Author's Accepted Manuscript
Why is progress in AML so slow? Elihu Estey M.D.
PII: DOI: Reference:
www.elsevier.com/locate/enganabound S0037-1963(15)00031-1 http://dx.doi.org/10.1053/j.seminhematol.2015.03.007 YSHEM50818
To appear in: Semin Hematol
Cite this article as: Elihu Estey M.D., Why is progress in AML so slow?, Semin Hematol , http://dx.doi.org/10.1053/j.seminhematol.2015.03.007 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Why is progress in AML so slow?
Elihu Estey, M.D. Division of Hematology University of Washington Medical Center and Fred Hutchinson Cancer Research Institute, Seattle WA
Address correspondence to Elihu Estey, M.D. 825 Eastlake Ave East, Seattle WA 98109; [email protected]
The author declares that he has no conflicts of interest or competing financial or personal relationships that could inappropriately influence the content of this article
Abstract Therapeutic progress in acute myeloid leukemia (AML) is generally acknowledged to have been slower than that in the other commonly occurring types of leukemia. To a very large extent this reflects a relative lack of understanding of AML “biology” and in particular an inability to identify genetic and/or molecular aberrations not found in normal myeloid precursors (“targets”). Here however I also point out that the pace of development/acceptance of new therapies may be retarded by continued adherence to past practices, although these may lack empirical support. Among these practices are reliance on pre-clinical models that do not accurately represent clinical AML, delay in combining targeted therapies with each other or with “chemotherapy”, and limitation of
eligibility for clinical trials to patients with relapsed/refractory AML or unfit patients with newly-diagnosed disease, and the stereotyped use of single-arm phase 2 trials followed by a very large randomized phase 3 trials. Finally I question whether improvement in survival should be the sole or even principal criterion for approval of new drugs in AML.
The ex vivo modification and subsequent therapeutic use of autologous T cells as exemplified by bi-specific antibodies such as blinatumomab(1) or chimeric antigen receptor (CAR) cells (2) has substantively changed management of acute lymphocytic leukemia (ALL). Likewise the introduction of the Bruton tyrosine kinase (BTK) inhibitor ibrutinib has improved the outlook of patients with relapsed or refractory chronic lymphocytic leukemia (CLL) and of older untreated patients with this disease (3). Beginning with imatinib several drugs that inhibit the BCR-ABL encoded tyrosine kinase characteristic of chronic myeloid leukemia (CML) have altered therapy of CML, resulting in likely cure in many patients (4).
Therapeutics in the fourth common leukemia, acute myeloid leukemia (AML), has not been entirely stagnant. Arsenic trioxide (ATO) + all-trans retinoic acid (ATRA) has replaced ATRA + chemotherapy in patients with newly diagnosed acute promyelocytic leukemia presenting with a white cells count < 10,000/μl (5) . Advances in supportive care especially the introduction of newer anti fungal agents have reduced treatmentrelated mortality (TRM) and extended survival for patients even if remission is not attained (6). Nonetheless it is generally accepted that advances in therapy of AML have not been as impressive as those in ALL, CML, and CLL (7), raising the question as to why progress in AML has been so slow. The most obvious explanation is that abnormalities present in AML blasts (or in the small fraction of AML “stem cells” that may determine outcome) but not in their normal counterparts (i.e. “targets”) have been more difficult to define in AML than in other leukemias. Absent such targets bi-specific antibodies or CAR T-cells would exert much less selectivity in AML than in ALL; for
example use CAR T-cells directed against an antigen such as CD123 might require subsequent allogeneic hematopoietic cell transplant (HCT) because CD123 is found both on AML and normal immature progenitors (8).
While progress in therapy of AML will likely not materially accelerate until identification of better targets, there are various problems that may be retarding therapeutic development today and may continue to do so in the future even after more selective targets are found. These problems revolve around what appears a general reluctance to change approaches that have not obviously served us particular well. What follows are examples and proposed alternative approaches.
