Prospective evaluation of the pharmacogenetics of azathioprine in the treatment of inflammatory bowel disease


Dr J. Sanderson, Department of Gastroenterology, 1st Floor, College House, St Thomas’ Hospital, London SE1 7EH, UK.


Background  One-third of patients with inflammatory bowel disease (IBD) receiving azathioprine (AZA) withdraw treatment due to side effects or lack of clinical response.

Aim  To investigate whether pharmacogenetic loci or metabolite concentrations explain clinical response or side effects to AZA.

Methods  Patients with IBD were given 2 mg/kg of AZA without dose escalation or adjustment. Serial clinical response, thiopurine methyl transferase (TPMT) activity and thioguanine nucleotide (TGN) concentrations were measured over 6 months. All patients were genotyped for inosine triphosphatase (ITPase) and TPMT. Clinical response and side effects were compared to these variables.

Results  Two hundred and seven patients were analysed. Thirty-nine per cent withdrew due to adverse effects. Heterozygous TPMT genotype strongly predicted adverse effects (79% heterozygous vs. 35% wild-type TPMT, P < 0.001). The ITPA 94C>A mutation was associated with withdrawal due to flu-like symptoms (P = 0.014). A baseline TPMT activity below 35 pmol/h/mg/Hb was associated with a greater chance of clinical response compared with a TPMT above 35 pmol/h/mg/Hb (81% vs. 43% respectively, P < 0.001). Patients achieving a mean TGN level above 100 were significantly more likely to respond (P = 0.0017).

Conclusions  TPMT testing predicts adverse effects and reduced chance of clinical response (TPMT >35 pmol/h/mg/Hb). ITPase deficiency is a predictor of adverse effects and TGN concentrations above 100 correlate with clinical response.


The thiopurine drugs, azathioprine (AZA) and mercaptopurine (MP) are an important mainstay in the effective treatment of Crohn’s disease (CD) and ulcerative colitis (UC) with efficacy rates of 55–70%.1–4 Indeed, with greater emphasis now placed on reduction to steroid exposure, usage of these drugs for inflammatory bowel disease (IBD) has steadily increased in recent years.5 In general, thiopurines are well tolerated and safe. However, as many as one-third of patients may suffer adverse effects2, 6, 7 prompting drug withdrawal and a switch to alternative treatments. The metabolism of AZA and pharmacogenetic influences are represented in Figure 1 (Data S1 and S2).

Figure 1.

 Individuals with high TPMT can generate greater methylated metabolites of MP as opposed to cytotoxic TGNs. Those with ITPA deficiency will trap TITP also a possible cause for toxicity. AZA, azathioprine; MP, mercaptopurine; 6-TU, 6-thiouric acid; TITP, 6-thioinosine triphosphate; 6-TGNs, 6-thioguanine nucleotides; HPRT, hypoxanthine phosphoribosyltransferase; ITPA, inosine triphosphate pyrophosphohydrolase; TPMT, thiopurine S-methyltransferase; XOD, xanthine oxidase.

Considerable interest has focused on the metabolism of thiopurines as a means of identifying ways of individualizing therapy to minimize adverse effects and maximize clinical response. Genetic variation of the activity of thiopurine methyl transferase (TPMT) and its influence on how an individual metabolizes AZA is an important pharmacogenetic model. Individuals with TPMT deficiency who receive standard doses of thiopurines are at significant risk of toxicity, primarily as a result of unchecked production of thioguanine nucleotides (TGNs).8–11

The majority of our understanding of thiopurine biology stems from work in acute lymphocytic leukaemia in which TPMT activity has been found to be inversely proportional to TGNs and a positive clinical response.8, 12 The opinion about the influence of these two parameters in IBD therapy is still divided, probably due to a lack of prospective data. However, some studies which are also supported by a meta-analysis of retrospective and cross-sectional studies have confirmed a strong association between TGN concentrations and induction of remission.13 These indicate that a TGN level consistently above 230–260 pmol/8 × 108 RBC 14, 15 is associated with a favourable response. Considering TPMT, only retrospective studies in IBD have addressed (and confirmed) that pretreatment measurement of TPMT might predict clinical response to AZA.11, 16 The majority of the remaining studies have looked for a correlation between adverse reactions and TPMT activity with a recent prospective study confirming a fourfold risk of myelotoxicity in patients with intermediate TPMT activity.17 However, variation in TPMT activity accounts for no more than 10% of overall thiopurine toxicity 18 and perhaps around one-third of myelotoxicity.19 Whilst other factors account for much of the remaining toxicity (e.g. hypersensitivity), genetic variation of other enzymes metabolizing thiopurines is also likely to be important.

