Thiopurine methyltransferase genotype and activity cannot predict outcomes of azathioprine maintenance therapy for antineutrophil cytoplasmic antibody associated vasculitis: A retrospective cohort study

Objective Azathioprine is a widely used immunosuppressive drug. Genetic polymorphisms and activity of the enzyme thiopurine methyltransferase (TPMT) have been associated with azathioprine efficacy and toxicity in several populations. We investigated whether these associations also exist for ANCA associated vasculitis (AAV) patients, who receive azathioprine maintenance therapy after remission induction with cyclophosphamide. Methods 207 AAV patients treated with cyclophosphamide induction and azathioprine maintenance therapy were included and followed for 60 months. TPMT genotype and tertiles of TPMT activity were compared to relapse free survival and occurrence of adverse events, particularly leukopenia. Multivariable regression was performed to account for confounders. Results In univariable analysis, relapse free survival was not significantly associated with TPMT genotype (P = 0.41) or TPMT activity (P = 0.07), although it tended to be longer in lower tertiles of TPMT activity. There was no significant association of TPMT genotype and activity with occurrence of any adverse event. In multiple regression, leukocyte counts at the end of cyclophosphamide induction were related to risk of leukopenia during azathioprine therapy [P<0.001; OR 0.54 (95% CI 0.43–0.68)] and risk of relapse during follow-up [P = 0.001; HR 1.17 (95% CI 1.07–1.29)] irrespective of TMPT genotype or activity. Conclusion TPMT genotype and activity were not independent predictors of relapse, and could not predict leukopenia or other adverse effects from azathioprine. Leukocyte counts after cyclophosphamide induction were related to both outcomes, implying a greater influence of cyclophosphamide response compared to azathioprine and TPMT in AAV patients.


Results
In univariable analysis, relapse free survival was not significantly associated with TPMT genotype (P = 0.41) or TPMT activity (P = 0.07), although it tended to be longer in lower tertiles of TPMT activity. There was no significant association of TPMT genotype and activity with occurrence of any adverse event. In multiple regression, leukocyte counts at the end of cyclophosphamide induction were related to risk of leukopenia during azathioprine therapy [P<0.001; OR 0.54 (95% CI 0.43-0.68)] and risk of relapse during follow-up [P = 0.001; HR 1.17 (95% CI 1.07-1.29)] irrespective of TMPT genotype or activity. PLOS