Problems with pre-clinical models Pre-clinical models of AML have consistently overestimated the likely effectiveness of new drugs when tested in patients. These models are, to a considerable extent, artificial constructs. For example, the leukemia cells from patient-derived specimens that grow out in culture or in an experimental animal outside their ‘normal’ environment may descend from a selected subpopulation of cells and be insufficiently representative of the entirety of the disease. Here it is important to note the genomic complexity of AML and the presence of multiply co-existing molecularly defined clones or subclones(9-13). Xenotransplantation experiments indicate that sub-clones differ in in their ability to engraft and produce AML leading to decreased sub-clone complexity in the AML seen in xenotransplant recipients, with no consistent relation between a sub-clone’s engraftment potential and its likely contribution to clinical relapse(11) Likewise, the clinical
relevance of assays examining drugs tested in cell lines passaged for many years in vitro is highly suspect. These issues suggest caution in assuming a relation between a drug’s clinical potential and its activity in reversing/preventing AML in xenotransplantation experiments and/or in cell lines, recalling that the rationale for clinical testing of many new drugs rests on pre-clinical results in cell line and/or xenotransplantation models.
Possible approaches to improve the relevance of pre-clinical models include the use, as in solid tumors, of organoids derived from primary tumors in an attempt to develop assays of tumor growth that more closely mimic the in vivo behavior of the whole tumor (1415). It might also be valuable to perform drug sensitivity assays in blasts in the presence of supporting stroma, which has been proposed to be protective against the anti AML effects of various drugs (17).
Additionally, most preclinical studies are limited in duration (usually 3-6 weeks maximum), and examine only a small number of dosing approaches and a single efficacy endpoint. Yet because for instance some epigenetic therapies may take weeks to show a therapeutic effect in patients (18), circumscribed preclinical in vitro testing may not permit a realistic prediction of the therapeutic effectiveness of these compounds. Finally, there are many AML genetic subtypes for which there are simply no good in vitro or in vivo models.
Delays in testing combinations
A focus on single-agent trials of “targeted therapies” provides an example of excessive conservatism despite both basic and clinical evidence arguing this approach. Thus during AML evolution it is likely that multiple new critical abnormalities may be acquired that will then act to drive disease progression(10-13). These additional mutations, whose emergence may be hastened by therapy, may be downstream of the initial (founding) mutation(s) or they may bypass these founding mutations by using a parallel cellular pathway. In this fashion the founding clone frequently give rise to a variety of sub-clones potentially resistant to therapy. It seems improbable that a single targeted agent will be effective in this setting. Clinical experience suggests the same. Given the potent kinase inhibition achieved with imatinib and its successors and the key role of activated BCRABL kinase signaling in the biology of CML, these drugs produce durable remissions and even the potential of cure (e.g. cessation of therapy) in patients with chronic phase CML. The operative word here is “chronic”. Because when used alone in the blastic phase of CML, imatinib and congeners are much less effective with cure largely dependent on the use of intensive chemotherapy and allogeneic hematopoietic cell transplantation (HCT)(4,18-22). Clinically, the blast phase of CML is more similar to AML than to chronic phase CML. If such a precisely targeted and successful drug as imatinib does not show durable, potent efficacy as a single agent in blast phase CML, is it credible that other targeted drugs used as single agents will be successful in AML? Likewise despite the impressive single-agent activity neither ATO nor ATRA result in as high a cure rate in APL as combination of ATO+ATRA with each other or with chemotherapy (5). Likewise although targeted therapies such as inhibitors of FLT3, histone deacetylases and proteasomes have almost inevitably been combined with
conventional chemotherapy, a median of 4-5 years have elapsed from initial reports of single- agent trials to initial reports of targeted therapy-chemotherapy combinations.
At least two factors appear to be at work here: concerns for toxicity and for the difficulty in distinguishing effects of targeted agents from each other or from chemotherapy unless a relatively large number of patients are first treated with the single agent. Regarding the first, although a “first do no harm” is a fundamental principle of medicine, AML is not a condition that resolves spontaneously. The vast majority of deaths do not result from TRM even in patients aged 70-79, at least those reported in the literature, but from failure to enter CR despite not incurring TRM or from relapse (23-24). Regarding the latter, patients are likely to be less concerned about the relative contributions of the targeted therapy and chemotherapy to efficacy and toxicity than in pursuing single-agent approaches with limited past records of success. Although rarely used, statistical designs exist that specifically permit dose- finding for multi-agent combinations (e.g. a targeted agent and chemotherapy), allow patient-specific dose finding for example based on covariates such as age that are often more important predictors of toxicity than dose, and permit simultaneous monitoring of both efficacy and toxicity (25-26).
