Potential conflict of interest: Nothing to report.
Financial support for the laboratory consumables used in this study was provided by Bardhan Research and Education Trust of Rotherham Ltd. (Barnsley, UK).
Azathioprine (AZA) is used to maintain remission in autoimmune hepatitis (AIH), but up to 18% of patients are unresponsive. AZA is a prodrug, and the formation of active thioguanine nucleotide (TGN) metabolites varies widely. We aimed to assess the relationship between AZA metabolite concentrations (i.e., TGNs and methylmercaptopurine nucleotides [MeMPNs]), thiopurine methyltransferase (TPMT) activity, therapeutic response, and toxicity in adult patients with AIH prescribed a stable dose of AZA for the maintenance of remission. Red blood cell (RBC) TGNs and MeMPNs were measured in serial blood samples over a 2-year period. The average TGNs (avTGNs) and MeMPNs (avMeMPNs) concentrations for each patient were used for analysis. Therapeutic response was defined as the ability to maintain remission, defined as a normal serum alanine aminotransferase (ALT) level (ALT <33 IU/mL). Patients who maintained remission (n = 53), compared to those who did not (n = 17), tended to be on lower doses of AZA (1.7 versus 2.0 mg/kg/day; P = 0.08), but had significantly higher concentrations of avTGN (237 versus 177 pmol/8 × 108 RBCs; P = 0.025). There was no difference in MeMPN concentrations or TPMT activities between the two groups. There was a negative correlation between ALT and avTGN (rs = −0.32; P = 0.007). An avTGN concentration of >220 pmol/8 × 108 RBCs best predicted remission, with an odds ratio of 7.7 (P = 0.003). There was no association between TGN, MeMPN, or TPMT activity and the development of leucopenia. Two patients developed AZA-induced cholestasis and the avMeMPN concentration was higher in those patients, compared to those who did not (14,277 versus 1,416 pmol/8 × 108 RBCs). Conclusion: TGN concentrations of >220 pmol/8 × 108 RBCs are associated with remission. TGN measurement may help identify inadequate immunosupression. AZA-induced cholestasis was associated with increased MeMPN concentrations. (HEPATOLOGY 2012)
Autoimmune hepatitis (AIH) is a chronic inflammatory liver disease, characterized by a relapsing and remitting course and a high mortality if left untreated. Azathioprine (AZA), a potent immunosuppressant, is commonly used as a “steroid sparing” agent for the induction and maintenance of remission in autoimmune hepatitis. Approximately 10% of patients with AIH experience AZA-related side effects necessitating drug withdrawal,1 and a further 18% of patients are unable to maintain remission with AZA alone.2
AZA is a prodrug that undergoes extensive intracellular activation by a multienzymatic process to form the active metabolites, 6-thioguanine nucleotides (TGNs) (Fig. 1). The immunosuppressive mechanism of action of AZA metabolites is not fully understood. TGN metabolites act as purine antagonists. Incorporation of TGN into DNA is the primary mode of overt cytotoxicity.3, 4 TGN induces apoptosis of activated T lymphocytes by the inhibition of intracellular signalling pathways, specifically the blockade of Rac1 activation,5 and selectively inhibits inflammatory gene expression.6 Methylmercaptopurine nucleotide (MeMPN) is a potent inhibitor of de novo purine synthesis, but its exact role in immunosupression is not fully understood.7
The ability to form TGN active metabolites after a dose of AZA is partially determined by variations in thiopurine methyltransferase (TPMT) activity, which is under the control of a common genetic polymorphism inherited in an autosomal codominant pattern. In Caucasians, the frequency of distribution of TPMT activity is trimodal: 89% of the population are homozygous for the wild-type allele and have high enzyme activity, 11% are heterozygous with intermediate enzyme activity, and 0.3% inherit two variant alleles, resulting in no functional activity.8, 9 Individuals who lack functional TPMT activity produce grossly elevated concentrations of TGNs and experience profound myelosuppression on standard doses of thiopurines.10, 11
