|Home | About | Journals | Submit | Contact Us | Français|
Acute leukemias are complex diseases on multiple levels, and laboratory efforts over the past 3 decades have focused on better understanding of the molecular underpinnings and their stem cell biology. We now have a panoply of technologic advances that allow us to characterize individual leukemias by molecular profiles that relate directly to clinical behavior, to detect minimal residual disease, and to begin to develop “targeted” therapeutic strategies based on molecular considerations. There are a number of challenges surrounding this task: first, how to combine these agents with traditional chemotherapeutics and/or with each other to maximize leukemic cell kill and increase the cure rate; second, how to use these targeted agents in the minimal residual disease with potential curative intent; third, for patients unable to tolerate or unlikely to benefit from aggressive approaches, how to use one or more of these agents to reduce tumor bulk and either permit some restoration of normal marrow function or induce morphologic and functional differentiation of the leukemic clone to overcome the leukemia-associated bone marrow failure; and lastly, how to measure the effects of these agents on the molecular and cellular biologic levels in ways that correlate with and might even predict overall clinical outcome. These challenges are further complicated by the inherent heterogeneity in host biology; disease etiology and biology; and interactions among host, disease, and treatment that ultimately determine individual clinical outcomes. Toward this end, we will discuss selected issues surrounding new clinical trial designs and the development of clinically relevant molecular endpoints that might facilitate the development of new treatment approaches that will improve the outlook for adults with acute leukemias.
Acute leukemias are complex diseases on multiple levels, and laboratory efforts over the past 3 decades have focused on better understanding of the molecular underpinnings and their stem cell biology. The diversity of these malignancies is manifest by a wide variety of morphologic subclasses, highly varied clinical presentations, and significant variation in the responses seen clinically. Unfortunately, acute leukemias have a generally poor overall clinical outcome in adults. Thirty years ago, the achievement of complete remission (CR) was determined by morphologic assessment of the individual marrow and in itself considered to be a victory. Clinically, it was realized very early on that while morphologic CR remained the first goal post of response, it provided little prognostic information and was clearly insufficient to determine a cure. Technologic advancements in marking and measuring the leukemic population have provided a clearer determination of minimal residual disease (MRD)1-3 and offer a first-order assessment to discriminate between patients with drug-sensitive disease who might be cured with traditional cytotoxic chemotherapy, possibly use lower total doses, and those with inherent drug-resistant leukemia who will not be cured.
In addition, technologic advancements allow us to characterize individual leukemias by molecular profiles that not only relate to clinical behavior4-10 but also provide potential targets to direct therapeutic strategies that include blocking upstream11,12 and downstream13-17 signal transduction intermediaries, reversing epigenetic gene silencing,18 and evoking potential immunomodulatory approaches.19,20 Each of these targets provides the potential to circumvent traditional drug resistance, and each might hold promise to improve therapeutic outcomes for our patients. However, the clinical effect of such strategies might not be easily measured using traditional response criteria, and the development of criteria to better assess biologic activity and early efficacy endpoints will be critical to advancing novel agents and clinical trial designs. Moreover, our ability to define surrogate endpoints that reflect clinical response is complicated by the inherent heterogeneity in host biology; disease etiology and biology; and interactions among host, disease, and treatment that ultimately determine individual clinical outcomes. Below and as summarized in Table 1, we will discuss some critical issues surrounding our current ability to measure net clinical drug effect and accurately compare the results of these new approaches with the results being achieved with current treatment or supportive interventions.
The overarching endpoint by which we judge the relative success or failure of new treatment strategies is the overall duration of survival. Yet, the actual quality of the noted survival, often considered a surrogate for therapy-related toxicity, must also be evaluated and integrated into the evaluation of a treatment's success. It would seem intuitive that the effect of a treatment on duration of overall survival (OS) should directly relate to the specific measurements of treatment success such as achievement and duration of CR. Unfortunately, this is not always the case. As an example, CRs of short duration have little to no effect on OS despite achieving the time-honored endpoint of normalization of peripheral blood counts and < 5% marrow blasts by morphology. Such responses are more consistent with primary refractory disease. Moreover, the price to be paid for an intensive regimen is often significant and directly effects the patient's quality of life (QOL), particularly in the case of older adults with acute myeloid leukemia (AML). Alternatively, an exclusive endpoint of CR might not assess the full potential of an experimental agent or regimen to induce hematologic improvement that, in turn, might translate into survival advantage and/or enhancement of QOL.
To begin to define and address these burgeoning challenges to assessment and development of antileukemia agents, the US Food and Drug Administration (FDA) and the American Society of Hematology held a joint workshop on endpoints to establish efficacy of new agents in the treatment of acute leukemia.21 Among the central issues are patient heterogeneity, the ability to detect and measure MRD as a surrogate of drug resistance, the clinical implications of a “less-than-CR” response or a stabilization that affords the patient a chance to move to allogeneic stem cell transplantation (a so-called bridge to transplantation) or another potential clinical intervention (ie, maintenance strategies, immunotherapy), and the individual patient's QOL assessment.