Inosine triphosphopyrohydrolase (ITPase) catalyses the breakdown of inosine triphosphate (ITP) as part of a futile cycle in the purine metabolic pathway.20 Genetic ITPase deficiency is thought to be a benign disorder of no known detriment to health under normal circumstances. However, following exposure to thiopurines, ITPase deficiency results in the cellular accumulation of thio-inosine triphosphate (thioITP) [Arenas, personal communication], a ‘rogue nucleotide’ with the potential of cell toxicity.21 Indeed, three studies, one prospective, have reported association of thiopurine toxicity with the presence of the ITPA 94C>A mutant allele.22, 23 However, no consistent pattern of toxicity has been reported, with three retrospective studies reporting no association.24–26

The value of pretreatment assessment of TPMT in avoiding potentially life-threatening toxicity in patients with zero activity is well proved and increasingly taken up in clinical practice.27 However, the true clinical value of prior knowledge of the more frequent heterozygous TPMT state is less well defined. Furthermore, whether TPMT activity can provide useful information about likely clinical response has only been addressed retrospectively. Finally, the role of ITPase deficiency as an additional pharmacogenetic marker in thiopurine therapy remains unclear.

We therefore sought to study the role of TGNs, TPMT and ITPase as predictors of adverse effects and clinical response to AZA, using a well-powered prospective study of patients with IBD. To our knowledge, the cohort presented here constitutes the largest prospective analysis of thiopurine pharmacogenetic and pharmacokinetic parameters in IBD.



Patients between the ages of 18 and 80 years with CD or UC were enrolled into a prospective multi-centre study of AZA undertaken in a number of institutions who were members of the London IBD Forum. The decision to initiate AZA was taken at the discretion of the treating doctor in each centre. Exclusion criteria included age (<18 or >80), previous thiopurine exposure and previous use of biologics. The study protocol was approved by the multi-centre and local research ethics committee at the primary institution and all other institutions recruiting patients (MREC number 00/1/33). Written informed consent was obtained from each patient entering the study. Scientific methodology for TPMT phenotyping, TGN measurements and all genetic analysis have either been developed or are routine practice in the Purine Research Laboratory (PRL). All these tests were performed in the PRL by the authors, M. A., A. M., J. D. and A. A.

Study protocol

At enrolment, TPMT activity was measured. Treating doctors were blinded to these TPMT results unless a ‘very low’ result was obtained as it was decided that these patients would be excluded from the study (although none were encountered). Permission was then given to commence treatment and in all cases, AZA was started as near 2 mg/kg daily as possible and without dose alteration.

The study period was 6 months to permit a reasonable assessment of clinical response to AZA. Clinical activity indices (Harvey Bradshaw index for CD and Truelove and Witts score for UC) were recorded at initiation and then at weeks 4, 12 and 24. To correlate these with TPMT and TGN concentrations, blood was also taken at these time intervals. In addition an endoscopic assessment of all patients with UC at the start and end of the study was available to corroborate the Truelove and Witts index (data not shown). Steroids were tapered according to clinical response with the target of achieving complete withdrawal at 12 weeks. Full blood count, erythrocyte sedimentation rate (ESR), standard biochemistry, liver function tests and C-reactive protein were measured at baseline and at each study visit (2, 4, 6, 12, 16 and 24 weeks). Patients were asked to report any adverse events at each visit or, where necessary, between visits by contacting the local treating doctor or study nurse. After the study period all the study centres were asked about any further side effects at 9 months. Definitions for adverse events and criteria for withdrawal or dose reduction are given in Table 1.

Table 1.   Definitions and frequencies of adverse drug reactions (ADR) encountered in the study
ADRDefinition of ADRNumber% of ADR group% of ADRs in overall groupNumber of TPMT heterozygotes with ADRs
  1. ALT, alanine transaminase.