Studies investigating the clinical use of measuring TPMT activity and intracellular thiopurine metabolites in childhood acute lymphoblastic leukemia found an inverse relationship between red blood cell (RBC) TGN concentrations and TPMT activity, and those who accumulated higher TGN concentrations had a significantly improved leukemia-free survival, but a higher incidence of neutropenia.10, 12 Parallel studies in inflammatory bowel disease (IBD) have reported conflicting results, but a meta-analysis concluded that patients in remission had significantly higher TGN concentrations than those with active disease, and TGN concentrations above threshold values of 230-260 pmol/8 × 108 RBCs were significantly associated with remission with an odds ratio (OR) of 3.27.13
AZA can cause hepatotoxicity, and this may present with asymptomatic elevation of aminotransferases, cholestasis, or vascular liver disease (e.g., nodular regenerative hyperplasia, veno-occlusive disease, or peliosis hepatis).14 It has been suggested that the occurrence of the former two may be associated with elevated MeMPN concentrations,15, 16 whereas nodular regenerative hyperplasia may be associated with elevated TGN concentrations.17
Sparse data are available on AZA metabolism in AIH and on the utility of such measurements in predicting AZA toxicity. The primary aim of this study was to assess the relationship between AZA metabolite concentrations (i.e., TGNs and MeMPNs), TPMT activity, therapeutic response, and toxicity in adult patients with AIH prescribed a constant dose of AZA for the maintenance of remission.
6MP, 6-mercaptopurine; AIH, autoimmune hepatitis; ALT, alanine aminotransferase; avALT, average ALT; avMeMPN, average MeMPN; avTGN, average TGN; AZA, azathioprine; CI, confidence interval; CV, coefficient of variation; IAIHG, International Autoimmune Hepatitis Group; IBD, inflammatory bowel disease; LoQ, limit of quantification; MeMPN, methylmercaptopurine nucleotide; OR, odds ratio; RBC, red blood cell; TGN, thioguanine nucleotide; TPMT, thiopurine methyltransferase; ULN, upper limit of normal.
Patients and Methods
The study was approved by the Sheffield Research Ethics Committee (09/H1308/102), and full informed written consent was obtained from participating patients.
Patients were eligible for inclusion if they had a diagnosis of probable or definite AIH based on the revised International Autoimmune Hepatitis Group (IAIHG) criteria18 and were at the maintenance of remission stage of treatment. Patients were recruited from the liver clinic at the Royal Hallamshire Hospital (Sheffield, UK) between 2009 and 2011. For the induction of remission, patients were treated with a combination of prednisolone starting at a dose of 30-60 mg/day tapered down as the serum transaminases normalized to 10 mg/day and AZA (1 mg/kg/day) for approximately 2 years.19 AZA was then increased up to 2 mg/kg/day,20 and prednisolone was gradually tapered with the aim to maintain remission on AZA monotherapy. If relapse were to occur, prednisolone was recommenced and tapered down to the minimum dose required to maintain normal serum transaminases, and, if appropriate, the dose of AZA was further increased. All recruited patients had achieved a complete response, as defined by the IAIHG,18 before AZA dose escalation and prednisolone taper.
At recruitment, blood samples (10 mL of lithium heparin) were taken for the measurement of TPMT activity and AZA metabolite concentrations. AZA metabolite concentrations were subsequently measured at each clinic visit throughout the study period, which ran from August 2009 to November 2011. At each clinic visit, body weight, dose of prednisolone and of AZA, full blood count, and liver function tests were also recorded. Demographic data on all patients were collected, including, gender, age, age at diagnosis, fibrosis stage at most recent liver biopsy, and duration of AZA therapy.
Outcome was characterized in two ways: (1) ability to maintain remission, defined in accord with the 2010 American Association for the Study of Liver Diseases guidelines,21 as a normal serum alanine aminotransferase (ALT) level (ALT, <33 IU/L) and (2) occurrence of relapse, defined in accord with the IAIHG criteria,18 as elevation of ALT to more than twice the upper limit of normal (ULN) or liver histology showing active disease. These two outcomes were considered separately because patients with ALT between 1 and 2× ULN are not considered to be in remission, but do not meet the criteria for relapse.