Although combination studies would require increased collaboration among pharmaceutical companies and perhaps a novel attitude on the part of regulatory agencies, the data discussed above will hopefully foster such change. FLT3 inhibitors afford a simple example. Thus D835 point mutations in the FLT3 gene that are associated
with development of resistance to quizartinib can be targeted by crenolanib, providing a rationale for using the drugs in combination (27).
Restriction of eligibility Trials of new therapies are stereotypically largely limited to patients with relapsed/refractory AML or, those newly diagnosed patients considered unfit for conventional chemotherapy. While it is it is plausible that truly effective drugs would work even in very advanced disease (as is the case with ATRA) it seems similarly plausible that drawing conclusions about the value of a targeted therapy based solely on testing in relapsed/refractory patients incurs a significant risk of reaching a falsely negative conclusion. Likewise, restricting testing to newly - diagnosed unfit patients makes it difficult to combine targeted agents with chemotherapy, despite the possible merit of this approach noted above.
To avoid false negatives of trials of targeted therapies should include other patients. One possibility are fit, newly diagnosed patients with high-risk disease based on cytogenetics (complex/monosomal karytotype) or such molecular features as high allelic ratio FLT3 internal tandem duplications (28). Including fit elderly patients with such features might be advantageous because they are less often candidates for HCT and thus there would be less confounding between the effect, particularly on survival, of the targeted agent or targeted agent chemotherapy combination and the effect of HCT. Since such patients might also be inherently resistant leading to falsely negative conclusions, trials might eventually include better prognosis patients. Thus a full spectrum of patients from
relapsed/refractory worst to best prognosis newly diagnosed would be included. Initial trials would begin in (a) relapsed/refractory patients and would then proceed to (b) worse prognosis newly-diagnosed patients and then (c) bestr prognosis newly diagnosed patients. The number of group (b) patients treated would be predicated on results of group (a): the worse the results the fewer patients. Similarly, although group (c) patients would always be included, the number of such patients would depend on results in groups (b) and (a).
Very few trials are undertaken in patients in complete remission. A potential disadvantage of the CR setting is the potential predominance in CR of resistant subclones that either were present at diagnosis or emerged under therapy. However the lower volume of disease in the CR setting might be advantageous in avoiding false negative conclusions. Similarly, observations that reduced intensity HCT, in effect a targeted therapy relying on an immunologically mediated graft-vs. - AML effect is more successful when used in patients in remission than in patients with active disease suggest that targeted therapies might be effective in the former but not the latter setting (29). Hence trials of new therapies might be designed to always include patients in CR, particularly those at high risk of relapse as based on pre-treatment cytogenetic or molecular features and especially post-treatment features such as achievement of CRp or CRi rather than CR and particularly presence of minimal residual disease (MRD) (30) in much the way these trials would always include both relapsed and newly-diagnosed patients. A logistical disadvantage of the CR (or CRp/CRi setting) is the much longer
time needed to observe relapse than to observe CR. This time might be shortened if reduction in MRD was used as an endpoint, as discussed below.
A feature of current trials of targeted therapies is restriction of eligibility to patients who have the presumed target. This certainly has virtue in cases where there is confidence that affecting the target is necessary and sufficient for clinical efficacy. In this scenario allowing entry of a wide range of patients, although response may only occur in specific patient subsets, is a recipe for a false negative conclusion. However in circumstances where the relevance of the presumed target is less obvious limiting eligibility risks another false negative, namely missing subsets where the therapy might be effective. For example, although sorafenib is often considered a “FLT3 ITD inhibitor”, the drug is actually a multikinase inhibitor (31) and a trial randomizing patients aged 18-60 with newly diagnosed AML to 3+7+/- sorafenib found that an improvement in event free survival (EFS) in the sorafenib group largely resulted from patients who were FLT3 ITD negative (32). And although it is a more selective FLT3-ITD inhibitor with less “offtarget” effects, quizartinib also has been reported to have activity in FLT3-ITD negative AML, suggesting it affects other clinically relevant targets (33). Another example is an ongoing North American trial evaluating dasatinib as a KIT inhibitor in core-binding factor AML; the trial allows entry of patients with and without a KIT mutation, hypothesizing that overexpressed wild type KIT may also be a valid target for dasatinib hypothesizing that overexpressed wild type KIT may also be a valid target for dasatinib in some patients (34).