Gastric intoleranceOccurrence of any or a combination of nausea, vomiting and abdominal pain with normal amylase and abdominal ultrasound344115.78
Flu-like symptomsGeneral malaise, temperature and muscle and joint pains1113.35.10
PancreatitisClinical evidence and amylase ×4 upper limit of normal. Supportive radiological findings89.63.70
HepatitisRise in ALT > ×2 upper limit of normal89.63.70
RashNew onset of rash after starting AZA which resolves on withdrawal of AZA89.63.72
MyelotoxicityNeutrophil count <1.5 × 109 or total white cell count <3.5 × 10978.43.25
OtherIncludes infection and nonspecific side effects during treatment but requiring withdrawal of AZA78.43.20
Total 831003815 of 19

Clinical response

Indications for starting AZA and the definition of clinical response for each indication are given in Table 2. Clinical response was assessed at 6 months to allow for the known slow therapeutic onset of thiopurines and to ensure response, especially steroid sparing, was sustained. This was divided into a target of achieving the given definition of response by 12 weeks (e.g. complete steroid withdrawal) followed by documented maintenance of this response for the remaining 3 months. Surgery and initiation of biological or other therapy were also considered as treatment failure.

Table 2.   Definitions of clinical response used in the study according to indication for azathioprine
  1. HBI, Harvey Bradsaw Index; T&W, Truelove and Witts score; CD, Crohn’s disease; UC, ulcerative colitis.

Steroid sparingComplete response was defined as complete withdrawal of all corticosteroids by week 12 and no further steroid use to week 24. Partial response in this group defined as a reduction to no greater than 5 mg (prednisolone) by week 12 and no increase to week 24
Maintenance of remissionComplete response defined as no steroid use or rise in HBI or T&W score (where appropriate) throughout the 24 week study period
Induction of remission/ active diseaseCD: Reduction of HBI to 3 or below by week 12 and no subsequent increase in HBI or steroid use to week 24. UC: Reduction in T&W score to normal by week 12 and no subsequent increase or steroid use to week 24
Biological useThe use of biological therapy was regarded as treatment failure at any point in the 24 week study period

Pharmacogenetic profiling

At baseline, blood was taken for measurement of red cell TPMT enzyme activity and for assessment of TPMT and ITPA genotype in each patient. TPMT activity and TGN concentrations were then also measured at 2, 4, 12 and 24 weeks of the study. TPMT activity (phenotype) was assayed by the mass spectroscopy method as previously described.28 Reference ranges were: ‘very low’, 0–10 pmol/h/mg Hb, intermediate activity, 11–25 pmol/h/mg Hb and normal, 26–50 pmol/h/mg Hb. Patients were genotyped for the three common TPMT polymorphisms: TPMT*3B (460 G>A, Ala154Thr in exon 7), TPMT*3C (719 A>G, Tyr240Cys in exon 10) and TPMT*3A (presence of both polymorphisms). The two functional ITPA polymorphisms were also genotyped: the exonic ITPA 94C>A [Pro32Thr] mutation and the intron mutation ITPA IVS2 + 21A>C, as previously described.20 All genotypes were checked for Hardy–Weinberg equilibrium.

Measurement of TGNs

Blood was collected in 4.5 mL EDTA tubes and transported to the laboratory within 5 days of collection. TGNs were extracted and oxidized from 100 μL saline-washed packed RBCs following the method of Rabel et al.29 An aliquot of 75 μL of the oxidized TGN was injected onto a Waters Alliance HPLC system (Milford, MA, USA) with fluorescent detection. Separation of thioguanine mono-, di- and triphosphate nucleotides (were eluted from a C18 column) were separated by ion-pair reverse phase HPLC (Genesis C18 250 × 4.6, 4 μm; ChromTech, Congleton, UK). Buffer A: 40 mm KH2PO4, 4.7 mm tetrabutylammonium hydrogen sulphate (TBA), pH 3.0. Buffer B: 2 mm KH2PO4, 60% (v/v) methanol, 4.7 mm TBA, pH 3.4. The gradient developed from initial conditions of 95% buffer A and 5% buffer B, going to 75% buffer B over 24 min, then back to initial conditions. The run time was 35 min. The fluorescent detector (Waters 474 scanning fluorescent detector) was set at a λex = 329 nm and λem = 410 nm. A thioguanine mono-, di- and triphosphate standard used for quantitation was a gift from Dr R.A. De Abreu (University Medical Center, Nijmegen, The Netherlands). The method did not allow evaluation of methyl-thiopurine concentrations. Bench side stability and intra-individual reproducibility of TGNs were confirmed by MA and AA before patients were recruited.