AZA Metabolites and TPMT Activity Determination.
RBC TGN and MeMPN concentrations and TPMT activity were measured by high-performance liquid chromatography assay.22, 23 To avoid interference by AZA drug metabolites, the TPMT assay was modified as previously described.24 The lower limit of quantification (LoQ) for TGN was 30 pmol/8 × 108 RBCs and for MeMPN was 60 pmol/8 × 108 RBCs. Assay precision was measured using a patient quality control, prepared using pooled RBCs, and stored at −30°C. For each assay, an aliquot of the quality control was defrosted and treated in parallel with each batch of patient samples. The break-point separation between high and intermediate TPMT activity distributions was made at 9.5 IU/mL.9, 10
Statistical analysis was performed using Predictive Analytics Software (PASW) Statistics 18.0 for Windows (SPSS, Inc., Chicago, IL) and Minitab (Minitab Inc., State College, PA). The average of the TGN (avTGN), MeMPN (avMeMPN), and serum ALT (avALT) concentrations for each patient over the study period were used in the analysis. Where there was a dose change during the study period, only metabolite measurement before the dose change was used in analysis. All results are presented as median (range). Statistical comparisons were performed using Mann-Whitney's test for two unpaired continuous variables and Fisher's exact test or the chi-squared test for dichotomous variables. Median differences (and the 95% confidence interval; CI) were calculated from the point estimates of all the median differences. Correlations were assessed by Spearman's rank-correlation coefficient. All tests were two-tailed, and P value was significant at 0.05.
Characteristics of the 70 patients studied are given in Table 1. A total of 355 samples were analyzed, with a median of 5 (range, 2-9) samples per patient. The median avTGN concentration was 222 (range, 66-888) pmol/8 × 108 RBCs, and avMeMPN concentration was 1,416 (range, 175-23,798) pmol/8 × 108 RBCs. There was no correlation between dose of AZA (mg/kg/day) and avTGN concentration, but there was a positive correlation with avMeMPN concentration (Fig. 2) (rs = 0.4; P = 0.001).
Table 1. Characteristics of 70 Patients Studied
Results are expressed as median (range).
Number with cirrhosis
Age at diagnosis of AIH, years
Age at start of study, years
Duration of AIH at start of study, years
Dose azathioprine studied, mg/kg/day
Time on studied dose of azathioprine, months
Average prednisolone dose over study period, mg/day
TPMT activities ranged from 5.7 to 16.8 IU/mL. Five (7%) patients had intermediate TPMT activity. When patients with an intermediate TPMT activity were compared with those with a high activity, patients with an intermediate activity were on a lower dose of AZA (1.5 [range, 0.4-2.0] versus 1.8 [range, 0.7-23.2] mg/kg/day; P = 0.4) and had higher avTGN (345 [range, 66-888] versus 221 [range, 67-636] pmol/8 × 108 RBCs; P = 0.7) and lower avMeMPN (570 z[range, 189-2,050] versus 1,698 [range, 175-23,798] pmol/8 × 108 RBCs; P = 0.1), respectively. The small size of the intermediate cohort probably contributed to the observed lack of statistical significance of these observations.
Intrapatient Variation in Metabolite Formation.
To assess the intrapatient variation of metabolite formation on a constant dose of AZA, the coefficient of variation (CV) (standard deviation/mean × 100%) for metabolite concentrations was calculated for patients with four or more measurements (n = 52).
The median intrapatient CV for TGN and MeMPN was 18.3% (range, 6.5-90.5) and 37.1% (range, 7.3-164), respectively. For comparison, the CV of TGN and MeMPN for the quality control pooled patient sample was 8.6% and 7.7%, respectively, from 55 assay runs performed over the study period. Three patients had strong evidence of noncompliance, with one or more measured TGN concentration below the LoQ.
Metabolite Concentrations and Therapeutic Response.
