Improving consolidation therapy in acute myeloid leukemia - a tough nut to crack

After intensive induction therapy, 60% to 80% of younger (≤60 years) and 40% to 60% of older (>60 years) patients with acute myeloid leukemia (AML) achieve a complete remission.1 However, despite intensive consolidation therapy including intensive chemotherapy, autologous or allogeneic hematopoietic cell transplantation approximately half of younger and 80% to 90% of older patients relapse and the majority of relapsed patients succumb to their disease.2 Based on these figures, it is expected that prevention of relapse by better consolidation therapy would immediately translate into better overall survival. This interrelationship appears to be simple but it is now 24 years ago that this could be demonstrated in a randomized clinical trial that showed a dose-response effect for cytarabine, with improved relapse-free and overall survival among the patients receiving high-dose cytarabine,3 which remains a cornerstone of consolidation chemotherapy in AML.541

In this issue of Haematologica, Burnett and colleagues report on a randomized comparison evaluating the addition of the mammalian target of rapamycin inhibitor, everolimus, to consolidation therapy in AML.6 Everolimus was given for a maximum of 84 days between chemotherapy courses in the experimental arm of the study. Despite the pre-clinical in vitro and in vivo rationale for everolimus, further supported by promising clinical data from phase I/II trials, the independent Data Monitoring Committee (DMC) advised study termination after randomization of 339 patients (2:1 ratio) due to excessive mortality in the everolimus arm. Toxicity of everolimus was primarily gastrointestinal (mucositis and diarrhea) and biochemical evidence of liver toxicity. The primary reason for increased mortality was infection-related deaths within the first 6 months of treatment mainly due to the immunosuppressive effects of everolimus, which reflects what has been seen with the use of this drug in solid tumors.7

This is a remarkable study and we would like to highlight two aspects: (i) the important role of the DMC in taking care of patients’ safety, (ii) the issue of whether we have the right strategies to improve results in consolidation therapy.

Overall, the DMC played a very active role in the study by first recommending dose-reduction for the starting dose from 10 mg to 5 mg after randomization of 146 patients, based on the observation of increased side effects and reduced compliance; the DMC then recommended stopping the trial prematurely after randomization of 339 of the intended 600 patients. These DMC decisions were based, at those time points during the study, on incomplete datasets and were associated with some uncertainty and thus the decisions were not easy to take.8 In hindsight, with more trial data available, these decisions were clearly justified and prevented exposure of additional patients to an increased risk of death. This trial is, therefore, a good example of successful DMC work with useful recommendations at the right time points during a study. This underlines the importance of anticipating risks already in the planning phase of a clinical trial, incorporating the identified risks into the statistical design and structure of a study with predefined interim analyses, selecting appropriate DMC members with engagement at the interim analyses as well as unscheduled analyses if necessary and, finally, of having an experienced and alert study team.

The second interesting aspect of the study which we like to focus on is whether we are using the right strategy to improve consolidation therapy in AML. Based on favorable results with mammalian target of rapamycin inhibitors in preclinical models and clinical phase I/II studies in patients with active disease (newly diagnosed or relapsed AML), everolimus was added in the UK NCRI AML17 trial to consolidation therapy for patients who were in first complete remission. One assumption behind this approach is that effective biological mechanisms of action are similar during induction treatment and during consolidation therapy with residual leukemic cells hidden in bone marrow niches. It became apparent that this was not the case: in contrast to the encouraging results of everolimus in patients with active disease, the use of this inhibitor as an add-on to standard consolidation chemotherapy was associated with increased toxicity and a significant excess of deaths in remission. Even more disappointing was the fact that there was no evidence that everolimus is effective in preventing relapses. Interestingly, this mirrors data from clinical trials evaluating midostaurin and gemtuzumab ozogamicin (GO).

