- Top of page
Purine analogues, azathioprine (AZA) and its active metabolite mercaptopurine (MP), are potent immunomodulators that are widely used in the treatment of IBD as a steroid-sparing agent to induce and to maintain remission,1 although its mechanism of action is not completely understood.2
Both AZA and MP are prodrugs, which require intracellular activation by a multienzymatic process involving three competing pathways via three different critical enzymes: Hypoxanthine phosphoribosyl transferase (HPRT), thiopurine methyl transferase (TPMT) and xanthine oxidase (XO).2, 3 The HPRT converts AZA/MP into thioinosine monophosphate that is later converted into the active cytotoxic metabolite 6-thioguanine nucleotide (6TGN), which has immune modifier activity, but may lead to myelosuppression. The XO metabolises an important part of the MP into the inactive agent 6-thiouric acid,2, 4 and TPMT catalyses two reactions resulting in the formation of 6-methyl-mercaptopurine ribonucleotide (6MMP), which is also inactive, but may be responsible for some toxic effects like hepatotoxicity.2, 5, 6
The TPMT gene has an autosomal codominant inheritance; occasional genetic polymorphism has been described regarding the TPMT activity, which results in a trimodal distribution7, 8; those patients, heterozygous or homozygous for the ‘low activity’ mutation gene may have a special susceptibility for myelotoxicity with thiopurine therapy.9, 10 Approximately 90% of the population are homozygous for the wild-type TPMT gene and have normal to high activity of this enzyme; it is a commonly held belief that this characteristic provides a safer condition regarding the intake of thiopurines. On the other hand, higher TPMT activity has been related to lower values of active metabolites that may reduce the probability of achieving clinical response.11–13
Approximately 30% of patients fail to respond to standard doses of thiopurines.14, 15 Since 6TGNs are thought to be both active and toxic metabolites, there have been many attempts to define a therapeutic range and a toxic threshold; however, most of these studies are small, retrospective or cross-sectional, with short-term follow-up and heterogeneity in their inclusion criteria. Despite this, two meta-analyses have been conducted to examine the relationship between red blood cells (RBC) levels of 6TGN and the clinical response to AZA/MP.16, 17 Both conclude that IBD patients in remission have significantly higher 6TGN levels than those with active disease; however, the clinical usefulness of this conclusion is limited due to the lack of clearly defined thresholds (230 or 260 pmol/8 × 108 RBC) and their low sensitivity and specificity to predict remission.16, 17
The proposed threshold values have not been confirmed by other groups and there is no unique and reliable threshold value for clinical practice.18–21 Moreover, the American Gastroenterological Association suggested that the usefulness of thiopurine metabolite monitoring in IBD patients when trying to optimise dose or to monitor toxicity still remains unclear (cited as recommendation grade C).22
A prospective multicentre study was designed aiming to elucidate if there is any useful relationship between the RBC levels of AZA metabolites and the achievement of clinical response or the development of toxicity. The first aim was to establish if early determination of TPMT activity or the RBC levels of these metabolites was effective in predicting the outcome of the AZA/MP therapy. The identification of clinically significant 6TGN threshold values or a therapeutic range for clinical efficacy or safety, to be used in clinical practice, was also performed. Unless this therapeutic range is clearly defined, a thiopurine dose adjustment according to those levels of metabolites should not be firmly recommended.
- Top of page
From October 2005 to December 2008, 20 hospitals throughout Spain participated in the study and 154 patients were initially considered for inclusion, but one patient with undetectable TPMT activity and 17 patients lacking at least one of the required clinical evaluations or metabolite determinations during follow-up, and thus could not be considered for the final analysis, were excluded. Therefore, patients considered for analysis were from 16 hospitals, of which five hospitals each recruited 10 or more patients, and less than 10 patients were recruited at each of the remaining 11 hospitals, of which seven hospitals each recruited between one and three patients. Moreover, 23 patients were not included in the initial analysis because they were under combo therapy with thiopurines and infliximab, due to previous thiopurine treatment failure (17 had CD, and only 50% responded to infliximab); this group of patients was the matter of a different analysis.