Analysis of results and statistical methods

The baseline pretreatment TPMT activity and genotypes of TPMT and ITPA were each compared to withdrawal due to adverse effects, to individual groups of adverse effects and to clinical response (complete response or nil/partial response). Likewise, mean and highest TGN concentrations over the 6 month study period were correlated with clinical response. Analyses of total white cell, neutrophil and lymphocyte counts (mean and lowest), and highest and change in mean cell volume (MCV) were also made against clinical response.

The variables were examined and (apart from TPMT) the distribution was not normally distributed. Fisher’s exact test was used to determine whether there was any dependence between TPMT genotype and risk of withdrawal due to adverse effects. An ROC curve was constructed to determine a suitable cut-off point for TPMT activity in an effort to predict clinical response. The cut-off point was visually investigated using overlayed histograms of TPMT activity for the different clinical response categories. The efficacy of this value was examined by cross tabulation. The Student’s t-test was used to compare TPMT activity and clinical response; the efficacy of the threshold value for predicting clinical response was tested using Fisher’s exact test. A threshold value of TGNs was also determined by the same technique and the Fisher’s exact test was also used to compare TGN concentrations and response to AZA. The chi-squared test was used to compare individual types of adverse effects with the ITPA genotype. The TGN concentrations at 4, 12 and 24 weeks were compared to clinical response with the chi-squared test for trends. The effect of the use of 5ASA on TGN concentrations was performed with the Mann–Whitney U-test.

The haematological parameters were compared to clinical response using the t-test. Tests were deemed statistically significant at the 0.05 level. Statistical calculations were performed using the R V2.2.0 program (


A total of 215 patients were recruited into the study from 11 centres. Of these two patients broke the protocol, three had incomplete data, two had surgery before week 12 and one withdrew from the study leaving 207 for final analysis. The mean age was 40.3 years (range 18–80 years) and 115 (54%) were female. One hundred and seventeen patients had CD [65/117 (56%) female] of whom 25 (21%) had ileal disease, 52 (44%) ileocolonic, 31 (27%) colonic disease, nine (8%) disease limited to the proximal small bowel/upper gastrointestinal (GI) tract. Eleven (9%) had coexistent perianal or enterocutaneous fistulating disease which was not the primary indication for AZA use. Of the 90 patients [46/90 (51%) female] with UC, 50 (55%) had pancolitis, six (7%) subtotal colitis and 34 (38%) distal colitis.

In the majority of patients (85%), the indication for AZA was as steroid-sparing therapy. Induction of remission of active disease was the indication in six (3%), maintenance of remission in 26 (12%). The median starting dose of AZA was 1.95 mg/kg (range 1.88–2.38 mg/kg). The TPMT distribution conformed to the expected population distribution (data not shown). No patients with zero TPMT were detected and 19 of 219 (9%) had a heterozygous TPMT genotype. No significant induction of TPMT was observed from baseline to subsequent measurement. The correlation between TPMT genotype and phenotype was 100%, whereas between phenotype and genotype it was 99% (data not shown).

There was a high overall withdrawal rate of 83 (out of 215) patients (39%) due to adverse effects on an intention-to-treat analysis. A further 12 (6%) reported side effects but continued treatment over the 6 month period. The details of adverse effects leading to drug withdrawal are given in Table 1. In particular, there was an unexpectedly high occurrence of gastric intolerance. One patient had asymptomatic amylasaemia which persisted despite withdrawal of AZA, and a diagnosis of macroamylasaemia was made.

Amongst the whole cohort of 207 patients, a complete clinical response to AZA occurred in 79 of 207 patients (38%). Clinical response amongst the 124 patients completing the 6 months required for inclusion in the analysis was 64% (79/124).

TPMT vs. adverse effects

A heterozygous TPMT genotype strongly predicted withdrawal of AZA due to adverse effects, such that 15 of the 19 (79%) heterozygotes did not tolerate 6 months of AZA, compared with 66 of 188 (35%) who were TPMT wild type (P = 0.000265). Gastric intolerance was the most frequent reason for withdrawal amongst heterozygotes and was significantly more common than in those with normal TPMT. Twenty patients had GI intolerance: seven (37%) were TPMT heterozygotes while only 13 (7%) were wild type (χ2 = 13.9, P < 0.001). Myelotoxicity was also more frequent in those with a heterozygous TPMT genotype (26% vs. 0.5%, P < 0.01). Importantly, gastric intolerance occurred early, within 6 weeks and only those remaining on AZA for longer went on to develop myelotoxicity (Figure 2).