Introduction
Anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) refers to a group of primary small-vessel vasculitides. The most common forms are granulomatosis with polyangiitis (GPA, formerly Wegener's Granulomatosis) and microscopic polyangiitis (MPA). [1] Induction treatment with cyclophosphamide, rituximab or mycophenolate mofetil combined with corticosteroids can achieve remission in most AAV patients and reduce mortality, but is associated with considerable toxicity. [2][3][4] For this reason, patients switch to less toxic maintenance therapy after achieving remission, most frequently azathioprine. [3][4][5] Even with azathioprine maintenance therapy, there is a risk of potentially severe adverse effects, most frequently leukopenia and infection. [5,6] In recent years, there has been increasing interest in personalised medicine whereby treatment is adjusted based upon characteristics of an individual patient, thereby optimizing efficacy and reducing toxicity. Azathioprine and its metabolite 6-mercaptopurine are converted via several enzymatic steps into 6-thioguanine nucleotides (6-TGN), the active metabolites responsible for the immunosuppresive effect and myelotoxicity. The enzyme thiopurine methyltransferase (TPMT) methylates several metabolites along the enzymatic pathway, thereby reducing the amount of 6-TGN formed. [7,8] Several polymorphisms of the gene encoding TPMT have been identified, each resulting in decreased activity of the enzyme. [7][8][9] Approximately 89% of Caucasians are homozygous for wildtype TPMT alleles corresponding with normal or high activity.11% carry a wildtype and a variant allele corresponding with intermediate TPMT activity. Very few individuals (0.3%) are homozygous or compound heterozygous for variant alleles resulting in absence of TPMT activity. [8,10] Studies in several populations, mainly inflammatory bowel disease (IBD), have shown that TPMT variant alleles and lower TPMT activity are associated with a higher risk of bone marrow toxicity. [7,[11][12][13] Patients carrying two variant TPMT alleles are especially at risk for severe myelotoxicity [11] and require either a 10-fold lower dose or alternative therapy (e.g. methotrexate). [6,8,14] For patients with intermediate TPMT activity carrying one variant allele, more controversy exists. Several meta-analyses have shown an increased risk of myelotoxicity in these patients. [12,15] While clinical trials did not find a significant reduction of toxicity when adjusting azathioprine or 6-mercaptopurine dose on TPMT genotype or activity, [14,16,17] post-hoc analysis showed a significant reduction of myelotoxicity within carriers of a variant allele. [14] The aforementioned studies mainly involve patients with IBD, who receive azathioprine as their main treatment drug. Since AAV patients receive azathioprine after an induction phase with cyclophosphamide, [4] the influence of TPMT might be smaller and less relevant in this population.
The aim of this study was to see whether TPMT genotype and activity are associated with bone marrow toxicity and risk of relapse in AAV patients treated with azathioprine maintenance TPMT and azathioprine maintenance therapy for ANCA associated vasculitis PLOS ONE | https://doi.org/10.1371/journal.pone.0195524 April 9, 2018 2 / 14 requests: Dr. Elisabeth Brouwer (a researcher who did not collaborate in the study and is not an author on the article) will hold the data and will serve as the point of contact for data access requests. Interested, qualified researchers can contact Dr.
Brouwer at e.brouwer@umcg.nl. On request, Dr. Brouwer will contact the "Loket Contract Research", a legal department within the UMCG concerned with research contracts. They will review whether the data request can be honoured based on the nature of the request, the Informed Consent of the study and current Privacy Laws in the Netherlands. If the request is accepted, Dr. Brouwer will send a Data Transfer Agreement to the requestor ensuring responsible use of the data. After that, data will be transferred to the requestor for the required and agreed upon period of time.

Funding:
The authors received no specific funding for this work.
therapy. We expanded on an earlier study in our population [18] by taking into account the influence of cyclophosphamide induction therapy on these outcomes.

Patients
For this retrospective cohort study, 377 patients, diagnosed with GPA, MPA or Renal Limited Vasculitis (RLV) between September 1984 and August 2013 in the University Medical Center Groningen (UMCG) and treated with oral cyclophosphamide following diagnosis, were considered for inclusion. Patients were included if they switched to azathioprine after induction of remission and had a follow-up of at least a year. All patients have given written informed consent according to the Declaration of Helsinki for participation in a large cohort study investigating biomarkers (including TPMT) in relation to disease outcome in AAV. Ethical approval for the study was granted by the local Medical Ethical Committee of the University Medical Center Groningen (NL29354.042.10).

Treatment protocol
Following diagnosis, all patients were treated with oral cyclophosphamide (1.5-2.0 mg/kg/ day) combined with prednisolone (1mg/kg/day, max 60mg/day). Prednisolone dose was reduced according to a standard schedule (S1 Table). After 3 months of stable remission, all patients switched to maintenance therapy with azathioprine. The starting dose was a conversion from cyclophosphamide dose to the same azathioprine dose. The target azathioprine dose was 1.5-2.0mg/kg/day. Starting 12 months after diagnosis, azathioprine dose was reduced by 25 mg/day every 3 months. Leukocyte counts were measured 1 week after starting azathioprine and at least every 4 weeks thereafter. During treatment,cyclophosphamide and azathioprine dose were adjusted based on leukocyte counts (goal: leukocytes !4.0 Ã 10 9 /l) in accordance with the CYCAZAREM protocol, [5] and occurrence of infections.

