Thiopurine dose intensity and treatment outcome in childhood lymphoblastic leukaemia: the influence of thiopurine methyltransferase pharmacogenetics

Summary The impact of thiopurine methyltransferase (TPMT) genotype on thiopurine dose intensity, myelosuppression and treatment outcome was investigated in the United Kingdom childhood acute lymphoblastic leukaemia (ALL) trial ALL97. TPMT heterozygotes had significantly more frequent cytopenias and therefore required dose adjustments below target levels significantly more often than TPMT wild‐type patients although the average dose range was similar for both genotypes. Event‐free survival (EFS) for patients heterozygous for the more common TPMT*1/*3A variant allele (n = 99, 5‐year EFS 88%) was better than for both wild‐type TPMT*1/*1 (n = 1206, EFS 80%, P = 0·05) and TPMT*1/*3C patients (n = 17, EFS 53%, P = 0·002); outcomes supported by a multivariate Cox regression analysis. Poor compliance without subsequent clinician intervention was associated with a worse EFS (P = 0·02) and such non‐compliance may have contributed to the poorer outcome for TPMT*1/*3C patients. Patients prescribed escalated doses had a worse EFS (P = 0·04), but there was no difference in EFS by dose intensity or duration of cytopenias. In contrast to reports from some USA and Nordic trials, TPMT heterozygosity was not associated with a higher rate of second cancers. In conclusion, TPMT*1/*3A heterozygotes had a better EFS than TPMT wild‐type patients. Thiopurine induced cytopenias were not detrimental to treatment outcome.

Mercaptopurine and thioguanine (2-amino 6-mercaptopurine, tioguanine) are purine analogue drugs that have been used as cytotoxic agents in the treatment of leukaemias for over 50 years (Burchenal et al, 1953;Murphy et al, 1955). Both these thiopurine drugs are metabolized to form thioguanine nucleotide metabolites (TGNs), which have cytotoxic and immunosuppressive properties. Incorporation of TGN into DNA initiates subsequent cytotoxicity (Tidd & Paterson, 1974;Warren et al, 1995;Karran, 2006) and TGN inhibition of intracellular signalling pathways induces apoptotic cell death whilst one TGN, thioguanine triphosphate, suppresses lymphocyte activation by binding to RAC1 in place of endogenous guanine triphosphate (Tiede et al, 2003;Poppe et al, 2006;Marinkovic et al, 2014).
TGN production from the inactive thiopurine drug is regulated by the polymorphic enzyme thiopurine methyl-transferase (TPMT). Thioguanine forms TGNs directly but mercaptopurine forms additional intermediate nucleotide metabolites; one of these, mercaptopurine nucleotide (mercaptopurine riboside 5 0 -monophosphate or 6-thioinosinic acid) is a good substrate for TPMT, and the resulting methyl-mercaptopurine nucleotide metabolites (MeMPNs) are formed at the expense of TGNs. TPMT deficiency (1 in 300 individuals) is associated with bone marrow failure when such patients are treated with standard doses of thiopurine drugs Evans et al, 1991;McBride et al, 2000); a less severe myelosuppression can develop in TPMT heterozygotes ('intermediate' activity, 11% of subjects), (Weinshilboum & Sladek, 1980;Relling et al, 1999a).
For thiopurine drugs the therapeutic window between toxicity and efficacy is narrow. For children with acute research paper First published online 29 November 2014 doi: 10.1111/bjh.13240 lymphoblastic leukaemia (ALL) treated with mercaptopurine, those individuals with lower TPMT activities and/or higher TGN levels have a lower relapse-risk (Lennard & Lilleyman, 1989;Schmeigelow et al, 1995;Relling et al, 1999a). Traditionally, daily oral mercaptopurine is taken during the longterm maintenance phase of chemotherapy in childhood ALL. Within the United Kingdom (UK) ALL protocols this phase of treatment lasts for 2 to 3 years and maintains the disease remission that has been induced by other potent cancer chemotherapeutic agents; the central role of mercaptopurine in maintaining disease remission is illustrated by the worse outcome observed in all attempts to shorten the duration of mercaptopurine-containing maintenance chemotherapy such that the duration of total therapy falls below 2 years (Schrappe et al, 2000;Richards et al, 1996).