In our current configurations, it takes a long time to determine whether a new agent is a worthwhile addition to the treatment arsenal for acute leukemias and can jump the hurdles of FDA approval. Indeed, the proper selection of a “comparator” or control arm is made especially difficult by the lack of effective “standard of care,” given that the current 5-year survival rates for the “best” patients subgroups is ≤ 40% and drops to ≤ 20% for the remaining patients. The goal of achieving a CR in order to improve OS is further confounded by the toxicity traditionally seen with the “CR-driven” intensive chemotherapy-based induction and consolidation. In addition, our ability to understand if a new agent or approach can actually improve overall clinical outcome without increased toxicity (or possibly with less toxicity) is challenged by the heterogeneity of the biology of acute leukemias and the heterogeneity of the patients with the disease.21-23
In this regard, we need to embrace trial designs that will hasten the drug development process and, at the same time, allow for treatment of more patients at optimal doses and fewer patients at suboptimal or toxic doses.24 To allow for the treatment of more patients at or near a clinically predicted optimal dosing and fewer at suboptimal dosing, Estey and Thall have proposed the concept of adaptive randomization according to “real-time” clinical results to allow for faster and perhaps more accurate assessment of the effect of a particular therapeutic manipulation for definable subsets of patients within a heterogenous disease population.23 This type of Bayesian design allows more patients to be randomized to the therapeutic arm that stands the greatest chance for benefit (or the least chance for harm) and can be applied to both phase I and phase II trials, especially in the setting where multiple new agents or combinations are being compared. Moreover, the confounding effects of the inherent biologic heterogeneity of AML may be minimized by using a “biologically relevant” stratification within the randomized arms according to combinations of critical host and disease features such as patient age, underlying myelodysplasia (MDS) or treatment-related AML, and genetics.
It is unfortunate that “cure” remains an ephemeral concept for most adults with acute leukemias and others with primitive “stem cell” leukemias that are characterized by genetic complexity and inherent drug resistance. These diseases not only are considered incurable outside of the allogeneic transplantation setting but are also clearly less responsive to traditional induction approaches, with less than 50% achieving CR in most series.
This is not to say that individuals with such leukemias may not reap benefits from treatments, and in fact, it is this population that may benefit from diverse therapeutic interventions (Table 2). In this regard, responses that are “less than CR” on both morphologic and molecular grounds may translate into improvements in survival duration and/or QOL but may do so in a nonlinear fashion. The ability to both recognize and define such “less-than-cure” outcomes clinically and molecularly would greatly benefit clinicians and could easily be enlisted to help determine the early efficacy for new agents under development and may help sort out the best clinical situations to use them. Ideally, such data should be available to those tasked with the arduous responsibility of determining FDA approval of such new agents. To date, quantifying “less-than-CR” responses and their effect on OS and quality of survival is a work in progress. The effect of “less-than-CR” responses has been followed most closely using gemtuzumab ozogamicin for relapsed AML in patients over the age of 60 years where CR without full recovery of the platelet count (CRp) appears to result in similar OS compared with a traditional CR.25,26 A similar “equivalency” for CR and incomplete CR, of which CRp is a subclass, has been noted for relapsed/refractory AML patients treated with clofarabine.27 Such is not the case, however, for newly diagnosed patients undergoing initial induction therapy, where survival is longer for those achieving true CR.28,29
On the other hand, CR may not be required for clinical survival benefit when determining the effect of epigenetic modulatory therapies such as DNA methyltransferase (DNMT)-1 inhibitors (5-azacytidine and decitabine)30,31 or histone deacetylase (HDAC) inhibitors,32,33 and the immunomodulatory thalidomide derivative lenalidomide34,35 on disease progression and OS in myeloid malignancies, especially myelodysplastic syndromes (MDS). In these settings, the enhancement of both OS and QOL may relate to an improvement in blood counts that result in decreased transfusion requirements and an attendant decrease in complications such as iron overload and development of anti–red blood cell and/or platelet antibodies. This success might be in part because of the ability of MDS cells' ability to retain differentiation pathways that in turn can lead to muitilineage hematopoiesis, which is typically lost in full-scale AML.
