Twenty five years of UK trials in acute myeloid leukaemia: what have we learned?

There have been national trials in acute leukaemia in the United Kingdom since the 1960. As haematology developed into a clinical specialty direct care move into the ambit of clinical haematology. Participation in trials reflected this evolution as did the expansion to include both Teaching and District General hospitals. More intensive chemotherapy combinations have been facilitated by better supportive care, including support from other medical disciplines, and nursing expertise. With the median age of acute myeloid leukaemia patients being 68 years, the intensive approach is not suitable for the majority of older patients where a different approach is needed. The UK has a creditable record in delivering improved care in recent years through a portfolio of prospective randomised trials. In a sequence from 1988 to 2014 of many of the largest trials ever undertaken in acute myeloid leukaemia (AML), a number of improvements to treatment have developed. Of major parallel importance has been the improvement in supportive care which, particularly in younger patients, has permitted more intensive treatment including transplantation to be assessed. The large recruitment has enabled several questions to be asked in a single trial. This experience is reviewed here exclusively in patients considered to be fit for an intensive chemotherapy approach. An additional portfolio in elderly unfit patients is not included. In the first 10 years children could participate, but the introduction of unlicensed drugs to the protocols was not permissible in younger patients. The sequence of treatments tested in the trials within the programme evolved over the years (Fig 1A, B). This includes a review of 16 959 patients who participated a programme of seven randomised trials which delivered a total of 33 249 randomisations from, at different times, >190 institutions in the UK, Ireland, Denmark, New Zealand and, for children, the Netherlands. Several treatment schedules were assessed (Table 1). An additional 1758 young patients, who were high-risk or in relapse entered six randomisations in sequential trials, 2394 older patients who were not considered fit for intensive treatment entered 12 randomised comparisons. These interventions are not discussed here.These trials of intensive treatments have delivered important improvements in survival in children and younger adults, (Fig 2A, B) but less so in older patients (Fig 2C). The first major trial was AML10 which recruited 1966 patients generally under 55 years of age, including 143 children, from May 1988 to April 1995. By subsequent standards the design was simple in that it was a comparison of two drugs, thioguanine and etoposide, in combination with daunorubicin and twice daily cytarabine (Ara-C). In this and subsequent trials the intention was to give two induction courses which differed in that the Ara-C was given b.i.d. for 10 days in course 1 and for eight days in course 2. The daunorubicin dose was 50 mg/m2 on days 1, 3 and 5. The overall response rate (ORR) was 81% with 70% of remissions achieved with the first induction course. The ORR was age-dependent, being 91% in children to 77% in patients >35 years. This was not a comparison of the addition of a third drug, which in retrospect might have been a more informative option. No difference in any measured endpoint was observed including the monocytic subtype which was postulated at the time (Hann et al., 1997). The age threshold of 55 years was set because a major aim of the trial was to test a total body irradiation-based autologous stem cell transplant, and it was not thought to be safe above this age. The results are discussed below. In November 1994, the AML12 trial opened and represented increasing interest in participation. In total 2934 patients were recruited from 198 centres, including valuable inclusion of centres in New Zealand, Ireland and the Netherlands. Of importance to the way the portfolio was to develop was that the high level of recruitment,which was a feature of the AML12 and subsequent trials, enabled postinduction questions to be posed, but also permitted additional induction randomisations to be introduced. The risk of such an approach is that treatment interactions may occur which confuse interpretation of the data; however, expert statistical supervision can check this. With respect to AML12 the decision was to carry forward the etoposide option from AML10 as the control arm and against it test the substitution of mitoxantrone for daunorubicin (Burnett et al., a,b). Due to the large recruitment it was possible to introduce alternative induction questions, so after 1658 patients were recruited to test cytarabine, daunorubicin and etoposide (ADE) versus mitoxantrone, cytarabine and etoposide (MAE), a new randomisation was introduced, namely the standard Ara-C dose of 100 mg/m2 b.i.d. (S-DAT) versus 200 mg/m2 b.i.d. (H-DAT). Since the result of the MAE versus ADE randomisation was not yet known this Ara-C randomisation was set in the context of the thioguanine-containing DAT combination and recruited 1193 