In general value might be found in an approach analogous to that described for relapsed/refractory and newly diagnosed patients. Specifically initial studies would be conducted in patients possessing the target and subsequently expanded to other patients.
Problems with conventional clinical trial designs The typical phase 3 trial in AML enrolls about 400 patients into a standard and an investigational arm. This often requires several years to accomplish, not counting the time needed for follow-up of the last patients enrolled. During these years, other therapies are not investigated, which may be problematic if there are several new therapies of interest. Phase 3 trials enroll so many patients because they aim to detect relatively small differences between standard and investigational regimens with 80-90% power (corresponding to a false-negative rate of 10-20%) and P = 0.05 (corresponding to a false-positive rate of 5%). The first question is whether the differences we aim to detect are truly meaningful clinically. For example, an otherwise healthy 68-year-old with newly diagnosed AML would live an additional 15 years (180 months) if he or she did not have AML randomized trial (35). Standard treatment results in a median survival of 12 months, and thus the patient loses 168/180 = 93% of anticipated life expectancy. A new treatment (azacitidine is an example) prolongs survival to 18 months, thus resulting in a loss of 162/18O = 90% of life expectancy. Many patients might consider this improvement, which is quite similar to those often targeted, medically insignificant. A second question is whether a false-negative rate of 20% and false- positive rate of 5% are acceptable for AML. In diseases where good treatment exists, the consequences of a false positive are much greater than in a disease such as AML where standard treatment is
routinely unsuccessful. Hence, revision of phase Ill trials for AML to aim for larger differences, with P = 0.10, would better fit clinical reality and would permit investigation of a larger number of treatments.
Phase 3 trials are conventionally preceded by single-arm phase 2 trials. It seems paradoxical that randomization is a fundamental part of phase III trials but not of the phase II trial that determines whether the phase III trial will be undertaken. This has led several authors to propose the use of randomized phase 2 trials whose intent is to select the best therapy to take into a larger trial (36-37). These selection design trials are often criticized as "underpowered." And, indeed, consequent to their small sample sizes relative to phase3 trials, their power is only 60% to detect a case where three drugs have a true CR rate of 50% and a fourth has a true CR rate of 65%. This 60% power is can be contrasted with the 80% common to many large phase III trials. However, this 80% is only nominal. Consider a case where there are four candidate new therapies each of which might be compared with standard therapy in phase 3. However as often the case pre-clinical rationale is insufficient to know which of the four is best. Accordingly there is only a 25% probability that the best of the four will be selected. Thus the 60% power of a selection design competes with a power of 25% X 80% = 20%. Simply put the worse false negative may occur when a therapy is never investigated at all. Drug development might be improved with use of randomized selection designs, for example the play-the-winner design employed by the Medical Research Council/ National Cancer Research trials in the United Kingdom (37).
Criteria for drug approval Currently the U.S. Food and Drug Administration’s (FDA) principal criterion for approval of drugs in AML is extension of survival (OS) or improvement in “quality of life” (QOL) as demonstrated in a randomized trial. “Accelerated approval” can be given based on outcomes that are likely surrogates for OS or QOL, but a subsequent randomized trial demonstrating improved OS or QOL has generally been required.
One problem with this approach is that it ignores the heterogeneous nature of AML and the resultant possibility that a new treatment might be effective in some subsets but not others. For example, after receiving accelerated approval gemtuzumab ozogamicin (GO) was withdrawn from the U.S. market by its manufacturer because a South West Oncology Group (SWOG) trial randomizing patients to standard 7+3 therapy +/- GO did not show an OS benefit (38). However other similar randomized trials did show an OS benefit in patients with “favorable” and “intermediate” prognosis cytogenetics, suggesting that it may be unrealistic to expect that any therapy will produce a general prolongation in OS (39-40). Of course if randomized trials are done in specific but relatively small subsets, many years may be required to complete the trial unless we are prepared to either look for larger differences or accept less power or larger p-values; the latter would be a welcome development in my opinion since p
Why is progress in AML so slow? Elihu Estey M.D.