Figure 2.

 Timing of adverse drug reactions leading to withdrawal of azathioprine (AZA) amongst patients with heterozygous thiopurine S-methyltransferase (TPMT) deficiency. Patients all received 2 mg/kg of AZA as per study protocol. Overall, 15 of the 19 patients (79%) with heterozygous TPMT deficiency withdrew treatment due to ADRs during the 6 month study period. Of these, most (n = 11) withdrew treatment due to non-haematological ADRs (mainly gastrointestinal intolerance) with all but one case occurring within 6 weeks of starting AZA. Conversely, myelotoxicity (n = 4) did not occur until week 12 or later. Individuals with heterozygous TPMT deficiency are most likely to withdraw therapy due to gastric intolerance before they are able to develop myelotoxicity.

ITPA genotype and adverse effects

In the study group 202 were genotyped for the ITPA 94C>A and IVS2 + 21A>C polymorphisms. The allele frequencies were 0.089 for the ITPA 94C>A mutation and 0.093 for the ITPA IVS2 + 21A>C mutation. Overall, withdrawal due to adverse effects was not predicted by the ITPA genotype (ITPA 94C>A: P = 0.31 and ITPA IVS2 + 21A>C: P = 0.46). However, analysis by specific side effects and clusters of ITPA polymorphisms compared with patients without side effects revealed that the ITPA 94C>A mutation was strongly associated with flu-like symptoms (P = 0.014, 95% CI = 1.23–13.94, OR = 4.13). There was no association with any other side effect (gastric intolerance, P = 0.50 for ITPA 94C>A).

TPMT vs. clinical response

The activity of TPMT was strongly predictive of clinical response. There was a 43% (24/56) response in those with a red cell TPMT activity greater than 35 pmol/h/mg Hb, compared with 81% (55/68) in those with red cell TPMT activity below 35 pmol/h/mg Hb (Fisher’s exact P < 0.001). Likewise, mean TPMT activity was significantly higher in non-responders vs. responders to AZA (38.6 pmol/h/mg Hb vs. 32 pmol/h/mg Hb respectively, P < 0.001). Dose of AZA (mg/kg) received did not differ either between responders and non-responders (1.95 mg/kg in both groups, t-test P = 0.65) or in each TPMT group (above and below 35 pmol/h/mg Hb) (1.94 mg/kg vs. 1.96 mg/kg, t-test P = 0.54).

TGN measurements vs. clinical response

The TGN values took 4 weeks to reach steady state. TGN concentrations at 2 weeks were therefore not included in the analysis which comprised an average of the concentrations at 4, 12 and 24 weeks. All three measurements were available for 75 patients, at least two TGN concentrations in 112 patients and at least one measurement in 124 patients. Overall, there was no relationship between mean TGN level and clinical response (P = 0.14). However, an analysis comparing patients with TGN values above and below 100 pmol/8 × 108 RBC showed that there was a significant difference in clinical response between these two groups: 74% (59/80) response in those with TGN > 100 pmol/8 × 108 RBC vs. 46% (20/44) in those with TGN < 100 pmol/8 × 108 RBC (P = 0.0017). There was also a significant trend (P = 0.018) towards increasing the likelihood of successful clinical response with greater numbers of TGN values above 100 in any particular patient (Figure 3).

Figure 3.

 Relationship between clinical response to azathioprine (AZA) and frequency of thioguanine nucleotide (TGN) concentrations above 100 pmol/8 × 108 RBC during study. Bar chart showing the relationship between clinical response to AZA and the numbers of TGN concentrations above 100 pmol/8 × 108 RBC. TGN concentrations taken during study at 4, 12 and 24 weeks. Clinical response assessed after 6 months of AZA 2 mg/kg. Patients with no TGN level over 100 pmol/8 × 108 RBC had a very low likelihood of successful clinical response whilst those with all three TGN values above this level had a very high likelihood of response. There was a significant trend towards increasing chances of clinical response with numbers of TGN values above 100 pmol/8 × 108 RBC (P = 0.018).