In the UK Medical Research Council national ALL trial ALL97, children and young adults were randomized to receive either mercaptopurine or thioguanine in the maintenance phase. The clinical results of the ALL97 thiopurine randomization have been previously reported and showed no difference in efficacy between thioguanine and mercaptopurine. A lower risk of central nervous system (CNS) relapse for thioguanine was offset by an increased risk of death in remission, (mainly due to infections), and an increased risk of thioguanine-induced veno-occlusive disease (VOD) Vora et al, 2006). We have previously reported the TPMT genotype-phenotype discordance and the superiority of TPMT genotyping for the detection of the TPMT heterozygote in this patient cohort, together with the influence of TPMT genotype on thiopurine metabolite accumulation (Lennard et al, 2013). In this paper we report on the influence of the TPMT genetic polymorphism on thiopurine dose intensity and myelosuppression, and the resulting impact on treatment outcome.

Patients and chemotherapy
ALL97 [International Standard Randomized Controlled Trial Number (ISRCTN) registration number ISRCTN26727615] was a randomized comparison of dexamethasone versus prednisone and mercaptopurine versus thioguanine in patients aged 1 to 18 years. The patient cohort has been previously described . Treatment centres obtained local ethics committee approval and informed consent from patients and/or parents before entering children into the trial. The trial had an add-on thiopurine biological study. ALL97 was modified in November 1999 and termed ALL97/99, but the randomizations and biological studies were retained, as was the registered ALL97 trial name. Details of the treatment regimens and modifications have been previously reported (Mitchell et al, 2005(Mitchell et al, , 2009Vora et al, 2006). ALL97 closed to accrual in June 2002. At closure, all of the children randomized to thioguanine who had not finished maintenance chemotherapy were transferred to mercaptopurine.
During maintenance patients received daily oral randomized thiopurine, weekly methotrexate, monthly intravenous vincristine and 5 days of randomized steroid. Maintenance was given for 2 years in ALL97 but was increased to 3 years, for boys only, in ALL97/99. When ALL97 was superseded by ALL97/99, all boys had maintenance increased to 3 years; effectively only boys diagnosed in 1997 received 2 years maintenance. The thiopurine dose was titrated to toxicity from a standard protocol dose (thioguanine 40 mg/m 2 and mercaptopurine 75 mg/m 2 ; 100% protocol dose) for both TPMT heterozygous and wild-type patients. Patients with TPMT deficiency (homozygous for two variant low activity alleles) were titrated from a starting dose of 10% protocol dose (7Á5 mg/m 2 mercaptopurine, 4Á0 mg/m 2 thioguanine). The dose titration protocols for the ALL97 and ALL97/99 phases of the trial have been described elsewhere . Briefly, ALL97 had a more aggressive titration protocol, with dose adjustments every 4 weeks if the neutrophil counts remained above 1Á0 9 10 9 /l and platelet counts above 100 9 10 9 /l, whilst ALL97/99 dose adjustments occurred at the start of each 12-week maintenance cycle to maintain the neutrophil count between 0Á75 and 1Á5 9 10 9 /l and platelet count over 75 9 10 9 /l. For both ALL97 and ALL97/99, the thiopurine dose was reduced if the neutrophil count fell below 0Á75 9 10 9 /l (or platelet count 75 9 10 9 /l) and withdrawn if neutrophil counts fell <0Á5 9 10 9 /l (or platelet counts <50 9 10 9 /l).

Thiopurine dosage calculations
Throughout treatment, drug dosage was recorded weekly and cell counts were recorded at monthly or fortnightly intervals (or more frequently) as appropriate. These hand-written forms were forwarded and collated by the Clinical Trials Service Unit, Oxford. Databases were designed to capture the thiopurine dosage and cell count information. The database started at Week 8, after the induction and consolidation blocks. High-risk patients (ALL97 protocol HR1 and ALL97/ 99 regimen C, which contained additional multi-drug chemotherapy during the first year) were excluded from the thiopurine dosage analysis, as were children who relapsed or died during the first year of chemotherapy.