PII: DOI: Reference:
www.elsevier.com/locate/enganabound S0037-1963(15)00031-1 http://dx.doi.org/10.1053/j.seminhematol.2015.03.007 YSHEM50818
To appear in: Semin Hematol
Cite this article as: Elihu Estey M.D., Why is progress in AML so slow?, Semin Hematol , http://dx.doi.org/10.1053/j.seminhematol.2015.03.007 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Why is progress in AML so slow?
Elihu Estey, M.D. Division of Hematology University of Washington Medical Center and Fred Hutchinson Cancer Research Institute, Seattle WA
Address correspondence to Elihu Estey, M.D. 825 Eastlake Ave East, Seattle WA 98109; [email protected]
The author declares that he has no conflicts of interest or competing financial or personal relationships that could inappropriately influence the content of this article
Abstract Therapeutic progress in acute myeloid leukemia (AML) is generally acknowledged to have been slower than that in the other commonly occurring types of leukemia. To a very large extent this reflects a relative lack of understanding of AML “biology” and in particular an inability to identify genetic and/or molecular aberrations not found in normal myeloid precursors (“targets”). Here however I also point out that the pace of development/acceptance of new therapies may be retarded by continued adherence to past practices, although these may lack empirical support. Among these practices are reliance on pre-clinical models that do not accurately represent clinical AML, delay in combining targeted therapies with each other or with “chemotherapy”, and limitation of
eligibility for clinical trials to patients with relapsed/refractory AML or unfit patients with newly-diagnosed disease, and the stereotyped use of single-arm phase 2 trials followed by a very large randomized phase 3 trials. Finally I question whether improvement in survival should be the sole or even principal criterion for approval of new drugs in AML.
The ex vivo modification and subsequent therapeutic use of autologous T cells as exemplified by bi-specific antibodies such as blinatumomab(1) or chimeric antigen receptor (CAR) cells (2) has substantively changed management of acute lymphocytic leukemia (ALL). Likewise the introduction of the Bruton tyrosine kinase (BTK) inhibitor ibrutinib has improved the outlook of patients with relapsed or refractory chronic lymphocytic leukemia (CLL) and of older untreated patients with this disease (3). Beginning with imatinib several drugs that inhibit the BCR-ABL encoded tyrosine kinase characteristic of chronic myeloid leukemia (CML) have altered therapy of CML, resulting in likely cure in many patients (4).
Therapeutics in the fourth common leukemia, acute myeloid leukemia (AML), has not been entirely stagnant. Arsenic trioxide (ATO) + all-trans retinoic acid (ATRA) has replaced ATRA + chemotherapy in patients with newly diagnosed acute promyelocytic leukemia presenting with a white cells count < 10,000/μl (5) . Advances in supportive care especially the introduction of newer anti fungal agents have reduced treatmentrelated mortality (TRM) and extended survival for patients even if remission is not attained (6). Nonetheless it is generally accepted that advances in therapy of AML have not been as impressive as those in ALL, CML, and CLL (7), raising the question as to why progress in AML has been so slow. The most obvious explanation is that abnormalities present in AML blasts (or in the small fraction of AML “stem cells” that may determine outcome) but not in their normal counterparts (i.e. “targets”) have been more difficult to define in AML than in other leukemias. Absent such targets bi-specific antibodies or CAR T-cells would exert much less selectivity in AML than in ALL; for
example use CAR T-cells directed against an antigen such as CD123 might require subsequent allogeneic hematopoietic cell transplant (HCT) because CD123 is found both on AML and normal immature progenitors (8).
While progress in therapy of AML will likely not materially accelerate until identification of better targets, there are various problems that may be retarding therapeutic development today and may continue to do so in the future even after more selective targets are found. These problems revolve around what appears a general reluctance to change approaches that have not obviously served us particular well. What follows are examples and proposed alternative approaches.
