerative hyperplasia. Patient 5 (Table 2) displayed marked gastric intolerance and stopped treatment.

Three patients, two normal TPMTs and one intermediate, had side effects that resolved with dose reduction. Of the 11 patients started on this combination, nine (82%) tolerated the treatment long-term. One patient required a single infusion of infliximab after 18 of 62 months of therapy (Table 3). The remaining eight of 11 patients have remained in full remission (clinical and endoscopic) without any intervention using steroids or biologics. Clinical remission has been correlated to endoscopic remission in all the patients studied. Of particular interest is the complete regression of a potentially surgical large semi-obstructing, inflamed caecal mass (Table 2, patient 6).

Table 3.   Outcome of allopurinol and AZA/MP co-therapy
Initial responses to allopurinol plus AZA/MPInterventionsLong-term outcomes
  1. AZA, azathioprine; MP, mercaptopurine; ASA, aminosalicylic acid.

6 Complete response, no side effects1 Given infliximab at 18 months5 Complete responders
3 Myelotoxicity within 6 weeksDose reduction of thiopurineComplete response
1 Intense nauseaStopped after 2 weeksNow successfully on 5ASAs only
1 HepatitisStopped after 8 weeksSteatosis and mild liver fibrosis on biopsy


We tested and confirmed the hypothesis that allopurinol could protect against acute thiopurine-induced hepatotoxicity in humans. There were several aspects of this study that are either unique or noteworthy compared to our previous study and the other co-therapy studies of Chzanowska, Sparrow and Witte.12, 14, 15, 17 In our group, we included two patients who had intermediate TPMT, where the AZA/MP dosing was tailored to pre-treatment TPMT phenotype and then further refined for allopurinol co-therapy (Table 1). We also provide evidence that this co-therapy is both effective (82% response) and safe in long-term use (median 42 months), even in the TPMT intermediates who were calculated to require very low thiopurine doses (0.25 mg/kg of AZA and 0.125 mg/kg of MP). Our cohort was not influenced by supra-normal doses of AZA/MP or raised TPMT activity (median TPMT was 34 pmol/h/mg Hb), indicating that this co-therapy is effective for hepatic injury from more than hypermethylation. The benefit of close haematological monitoring is also highlighted because thiopurine doses as low as 25% of normal could still result in myelotoxicity, which was simply improved by lowering the AZA/MP further, without loss of response. An incidental finding that points to an additional potential useful role of this combination was the bypassing of flu-like symptoms to AZA in a patient with an inosine triphosphatase mutation.10 A possible explanation of this benefit may lie in the fact that this co-therapy is likely to reduce other methylated toxins such as methyl thio-inosine triphosphate10 (Figure 2). Finally, we adopted a higher dose of allopurinol (200 mg/day) compared with the very low dose of 10 mg daily of our previous study11 and the 100-mg allopurinol of Chzanowska,12 Sparrow14, 15 and Witte17– this was to theoretically maximize the potential anti-inflammatory effect of this agent.

Figure 2.

 Metabolites of azathioprine/mercaptopurine and their outcomes. Individuals with high TPMT and/or given higher doses of AZA can generate greater hepatoxic methylated metabolites of MP. Those with ITPase deficiency will trap TITP, also a possible cause for toxicity. AZA, azathioprine; MP, mercaptopurine; 6-TU, 6-thiouric acid; TITP, 6-thioinosine triphosphate; 6TGNs, 6-thioguanine nucleotides; HPRT, hypoxanthine phosphoribosyltransferase; ITPase, inosine triphosphate pyrophosphohydrolase; TPMT, thiopurine S-methyltransferase; XOD, xanthine oxidase.

Another therapeutic manoeuvre that attempts to bypass thiopurine side effects includes swapping between AZA and MP. For example, nitro-imidazole, a breakdown product of AZA metabolism, has been reported as hepatotoxic18 and some protocols advocate the swap to MP in the event of hepatitis. However, similar incidences of hepatotoxicity for AZA (2.1%) and MP (2.7%)6 do not favour this view. In fact, the nitro-imidazole moiety has been more convincingly related to gastric intolerance rather than hepatotoxicity.19 In our study, two patients experienced significant nausea as well as the hepatitis on AZA, both of which resolved with switching to low-dose MP and allopurinol.