Data collection
All information was collected from the patients' records. For all patients, demographic, disease and treatment characteristics, as well as clinical outcome data were registered. Diagnosis was based on the 2012 Chapel Hill Consensus Conference definitions. [1] Disease activity at diagnosis was scored using the Birmingham Vasculitis Activity Score 1 (BVAS-1). [19] Patients were screened for the presence of ANCA using indirect immune fluorescence (IIF), and ANCA-specificity was determined using ELISA. The primary endpoints of the study were relapse-free survival in months and leukopenia. Relapse was defined as new or worsening disease activity requiring dose increase or switch of immunosuppressive medication. Leukopenia was defined as leukocyte count <4.0 Ã 10^9/l. [20] Secondary categorical endpoints were moderate leukopenia (leukocyte count <3.0 Ã 10^9/l), [20] macrocytic anemia (Hb <7.5 for females and <8.0 for males; MCV>96fl), hepatotoxicity (ASAT and/or ALAT >2x upper limit of normal, or AF >125 U/l), infection (requiring hospitalisation and/or antibiotics, or opportunistic e.g. CMV, VZV, HSV, and/or pneumocystis jirovecii pneumonia). These endpoints were scored if they occurred at any time during azathioprine therapy. Secondary continuous endpoints were leukocyte counts 3,6,9 and 12 months after switch to azathioprine, and the [leukocyte( Ã 10 9 /l)] Ã [azathioprine(mg/kg/d)] product 3,6,9 and 12 months after switch as a measure of sensitivity for azathioprine-induced bone marrow depression. [18] Diagnosis, ANCA specificity, age at diagnosis, baseline serum creatinine, co-trimoxazole use at switch to azathioprine (none, prophylactic or therapeutic dose), leukocyte count at switch and duration of azathioprine therapy were registered for their potential influence on relapse. Factors registered for their potential influence on risk of leukopenia include prednisolone dose at switch, cyclophosphamide dose at switch, leukocyte count at switch and azathioprine dose at switch. Prednisolone dose during azathioprine therapy was registered to account for its influence on leukocyte counts.
TPMT activity was determined by adding 6-thioguanine to human erythrocytes in vitro, and measuring the amount of 6-methylthioguanine formed (TPMT catalyses this reaction), expressed in nmol 6-methylthioguanine formed per gram haemoglobin per hour (nmol/gHb/ hr). [21] In the majority of patients (67%), TPMT genotype and activity were measured after starting azathioprine treatment. In some patients (33%), these were measured before starting azathioprine. The date of blood withdrawal for TPMT measurement was registered for all patients.

Statistics
Statistical analysis was done using SPSS Statistics 22 (IBM Corporation, New York, US).
Data are shown as median + interquartile range (IQR) or number + percentage. A twosided P<0.05 was considered statistically significant. Univariate analysis was performed for TPMT genotypes and tertiles of TPMT activity (tertiles determined based on equal numbers of patients per group) using a Log Rank test for relapse free survival (up to 60 months after diagnosis), Fisher's exact test or Chi Square test for risk of adverse events, and Mann-Whitney or Kruskal-Wallis test for leukocyte count and [leukocyte] Ã [azathioprine] product. Multivariate analysis was performed with relapse-free survival, risk of leukopenia and leukocyte counts as outcome variable, using Cox regression, logistic regression and linear regression, respectively. Possible predictors in the analysis were TPMT genotypes, tertiles of TPMT activity and potential influencing factors for the respective outcomes mentioned under 'data collection'. A forward stepwise model was used, where variables were included as covariates based on a univariate P<0.05 and excluded on a multivariate P-value >0.10. Non-proportional hazards for predictors in Cox regression were accounted for by adding a time-by-predictor interaction variable to the model. [22]

Patients and TPMT genotype and activity
207 patients were included in the analysis. (Fig 1) Demographic and disease characteristics, as well as distribution of TPMT genotypes and TPMT activity, are shown in Table 1.

Relapse free survival
Within 5 years after diagnosis, 6 of 19 patients (32%) with heterozygous TPMT genotype experienced a relapse, compared to 84 of 188 patients (45%) with normal TPMT genotype. There was no significant difference in relapse-free survival (P = 0.30) between TPMT genotypes ( Fig  3A).
Tertiles of TPMT activity showed a negative trend with relapse-free survival (P = 0.07) ( Fig  3B). In the lowest tertile (T1), 34% of patients experienced relapse within 5 years, compared to 47% in the middle tertile (T2) and 50% in the highest tertile (T3). Relapse free survival was still not significantly related to TPMT genotype (P = 0.39) and tertiles of TPMT activity (P = 0.21) after exclusion of patients intolerant to azathioprine.