patients. There was preclinical evidence that all-trans retinoic acid (ATRA) could sensitise leukaemic blasts to chemotherapy (Hassan & Rees, 1988; Lishner et al., 1989), so an additional randomisation was also introduced which compared the addition of ATRA (45 mg/m2), or not, daily from day 1 to 60. Both induction options of MAE and ADE resulted in an overall response rates of 86% [78% complete remission (CR); 8% incomplete haematologic recovery (CRi)], and 7% resistant disease. MAE was associated with a reduced risk of relapse (48% vs. 57% at eight years) but this did not improve survival (42% vs. 39%) because it was offset by greater myelosuppression after induction which was associated with some increased deaths in CR. Similarly there were no differences between S-DAT and H-DAT (Burnett et al., 2010a). It is perhaps of interest that the DAT-based treatments had a larger component of CRi in the overall response (14%). The addition of ATRA (n = 1075) was not beneficial in this trial, which was consistent with a similar trial conducted by the MD Anderson Group (Estey et al., 1999) whereas the German SAL group did show benefit, in particular for the patients with an NPM1c mutation (Schlenk et al., 2004), which was not confirmed in the UK study (Burnett et al., b). Although the ATRA was given in a different schedule in the SAL study, including etoposide, there is no obvious explanation for this difference, except that in a subsequent German phase 2 study which incorporated etoposide the benefit in the induction of the NPM1c patients, was longer present. The use of growth factors in the treatment of AML was gaining momentum. The role could be as a supportive mechanism to curtail the length of neutropenia after induction treatment, or to give before chemotherapy to induce the leukaemic cells into cycle, thus 'priming' them for chemotherapy. Altogether 480 patients were randomised to receive, or not, 10 days of granulocyte-colony stimulating factor G-CSF starting on day +8 of induction chemotherapy. There was a significant reduction of four days of neutropenia and two days of hospitalisation, but there was no benefit with respect to CR, relapse or survival (Wheatley et al., 2009a). The experience of AML12 suggested that the trial recruitment could be sufficient to support several randomisations in factorial design. Three induction chemotherapy combinations were compared in AML15 [daunorubicin and cytarabine (DA) versus cytarabine, daunorubicin and etoposide (ADE) versus fludarabine, cytarabine, idarubicin and G-CSF (FLAG-IDA)]. (n = 3106). The intention was to assess the value of the third drug (etoposide) in combination with DA, and to include the first assessment of FLAG-IDA as first-line treatment, which was encouraged by promising experience in relapsed disease (Parker et al., 1997). In addition, the first targeted antibody-directed chemotherapy [gemtuzumab ozogamicin (GO)] was being developed as monotherapy for relapse, but AML15 was the first randomised trial to combine it with chemotherapy. Fortunately a pilot trial was undertaken because of potential toxicity, the conclusions of which were that a combination was safe if a reduced dose of GO (3 mg/m2) or possibly 6 mg/m2 was given: inclusion of thioguanine caused liver toxicity. Administration of GO with each course was not feasible but it could be given with induction and the first consolidation course (course 3) (Kell et al., 2003). This would be the programme's first incorporation of an unlicensed drug. The induction treatment options were therefore ADE versus DA or ADE versus FLAG-IDA. Patients in the DA or FLAG-IDA arms could additionally be randomised to GO (3 mg/m2) on day 1 of course 1. The use of GO in the etoposide arms was delayed until June 2005, because of insufficient experience in the pilot trial. So this complicated induction randomisation tested FLAG-IDA versus ADE (n = 1268) and DA versus ADE (n = 1983), and subsequently the addition of GO (n = 1113). At the start of the trial prolonged thrombocytopenia was observed in the FLAG-IDA arm following the second induction course. This was not due to prior GO, and was resolved by reducing the idarubicin dose in FLAG-IDA from 10 to 8 mg/m2. With respect to the etoposide question (DA vs. ADE) the overall responses were similar (84% vs. 86%) but more patients on ADE achieved remission with one course (70% vs. 63%). The toxicities were similar except for more gastrointestinal toxicity on ADE, but the survivals were not different. FLAG-IDA emerged as the most powerful induction schedule with an ORR of 86% (84% CR; 2% CRi) and 90% of the remissions were achieved after the first course (Burnett et al., b). The 30- and 60-day mortalities were not different from ADE. There was significantly more myelosuppression with FLAG-IDA which emerged after the second course where neutrophil recovery took 12 more days and platelet recovery to 100 × 109/l took 48 days compared with 21 days for ADE. The risk of relapse was however significantly reduced (38% vs. 55%), but excess deaths in remission (17% vs. 11%) cancelled a survival benefit. Of potential interest was the possible effect intensive induction