Based on the pivotal large international multicenter randomized double-blinded phase III trial (CALGB 10603, RATIFY, clinicaltrials.gov: NCT00651261) in young adults (18–59 years) with FLT3-mutated AML, the US Food & Drug Administration (FDA) and the European Medicines Agency (EMA) recently approved midostaurin as an adjunct to conventional chemotherapy, including induction and post-remission therapy, without an upper age limit.9 A debate is ongoing how midostaurin affects overall survival and about the role of midostaurin in consolidation and maintenance therapy. A recent exploratory analysis of the trial revealed that midostaurin most effectively prevented relapse in patients who underwent allogeneic hematopoietic cell transplantation in first complete remission. These patients had a trend to better survival (P=0.07) and a significantly lower cumulative incidence of relapse (P=0.02).10 In contrast, patients in first complete remission who received high-dose cytarabine consolidation therapy had a comparable cumulative incidence of relapse rate whether they received midostaurin or not. Thus again, there is no clear evidence that midostaurin given as add-on therapy to high-dose cytarabine prevents relapse. In fact, the addition of midostaurin to first induction therapy seems to have the greatest impact on the observed beneficial effect on event-free and overall survival.9

Based on the results of the ALFA-0701 (NCT00927498) study in newly diagnosed patients,11 the AML-19 study in patients with newly diagnosed AML unsuitable for intensive chemotherapy,12 and the MyloFrance-1 study in relapsed/refractory AML,13 GO was reapproved by the FDA in 2017 for the treatment of newly-diagnosed CD33 AML in adults and treatment of relapsed or refractory CD33 AML in adults and in pediatric patients 2 years and older. In Europe, GO was approved in 2018 for the treatment of patients aged 15 years and older with previously untreated, de novo, CD33 AML. Both approvals (FDA, EMA) for newly-diagnosed CD33 AML in adults included the addition of GO (3 mg/m, day 1) to consolidation therapy with daunorubicin and cytarabine. However, in two trials assessing GO administered on a randomized basis in post-remission therapy, no significant impact on survival was observed.1514 In the MRC AML-15 trial a total of 948 patients were randomly assigned to receive or not receive GO as an adjunct to first consolidation therapy.14 Once again, there was no evidence that relapses were prevented (P=0.20) and the overall survival rates of the patients in the two groups were nearly superimposable (hazard ratio, 1.02; 95% confidence interval: 0.82–1.27). In a study of the HOVON group, older patients achieving complete remission after intensive induction therapy were randomized to three cycles of GO (6 mg/m every 4 weeks) or no postremission therapy.15 The two treatment groups (113 patients receiving GO versus 119 control patients) were comparable with respect to age, performance status, and cytogenetics. There were no significant differences between the groups with regard to overall survival (P=0.52) and disease-free survival (P=0.40).

These examples consistently show that new drugs that are active as an adjunctive therapy to standard induction are not necessarily active in consolidation therapy. But why is this the case and how can we improve the situation? In AML patients with active disease (newly diagnosed, relapsed or refractory) there is usually a bulk population of leukemic cells which can be characterized in depth by sophisticated methods and treated with targeted drugs if the target is present, such as activating FLT3 mutations, CD33 expression, or probably active AKT signaling. In contrast, the clinical situation of consolidation therapy is currently difficult to model in vitro or in vivo. Therefore, the preclinical evidence available before the initiation of a clinical trial is often limited. For example, the senescence-associated reprogramming of non-stem bulk leukemia cells into self-renewing, leukemia-initiating stem cells,16 which may occur during the course of AML treatment, is currently not assessed within clinical trials. It is, therefore, of the utmost importance that methodologies of stem cell research are adapted to be fit for the purpose of use in clinical studies, particularly addressing consolidation research questions.

We also need to improve the sensitivity and specificity of our methods of describing the depth of remission during the consolidation treatment phase. The term molecular remission has been introduced in current guidelines.51 Nevertheless, consolidation therapy still resembles flying blind in that after each cycle complete remission is documented but frequently without taking measurable residual disease (MRD)17 assessment into account. In addition, more than 50% of relapses are not predicted by MRD assessment and occur in MRD-negative groups.4 The consequences of this are low levels of test sensitivity of real-time quantitative polymerase chain reaction-based methods,1918 whereas flow-cytometry and sequencing-based methods have been characterized by low levels of specificity.2220 Thus, MRD assessment during the course of AML treatment is essential and may help to improve clinical research in consolidation therapy of AML. Nevertheless, its assessment is currently only informative for the evaluation of consolidation treatment strategies if MRD is positive and declining or rising. Negative results are still difficult to interpret.4