Adverse events
In total, 35 patients (16%) were intolerant to azathioprine. 17 patients (8%) had gastro-intestinal complaints, 17 (8%) had a febrile hypersensitivity reaction, and 1 patient (1%) had a rash. Intolerance to azathioprine was not related to TPMT genotype (P = 0.11) or tertiles of TPMT activity (P = 0.39). There was no significant difference between TPMT genotypes in occurrence of mild or moderate leukopenia ( Table 2). Occurrence of mild or moderate leukopenia also did not significantly differ between tertiles of TPMT activity ( Table 3).
In logistic regression, TPMT genotype and activity were not significantly related to leukopenia during azathioprine therapy. Prednisolone dose and cyclophosphamide dose at switch were also not significant. Leukocyte count at switch (i.e. at the end of induction therapy with cyclophosphamide) remained in the model as a significant predictor of leukopenia (P<0.001), as well as azathioprine dose at switch (P = 0.04) with a higher risk of leukopenia during azathioprine therapy in patients with a lower leukocyte count at the end of cyclophosphamide therapy (Odds Ratio (OR) 0.54; 95% CI 0.43-0.68), and for patients with a higher starting dose of azathioprine (OR 2.2; 95% CI 1.0-4.6). See also S3 Table. Leukocyte counts 3, 6, 9 and 12 months after switch to azathioprine were not significantly different between TPMT genotypes or tertiles of TPMT activity. The [leukocyte] Ã [azathioprine] product was significantly lower for heterozygous patients 3 months after switch (P = 0.03) but not at later time points (Fig 4A). The [leukocyte] Ã [azathioprine] product did not significantly differ between tertiles of TPMT activity (Fig 4B).
In multivariate linear regression, after correction for prednisolone dose, TPMT genotype was still not a significant predictor of leukocyte count or   TPMT genotype was not related to occurrence of macrocytic anemia, liver toxicity, infection, or intolerance to azathioprine ( Table 2). Occurrence of these adverse events was also not significantly different between tertiles of TPMT activity (Table 3).