Finally, 113 patients completed the required follow-up and, overall, the mean follow-up was 25 weeks (range 2–48). Fifty per cent were male and the mean age was 36 years old (range 18–77). Seventy-one per cent of patients had CD, and the remaining had UC. The CD patients were classified regarding location of their disease as follows: terminal ileum (28.5%), colon (20%), ileocolonic (48.5%) and proximal digestive tract (3.2%); 23.5% also had fistulising disease. Sixty-five per cent of the UC patients included had extensive colitis, 32.5% had left-sided colitis and only 2.5% had proctitis. The mean duration of the disease at the inclusion in this study was 5.6 months (range 0–35 months).
Among all patients included, 90% had high (wild) TPMT activity (>13.7 U/mL RBC), and 10% had intermediate TPMT activity (5–13.7 U/mL RBC). Only one patient was excluded from the study due to low TPMT activity. Distribution of TPMT activity is represented in Figure 1. At inclusion, 78% of the patients were on steroid therapy (the remaining had just completed steroid tapering) and 41% were receiving aminosalicylates. At the moment of inclusion, mean Truelove index value was 13 points (range 11–17) for UC patients, and the mean CDAI was 173 points (range 120–421). Interestingly, there were more patients who were not on steroids at inclusion (20 patients, 80%) who achieved final response as defined, but only 44 patients (55%) of those who were on steroids at inclusion were considered as final responders (P = 0.01). On the other hand, aminosalicylates therapy did not show such effect: 28(64%) patients who were under aminosalicylates achieved final response as defined, whereas 38(61%) patients who were not receiving aminosalicylates therapy reached the treatment goal (P = 0.8).
Regarding thiopurines, 91% of the patients were prescribed AZA and the remaining received MP. Although all patients started thiopurines in a dose as close as possible to 2.5 mg/kg of AZA or 1.5 mg/kg of MP, mean final dose of AZA was 2.3 mg/kg (range 1–2.8) and mean final dose of MP was 1.2 mg/kg (range 1–1.4). The choice between AZA and MP was made based on each investigator’s own criteria. Therapeutic success of thiopurines (clinical response), as previously defined, was achieved in 70 (62%) of the patients included.
According to the design of the study, the follow-up of the responding patients lasted for 6 months after steroid withdrawal, but was shorter in those patients who did not respond and needed new treatments to induce clinical response or had an adverse event that forced to discontinue the drug. On the other hand, those patients who were under steroid therapy at inclusion were followed up also during the steroid tapering. Eighty-eight patients belonged to this latter group; of those, 12(11%) patients were unable to withdraw steroids and were considered to have lack of response and six patients experienced early adverse events that forced to suspend thiopurines before steroids withdrawal. Ninety-five (84%) patients were followed up after steroid withdrawal, and they were evaluated at the second, fourth and sixth month; responding patients at each monitoring point were 86 (76%), 73 (65%) and 70 (62%), respectively. Regarding the remaining 25 patients, two of them suffered adverse events that forced to finish the thiopurine therapy and 23(20%) needed new medical treatment for IBD at some point during the follow-up, what was considered as treatment failure. The flow chart of the patients is expressed as a CONSORT diagram in Figure 2.
Efficacy of TPMT activity or AZA metabolite determination to predict the outcome of therapy with AZA/MP
No differences were found in TPMT activity values, which were determined before starting AZA/MP, among those patients who were responders or nonresponders at the end of follow-up. Mean TPMT activity in responders was 21 ± 4.3 U/mL and 21 ± 5.5 U/mL (P = 0.9) in nonresponders. No significant TPMT threshold values with enough sensitivity and specificity to distinguish responders from nonresponders were found. Moreover, taking all possible cut-off values into account, the AUC was 0.52 (Figure 3).