Patients included in the thiopurine dosage and cell count analysis had achieved remission by the end of induction and had completed forms detailing at least 95% of maintenance treatment received from Week 8. The total number of weeks that thiopurine could have been prescribed during maintenance cycles was calculated. Some children required a period of cell count recovery following the delayed intensive blocks given during year 1, at the start of a maintenance phase of thiopurine treatment. This time was calculated and subtracted from the total number of weeks that thiopurine could have been prescribed as these cytopenias were influenced by other chemotherapy. The daily thiopurine dose (mg/m 2 ) was totalled and the number of weeks that each child was prescribed the standard protocol dose (100% dose = 75 mg/m 2 mercaptopurine or 40 mg/m 2 thioguanine), escalated doses (>100%), reduced doses, or that thiopurine was withdrawn, was calculated. To calculate the average daily thiopurine dose, the totalled daily dose (per m 2 ) throughout maintenance was divided by the time (days) that thiopurine could have been prescribed. To enable comparisons between thiopurines and the use of dosage data from those children who switched from thioguanine to mercaptopurine during treatment, the percentage (%) standard protocol dose for each thiopurine was calculated. The % time when thiopurines were withdrawn, or the dose reduced or escalated, was also calculated, as was the % time with neutropenia (neutrophil count <1Á0 9 10 9 /l and <0Á5 9 10 9 /l) and thrombocytopenia (platelet count <100 9 10 9 /l).

Blood samples
TPMT genotype was determined in a diagnostic lithium heparin blood sample (5 ml), and/or a blood sample taken during remission maintenance chemotherapy. The blood sample protocol has been previously described (Lennard et al, 2013). TPMT activity was also measured in these blood samples; values were previously reported by Lennard et al (2013). Clinicians were informed if the patient had TPMT activity (the thiopurine dose to be adjusted, based on cell counts, from the protocol standard dose) or if the patient was TPMT-deficient (start the thiopurine dose at 10% of the protocol standard dose, adjust on the basis of cell counts). During remission maintenance chemotherapy blood samples were requested at the earliest point in interim maintenance when patients were tolerating thiopurines at the standard protocol, or the maximum tolerated dose; thiopurine metabolites measured in this sample served as a reference value. Within the thiopurine study of the ALL97 trials, additional blood samples were forwarded from clinicians on an ad hoc basis when patients were either unduly sensitive to thiopurines, or tolerating high doses (Lennard et al, 2013). Thiopurine metabolite values were fed back to the clinician, as a check on compliance with oral chemotherapy, prior to dose escalation.

Thiopurine assays
Thiopurine metabolite concentrations were measured by high performance liquid chromatography (HPLC); the lower limit of detection for TGNs and MeMPNs was 6 and 15 pmol/ 8 9 10 8 red cells, respectively (Lennard & Singleton, 1992;Lennard et al, 2013). Blood samples were genotyped for TPMT*3A, TPMT*3B and TPMT*3C, by amplification of exons 7 and 10 of the TPMT gene (TPMT*3A is an exon 7 and 10 double mutant) as previously described (Lennard et al, 2013). TPMT *2 was detected by sequencing exon 5 of the TPMT gene and novel sequence variations were identified by sequencing the TPMT open reading frame from exon 3 to exon 10 as previously described (Otterness et al, 1997;Lennard et al, 2013).

Compliance
Non-compliance with oral chemotherapy was suspected when patients maintained high cell counts despite tolerating long-term thiopurines at ≥100% doses (Lennard et al, 1995). With respect to mercaptopurine, low concentrations of both TGN and MeMPN metabolites (both metabolites < lower quartile concentrations) have been used as an index of partial compliance; these metabolites are products of competing pathways and show an inverse correlation (Lennard et al, 1995). Higher TPMT activity is associated with lower TGN concentrations and higher MeMPNs whilst lower TPMT activity is associated with higher TGNs and lower MeMPNs. Overt mercaptopurine non-compliance was defined as both TGN and MeMPN metabolites at or below the lower limit of detection. For thioguanine, due to the rapid accumulation of thioguanine-derived TGNs in the red cell following an oral dose, non-compliance was suspected if metabolite concentrations were <750 pmol TGNs (Lennard et al, 2013).