Over the study period, 53 of the 70 patients (76%) maintained remission. Patients who maintained remission tended to be on a lower dose of AZA than those who did not (Table 2). Despite this, higher avTGN concentrations were observed in those who maintained remission, compared to those who did not (237 versus 177 pmol/8 × 108 RBCs; median difference, 60; 95% CI for difference: 7-114 pmol/8 × 108 RBCs; P = 0.025) (Fig. 3). Furthermore, there was a negative correlation between serum avALT and avTGN concentration (rs = −0.32; P = 0.007).
Table 2. Treatment, Metabolite Concentrations, and TPMT Activity of 70 AIH Patients in Remission Versus Those Not in Remission
In Remission (n = 53)
Not in Remission (n = 17)
Values are given as median (range).
AZA dose, mg/kg/day
Prednisolone dose, mg/day
Number of patients on prednisolone (%)
avTGN, pmol × 108 RBC
avMeMPN, pmol × 108 RBC
TPMT activity, IU/mL
A receiver operating characteristic curve was constructed to determine the cut-off concentration for avTGN that best discriminated between those in remission and those not in remission (Fig. 4). An avTGN concentration of >220 pmol/8 × 108 RBCs best discriminated between the two groups, with 62% of those in remission and only 18% of those not in remission achieving a TGN concentration of >220 pmol/8 × 108 RBCs (P = 0.001). The OR of a therapeutic response for an avTGN concentration higher than 220 pmol/8 × 108 RBCs cutoff was 7.7 (95% CI: 2.0-30.2; P = 0.003).
Twelve of the fifty-three (23%) patients who maintained remission received prednisolone in combination with AZA during the study period; prednisolone was continued arbitrarily in 5 patients and because of previous relapse in 7 patients. Even after excluding those who were in remission, but who were continued on prednisolone, the avTGN concentration was significantly higher in those who maintained remission (on AZA monotherapy; n = 41), compared to those who were not in remission (n = 17) (237 versus 177 pmol/8 × 108 RBCs; median difference, 52; 95% CI for difference: 5-104 pmol/8 × 108 RBCs; P = 0.028).
We also performed an analysis on all 355 individual samples. The negative correlation between serum ALT and TGN concentration (rs = −0.37; P < 0.001) was still apparent, as was the significant difference in TGN concentration between those with a normal serum ALT level (n = 261) and those with an elevated serum ALT level (n = 94) at time of metabolite measurement (236 versus 191 pmol/8 × 108 RBCs; median difference, 51 pmol/8 × 108 RBCs; 95% CI for difference: 28-74; P < 0.0001).
Fifteen patients (21%) experienced a relapse during the study period. Those who did not relapse, compared to those who relapsed, had higher avTGN concentrations (236 versus 177 pmol/8 × 108 RBCs; median difference, 55 pmol/8 × 108 RBCs; 95% CI for difference: 1-115; P = 0.04). An avTGN concentration of >220 pmol/8 × 108 RBCs was attained by 58% of those who did not relapse and 27% of those who relapsed (P = 0.03).
In contrast to avTGN, there was no correlation between serum avALT and avMeMPN concentrations (rs = 0.1; P = 0.4). There was also no difference in avMeMPN concentrations or TPMT activity between those in remission and those not in remission (Table 2) or between those who relapsed and those who did not relapse.
Leucopenia (white cell count: <3.5 × 109/L) developed in 4 (6%) patients, leading to dose reduction in 3 patients. The avTGN concentrations measured in the latter 3 patients were 236, 294, and 469 pmol/8 × 108 RBCs on an AZA dose of 2.2, 0.9, and 1.7 mg/kg/day, respectively. In the remaining patient, the leucopenia was mild and transient (avTGN: 167 pmol/8 × 108 RBCs on AZA dose of 1.2 mg/kg/day). All of the patients who developed leucopenia had high TPMT activity.