Figure 3. Receiver operating characteristic (ROC) curves expressing the sensitivity and specificity of all the possible 6-thioguanine nucleotide (6TGN) and thiopurine-methyl-transferase (TPMT) activity cut-off points to predict the final outcome during the early stages of the thiopurinic therapy. AUC, area under the curve.
Download figure to PowerPoint
Serum titres of active metabolites were determined at week 2 and at months, 1, 2 and 4 after starting thiopurine therapy, aiming to demonstrate whether the early determination of active metabolites had any response predictive property. Mean values of 6TGN (and also of 6MMP, 6TGN/6MMP or 6TGN/TPMT) throughout the early period of thiopurine therapy, were not different between those patients who were responders or those who were nonresponders at the end of the follow-up (Table 1). Previous literature proposes the threshold values of 230 or 260 pmol/8 × 108 RBC to predict response, but in this cohort, these values showed poor sensitivity, specificity, positive and negative predictive values, during early follow-up (Table 1).
Table 1. Accuracy of early determination of 6-thioguanine nucleotide to predict final response to therapy
| ||Time from the start of thiopurinic therapy|
|2 weeks||1 month||2 months||4 months|
|Number of patients||113||96||95||76|
|Mean 6TGN in R/NR (P)||247/276 (P = 0.4)||283/335 (P = 0.3)||304/305 (P = 0.9)||354/322 (P = 0.7)|
|6TGN = 230 (95% CI)|
| % Sensitivity||41 (26–54)||58 (42–74)||57 (42–72)||68 (54–82)|
| % Specificity||56 (37–75)||50 (30–70)||50 (30–70)||62 (35–85)|
| % PPV||61 (44–78)||64 (48–80)||67 (51–82)||84 (71–97)|
| % NPV||36 (22–50)||44 (25–62)||40 (22–58)||40 (18–61)|
|6TGN = 260 (95% CI)|
| % Sensitivity||35 (21–49)||49 (33–65)||53 (38–68)||57 (42–73)|
| % Specificity||65 (48–84)||53 (33–74)||57 (37–77)||75 (48–93)|
| % PPV||63 (44–82)||62 (44–79)||68 (52–84)||87 (70–96)|
| % NPV||37 (24–51)||40 (23–58)||41 (24–58)||37 (19–56)|
|Mean 6MMP in R/NR (P)||2890/4115 (P = 0.1)||3640/3588 (P = 0.9)||3246/3558 (P = 0.7)||3218/4005 (P = 0.5)|
|Mean 6TGN/6MMP in R/NR (P)||0.2/0.2 (P = 0.7)||0.2/0.3 (P = 0.2)||0.5/0.8 (P = 0.2)||0.4/0.1 (P = 0.1)|
|Mean 6TGN/TPMT in R/NR (P)||12/17 (P = 0.1)||15/23 (P = 0.1)||16/19 (P = 0.4)||16/12 (P = 0.2)|
Taking all possible cut-off values into account at every point during the early follow-up, the AUC for AZA metabolites (and 6TGN/6MMP or 6TGN/TPMT) was below 0.7 in all subjects, and no value of 6TGN (or 6TGN/6MMP, or 6TGN/TPMT) showed enough sensitivity and specificity to predict the final outcome (Table 1, Figure 3).
There were no differences in these results regarding the type of IBD (CD or UC) or the immunosuppressant that was administered (AZA or MP).
Identification of a therapeutic threshold value of serum AZA metabolites
Mean serum metabolite levels at months, 2, 4 and 6 after steroid withdrawal were obtained to define the relationship between serum metabolite levels and clinical response. There were no statistical differences in 6TGN, 6MMP, 6TGN/6MMP or 6TGN/TPMT levels, at any time during follow-up after steroid withdrawal, between patients who were in and out of remission (Table 2). The ROC curves describing the sensitivity and specificity of all the possible cut-off points to distinguish responders from nonresponders at every step during follow-up after steroid withdrawal were performed; the AUC relating 6TGN levels and clinical response for each monitoring point was less than 0.7 in all subjects (0.61 at the second month, 0.38 at the fourth month, 0.41 at the sixth month). Once again, no cut-off point with clinically relevant sensitivity, specificity, positive and negative predictive values to identify patients in remission, was observed (Table 2). There were no differences in these results regarding the type of IBD (CD or UC) or the immunosuppressant that was administered (AZA or MP).