Statistical analysis
The Anderson-Darling test was used to examine the fit of observations to a normal distribution. Differences between groups were compared by the Chi-square statistic, or the Mann-Whitney test; quartile analysis of the equality of medians was by the Kruskal-Wallis test, quartile analysis of survival was by the log-rank test for trend. Outcome analysis was of event-free survival (EFS), with an event defined as time to relapse or death, as used in previous ALL97 analysis . Kaplan-Meier curves were calculated and comparisons between groups were performed by the logrank statistic with stratification by age, gender, white blood cell (WBC) count at diagnosis, trial phase (ALL97 or ALL97/99) and steroid received (prednisone, dexamethasone). Overall EFS (from randomization at the start of treatment) was carried out for any analysis involving all patients. For the outcome analysis of dosing and cell count data, patients joined the analysis at the end of year one (i.e., patients with events in year 1 were excluded) and EFS was calculated from this starting point (termed jEFS), stratified by length of maintenance (two or three years). All P values are two-sided, a P value <0Á05 was considered statistically significant. Cox regression multivariate analysis was used to test whether the effects of variables were independent. Analyses were to the annual follow up of 30th April 2011; median follow-up for survivors 11Á3 years (range 9Á6 to 14Á3 years). Statistical analyses were by SAS (version 9.2; SAS, Cary, NC, USA) or Minitab 16.

Patient numbers
The patient numbers and data available are summarized in Fig 1. There was no difference with respect to gender, age, WBC count at diagnosis, steroid randomization or thiopurine randomization between patients with and without thiopurine data.
Thiopurine metabolites and TPMT data have been previously reported for the thioguanine versus mercaptopurine randomized cohort in the ALL97 trial (Lennard et al, 2013). This current paper contains an additional 14 children, who were recruited whilst the trial was still open but the mercaptopurine versus thioguanine randomization was closed, who received non-randomized mercaptopurine.

Thiopurine dosage and cell counts
The average daily dose was 77% of the standard protocol dose and this did not differ between the ALL97 and ALL97/ 99 phases of the trial (Table II). As expected in a protocol in which the drug dosage was adjusted based on cell counts, quartile analysis of average % daily dose showed that those patients with higher doses had significantly higher cell counts; from the lower to the higher quartile (Q) of average % dose (Q1 = 3Á0 to 65Á2%; Q2 = 65Á3 to 76Á7%; Q3 = 76Á8 to 86Á8%; Q4 = 86Á9 to 162Á4%) the median % weeks with a neutrophil count <1Á0 9 10 9 /l was 32, 28, 23 and 15, respectively (P < 0Á001), the median % weeks with a neutrophil count <0Á5 9 10 9 /l was 14Á3, 12Á5, 9Á0 and 5Á9, respectively (P < 0Á001) and the median % weeks with thrombocytopenia was 10Á0, 4Á9, 2Á3, 2Á1, respectively (P < 0Á001). There was no difference in the % time with neutropenia between the ALL97 and ALL97/99 phases, but the incidence of thrombocytopenia was greater in ALL97 (Table II). Throughout both trial phases, thioguanine caused more thrombocytopenia than mercaptopurine but the degree of thrombocytopenia was greater in ALL97 than ALL97/99 (Table III). Although thioguanine-related VOD and the resulting endotheliitis may partly account for the excess thioguanine-related risk of thrombocytopenia, the additional risk in the ALL97 phase is unlikely to be related to VOD as it was more prevalent in the ALL97/99 phase than in ALL97 Vora et al, 2006). The lower platelet counts were more evident in older children (P < 0Á0001, Table II) and was seen in both phases of the trial. There was no difference in the % time with a neutrophil count <0Á5 9 10 9 /l between the trials (Table II). There was no difference in % time with neutropenia between thioguanine or mercaptopurine within ALL97/99 (Table III). Within ALL97, those taking mercaptopurine experienced more neutropenia than those taking thioguanine whilst those taking thioguanine experienced more thrombocytopenia (Table III). This increased myelosuppression was reflected by an increased duration of drug withdrawal and increased time spent on escalated doses in ALL97 (Table II) and is a direct reflection of the differing dose titration protocols between the ALL97 and ALL97/99 phases of the trial; the former was more aggressive, adjusting the dose monthly if cell counts remained above threshold whist the latter aimed for controlled myelosuppression with dose adjustment at the start of a 3-month maintenance cycle to keep cell counts within a therapeutic window.

