Efficacy and safety of thiopurinic immunomodulators (azathioprine and mercaptopurine) in steroid-dependent ulcerative colitis


Dr A. Lopez-Sanroman, Servicio de Gastroenterología, Hospital Ramón y Cajal, E28034 Madrid, Spain.
E-mail: alopez.hrc@salud.madrid.org


Background : The efficacy of azathioprine in the management of steroid-dependent ulcerative colitis is taken for granted. However, study populations frequently include together steroid-dependent and refractory patients.

Aim : To assess the efficacy and safety of thiopurinic immunomodulators in strictly defined steroid-dependent ulcerative colitis.

Methods : Survey of 34 patients with steroid-dependent ulcerative colitis, treated with azathioprine according to protocol. Therapeutical success: glucocorticoid withdrawal within 12 months, without steroid requirements during another year.

Results : Mean age was 39.1 ± 17 years. Pancolitis and extensive colitis accounted for 50% of cases. Therapeutic success of immunomodulator treatment reached 70.6%, intention to treat analysis (confidence interval 95%: 52–84%) and 72.7%, as per protocol (confidence interval 95%: 54–86%). Mean time to steroid withdrawal was 4.6 months. In therapy successes, mean corpuscular volume and total serum bilirubin increased with treatment time (P = 0.0001). Fifteen adverse effects were observed in 13 patients (38%). Azathioprine was withdrawn in seven cases (20.6%); in four of them (with liver toxicity), treatment with mercaptopurine was indicated.

Conclusions : Therapy with thiopurinic immunomodulators (azathioprine) represents the first option in the management of steroid-dependent ulcerative colitis. Its efficacy (70%) and its acceptable safety support this view. Increasing mean corpuscular volume and serum bilirubin values may be a surrogate marker of a beneficial effect.

Background and aim

Glucocorticoids (GCS) offer efficient and prompt relief of symptoms in many patients with moderate to severe ulcerative colitis (UC) flare-ups. Nevertheless, after an initial response, up to 20% of cases enter a steroid-dependence situation.1 The use of GCS as maintenance therapy is hampered by their potential adverse effects and also by the absence of clear efficacy; this is why steroid dependent patients need an alternative treatment that allows withdrawal of GCS. In this situation, clinicians frequently initiate the use of thiopurinic immunomodulators like azathioprine (AZA) and mercaptopurine (MP).2 The efficacy of those drugs in this setting is usually assumed, but data are surprisingly scarce in UC patients. Furthermore, in many available studies, steroid-dependent and steroid-refractory patients are evaluated together, ignoring the well established clinical and therapeutic differences between those two situations, the main one being that refractory patients are best managed with ciclosporin, alone or in combination with AZA. Our aim was to assess the efficacy and safety of AZA therapy in patients with strictly defined steroid-dependent UC, investigating if any clinical factor could influence or predict therapeutic success.


We included 34 patients with steroid-dependent UC, prospectively followed as in- or out-patients by the same clinician (ALS), between January 2000 and December 2003, according to pre-established protocol. Steroid dependency was defined as recurrent flare-up on GCS reduction or withdrawal, or as the clinical need for GCS treatment twice within six consecutive months or three times within a year. Therapeutical success of thiopurinic derivatives was defined as complete GCS withdrawal in a maximal follow-up of 12 months, without new GCS requirements during one whole year of follow-up.

Erythrocyte levels of thiopurinemethyltransferase (TPMT) were determined by a previously described radiochemical method based in the conversion of MP to a methylmercaptopurine, using 3H-S-adenosyl-l-methionine as a methyl donor;3 this was carried out at the Unit of Liver Molecular Toxicology, Biochemistry and Molecular Biology Department, University of Alcalá, Spain, and financed by the Scientific Department of Celltech Pharma, Spain. Whenever AZA was needed, the starting dose was adjusted to the TPMT activity: 2 mg/kg/day if TPMT levels were lower than 18 IU/mL, 2.5 mg/kg/day for higher TPMT values. AZA would have been avoided if any patient had had TPMT levels under 5 IU/mL.