Many of the issues surrounding clinical endpoints apply to molecular endpoints, as welt. The ultimate endpoints of cell differentiation and/or death can be described more specifically by the effect of an agent on the net expression or activity of a single molecule, or one or more integrated molecular pathways (with modulation of downstream intermediaries),7,11,13 or an even more global effect such as modulation of gene or protein expression profiles in response to drug exposure.12 Ultimately, however, for molecular measurements to have clinical significance, they must reflect and/or predict clinical outcome in terms of both efficacy and toxicity. As such, specific molecular targets and the results of their modulation must be relevant to the overall cellular biology, not only for the malignant cell itself but also for the cell in the context of its microenvironment.36-38 This consideration might lead to a measurement that differs from the more conventional approach that we use for cytotoxic agents, namely, a “biologically effective dose,” which may not be the highest dose of a particular agent in terms of dose-limiting toxicities or maximal tolerated dose but rather a dose that modulates the activity of molecule(s) or pathway(s) that are critical to net cellular and clinical responses. This concept might be particularly important in epigenetic therapies, as exemplified in landmark trials of combination therapy with DNMT-1 inhibitors 5-azacytidine or decitabine and HDAC inhibitors phenylbutyrate,39 valproic acid,40-41 or the benzamide HDAC inhibitor entinostat (SNDX-275, formerly MS-275),42 and therapies aimed as inducing differentiation rather than direct apoptosis of the malignant clone.43
The clinical trials of DNMT-1 inhibitors in combination with HDAC inhibitors have been accompanied by elegant correlative studies examining baseline levels and posttreatment changes in the expression of one or more selected “tumor suppressor” genes whose dysregulation might be involved in leukemogenesis and/or perpetuation of the malignant clone, including p15, p73, E-cadherin, DAPK, CEBP-α, and SOCS1.39-42,44 Though changes in the expression of these diverse genes can be documented in response to treatment, the relationship of changes (either qualitative or quantitative) to clinical response has been inconsistent. Indeed, Fandy et al measured changes in the methylation of several tumor suppressor genes in CD34+ bone marrow cells from patients with MDS or poor-risk AML undergoing treatment with 5-azacytidine plus entinostat and was unable to demonstrate differences in day-0, day-15, or day-29 gene expression in responders versus nonresponders.44 The relevance of changes in specific gene expression is further complicated by seemingly contradictory findings regarding the relationship of DNA methylation to overall outcome,45,46 a finding that likely reflects the inherent heterogeneity in molecular pathogenesis and pathophysiology of this complex disease plus an incomplete understanding of the full spectrum of gene expression before and after therapy.
There is no question that all phases of clinical drug development would be well-served by the availability of molecular endpoints, or so-called biomarkers, that reflect disease activity and serve as reliable surrogate markers for drug effects (both efficacy and toxicity) on an individual patient basis.47-49 Ideally, these markers could be used in early-phase trials to guide the selection of optimal doses and schedules of investigational drugs for further studies and ultimately might even guide appropriate patient selection. Delineated in Table 3 are several requirements for using such biomarkers in an optimally informative fashion: (1) defining the full spectrum of molecules and pathways that are being targeted by a specific agent; (2) understanding how the agent in question modulates the net expression and activity of the molecular target; (3) being able to define a dose-response relationship in tumor tissue or a reliable surrogate tissue between “target modulation” and the agent under study; and (4) being able to correlate the presence and magnitude of biomarker modulation with clinical response. Identification of a surrogate tissue with biologic parameters similar to the tumor might be limited to skin or buccal mucosa for epithelial malignancies but is much less of a problem for leukemias, where the target tissue itself is easily obtained in a longitudinal fashion throughout all stages of therapy.
The ability to measure drug efficacy at least in part by defining the molecular consequences of drug exposure in a dose-related fashion, ie, pharmacodynamic endpoints, forms the basis for the new phase 0 trial design that integrates molecular pharmacology with traditional pharmacokinetics and offers a potentially more rapid approach to the clinical testing of novel combinations and movement of targeted agents toward FDA approval.48-50 Indeed, one of the major goals of the phase 0 “first-in-human” study is to establish the validity of one or more molecular endpoints whose behaviors are modulated by the study drug in target tumor and surrogate tissues.48 The phase 0 approach involves limited drug exposure in terms of both dose and time with the intent of obtaining longitudinal tumor biopsies to measure drug effect. To date, this trial design has been applied to the development of veliparib (ABT-888), an inhibitor of the DNA repair enzyme poly(ADP-ribose) polymerase (PARP) which, by itself, is not expected to have significant toxicity unless it is combined with DNA-damaging agents.49,51,52 The PARP inhibitor is well suited to the phase 0 design, where the primary objectives are to define the interaction of the study drug with its putative molecular target in human tissue in vivo and to characterize and validate the assay of that interaction in a clinically reproducible and useful fashion.