Clinical outcome
For the complete trial [n = 1948; 5-year EFS = 80%, overall survival = 89% ], there were significant differences by trial part, age group, presenting WBC and steroid, but no difference by gender or randomized thiopurine Mitchell et al, 2009). In the subset of patients used in the dosing analysis (n = 1082) the effects of these covariates on overall EFS were similar to the complete trial.
Thiopurine metabolites. Twenty patients taking mercaptopurine (approximately 3% of the mercaptopurine cohort) had overt drug non-compliance problems (nil metabolites) at some point during maintenance chemotherapy. Partial noncompliance problems (<Q1 both metabolites for the n = 707 mercaptopurine cohort = <276 pmol TGNs and <4698 pmol MeMPNs) were seen in 10% of patients taking mercaptopurine. Twenty-nine children taking thioguanine (6% of thioguanine patients) had TGNs <750 pmol/8 9 10 8 red cells and were considered to have problems with compliance, 12 of these patients (approximately 3% of thioguanine cohort) had TGNs <500 pmol/8 9 10 8 red cells (Lennard et al, 2013). After stratifying for trial, age group, sex, WBC and steroid used, patients accumulating <750 pmol TGNs/ 8 9 10 8 red cells on thioguanine had worse EFS than compliant patients (odds ratio [OR] = 2Á58, 95% CI: 1Á11-5Á7, P = 0Á04); the difference was more marked (OR = 3Á60, 95% CI: 1Á34-9Á69, P = 0Á02) when the definition of non-compliance was restricted to <500 pmol TGNs/8 9 10 8 red cells. However, there was no difference in EFS for non-compliant mercaptopurine patients for any endpoint examined. For all of the patients with trial metabolite data there was no rela-tionship between thioguanine-derived TGNs, mercaptopurine-derived TGNs or mercaptopurine-derived MeMPNs and EFS (which includes all patients from the start of treatment) or jEFS (which excludes all patients with events in year 1), when the thiopurine metabolites were analysed as either continuous variables or split into quartiles. The results for thiopurine metabolites remained unchanged when stratified by TPMT genotype, trial, age group, sex, WBC or steroid used.
Thiopurine dosage. Quartile analysis of the proportion of time spent on escalated doses showed that those patients who spent more time on escalated doses had a worse outcome (jEFS P = 0Á04). There was no heterogeneity of effect of time spent on escalated doses within subgroups defined by genotype, randomized thiopurine or thiopurine taken. Quartile analysis showed no significant trend in jEFS for average % dose thiopurine taken or the proportion of time spent at reduced or no dose. Likewise, there was no significant difference in outcome seen by proportion of time spent with neutropenia or thrombocytopenia. There was no difference in jEFS between those patients who did not experience any episodes of thiopurine-induced cytopenia and the majority of patients that did. There was no evidence of heterogeneity in sub-groups defined by TPMT genotype.
There was no significant difference between the TPMT*1/ *3C and TPMT*1/*3A patients with respect to mean daily dose or incidence of cytopenias, although the number of TPMT*1/*3C patients with full dose intensity data available was small (n = 9), (Table IV). One TPMT*1/*3C patient had difficulty tolerating mercaptopurine (average daily mercaptopurine dose 3%, 10% of time with drug withdrawn), DNA sequencing and measurement of the formation of MeMPN metabolites showed that this boy was not TPMT deficient. However, the TPMT activity (5Á6 units/ml packed red cells) was at the extreme lower end of the TPMT heterozygous distribution (Lennard et al, 2013). Despite experiencing thrombocytopenia for 25% of the time (14% with neutropenia also) and drug withdrawal/reduction due to other mercaptopurine 'sensitivities', this child (treated on ALL97/99 regimen B) remains in remission.

Discussion
This study focused on the complex relationships between TPMT genotype, thiopurine dose intensity and myelosuppression, and the effect of these variables on treatment outcome within the ALL97 trials. Within the MRC UKALL X trial, a study of mercaptopurine dose intensity showed that those children who had one or more episodes of neutropenia (cell counts <0Á5 9 10 9 /l and dose withdrawal) had a better prognosis than those who never became neutropenic (Chessells et al, 1997). Moreover, within UKALL X, those children randomized to receive intensive blocks were prescribed lower doses of mercaptopurine and became neutropenic more readily; the intensive blocks influenced the subsequent response to mercaptopurine maintenance chemotherapy. A reverse association of neutropenia with worse EFS was seen in a later U.S. study in which all children received intensive blocks of treatment. In a multivariate analysis, a higher dose intensity of mercaptopurine was a significant predictor of improved EFS. The worse EFS observed in those with a lower mercaptopurine dose intensity was due to neutropenia and the subsequent weeks of missed therapy due to low cell counts, rather than lower doses of mercaptopurine (Relling et al, 1999b).