Table 2. Accuracy of the determination of 6-thioguanine nucleotide to assess the efficacy of the therapy after steroid withdrawal
| ||Time from steroid withdrawal|
|2 months||4 months||6 months|
|Number of patients||86||73||70|
|Mean 6TGN in R/NR (P)||305/359 (P = 0.4)||377/361 (P = 0.9)||427/318 (P = 0.9)|
|6TGN = 100 (95% CI)||94 (81–99)/0 (0–40)||92 (79–98)/0 (0–70)||91 (77–98)/0 (0–60)|
|Sen/Spec; PPV/NPV||83 (70–96)/0 (0–84)||92 (79–98)/0 (0–70)||89 (74–97)/0 (0–71)|
|6TGN = 230 (95% CI)||71 (55–88)/57 (18–90)||69 (53–85)/0 (0–70)||57 (39–75)/25 (0.6–80)|
|Sen/Spec; PPV/NPV||89 (72–98)/28 (8–58)||90 (73–98)/0 (0–26)||87 (66–97)/6 (0.1–30)|
|6TGN = 260 (95% CI)||69 (52–85)/57 (18–90)||61 (45–78)/0 (0–70)||51 (33–69)/25 (0.6–80)|
|Sen/Spec; PPV/NPV||89 (71–98)/27 (8–55)||89 (71–98)/0 (0–22)||86 (64–97)/6 (0.1–27)|
|Mean 6MMP in R/NR||4275/3431 (P = 0.6)||2742/3736 (P = 0.7)||3418/3840 (P = 0.8)|
|Mean 6TGN/6MMP in R/NR (P)||0.3/0.4 (P = 0.7)||0.1/0.6 (P = 0.5)||0.3/0.6 (P = 0.8)|
|Mean 6TGN/TPMT in R/NR (P)||14/19 (P = 0.2)||18/19 (P = 0.9)||28/14 (P = 0.2)|
Finally, serum metabolite levels of those patients that were under combo therapy with thiopurines and infliximab (23 patients, 17% of the patients included) or with thiopurines and aminosalicylates (46 patients, 41% of the patients analysed) were examined. No statistical differences were found in serum 6TGN or 6MMP along the follow-up, between responders and nonresponders to infliximab. Moreover, no statistical differences in serum 6TGN or 6MMP along the follow-up between patients who also were or were not under infliximab therapy or under aminosalicylates were discovered.
Two patients were switched from AZA to MP due to gastro-intestinal intolerance, but were able to complete follow-up. In the safety analysis, only myelotoxicity, hepatotoxicity and events forcing treatment discontinuation were considered: Eight subjects (7%) were reported, including two subjects of myelotoxicity (2%), four subjects of acute pancreatitis (3.5%), one subject with cutaneous rash and one subject with gastro-intestinal symptoms.
There were statistical differences in TPMT activity neither between those patients who did suffer (21.2 ± 4.7 U/mL) and those who did not suffer (20.3 ± 2.9 U/mL) toxicity nor, specifically, between patients who did suffer (21.2 ± 4.7 U/mL) and those who did not suffer (18.7 ± 2.4 U/mL) myelotoxicity.
Mean metabolite levels (6TGN, 6MMP and also 6TGN/6MMP or 6TGN/TPMT) were not statistically different among those patients who did or did not suffer any toxicity, including myelotoxicity, at any time during follow-up.