with FLAG-IDA could have on postinduction treatment requirements, which will be discussed below. The GO component of the AML15 trial was the largest assessment (n = 1113) of antibody-directed therapy in any cancer. There was no impact on remission rates either morphologically or, in a limited number of cases, by immune flow measurable residual disease assessment (Burnett et al., 2011). However, as in every subsequent trial of GO in induction, there was a significant reduction in relapse. In the initial report of AML15 in 2010 there was no overall survival benefit at five years. When risk subgroups were examined separately there was a clear survival benefit in favourable, but not in unfavourable cytogenetics (79% vs. 51%) and a trend for benefit in intermediate-risk patients. When cytogenetics, blast % response to course 1, secondary disease and presenting white blood cell (WBC) counts were taken into account in a multiparameter risk score, 74% of all patients had at least a 10% absolute survival benefit (56% vs. 46%) (Burnett et al., 2011). When the survivals were updated to eight years for a meta-analysis, an overall survival benefit emerged (Hills et al., 2014). The AML15 trial did not explore the dose of GO. The pilot trial had a suggestion of potential toxicity at 6 mg/m2 (Kell et al., 2003) in a small number of patients, but the good tolerance of the 3 mg/m2 in AML15 suggested that a 6 mg/m2 dose should be tested. An updated individual patient-based meta-analysis of the five adult trials (Hills et al., 2014), to which AML15 and -16 contributed 67% of the patients, assessing the addition of GO to induction treatment suggested little difference, but possibly less toxicity with the 3 mg dose compared with 6 mg. No trial had directly compared these two doses, so this was the initial question in AML17. In all, 788 patients were randomised to GO (3 vs. 6 mg/m2) of whom two thirds in each arm received ADE chemotherapy and one third received DA. The ORR was similar although there were significantly more CRs (82% vs. 76%) with 3 mg/m2, but less CRi's (7% vs. 10%). Early deaths at 30 and 60 days were higher on 6 mg/m2 but there was no difference in relapse or survival either overall or in any risk subgroup (Burnett et al., b). The UK studies did not require CD33 expression as an entry criterion, and as a predictive biomarker for benefit the CD33 results have been mixed (Walter et al., 2007; Khan et al., 2017); similarly single-nucleotide polymorphism (SNP) genotyping of large numbers of GO-treated patients disagree about it predictive ability (Lamba et al., 2017; Gale et al., 2018). On conclusion of this randomisation, a new induction randomisation was introduced. Studies of the initial dose of, and exposure to, daunorubicin were being undertaken by other major collaborative groups which suggested that 90 mg/m2 should be the standard dose (Fernandez et al., 2009; Löwenberg et al., 2009; Lee et al., 2011). However, the control arm dose was 45 mg/m2 which was less than the widely used 60 mg/m2 dose, although in the UK to date 50 mg/m2 was the chosen dose. In the second half of AML17, 1206 patients were randomised to 90 vs. 60 mg on days 1, 3 and 5 of the first induction course (Burnett et al., 2015a). It should be noted that all patients were also scheduled to receive a second course which included three doses of daunorubicin (50 mg/m2), so the total exposure to daunorubicin could be 420 mg/m2 if in the 90 mg arm or 330 mg/m2 in the 60 mg arm, both of which exceeded the total exposure in the other collaborative group trials which usually only gave daunorubicin in the first course. The outcome was that there was no difference in overall response, relapse risk or survival, overall or in any risk subgroup at two years (Burnett et al., 2015a), although there was a trend for superior survival in the FLT3-mutated subset. On longer follow-up this observation was confirmed to give a significant survival benefit (54% vs. 34%) in this patient subset (; Burnett et al., b). This survival is similar to that achieved when the recently approved inhibitor midostaurin (Stone et al., 2017) is added to chemotherapy. The programme ran parallel trials in older patients, primarily aimed at patients over 55–60 years of age. The reason for this was that most of the early trials focussed on a transplant option in consolidation, which at that time was not thought to be feasible in older patients. This trial recruited 1314 patients from November 1990 and June 1998. The median age was 66 years and 20% had some form of pre-AML disorder to which AML was a secondary disease. Contemporary trials frequently excluded this patient subset. The induction comparison was DAT versus ADE versus MAC (mitoxantrone, Ara-C). The remission rate with DAT was significantly better than with ADE (62% vs. 50%) and MAC (62% vs. 55%), but the survivals at five years were not different and poor at 12%, 8% and 10% respectively (Goldstone et al., 2001). ADE in this population was associated with more induction deaths (26% vs. 16% for DAT and 17% for MAC). On the other hand, resistant disease was similar in all treatments. These observations were independent