The following demographic and clinical data were registered: age, gender, extent of disease, time since UC diagnosis, previous moderate or severe flare-ups and their response to treatment, AZA dose, and levels of TPMT.

Adverse effects attributable to AZA use were also prospectively noted. Every patient was submitted to the same follow-up schedule: clinical interview and complete blood count and liver function tests (alanine aminotransferase -ALAT-, aspartate aminotransferase -ASAT-, gammaglutamyltranspeptidase -GGT-, alkaline phosphatase and bilirubin) before the first AZA dose, and then every 2 weeks during the first month, monthly during the first trimester and every 3 months thereafter. Serum amylase and lipase were determined as clinically needed. The following adverse effects were especially investigated: pancreatitis, digestive intolerance, any infection, bone marrow toxicity (leucopenia < 3000/mm3, neutropenia < 1500/mm3, thrombopenia <100 000/mm3), and liver toxicity (ALAT, GGT or alkaline phosphatase above twice their upper normal level). Steroid withdrawal was progressive, starting 2 weeks after the first AZA dose.

Statistical analysis of qualitative variables included calculation of proportions and their 95% confidence interval (CI 95%). In quantitative variables, arithmetic mean and standard deviation were calculated. Categorical variables were compared by means of Fisher's exact test, because less of five registers were expected in some cases. Therapeutic success was analysed by ‘intention-to-treat’ (included all patients enrolled into the study) and ‘per protocol’ (excluded patients with poor compliance of treatment). Analysis of differences between laboratory values before treatment, and 1 and 3 months thereafter, was carried out by the Friedman's test. Wilcoxon's signed rank test was used for comparison between paired data. A statistical P-value <0.05 was considered statistically significant.


Mean age was 39.1 ± 17 years; 62% were male. Pancolitis and extensive colitis accounted for 50% of cases, 35% were left-sided colitis and 15% proctosigmoiditis. A patient with moderate mental retardation abandoned her therapy for unexplained reasons after a few days of AZA administration. She was considered a treatment failure for intention to treat analysis. Mean AZA dose reached 2.17 ± 0.24 mg/kg/day. Mean prednisone dose at the beginning of AZA therapy was 38.7 ± 17.5 mg/kg/day. Patients had been on steroids, a mean of 4.2 ± 1.8 months. Fifteen adverse effects were observed in 13 patients (38% of treated patients, CI 95%: 22–56%) (Table 1). In two cases (5.9%), this did not lead to a treatment modification (ophthalmic herpes zoster and digestive intolerance), and in a further four patients (11.8%). AZA dose was halved (three cases with bone marrow toxicity and one case with digestive intolerance). Treatment was interrupted in seven patients (20.6%), and in four of them (all with mild liver toxicity) alternative therapy with MP was started; those four patients could be evaluated after follow-up (two cases of therapeutic success and two failures). Mean TPMT activity reached 21.7 U/mL of erythrocytes (range 10.7–31.6). All four cases with bone marrow toxicity had TPMT levels within normal range (higher than 18 U/mL erythrocytes).

Table 1.  Azathioprine adverse effects observed during follow-up
Adverse effect% Casesn
Bone marrow toxicity (three neutropenia, one thrombopenia)8.83
Infections (two herpes, one endocarditis)8.83
Dyspepsia (nausea, epigastric pain)2.91
Hepatotoxicity + neutropenia2.91
Hepatotoxicity + dyspepsia2.91

Immunomodulator treatment was successful in 70.6% (52–84%) of cases, according to intention to treat, and in 72.7% (54–86%) as per protocol. In patients with therapeutic success, mean time until complete steroid withdrawal was 4.6 months (range 1–9). Withdrawal was achieved during the first 6 months in all, except in two cases.

The following variables did not influence therapeutic success: age (categorized as over or under 40 years), UC extent (extensive colitis vs. left-sided colitis or proctosigmoiditis), time since UC diagnosis (≤1 year vs. >1 year), AZA dose (2 mg/kg/day vs. 2.5 mg/kg/day), number of previous flare-ups (0–1 vs. >1) and previous response to steroids (yes/no) (Table 2). We observed a noteworthy albeit non-significant treatment success difference according to gender (success in 81% of males and only 53.8% of females). The low number of events prevented the performance of a model of multiple logistic regression.