No subject of hepatotoxicity was reported; however, greater than 5700 pmol/8 × 108 RBC 6MMP values were found, which is the commonly proposed threshold value for hepatotoxicity,6 in 119 determinations during follow-up, affecting up to 32% of the patients included.
- Top of page
Thiopurine drugs, AZA and MP, are the classic elective therapy in steroid dependency and the cornerstone of maintenance treatment in IBD, but as many as 1/3 of patients must discontinue treatment because of adverse events or lack of response.14, 15, 29–31 These two situations seem to be related to the two known factors of thiopurine metabolism that play an important role in this process: TPMT activity and the amount of active metabolites. The great interest that has been focused on the metabolism of thiopurines in the current era of pharmacogenetics, is related to a supposed capability to predict, to monitor, to individualise and to optimise treatment with thiopurines in patients with IBD, as the genetic variation of the TPMT activity and its influence on how an individual metabolises thiopurines is an important pharmacogenetic model that seems to be a matter of therapeutic intervention32–35
There have been many attempts in the literature trying to assign to the RBC levels of these active metabolites (6TGN, 6MMP), some kind of monitoring properties regarding both efficacy and safety.16, 17, 36 Moreover, threshold values have been proposed in current literature: 230–260 pmol/8 × 108 RBC of 6TGN for clinical efficacy16, 17 and 5700 pmol/8 × 108 RBC of 6MMP for hepatotoxicity,6 and more recently 11 400 pmol/8 × 108 RBC of 6MMP for myelotoxicity.36 Despite the shortage of prospective data addressing the topic, two recent meta-analyses of small, heterogeneous, retrospective and cross-sectional studies in most subjects, agree that there is an association between 6TGN concentrations and the achievement of clinical remission, although the low sensitivity and specificity to identify responding patients based on 6TGN metabolites, makes it hard to assume serum determination of these active metabolites as a useful tool for clinical practice at the present time.16, 17
This large study attempts to correlate serum determination of active metabolites with clinical response. Full dose of AZA (2.5 mg/kg) or MP (1.5 mg/kg) was employed, according to the results of a previous meta-analysis,31 to avoid the effect that a low or variable dose of the drug might have on the accurate evaluation of its efficacy and side-effects.13 Furthermore, the prospective evaluation lasted for 6 months after steroid withdrawal, which means almost a year of follow-up in many of the subjects. Because of the well-known slow therapeutic onset of thiopurines, which may take more than 4 months,25 a shorter study period may have resulted in an inaccurate estimation of the efficacy of the drug. The number of patients that achieved steroid-free clinical remission lasting at least 6 months represented approximately 2/3, which is consistent with current literature.14, 15, 29–31
First, the study tried to demonstrate whether the serum determination of active metabolites was effective in predicting the outcome of therapy. The clinician could minimise unsafe and unnecessary immunomodulatory therapy with this information. Nevertheless, the results of this study do not support the systematic determination of the active metabolites of AZA at the start of the therapy to distinguish those patients who will achieve sustained steroid-free clinical remission.