of age. This experience does not establish the need for a third drug, but that there was no difference between thioguanine and etoposide. Similarly the introduction of mitoxantrone, which was expected to be more effective, and possibly less cardiotoxic than daunorubicin, did not fulfil its promise, although the comparison could have been compromised by the fact that there were more days of exposure to Ara-C in the DAT arm. In 1994 an additional randomisation to G-CSF or placebo was introduced. At this time the introduction of growth factors [G- or granulocyte–macrophage (GM)-CSF] was expected to be safe and beneficial in AML. When given as supportive care, as opposed to a 'priming' strategy, the duration at neutropenia was reduced by five days; there was no impact on remission rate or overall survival (Wheatley et al., 2009a). Anxieties about stimulating leukaemic proliferation did not occur in this or other trials. In this trial three induction options were tested. A major objective was to test modulation of resistance mediated by P-glycoprotein using PSC-833, a derivative of cyclosporin. P-glycoprotein is an obvious target in older AML patients because it is associated with poorer response to treatment and it is more frequent in older patients (Leith et al., 1999). A study carried out by the SWOG group in relapsed patients achieved a better response rate and overall survival when ciclosporin was given with a five-day continuous infusion of daunorubicin (List et al., 2011). In our study (AML-R) in relapsed disease, a similar strategy but with standard bolus dosing of daunorubicin was attempted using ciclosporin, but it failed to improve outcomes partly because ciclosporin was more nephrotoxic and immunosuppressive (Liu Yin et al., 2001). However, a novel ciclosporin analogue, PSC-833, became available that had neither of these disadvantages. In a joint pilot study with the HOVON group, it was concluded that the dose of daunorubicin had to be reduced (to 35 mg/m2) (Sonneveld et al., 2000). This is necessary because the excretion of daunorubicin is delayed with a consequent increase in the area under the curve which contributed to toxicity. A potential pitfall of comparing daunorubicin 50 mg/m2 with daunorubicin 35 mg/m2 combined with PSC-833 if no difference was seen, could be that either PSC-833 did not work, or that the daunorubicin dose was sub-therapeutic. Since the issue had never been tested, an additional randomisation was included to compare a 35 mg/m2 dose with 50 mg/m2. Given the anticipated high recruitment, it was also possible to test whether doubling the daily induction dose of Ara-C to 400 mg/m2 was helpful. This ambitious randomisation recruited 1273 patients (Burnett et al., 2009). Among them, 170 elected to receive daunorubicin 50 mg/m2, randomised to daunorubicin 50 mg/m2 (n = versus daunorubicin 35 mg/m2 (n = and 200 randomised to daunorubicin 35 mg/m2 PSC-833 versus 200 on daunorubicin 35 mg/m2 one two patients were randomised to the Ara-C 200 versus 400 mg/m2 These treatments were given for two In addition, in this trial, and to be discussed an objective was to establish whether a total of four courses and cytarabine Ara-C etoposide were superior to a total of three courses Of interest in the context of the subsequent trials comparing daunorubicin doses of 90 vs. 45 mg/m2, no difference in remission survival or overall survival was in this comparison of 35 mg/m2 vs. 50 mg/m2. The PSC-833 arm had outcomes because of an increase in deaths in the first vs. the higher Ara-C dose in induction of which other treatment was combined and the addition of the course were of no benefit. with several treatments in there was no evidence of treatment A promising novel was in (Burnett et al., et al., so was to compare daunorubicin versus for induction in older patients with AML or high-risk A total of patients age, were randomised to two courses of each induction option (Burnett et al., Patients were also included in additional dose of GO (3 mg/m2) in course 1 first trial in older The postinduction questions of a total of two versus three courses and addition of or not as are discussed below. The was The ORR was CRi with no difference between DA and daunorubicin and There was no difference in or significantly more supportive care was in the DA arm platelet and neutrophil recovery was significantly with There were no differences in of relapse vs. = survival from relapse vs. = vs. = or vs. = It was concluded in of its results in phase 2 trials as the novel given at an dose of mg/m2 for five days with daunorubicin is not superior to standard Ara-C daunorubicin as induction for older patients with The second induction randomisation was to GO (3 mg/m2) and included patients (Burnett et al., The ORR was not 70% vs. with no increase in 30- or 60-day The expected of toxicity was As in AML15 and other GO studies the addition of GO significantly reduced the of relapse from to in a but benefit at three years vs. (Burnett et al., The results were not to any although the of benefit or to the of daunorubicin or There did not to be a benefit for the although there were only 33 patients Of interest was