During the course of the study, 2 patients developed cholestatic jaundice, with only mild elevations of ALT and liver biopsy suggestive of AZA-induced cholestasis, rather than an AIH flare. In 1 patient, AZA dose had been escalated to 3.2 mg/kg/day because of poorly controlled disease. Because of a lack of clinical response, the patient was admitted to the hospital, where he had directly observed treatment. During the admission, the TGN and MeMPN concentration increased from 150 to 254 and 1,698 to 9,675 pmol/8 × 108 RBCs, respectively, and he developed jaundice. The second patient was on a stable AZA dose, and avMeMPN concentration was 18,879 (range over study period: 13,693-20,084) pmol/8 × 108 RBCs. Both patients had high TPMT activity (12.7 and 13.7 IU/mL, respectively). In both patients, withdrawal of AZA resulted in clinical and biochemical improvement. There was no difference in the avTGN concentration, but the avMeMPN concentration was higher in the 2 patients who developed cholestasis, compared to those who did not (14,277 versus 1,416 pmol/8 × 108 RBCs) (Fig. 5).
One patient, who was asymptomatic and had a normal serum bilirubin and ALT level, developed nodular regenerative hyperplasia, diagnosed by routine liver biopsy performed to assess histological disease activity. The avTGN was 469 pmol/8 × 108 RBCs, which was in the upper quartile of the study population. The avMeMPN concentration was 1,003 pmol/8 × 108 RBCs on an AZA dose of 1.7 mg/kg/day, and the TPMT activity was 14.3 IU/mL.
Effect of Cirrhosis.
Liver biopsy specimens were available in 65 patients for assessment of fibrosis stage. Cirrhosis was present in 11 (17%) of patients. Patients with cirrhosis were on lower AZA dose (1.3 versus 1.9 mg/kg; median difference, −0.5; 95% CI for difference: −0.9 to −0.0001; P = 0.03) and had similar TGN (251 versus 220 pmol/8 × 108 RBCs; P = 0.4), but lower MeMPN (428 versus 1,877 pmol/8 × 108 RBCs; median difference, −952; 95% CI for difference: −2,627 to −263; P = 0.006) concentration, compared to patients without cirrhosis. The TPMT activity was similar in both groups (11.9 versus 11.7 U/mL RBC; P = 0.9). There was no difference in the response rate between the patients with cirrhosis and patients without cirrhosis (P = 0.7).
The main finding of this study was that in patients with AIH, successful maintenance of remission during treatment with AZA was associated with higher average TGN levels than was failure to maintain remission.
Specifically, TGN values of >220 pmol/8 × 108 RBCs were significantly associated with continued AIH remission, with an OR of 7.7.
Our results are consistent with those of studies in childhood leukemia and IBD, which also support an association between higher TGN levels and therapeutic efficacy of AZA.12, 13 However three previous studies in AIH have not demonstrated such a relationship.25-27
Our results might differ from those of these previous studies for several reasons. First, we considered the average of repeated TGN values in each patient, whereas previous studies reported single TGN measurements on an unselected cohort of patients. Second, in accord with recent American and UK guidelines,21, 28 we defined remission as normalization of serum ALT. One previous study also used this definition,26 but the other two defined remission as serum ALT of less than 1.5 times normal27 and less than twice normal.25 Third, we focussed specifically on patients who were at the maintenance of remission stage of treatment, with a target maintenance dose of 2 mg/kg/day (if tolerated), whereas previous studies were in unselected patients on more variable doses; therefore, the median dose of AZA and the median TGN concentrations (1.9 mg/kg/day and 222 pmol/8 × 108 RBCs respectively) were higher in our study than in previous reports. In one report of 52 patients, the mean dose of AZA was 1.6 mg/kg/day and median TGN concentration was 155 pmol/8 × 108 RBCs, with no difference in between those who did and who did not require prednisolone in addition to AZA to maintain remission.25 In a second report of 49 patients, the mean AZA dose was 1 mg/kg/day with TGN concentrations of 149 pmol/8 × 108 RBCs (27). A third, large, study (n = 143) also reported similar TGN concentrations between responders and nonresponders (113 versus 121 pmol/8 × 108 RBCs) on AZA dose of 1.19 and 1.64 mg/kg/day, respectively, but a high proportion of patients in this study (72%) were on concomitant prednisolone.26