Finally, drug development in the acute leukemias is complicated by the baseline morbidity of these diseases that relates directly to leukemia-associated bone marrow failure with an expectedly high risk for overwhelming infection and attendant multiorgan dysfunction. These complications do not accompany solid tumors without bone marrow involvement, and if such complications arise during a clinical trial, it is logical to attribute them to the study drug. Such attribution, however, is not the case for the acute leukemias. Thus, the current definitions of serious adverse events and dose-limiting toxicities (DLTs) that are customary for solid tumors may preclude full dose-escalation of new agents in the acute leukemias. To address this situation, Atallah and colleagues have proposed the establishment of a baseline toxicity rating for patients with acute leukemias that takes into account the inherent organ dysfunctions associated with leukemia and the expected toxicities superimposed by induction therapy with cytosine arabinoside and anthracyclines.53 In the phase I setting, this type of baseline might permit a clearer picture of what toxicities are truly related to the study drug and where DLTs actually occur. In the phase II and III setting, the ability to “subtract” the expected toxicities from those observed during the addition of a new agent to chemotherapy might decrease me obligatory reporting of so-called serious adverse events and therefore decrease some of the tremendous cost associated with regulatory oversight by institutional, governmental, and pharmaceutical agencies.
Our continuing challenge is to define the spectrum of clinical and molecular endpoints by which we can judge net drug efficacy and determine the optimal role of new agents in the therapeutic armamentarium for acute leukemias. The ability to recognize the heterogeneity of AML with respect to its malignant stem cell biology compared with normal hematopoietic stem cells, its cell kinetics, and its aberrant molecular pathophysiology may lead to specific curative strategies for most if not all of the AML subtypes. There are a panoply of new agents designed to target selected components of key signal transduction pathways, for instance, inhibitors of FLT-3, vascular endothelial growth factor, farnesyltransferases, components of the PI3K/Akt/mTOR pathway, or components of pathways aimed at repairing DNA damage such as CHK-1 or PARP.16,17,54-59 Moreover, there are agents that have been developed to target molecules, for instance, the epidermal growth factor receptor (EGFR) involved in “epithelial carcinogenesis” that might exhibit off-target effects in hematopoietic cells even though those cells lack EGFR receptors, as has been detected with erlotinib60 and gefitinib.61 There are a number of crucial issues surrounding the optimal incorporation of these molecularly targeted agents into the therapeutic armamentarium: first, how to combine these agents with traditional chemotherapeutics and/or with each other to maximize leukemic cell kill and increase the cure rate; second, how to use these targeted agents in the MRD with potential curative intent; third, for patients unable to tolerate or unlikely to benefit from aggressive approaches, how to use one or more of these agents to reduce tumor bulk and either permit some restoration of normal marrow function or induce morphologic and functional differentiation of the leukemic clone to overcome the leukemia-associated bone marrow failure; and lastly, how to measure the effects of these agents on the molecular and cellular biologic levels in ways that correlate with and might even predict overall clinical outcome.
An ideal MRD approach would be able to discriminate normal from leukemic stem cells and thereby permit the selective destruction of the latter. This can only arise from current efforts to identify the unique characteristics that characterize each myeloid leukemia subtype stem cell on molecular and biologic levels. It is in this MRD setting that epigenetic modulatory approaches might prevent recurrence by promoting apoptosis or by reversing the silencing of genes involved in differentiation, and that immunomodulatory approaches, including vaccines directed against one or more aberrantly expressed proteins might find their greatest efficacy. Along these lines, the finding that aberrant methylation of selected genes in AML might be heightened at relapse relative to initial diagnosis provides an intriguing rationale for the use of DNMT-1 inhibitors in the MRD setting, aimed at preventing disease recurrence or progression.62
Ultimately, as we come to deepen our understanding of the molecular pathogenesis of AML, particularly on the level of the leukemia-susceptible stem cell, we may be able to prevent the occurrence of AML occurring as a consequence of genomic toxicity, as in treatment-related AML. The lessons learned from treatment of patients in remission63 and in preleukemic states and the molecularly targeted approaches being tested in those settings might be directly applicable to the primary prevention setting, especially if we are able to define individuals at high risk for leukemogenesis. At all stages of disease and treatment, patients with acute leukemias should be considered for clinical trials accompanied by studies of leukemia cell biology in order to permit the fluent translation of molecular discoveries into clinical advances. It is only through such scientifically sound translation that we will be able to move the field forward in a clinically meaningful way.
Grant support was provided from the following National Cancer Institute grants: U01 CA70095, 2P30 CA06973-44.
Dr. Judith E. Karp has received grant or research funding from Genzyme Corp; Sunesis Pharmaceuticals, Inc; Aegera; and Kyowa; has received honoraria from Genzyme Corp; and has served on an advisory committee or review panel for Xanthus Pharmaceuticals, Inc.; Actinium; and Cerus.
Dr. B. Douglas Smith has received grant or research funding from Novartis Pharmaceuticals Corp; Syndax; and Eisai, Inc.; has served on a Speaker's Bureau for Celgene Corp; has served on an advisory committee or review panel for Bristol-Myers Squibb Company.