Problems with pre-clinical models Pre-clinical models of AML have consistently overestimated the likely effectiveness of new drugs when tested in patients. These models are, to a considerable extent, artificial constructs. For example, the leukemia cells from patient-derived specimens that grow out in culture or in an experimental animal outside their ‘normal’ environment may descend from a selected subpopulation of cells and be insufficiently representative of the entirety of the disease. Here it is important to note the genomic complexity of AML and the presence of multiply co-existing molecularly defined clones or subclones(9-13). Xenotransplantation experiments indicate that sub-clones differ in in their ability to engraft and produce AML leading to decreased sub-clone complexity in the AML seen in xenotransplant recipients, with no consistent relation between a sub-clone’s engraftment potential and its likely contribution to clinical relapse(11) Likewise, the clinical
relevance of assays examining drugs tested in cell lines passaged for many years in vitro is highly suspect. These issues suggest caution in assuming a relation between a drug’s clinical potential and its activity in reversing/preventing AML in xenotransplantation experiments and/or in cell lines, recalling that the rationale for clinical testing of many new drugs rests on pre-clinical results in cell line and/or xenotransplantation models.
Possible approaches to improve the relevance of pre-clinical models include the use, as in solid tumors, of organoids derived from primary tumors in an attempt to develop assays of tumor growth that more closely mimic the in vivo behavior of the whole tumor (1415). It might also be valuable to perform drug sensitivity assays in blasts in the presence of supporting stroma, which has been proposed to be protective against the anti AML effects of various drugs (17).
Additionally, most preclinical studies are limited in duration (usually 3-6 weeks maximum), and examine only a small number of dosing approaches and a single efficacy endpoint. Yet because for instance some epigenetic therapies may take weeks to show a therapeutic effect in patients (18), circumscribed preclinical in vitro testing may not permit a realistic prediction of the therapeutic effectiveness of these compounds. Finally, there are many AML genetic subtypes for which there are simply no good in vitro or in vivo models.
Delays in testing combinations
A focus on single-agent trials of “targeted therapies” provides an example of excessive conservatism despite both basic and clinical evidence arguing this approach. Thus during AML evolution it is likely that multiple new critical abnormalities may be acquired that will then act to drive disease progression(10-13). These additional mutations, whose emergence may be hastened by therapy, may be downstream of the initial (founding) mutation(s) or they may bypass these founding mutations by using a parallel cellular pathway. In this fashion the founding clone frequently give rise to a variety of sub-clones potentially resistant to therapy. It seems improbable that a single targeted agent will be effective in this setting. Clinical experience suggests the same. Given the potent kinase inhibition achieved with imatinib and its successors and the key role of activated BCRABL kinase signaling in the biology of CML, these drugs produce durable remissions and even the potential of cure (e.g. cessation of therapy) in patients with chronic phase CML. The operative word here is “chronic”. Because when used alone in the blastic phase of CML, imatinib and congeners are much less effective with cure largely dependent on the use of intensive chemotherapy and allogeneic hematopoietic cell transplantation (HCT)(4,18-22). Clinically, the blast phase of CML is more similar to AML than to chronic phase CML. If such a precisely targeted and successful drug as imatinib does not show durable, potent efficacy as a single agent in blast phase CML, is it credible that other targeted drugs used as single agents will be successful in AML? Likewise despite the impressive single-agent activity neither ATO nor ATRA result in as high a cure rate in APL as combination of ATO+ATRA with each other or with chemotherapy (5). Likewise although targeted therapies such as inhibitors of FLT3, histone deacetylases and proteasomes have almost inevitably been combined with
conventional chemotherapy, a median of 4-5 years have elapsed from initial reports of single- agent trials to initial reports of targeted therapy-chemotherapy combinations.
At least two factors appear to be at work here: concerns for toxicity and for the difficulty in distinguishing effects of targeted agents from each other or from chemotherapy unless a relatively large number of patients are first treated with the single agent. Regarding the first, although a “first do no harm” is a fundamental principle of medicine, AML is not a condition that resolves spontaneously. The vast majority of deaths do not result from TRM even in patients aged 70-79, at least those reported in the literature, but from failure to enter CR despite not incurring TRM or from relapse (23-24). Regarding the latter, patients are likely to be less concerned about the relative contributions of the targeted therapy and chemotherapy to efficacy and toxicity than in pursuing single-agent approaches with limited past records of success. Although rarely used, statistical designs exist that specifically permit dose- finding for multi-agent combinations (e.g. a targeted agent and chemotherapy), allow patient-specific dose finding for example based on covariates such as age that are often more important predictors of toxicity than dose, and permit simultaneous monitoring of both efficacy and toxicity (25-26).