Haematological parameters vs. response

There was a significant relationship between lowest neutrophil count and complete response to AZA. The mean lowest neutrophil count was less in those patients who responded to AZA compared to those who had no response (4.1 vs. 5.6 respectively, P = 0.024). However, the lowest lymphocyte count did not correlate with clinical response. A strong relationship for MCV and clinical response was found (responders 87.2 fl vs. non-responders 82.6 fl, P = 0.001). Change in MCV from baseline to 6 months, however, did not correlate with clinical response.

Influence of 5-aminosalicylates

Co-prescription of 5-ASA did not influence TPMT activity (33.4 pmol/h/mg/Hb in those receiving 5-ASA vs. 33.7 pmol/h/mg Hb in those not on 5-ASA, t-test P = 0.79) and was not associated with any difference in clinical response to AZA (64% vs. 63%, on and off 5-ASA respectively, χ2P = 0.93). Use of 5-ASA was, however, independently associated with a reduced risk of side effects (28% in those on 5-ASA vs. 48% in those not taking 5-ASA, χ2P = 0.007). There was no effect of 5ASA on TGN concentrations [mean TGN on 5ASA 137 (n = 76) vs. 130 (n = 48) off 5ASA] (P = 0.49, Mann–Whitney U-test.).


This is the largest prospective evaluation of using full-dose AZA (2 mg/kg) and without dose adjustment in IBD patients. The study design minimized the effect of variable dosing on accurate evaluation of the pharmacokinetics of TGNs and side effects. Previous studies have demonstrated the increased risk of thiopurine toxicity in patients with TPMT deficiency in a variety of clinical settings, mainly haematological malignancy,12 transplantation 30 and autoimmune disease.31 In these studies, the chances of adverse effects varied from 20% to 35%.32, 33 However, the retrospective nature of most of these studies may not have provided a true measure of individual risk. Our study is one of the few comprehensive prospective studies and one could argue that the risk demonstrated (80% withdrawal) is a more accurate reflection of the actual risk of full-dose AZA therapy in those with intermediate TPMT activity.

The influence of genetic variation in TPMT activity on individual metabolism of thiopurines remains one of the classical applications of pharmacogenetics in modern medicine. Despite this, the uptake of TPMT testing remains variable, particularly in Europe.34, 35 Whilst pharmacoeconomic arguments for pretreatment TPMT testing are favourable36–42 and numerous retrospective studies demonstrate the risks of exposure for TPMT deficiency individuals (‘very low and intermediate’) to standard doses of AZA or MP,43–47 some uncertainty may stem from the lack of prospective studies of the influence of TPMT. Furthermore, emphasis on the fact that TPMT deficiency only explains a small proportion of overall thiopurine toxicity19 has added to a misplaced belief that prior knowledge of TPMT activity adds little to conventional practice.

However, the results of the study confirm the major influence of TPMT carrier status on risk of AZA toxicity with most TPMT heterozygotes withdrawing due to early-onset adverse events. Only two TPMT heterozygotes continued AZA beyond the 6 month study period, one developed myelotoxicity at 9 months and the second was a self-confessed poor adherer. An important finding is that most patients with intermediate TPMT activity withdrew due to gastric intolerance within 6 weeks of treatment and this was particularly caused by nausea. This is likely to explain why myelotoxicity (which first occurred at 12 weeks) is not encountered more often amongst TPMT heterozygotes.

Equally, genetic variation in the activity of other enzymes may explain some of the toxicity to thiopurines that is not accounted for by variation in TPMT activity. In this study, we examined the influence of the two functional ITPase polymorphisms. We discovered a clear association between presence of the ITPA 94C>A mutation and drug withdrawal due to the flu-like illness. This relationship with adverse events,21 withdrawal of treatment23 and myelotoxicity48 is supported by other studies. The mechanism of toxicity due to ITPase deficiency is uncertain, but toxicity could be predicted, perhaps as a result of interruption of a variety of nucleotide-dependent reactions from thio-ITP accumulation. As such, general toxicity symptoms such as flu-like illness might be expected rather than other more specific toxicities. In fact, the benefit of thioguanine in patients ‘allergic’ to AZA/6MP49 could well be explained by the effective by-passing of ITPase with thioguanine. Whilst the influence of ITPase deficiency appears small, it is definite and may yet prove of sufficient value to merit pretreatment testing, if only to highlight a need for vigilance regarding toxicity and permit prompt an alternative therapeutic approach.