We found wide intrapatient variation in the TGN concentrations over the 2-year study period. This has previously been observed in IBD.29 Three of the study participants had overt evidence of noncompliance with treatment. However, because RBC TGN concentrations reflect accumulation over several doses, rather than the dose taken the day before,7 partial compliance may also be a contributing factor to the observed intrapatient variability. AIH is a chronic disease, with a relapsing and remitting course, and may be asymptomatic,30, 31 thus potentially making it more difficult for patients to fully comply with drug treatment. A study evaluating adherence to treatment in 14 children with AIH using an electronic monitoring device reported an optimal adherence of 28%-94%, with no patient taking the medication exactly as prescribed.32 A study in IBD, using a structured questionnaire, found that 44% of patients admitted to missing more than one dose a week, with a median of 2 days per week of missed medication.33
Although we have found that a TGN concentration of >220 pmol/10 × 108 RBC is associated with an increased likelihood of remission, remission can also be achieved with lower TGN concentrations in the individual patient. Therefore, we suggest that TGN measurements should be mainly considered for those who have active disease, because TGN concentrations may help identify patients who have inadequate immunosuppression and may also point to noncompliance. Care must be taken to exclude the latter if increasing the AZA dose. In those where noncompliance has been excluded and the TGN concentration is low, dose increase with TGN metabolite monitoring may be useful, because AZA-derived 6-mercaptopurine (6MP)34 and associated metabolites35 exhibit nonlinear kinetics on dose escalation. For some individuals, especially those with very high TPMT activities, increasing the dose of AZA may increase the TPMT-catalyzed methylation rate in preference to TGN formation.35 This has been documented in IBD, where AZA dose escalation, for some patients, caused only minimal increases in TGN levels, but selectively increased MeMPN levels.36
A further finding of this study is that elevated MeMPN concentration was associated with AZA-induced cholestasis. This is consistent with two previous case reports in AIH in which biopsy proven hepatotoxicity was associated with high MeMPN concentration.16, 25 The mechanism by which increased MeMPN may cause hepatotoxicity is unknown. In AIH, it can be difficult to distinguish hepatotoxicity from a flare of the disease, and furthermore, cholestasis does not universally develop in those with a high MeMPN concentration. Therefore, although measurement of MeMPN may be useful in suspected cases of AZA-induced cholestasis, we would not recommend that increased MeMPN concentrations be used as an alternative to a liver biopsy as a method of diagnosing hepatotoxicity.
In keeping with previous reports,25, 26, 37 we did not find an association between TPMT activity and AZA-induced toxicity. However, we must emphasis the selected nature of the cohort studied (i.e., AIH patients who tolerated AZA and were on long-term maintenance therapy). TPMT status determination before starting AZA is useful only for the detection of complete TPMT deficiency. These individuals are at risk of profound myelosuppression on standard doses of AZA, and alternative immunosupression should be considered or, if treating with AZA, the dose should be reduced by 10-fold with thrice-weekly, instead of daily, dosing.38
The presence of cirrhosis is associated with an increased risk of developing AZA-induced toxicity.25 We found that patients with cirrhosis, compared to patients without cirrhosis, produced similar TGN metabolites, despite being on a lower dose of AZA. Therefore, metabolite monitoring in patients with cirrhosis may be useful to guide dosing to minimize toxicity while maintaining the efficacy of AZA, although further studies are needed before this can be firmly recommended.
In conclusion, we have demonstrated that TGN metabolite accumulation is associated with maintenance of remission in AIH. For the few patients who experienced AZA toxicity, cholestatic hepatoxicity was associated with increased MeMPNs, whereas the single patient who developed nodular regenerative hyperplasia had elevated TGNs. This study has demonstrated a role for monitoring AZA metabolites in the clinical management of AIH and indicated that the effects of AZA dose escalation in AIH needs further evaluation.