In the current study, in which all children received intensive blocks, the thiopurine dose intensity (taken either as the % average daily dose prescribed or the time spent on reduced doses or dose withdrawal due to low cell counts) was not associated with treatment outcome. The duration of, or episodes of, neutropenia or thrombocytopenia were not associated with outcome.
The gender difference in dose tolerance and myelosuppression, initially reported in UKALL X (Chessells et al, 1997), remains within ALL97. However, unlike UKALLX, there is no gender difference in outcome. The latter is perhaps a reflection of the extra year of exposure to thiopurines experienced by the majority of boys. TGN concentrations, metabolite concentrations associated with thiopurine-induced myelosuppression, differ with respect to TPMT status in both boys and girls. For both genders and both thiopurines, TPMT heterozygous patients accumulated significantly higher TGN concentrations compared to the TPMT wild-type cohort (Relling et al, 1999a;Lennard et al, 2013). The TPMT heterozygotes tolerated a significantly lower average daily thiopurine dose than the TPMT wild-type patients and experienced more cytopenias. This is in keeping with recent studies within Berlin-Frankf€ urt-M€ unster (BFM) protocols that have reported an increased sensitivity of patients with TPMT heterozygosity to mercaptopurine-induced myelosuppression (Karas-Kuzelicki et al, 2009;Peregud-Pogorzelski et al, 2011).
There is some debate as to whether the TPMT heterozygote should be prescribed a lower thiopurine dose than the TPMT wild-type patient; not all TPMT heterozygotes are intolerant of thiopurine. The current general advice is a lower starting dose for the TPMT heterozygote with a titration-upwards approach to an acceptable degree of myelosuppression (Relling et al, 2011(Relling et al, , 2013. In the present study, however, the TPMT heterozygotes tolerated significantly lower average % dosages than the TPMT wild-type patients (70% vs 78% for TPMT wild-type, a daily-dose difference of 6 mg/m 2 per day mercaptopurine or 3Á2 mg/m 2 per day thioguanine). However, the range of thiopurine dosages tolerated was wide, with the upper and lower limits similar for both TPMT genotypes. These findings do not support any change in the prescribing criteria (both genotypes start at the same standard protocol dose and titrate to toxicity) for the UK ALL trials with respect to the TPMT heterozygous patient; a similar conclusion to that reached with respect to mercaptopurine dosages in the German BFM protocols (Stanulla et al, 2005).
TPMT *1/*3A patients had a better EFS than TPMT *1/*1 patients, the former also experienced more cytopenias and accumulated higher TGN concentrations than the latter. However, neither the reference TGN concentration nor the frequency of cytopenias throughout thiopurine chemotherapy were directly associated with EFS. Survival was inexplicably worse for patients with TPMT*1/*3C than for TPMT*1/*3A patients. Despite similar mercaptopurine dosages and TPMT activities, the TPMT*1/*3C patients accumulated significantly less TGNs and lower MeMPN concentrations than TPMT*1/*3A patients; this could indicate an increased frequency of non-adherence and suboptimal metabolite exposure in the TPMT*1/*3C cohort. The differences in survival for the TPMT*1/*3C with respect to TPMT*1/*3A patients could explain why, in a previous single nucleotide polymorphism analysis in a smaller cohort of ALL97 patients (Matimba et al, 2014), we found no difference in EFS between TPMT genotypes when analysed as wild-type or variant allele.