that measurable residual disease by flow was available in a small number of randomised patients (n = but did not show a difference between in remission who had received GO or The major question in the AML10 trial in younger patients was whether the addition of a total body irradiation-based autologous transplant improved The randomisation took after the third course of treatment and patients to a course or a course by the (Burnett et al., who had a could receive a in first remission (Burnett et al., Although the additional reduced the risk of relapse, this was offset by in of these deaths were before the was to the course of chemotherapy. The reduction in relapse however a for more than four This question therefore was tested in the subsequent AML12 trial in addition to the question of whether a transplant was of benefit either as a or course. randomised to transplant who had a would receive an and if not would receive an This the only trial in which transplant is tested by standard randomisation (Burnett et al., So three questions were will a total of five courses of chemotherapy be better than is an transplant better than chemotherapy given as a or and is an better than chemotherapy given as a or The transplant experience is discussed below. The four versus five randomisation was subsequently carried over to the AML15 trial to a total of patients. The conclusion was that the addition of a course did not improve and was possibly to older patients, and so that four courses was the standard of had additional important on the GO pilot trial it was feasible to GO to the first consolidation course and at the chosen dose (3 mg/m2) it would not the of a subsequent So GO was also combined with the first consolidation thus GO assessment as a 2 × 2 design in AML15 where patients were randomised in induction in This aimed to establish if GO was beneficial only in consolidation, only in induction, in both or in The conclusion was that the in course 3 (n = did not improve survival whether or not GO had been given with Ara-C at a dose of 3 is standard of care in In there were sufficient patients to compare the option used in UK trials of cytarabine and etoposide with Ara-C and that of to whether the dose should be 3 or In the comparison of with Ara-C patients were There was no difference in of deaths in or and more supportive However, in patients with cytogenetics (n = there was a significant benefit for vs. When comparing the two dose of Ara-C (n = there was a trend for a higher relapse risk in the mg/m2 arm, as more hospitalisation, days on and blood support was at the 3 dose (Burnett et al., b). The comparison by risk was not with no significant differences by a statistical test of but 3 was in risk and to be in and 3 in although both doses were to in An observation from this trial was that patients who only received the two courses of induction FLAG-IDA as as who received both the two induction ADE and the two courses of Ara-C This was not in patients only two induction courses of ADE. This question of whether two courses of FLAG-IDA is sufficient total therapy is being in the trial. The importance of in an increased risk of relapse by the time of the AML15 trial, et al., and with clinical potential were In of this a pilot trial was undertaken to assess the potential of the which was one of the with may be an in this may be it may be associated with toxicity. It was thought that should be because of the use of The first randomised trial was conducted in relapse combined with mitoxantrone, etoposide and cytarabine or Ara-C et al., which did not show a survival however, patients who had an level of drug as assessed by an in did have a better remission In AML15 and AML17, patients were randomised to or not, given following each course of chemotherapy for to days et al., were permitted if the dose of was 74% had an mutation and had a mutation in the a At five years there was no difference between the arms with respect to relapse vs. vs. or of On subset only in patients who received GO and but not in with GO there was a survival benefit. The level of had been to with response et al., In a subset of patients the was This did not show a with but the relapse rate was significantly reduced in the compared with the patient An associated feature of inhibitor treatment is that with of chemotherapy the level of and et al., 2011). In this study, in some the level did not the The beneficial effect of was not by increased of There was preclinical evidence and some early clinical experience in studies that there was a for target of in AML et al., 2009; et al., et al., In AML17, patients who entered who were not did not have who did not have an mutation and were therefore for the randomisation, were In all, patients were randomised to receive or not following each chemotherapy course. was permitted with the for a for patients years, or of for the patients years, and between these to the was given after each chemotherapy course for to days (Burnett et al., 2018). A similar of patients in each arm received a transplant either in first or subsequent The of relapse at five years was not reduced in the arm compared with control