Although combination studies would require increased collaboration among pharmaceutical companies and perhaps a novel attitude on the part of regulatory agencies, the data discussed above will hopefully foster such change. FLT3 inhibitors afford a simple example. Thus D835 point mutations in the FLT3 gene that are associated
with development of resistance to quizartinib can be targeted by crenolanib, providing a rationale for using the drugs in combination (27).
Restriction of eligibility Trials of new therapies are stereotypically largely limited to patients with relapsed/refractory AML or, those newly diagnosed patients considered unfit for conventional chemotherapy. While it is it is plausible that truly effective drugs would work even in very advanced disease (as is the case with ATRA) it seems similarly plausible that drawing conclusions about the value of a targeted therapy based solely on testing in relapsed/refractory patients incurs a significant risk of reaching a falsely negative conclusion. Likewise, restricting testing to newly - diagnosed unfit patients makes it difficult to combine targeted agents with chemotherapy, despite the possible merit of this approach noted above.
To avoid false negatives of trials of targeted therapies should include other patients. One possibility are fit, newly diagnosed patients with high-risk disease based on cytogenetics (complex/monosomal karytotype) or such molecular features as high allelic ratio FLT3 internal tandem duplications (28). Including fit elderly patients with such features might be advantageous because they are less often candidates for HCT and thus there would be less confounding between the effect, particularly on survival, of the targeted agent or targeted agent chemotherapy combination and the effect of HCT. Since such patients might also be inherently resistant leading to falsely negative conclusions, trials might eventually include better prognosis patients. Thus a full spectrum of patients from
relapsed/refractory worst to best prognosis newly diagnosed would be included. Initial trials would begin in (a) relapsed/refractory patients and would then proceed to (b) worse prognosis newly-diagnosed patients and then (c) bestr prognosis newly diagnosed patients. The number of group (b) patients treated would be predicated on results of group (a): the worse the results the fewer patients. Similarly, although group (c) patients would always be included, the number of such patients would depend on results in groups (b) and (a).
Very few trials are undertaken in patients in complete remission. A potential disadvantage of the CR setting is the potential predominance in CR of resistant subclones that either were present at diagnosis or emerged under therapy. However the lower volume of disease in the CR setting might be advantageous in avoiding false negative conclusions. Similarly, observations that reduced intensity HCT, in effect a targeted therapy relying on an immunologically mediated graft-vs. - AML effect is more successful when used in patients in remission than in patients with active disease suggest that targeted therapies might be effective in the former but not the latter setting (29). Hence trials of new therapies might be designed to always include patients in CR, particularly those at high risk of relapse as based on pre-treatment cytogenetic or molecular features and especially post-treatment features such as achievement of CRp or CRi rather than CR and particularly presence of minimal residual disease (MRD) (30) in much the way these trials would always include both relapsed and newly-diagnosed patients. A logistical disadvantage of the CR (or CRp/CRi setting) is the much longer
time needed to observe relapse than to observe CR. This time might be shortened if reduction in MRD was used as an endpoint, as discussed below.
A feature of current trials of targeted therapies is restriction of eligibility to patients who have the presumed target. This certainly has virtue in cases where there is confidence that affecting the target is necessary and sufficient for clinical efficacy. In this scenario allowing entry of a wide range of patients, although response may only occur in specific patient subsets, is a recipe for a false negative conclusion. However in circumstances where the relevance of the presumed target is less obvious limiting eligibility risks another false negative, namely missing subsets where the therapy might be effective. For example, although sorafenib is often considered a “FLT3 ITD inhibitor”, the drug is actually a multikinase inhibitor (31) and a trial randomizing patients aged 18-60 with newly diagnosed AML to 3+7+/- sorafenib found that an improvement in event free survival (EFS) in the sorafenib group largely resulted from patients who were FLT3 ITD negative (32). And although it is a more selective FLT3-ITD inhibitor with less “offtarget” effects, quizartinib also has been reported to have activity in FLT3-ITD negative AML, suggesting it affects other clinically relevant targets (33). Another example is an ongoing North American trial evaluating dasatinib as a KIT inhibitor in core-binding factor AML; the trial allows entry of patients with and without a KIT mutation, hypothesizing that overexpressed wild type KIT may also be a valid target for dasatinib hypothesizing that overexpressed wild type KIT may also be a valid target for dasatinib in some patients (34).