This study also attempted to identify any threshold values or therapeutic range of AZA metabolites for clinical efficacy. This could be beneficial to distinguish those patients who need higher doses of the drug from those who do not respond to thiopurines, and also could be used by the clinician as target of dose optimisation. However, no applicable threshold values regarding efficacy or safety were found. The most commonly proposed 6TGN threshold values (230 or 260 pmol/8 × 108 RBC)16, 17 showed poor sensitivity, specificity, positive and negative predictive values to identify responding patients in our study. Another recent prospective study was also unable to confirm the effectiveness of the commonly proposed cut-off values.13 As a matter of fact, a previous prospective, randomised and controlled trial failed to demonstrate that the strategy of adjusting AZA dose according to 6TGN concentrations (maintaining serum 6TGN between 250 and 400 pmol/8 × 108 RBC) was clinically superior to the standard dose of 2.5 mg/kg in patients with CD and normal TPMT activity.37
Finally, some authors have suggested a synergistic effect of aminosalicylates or infliximab on thiopurine metabolite RBC levels38, 39; we found no influence on metabolite RBC levels by any of those drugs. Even though we could not analyse the impact of different doses, our results regarding the impact of aminosalicylates on thiopurine metabolism is consistent with previous studies.8, 40
There are facts that must be taken into account to understand these disappointing results. First, regarding 6TGN levels and their reproducibility, a marked inter-individual and intra-individual variability has been described, up to fivefold variations in the same patient.41 Second, despite the results of a large meta-analysis which pointed out that the optimal dose to enhance the chance of achieving clinical response to AZA was the weighted full dose of 2.5 mg/kg, a poor correlation between the weight-adjusted oral dose and serum 6TGN levels has been documented.5, 42, 43 Finally, the optimal method to measure 6TGN RBC levels still remains an unresolved question13, 44–46; moreover, some authors have suggested that there should be some other metabolites that could be more useful in predicting poor response to thiopurines.47
Some authors have focused their interest in the predictive properties of TPMT activity for response. According to what is already known regarding AZA metabolism, higher values of TPMT activity are related to lower response rates as a consequence of a diversion to methylation of MP instead of bioactivation of TGNs.48 This has been confirmed by previous studies, and even a number of TPMT activity threshold values have been proposed to predict clinical efficacy.5, 11–13, 19, 20, 48–50 Once again, this is an interesting idea and it could theoretically help to identify those patients who would surely be nonresponders early despite dose escalation, and switch them to a different immunomodulator. Despite of all that, TPMT activity has not shown to be effective in predicting the final response to thiopurines in our experience.
Most studies have focused their interest in the prediction of toxicity, especially myelotoxicity, in patients with TPMT activity deficiency. Even though it seems clear that patients with severe deficiency should not receive thiopurines because of the high risk of myelotoxicity,22 it is also well known that TPMT deficiency only explains a portion of all subjects of thiopurine-related myelotoxicity.10 Consequently, opinion is still divided as to whether TPMT activity determination is needed prior to starting AZA therapy.1 Despite that approximately 10% of patients had intermediate TPMT activity, which is consistent with current literature,7, 8 only a few patients (2%) developed relevant myelotoxicity in our experience. On the other hand, some authors have proposed that patients with high TPMT activity are predisposed to suffer hepatotoxicity because they accumulate methylated metabolites.50 No subject of significant hepatotoxicity was documented in this study, even though there were patients with high TPMT activity and a relevant proportion of patients had 6MMP serum values above 5700 pmol/8 × 108 RBC at least once during follow-up. Once again, as we were unable to find any predictive clue in TPMT activity or in the RBC levels of active metabolites to identify those patients suffering adverse events, we found no effective tool to alert the clinician of the development of serious adverse events in our experience. This would not have changed even if the only patient that was excluded because of a low TPMT activity had been included and suffered myelotoxicity. It is noteworthy that a previous study showed that TPMT distribution in Spanish population is similar to others, and therefore our conclusions could be assumed by other populations.51
Finally, this study has important limitations that must be taken into account before considering its results or conclusions; the most relevant are the following. The population studied is somehow heterogeneous, the management of the patients, including the concomitant drugs, the choice of the thiopurine agent or even its dose, is left to the investigators criteria, and the evaluation of response is only clinical, which has an undoubted subjective component. Despite the concern about the relevance of those weaknesses, that can cast some doubts on the conclusions of the study, we believe that it reliably reproduces the real life clinical practice scenario and therefore its results might help to understand thiopurine treatment response in clinical routine.
In conclusion, and according to the results of this study, early determination of TPMT activity or AZA metabolites seem not to be useful to predict response to or safety of thiopurines. In addition, as no useful serum azathioprine metabolites threshold values for efficacy or safety were found at any point during the follow-up, the results of this study suggest that further research is needed before recommending dose adjustments of thiopurines based on the RBC levels of metabolites.