Compliance with oral chemotherapy was a confounding factor in this study. Although low or absent metabolite concentrations may be due to other factors, non-compliance with oral chemotherapy is the most likely cause in this group of patients. It is possible that some parents may not have followed the clear guidance on evening dosing and avoiding food and milk products within 1 h of the thiopurine dose, and some patients may have had poor absorption due to other causes. However, in such cases, the apparent drug dose is reduced but the inverse relationship between TGN and MeMPN metabolite formation should be maintained. This is in contrast to the very low levels of both metabolites seen in non-compliance. There are few studies on non-compliance with oral anticancer therapy in children or adolescents, most studies focus on adults (Verbrugghe et al, 2013). Assessment of mercaptopurine non-compliance in previous UK ALL trials, by structured interview, associated low metabolite concentrations at high drug doses with admitted failure to take the tablets (Davies et al, 1993). A similar evaluation of noncompliance in ALL children by clinical (structured interviews and evaluation of medical charts) and laboratory (mercaptopurine metabolite monitoring) indices associated lower non-compliance with adverse socioeconomic factors (De Oliveira et al, 2004). Non-compliance perhaps explains the worse outcome for those who spent a longer time on escalated thiopurine dosages; an effect independent of TPMT status. Non-compliance with thioguanine therapy, as previously defined by low TGN levels (Lennard et al, 2013), was also associated with a worse EFS whilst non-compliance with mercaptopurine was not. This is perhaps explained by action taken by the clinician to address the non-compliance identified by the metabolite results in the case of mercaptopurine but not thioguanine, because thioguanine-derived TGN levels suggestive of non-compliance were derived after the trial closed (Lennard et al, 2013). If unaddressed, a lower compliance to mercaptopurine is known to increase the relapse risk, and ethnic differences in relapse risk have been associated with increased non-adherence (Bhatia et al, 2012).
TPMT heterozygotes taking mercaptopurine were found to have a lower relapse risk in the Nordic Society for Paediatric Haematology and Oncology (NOPHO) ALL-92 trial (Schmiegelow et al, 2009a) but the benefit was offset by a higher incidence of second cancers (Schmiegelow et al, 2009b). With a median follow-up of 11Á3 years, we have not observed an excess of second cancers in TPMT heterozygotes as reported in the NOPHO and St Jude total therapy trials; the latter studies linking lower TPMT activity with second brain tumours in children who received radiotherapy and with etoposide mediated myeloid leukaemias (Relling et al, 1998(Relling et al, , 1999cSchmiegelow et al, 2009b). Neither NOPHO nor ALL97 include etoposide. In the NOPHO and St Jude trials, high-dose methotrexate and cranial irradiation are given alongside oral thiopurines, the former is not included in the ALL97 trials whilst the latter was reserved for patients (<5%) with overt CNS disease at diagnosis . The BFM studies report that TPMT status is not a risk factor for the development of second cancers (Stanulla et al, 2009). In the BFM protocols, some of which contained cranial irradiation, mercaptopurine was given alongside highdose methotrexate but at much lower dosages (25 mg/m 2 per day) than the NOPHO and St Jude trials (75 mg/m 2 ). The higher doses of mercaptopurine in the latter trials, alongside cranial irradiation and/or high-dose methotrexate, could have contributed to the development of second malignancies. A subsequent NOPHO trial (NOPHO ALL-2000) reported that using reduced mercaptopurine dosages for the TPMT heterozygote patient, alongside high-dose methotrexate, reduced the risk of developing second cancers but this was counterbalanced by an increased risk of relapse for the TPMT heterozygote (Levinsen et al, 2014).
The data reported here will allow a more informed use of thiopurine drugs. Within the UK ALL trials, thiopurineinduced cytopenias did not have a detrimental effect on EFS. A reference TGN metabolite concentration, taken at an early stage in thiopurine therapy when the patient is tolerating the drug, is useful for subsequent comparisons of metabolite exposure but is, in itself, not predictive of EFS. Some patients exhibited variable compliance with their oral thiopurine therapy, which was shown to have a negative effect on EFS for thioguanine patients. Other than possible increased non-compliance problems in TPMT *1/*3C patients, a cohort with a higher proportion of ethnic minorities, the worse survival for the TPMT *1/*3C patients is unexplained. The worse outcome for TPMT *1/*3C will be re-examined within the UK ALL 2003 trial, the successor to ALL97, in which treatment intensity was adjusted based on minimal residual disease risk stratification. The EFS for ALL2003 (5-year EFS for low-risk patients >94%)  is far superior to the ALL97 trials and this may negate any impact of TPMT.