Table 2.  Characteristics of the analysed variables according to immunomodulator treatment (success or failure)
VariableTherapy successTherapy failureP-value
Age (years)37.2 ± 18.143.7 ± 17.70.35
Gender (% males)70%40%0.13
Extent of disease (% extensive colitis)50%50%1
Time since diagnosis of colitis (years)3.2 ± 3.23.1 ± 3.30.89
Azathioprine dose (% with 2.5 mg/kg/day)30%40%0.69
Previous flare-ups (% with more than one flare-up)62%60%1
Steroid responders in previous flare-ups (%)70%80%0.69

In patients with therapeutic success, Friedman's test showed that the mean corpuscular value (MCV) significantly differed at each control (initial, at 1 month and at 3 months, P = 0.0001). Mean MCV before treatment was 84.7 ± 5.1 fl, whilst 1 month thereafter it had significantly increased to 88.7 ± 6 fl, in patients with therapeutic success (P < 0.0001). In these patients, at the third month, MCV had further increased up to 92 ± 5.8 fl (P < 0.0001, when compared with MCV values after 1 month of therapy). Conversely, in patients with eventual therapeutic failure, mean MCV did not significantly change over time, in spite of maintained AZA administration (P = 0.63: 85.2 ± 3 fl before therapy, 87 ± 4.8 fl 1 month later, and 87.7 ± 5 fl after 3 months of therapy).

A similar evolution was observed in bilirubin levels. In patients with therapeutic success, Friedman's test showed that its values changed significantly over time (before therapy, and 1 and 3 months thereafter, P =0.0001). Mean pre-treatment value was 0.43 ± 0.2 mg/dL, 1 month later it had gone up to 0.61 ± 0.3 mg/dL (P < 0.0001), and at the third month, again to 0.71 ± 0.3 mg/dL (P = 0.03, when compared with the previous control). This was not observed in patients with therapeutic failure (bilirubin levels 0.51 ± 0.3 mg/dL before treatment, 0.61 ± 0.3 mg/dL at month +1, and 0.51 ± 0.2 mg/dL at month +3, P = 0.21). Absolute leucocyte count did not differ between responders and non-responders.

After steroid withdrawal, and during a mean follow-up of 25 months (range 8–44), we observed 20% of clinical recurrences (n = 5): four cases (16%) with mild to moderate activity (one treated with topical budesonide and three with oral prednisone) and one case (4%) with a severe flare-up, that did not respond to steroids, ciclosporin nor infliximab, and finally needed proctocolectomy to achieve disease control.

Alternative treatment in AZA/MP failures (n = 10) included five proctocolectomies with ileal pouch and ileoanal anastomosis, four cases with successful off-label infliximab treatment,4 and one poorly controlled patient with irregular oral and topical mesalazine therapy.


The usefulness of AZA/MP as maintenance therapy in UC is generally accepted, based on the results communicated by different authors.5–22 Nevertheless, inclusion criteria have not always been uniform, counting together apples and oranges (steroid dependent and steroid refractory patients). The main prerequisite for data analysis is a strict definition of the treated population, which has to be as homogeneous as possible, and of therapeutic success. Our series is composed only by strictly defined steroid-dependent patients. We think that this population is completely different to that of steroid-refractory patients treated with AZA; these are currently managed first with ciclosporin and then or even simultaneously with AZA. The situation is not comparable with that of stable steroid-dependent patients who are exposed to a single immunosuppressor.

We defined AZA efficacy as complete steroid withdrawal,16, 19 the main therapeutic objective in steroid-dependent UC. The use of a clinical score as an endpoint did not seem appropriate: many of those patients would have had equally low scores on entering the protocol, but under steroid therapy. This justifies complete steroid weaning as a better endpoint than a clinical index. Therapeutic response may take a significant latency;23 thus, to ensure that AZA had its maximum clinical effect, we predefined an extended 12-month follow-up period, during which patients could be progressively weaned off GCS. A further observation period of 1 year without new flare-ups was required to finally declare complete steroid withdrawal. We therefore consider that our therapeutic success criteria are strict and methodologically sound.