Most studies of the influence of TPMT have focused on the prediction of toxicity in those with genetic TPMT deficiency. However, high TPMT activity could also predict poor clinical response as a consequence of diversion to methylation of MP in preference to bioactivation to TGNs.50 In this study, we have confirmed that prior knowledge of TPMT is an important predictor of clinical response. Patients with a high TPMT (>35 pmol/h/mg/Hb) have a significantly reduced chance of responding to AZA. This cut-off echoes earlier finding by our group and others of the influence of TPMT activity on AZA efficacy in a both a renal transplant setting51 and IBD.11, 16 Prior knowledge of TPMT activity in the higher range indicates that empirical dosing (2–2.5 mg/kg) may be inadequate prompting earlier use of higher doses (i.e. 2.5–3 mg/kg) or alternative strategies to avoid lengthy periods of non-response. However, in a proportion of these patients, methylation may remain dominant making dose escalation futile or result in hepatotoxicity.15, 52 Hence, a very high TPMT might be a more reliable predictor of a need for co-therapy with allopurinol52 or an alternative immunosuppressive.

We measured TGN concentrations throughout the study period and discovered a significant correlation between good clinical response and concentrations greater than 100 pmol/8 × 108 RBC. This level is considerably lower than that determined in other published series which have reported response cut-off values of about 250 pmol/8 × 108 RBC.13 The difference is related to extraction and processing of TGNs (intact nucleotides by ion-pair HPLC) that was used in our study.29 This method is considerably different from other reported TGN assays53, 54 and results in different reference ranges.55 The optimal method for measurement of TGNs remains an unresolved question.

Our large prospective data set indicates that TGN level of above 100 pmol/8 × 108 RBC should predict a reasonable chance of successful clinical response. In practice, TGN concentrations are difficult to use as a guide to optimal dosing and uptake has therefore been poor in the UK. We would propose that the best use is still as an indicator of drug adherence and secondly, as a guide to the feasibility of dose escalation in the event of non-response (Data S3).

In our study there were several potential weaknesses. This includes the unexplained high incidence of adverse events (38%) compared with earlier studies.6 However, recent large studies report a similar frequency of adverse events as ours,17, 56 and the reason for this remains uncertain (Data S4).

In conclusion, this large prospective study has demonstrated very clearly that patients with intermediate TPMT activity are highly likely to withdraw from standard-dose AZA treatment due to early side effects. This is frequently due to nausea and gastric intolerance prior to the onset of myelotoxicity (median time of onset 12 weeks). We strongly recommend that these individuals receive appropriately reduced AZA dose (1 mg/kg) to avoid toxicity. Furthermore, we have shown that, as one would expect, pretreatment TPMT activity is a strong predictor of clinical response. A significant association between adverse events, specifically flu-like symptoms, was found for the ITPA 94C>A polymorphism. Finally, the study also shows a strong correlation with successful outcome on AZA if TGN concentrations exceed 100 pmol/8 × 108 RBC. Pretreatment TPMT testing offers significant value to doctor and patient. The role of TGN monitoring is less certain but the results of this study add weight to the view that TGN monitoring is of clinical value.


The authors would like to thank all other members of the London IBD Forum who assisted with the recruitment of patients to the trial. These include: Ms Jo Hirst, Guy’s and St Thomas’ Foundation Trust; Dr Jervoice Andreyev, Formerly at The Chelsea & Westminster Hospital; Dr Marta Carpani, Dr Andrew Thillainayagam & Ms. Lynn Evans, Charing Cross and Hammersmith Hospital; Dr Alastair McNair, Queen Elizabeth Hospital Woolwich; Dr Stuart Cairns & Ms Linda Dyer, The Royal Sussex County Hospital; Prof. David Rampton, St Bartholomew’s and The Royal London Hospitals; Dr Junaid Mehmood. Declaration of personal interests: None. Declaration of funding interests: This work was funded by the Trustees of Guy’s and St Thomas’ Hospital Foundation Trust. The genotype for inosine triphosphatase (ITPase) has been patented by Prometheus laboratories and Guy’s and St Thomas’ Hospital.