Discussion
In this study, we found no significant association of TPMT genotype and TPMT activity with relapse free survival. TPMT genotype and activity were not related to occurrence of azathioprine related adverse events. Leukocyte counts at the end of cyclophosphamide induction therapy were significantly associated with both relapse free survival and occurrence of leukopenia during azathioprine maintenance therapy. Because azathioprine therapy is associated with a risk of potentially severe adverse events such as bone marrow toxicity, several studies have focused on TPMT genotypes and activity as predictors of these adverse events. [8,15] It has been established mainly in IBD patient populations that patients with a heterozygous TPMT genotype and lower TPMT activity are at increased risk of developing adverse events, [8] and that pretesting for TPMT has a beneficial effect specifically in the group of patients with one or several TPMT variants. [14] Although we found that a lower TPMT activity was associated with a lower [leukocyte] Ã [azathioprine] product, indicating an increased sensitivity to azathioprine-induced leukopenia, we did not find an association of TPMT genotype and activity with bone marrow toxicity in our population of AAV patients. This might be explained by the fact that AAV patients do not receive azathioprine as the main treatment drug, but as maintenance therapy after induction therapy with cyclophosphamide. [4] The effect of cyclophosphamide on bone marrow toxicity during azathioprine therapy may be greater than the effect of TPMT, as evidenced by the strong association of leukocyte counts after cyclophosphamide treatment with both relapse free survival and leukopenia during azathioprine therapy in this study, and the lack of a significant difference in azathioprine starting dose between patients with and without leukopenia. Another explanation may be that leukocyte counts after cyclophosphamide therapy reflect an overall bone marrow susceptibility to the effects of both drugs.
The frequency of leukopenia in or population was relatively high compared to previous reports, such as in the CYCAZAREM trial (30%). [5] The first reason for this is that patients with leukopenia at the start of azathioprine therapy were also scored as having leukopenia during azathioprine therapy. The second reason is that any leukopenia during the full duration of azathioprine therapy was scored, compared to only the first 15 months after switch in the CYCAZAREM trial. [5] When counting only patients that developed leukopenia within 15 months after start of azathioprine, the frequency of leukopenia (<4.0 Ã 10 9 /l) in our population was 31%, similar to the frequency previously reported. [5] Theoretically, patients with lower TPMT activity can achieve a higher efficacy of azathioprine. [8,23] Some studies indeed found an association of TPMT activity with clinical response. [24,25] In a recently published RCT, adjusting azathioprine dose based on TPMT genotype did not result in a difference in treatment response between intervention and control groups. [14] Although we found that patients with TPMT variant alleles and patients with lower TPMT activity had a higher relapse free survival, these differences were not significant, especially when taking other predictors of relapse into account. Interestingly, higher leukocyte counts after cyclophosphamide therapy were a strong predictor of relapse. This indicates that response to cyclophosphamide may be a stronger predictor of clinical efficacy than TPMT.
This study has several limitations. First, although 207 patients is an impressive number for a single center study on a rare disease such as AAV, the sample size may be insufficient to detect relevant associations with sufficient power. This is especially true for the analyses on TPMT genotype, since there are only 19 patients with a heterozygous TPMT genotype in our study. Second, treating physicians were not blinded to a patient's TPMT genotype and activity. On the other hand, adjustment of azathioprine dose based on TPMT status was not included in the treatment protocol, and the initial azathioprine dose of patients with heterozygous TPMT genotype did not differ between patients whether their TPMT genotype and activity were measured before or after azathioprine therapy. Third, the study was conducted in a tertiary referral center, with some patients receiving part of their follow-up elsewhere. This resulted in missing values on adverse effects of 11 patients. The baseline characteristics and induction treatment of these patients did not significantly differ from those of patients with follow-up during azathioprine therapy. Lastly, the ethnicity for patients was not registered, although over 95% of patients in our study population are estimated to be Caucasian. As we only genotyped for TPMT variants common in Caucasians, some non-Caucasian patients might have reduced TPMT activity resulting from an untyped TPMT variant. This could theoretically result in underestimation of the effects from TPMT genotype.
The study also has several strengths. It was performed in a single center and all patients were treated according to the same protocol, thereby eliminating between-center differences in treatment and followup measurements. Also, compared to an earlier study on TPMT genotype and activity in AAV patients from our population, [18] this study has a larger sample size (207 compared to 108), has a longer duration of follow-up and includes multivariable analyses to account for induction treatment and other factors influencing disease free survival and risk of adverse events.
In conclusion, TPMT genotype and activity were not related to azathioprine efficacy and toxicity in our retrospective cohort of AAV patients receiving azathioprine maintenance therapy. Response to cyclophosphamide, on the other hand, may have a stronger predictive value on these outcomes. This should be confirmed in a sufficiently large multicenter study.
Supporting information S1 Table. Tapering scheme for prednisolone. † Start tapering earlier when in full remission for 2weeks, <6wks of therapy. (DOCX) S2 Table. Cox regression for 5 year relapse free survival. Cox regression analysis for 5 year/ 60 month relapse free survival for non-azathioprine intolerant patients (n = 172). Variables for the final model were selected using a forward stepwise method (inclusion if univariate P<0.05, exclusion if multivariate P>0.1). ANCA specificity, duration of azathioprine therapy, creatinine at baseline and leukocyte count after cyclophosphamide induction therapy were significantly associated with risk of relapse. Ã P<0.05; ÃÃ P<0.01; ÃÃÃ P<0.001. (DOCX) S3 Table. Logistic regression for risk of leukopenia. Logistic regression for risk of leukopenia (leukocyte count <4.0 Ã 109/l) for non-intolerant patients (n = 172). Variables for the final model were selected using a forward stepwise method (inclusion if univariate P<0.05, exclusion if multivariate P>0.1). A higher leukocyte count after cyclophosphamide induction therapy was associated with a lower risk of leukopenia, and a higher starting dose of azathioprine was associated with a higher risk of leukopenia during azathioprine therapy. Ã P<0.05; ÃÃ P<0.01; ÃÃÃ P<0.001. (DOCX)