Steroid dependency is relatively frequent in UC, appearing in every fifth patient treated with GCS.1 The efficacy of AZA in allowing steroid withdrawal in this clinical situation is impressive (70% of our patients achieved this goal). Our results agree with those of previous studies, that show clinical remission after 6–12 months of AZA/MP therapy in 50–75% of patients.7, 11, 15–19, 21 Taking into account, that the only alternative to immunosuppressor therapy would be proctocolectomy, a procedure not devoid of morbidity and mortality,24 we think that the therapeutic usefulness of thiopurine derivatives is clinically relevant. In our series, only five of 34 patients (14%) finally required surgery. Thus, AZA is effective in avoiding colectomy in patients with steroid-dependent UC.

Safety profile of AZA can be described as acceptable. Adverse effects observed during follow-up were similar to those previously described.25, 26 Bone marrow toxicity was observed in patients with normal TPMT activity, as previously described.27, 28 The finding of a normal TPMT activity does not prevent periodical clinical and laboratory review,29 it allows to start treatment with an adjusted dose.30

Keeping in mind that AZA therapy is frequently the last and only alternative to proctocolectomy, mild adverse effects (slight neutropenia, tolerable dyspepsia…) can be managed by AZA dose reduction before deciding on a complete withdrawal. Another option is to stop AZA and start the patient on MP therapy, as suggested by some, although published data only include isolated cases of liver toxicity.31, 32 The non-enzymatic reaction that converts AZA into MP, produces free imidazolic derivatives, that could be partially responsible for AZA hepatotoxicity.33 Although more data are clearly needed, if mild liver function tests alteration does not resolve after halving the AZA dose, a possibility is to use MP instead, we did so in four of our patients, achieving complete resolution of liver test alteration.

The AZA therapy is accompanied by a progressive erythrocyte macrocytosis. Our data suggest that a significant increase in the MCV and bilirubin could be a marker of probable therapy success. Changes in MCV and absolute MCV value, have been proposed as surrogate indicators of intraerythrocytary tioguanine (thioguanine) nucleotide levels.34, 35 The latter assay has been used as an indicator of optimal AZA dosage, permitting a finer adjustment and a monitoring of therapeutic adherence.36 MCV measurement could be a cheaper and readily available alternative. The hypothesis goes as follows: if MCV values fail to rise after 3 months of AZA treatment, this could reflect low levels of intraerythrocyte tioguanine (thioguanine) nucleotides levels, probably related to subtherapeutical AZA doses.35 In turn, bilirubin levels only rose in our responding patients, which could be related to an increased destruction of macrocytic erythrocytes in the spleen, or maybe to a certain degree of defective erythropoiesis. It is widely available and easy to assess, and its increase could also predict an adequate therapeutic response. Nevertheless, all these theories have to be adequately confirmed, and those determinations should never replace usual clinical and laboratory parameters on response and safety assessments. If adequate tioguanine nucleotide levels36 and perhaps MCV increments (3–8 fl at the third month)34 are not paralleled by a therapeutic response, this could reflect a true AZA resistance and call for alternative treatments.

Finally, we did not find any influence on response to AZA of the previous medical history (disease extent, time since diagnosis, previous flare ups, previous steroid response), which is not in agreement with previous experiences.5, 16 AZA doses of 2 and of 2.5 mg/kg/day are equivalent, in our experience.

Therapy with thiopurinic immunomodulators (AZA/MP) is the first option for the management of steroid-dependent UC, as confirmed by its efficacy (70%), and an acceptable degree of safety. It is important to patiently wean the patient off steroids. Increasing MCV and bilirubin levels are possible predictors of response, although traditional clinical and laboratory markers are still standard procedure. TPMT determination avoids exposure to AZA/MP of patients with absent or very low activity, but normal levels do not exclude the possibility of serious bone marrow toxicity.


Thiopurinemethyltransferase determination was financed by Celltech Pharma Spain. Dr Bermejo was assigned an investigational grant by Schering-Plough Spain, during the elaboration of this paper. Dr López-San Román has been advisor and lecturer for Schering-Plough Spain.

We would also like to thank our patients and coworkers.