Despite the well-known contra-indication of allopurinol and AZA, these two drugs have historically been given in combination to renal transplant recipients who had hyperuricaemia prior to transplant. Arising from this long-term practice, the recommended ‘rule of thumb’ dose reduction for AZA while on allopurinol is 30% of the normal dose to avoid severe myelotoxicity.20 Our previous renal transplantation study11 successfully used a co-therapy strategy to improve renal graft survival significantly. This employed a very low-dose allopurinol (effectively 10 mg/day) but still required a substantially reduced AZA dose – approximately half. This was lower than would have been predicted on a simple pharmacokinetic basis, and illustrated that the synergism of allopurinol with thiopurines is probably not simply caused by inhibition of XOD.

A curious aspect of the effect of allopurinol co-therapy is that not only have 6-TGNs been found to be increased, but also the 6-MMPs are reduced.14, 15 However, the observed reduction in the levels of 6-MMPs seemingly cannot be explained unless allopurinol also inhibits the activity of TPMT.15, 21 As it is known that allopurinol is not an inhibitor of TPMT,15 an alternative explanation is required. We propose that the apparent ‘diversion’ of AZA/MP metabolism away from methylation is illusory: the reduction in the levels of 6-MMPs occurs simply because the thiopurine dose is lowered to a concentration that is sub-optimal for TPMT activity. An acceptable explanation of the increased 6-TGNs is simple inhibition of oxidative first-pass metabolism by XOD, with the resultant increased bioavailability of low-dose thiopurine (Figure 2).

The heterogeneous nature of the clinical (hypersensitivity, idiosyncratic cholestatic reaction or presumed endothelial injury) and biochemical presentation (mixed rise of cholestastic and hepatic enzymes within the first 2 months)22 points to multiple mechanisms of thiopurines causing liver injury. One mechanism is hypermethylation, which usually results in both poor response and high levels of 6-MMP from dose escalation.9, 15 In contrast to this, our cohort included two TPMT intermediate patients and an overall group median TPMT well within the median of the normal TPMT range (34 pmol/h/mg Hb), without dose escalation. It is therefore unlikely that high levels of 6-MMPs were a significant cause of hepatotoxicity in our cohort. Consequently, we suggest that while thiopurine hepatoxicity may arise in some cases from overproduction of methylated thiopurines, this is probably just one of many mechanisms.

There are further studies that can be conducted. Analyses of 6-MMP and 6-TGN levels pre- and postallopurinol may have provided additional useful data. However, co-therapy and its effect on lowering 6-MMP and raising 6-TGN levels have been already well-documented.12, 14, 15 Liver biopsies may help to rule out subclinical hepatotoxicity in those patients on long-term co-therapy, although there we have found no reason to suspect this in our patient group.

Therapeutic strategies that may overcome poor response or side effects can be extrapolated from our data, and that of Sparrow, Dubinsky and others. ‘Fast methylators’ (very high TPMT) have recently been considered the ideal subjects for allopurinol co-therapy, but our study showed that patients with intermediate (heterozygous) TPMT who are prone to thiopurine hepatitis are also good candidates for co-therapy. Patients with flu-like symptoms associated with ITPA mutations should also be considered for co-therapy because they may benefit from the lower thiopurine dose.

In conclusion, co-therapy may be indicated to avoid poor response, side effects (hepatotoxicity and possibly flu-like symptoms) or delay in time to response. Careful white cell monitoring is necessary to avoid myelotoxicity. It is interesting that the combination of low-dose AZA/MP and allopurinol is finally being exploited to provide a safe and effective way of delivering thiopurine therapy: this would fulfil Gertrude Elion’s original vision for allopurinol.1


Declaration of personal and funding interests: None.