In general value might be found in an approach analogous to that described for relapsed/refractory and newly diagnosed patients. Specifically initial studies would be conducted in patients possessing the target and subsequently expanded to other patients.
Problems with conventional clinical trial designs The typical phase 3 trial in AML enrolls about 400 patients into a standard and an investigational arm. This often requires several years to accomplish, not counting the time needed for follow-up of the last patients enrolled. During these years, other therapies are not investigated, which may be problematic if there are several new therapies of interest. Phase 3 trials enroll so many patients because they aim to detect relatively small differences between standard and investigational regimens with 80-90% power (corresponding to a false-negative rate of 10-20%) and P = 0.05 (corresponding to a false-positive rate of 5%). The first question is whether the differences we aim to detect are truly meaningful clinically. For example, an otherwise healthy 68-year-old with newly diagnosed AML would live an additional 15 years (180 months) if he or she did not have AML randomized trial (35). Standard treatment results in a median survival of 12 months, and thus the patient loses 168/180 = 93% of anticipated life expectancy. A new treatment (azacitidine is an example) prolongs survival to 18 months, thus resulting in a loss of 162/18O = 90% of life expectancy. Many patients might consider this improvement, which is quite similar to those often targeted, medically insignificant. A second question is whether a false-negative rate of 20% and false- positive rate of 5% are acceptable for AML. In diseases where good treatment exists, the consequences of a false positive are much greater than in a disease such as AML where standard treatment is
routinely unsuccessful. Hence, revision of phase Ill trials for AML to aim for larger differences, with P = 0.10, would better fit clinical reality and would permit investigation of a larger number of treatments.
Phase 3 trials are conventionally preceded by single-arm phase 2 trials. It seems paradoxical that randomization is a fundamental part of phase III trials but not of the phase II trial that determines whether the phase III trial will be undertaken. This has led several authors to propose the use of randomized phase 2 trials whose intent is to select the best therapy to take into a larger trial (36-37). These selection design trials are often criticized as "underpowered." And, indeed, consequent to their small sample sizes relative to phase3 trials, their power is only 60% to detect a case where three drugs have a true CR rate of 50% and a fourth has a true CR rate of 65%. This 60% power is can be contrasted with the 80% common to many large phase III trials. However, this 80% is only nominal. Consider a case where there are four candidate new therapies each of which might be compared with standard therapy in phase 3. However as often the case pre-clinical rationale is insufficient to know which of the four is best. Accordingly there is only a 25% probability that the best of the four will be selected. Thus the 60% power of a selection design competes with a power of 25% X 80% = 20%. Simply put the worse false negative may occur when a therapy is never investigated at all. Drug development might be improved with use of randomized selection designs, for example the play-the-winner design employed by the Medical Research Council/ National Cancer Research trials in the United Kingdom (37).
Criteria for drug approval Currently the U.S. Food and Drug Administration’s (FDA) principal criterion for approval of drugs in AML is extension of survival (OS) or improvement in “quality of life” (QOL) as demonstrated in a randomized trial. “Accelerated approval” can be given based on outcomes that are likely surrogates for OS or QOL, but a subsequent randomized trial demonstrating improved OS or QOL has generally been required.
One problem with this approach is that it ignores the heterogeneous nature of AML and the resultant possibility that a new treatment might be effective in some subsets but not others. For example, after receiving accelerated approval gemtuzumab ozogamicin (GO) was withdrawn from the U.S. market by its manufacturer because a South West Oncology Group (SWOG) trial randomizing patients to standard 7+3 therapy +/- GO did not show an OS benefit (38). However other similar randomized trials did show an OS benefit in patients with “favorable” and “intermediate” prognosis cytogenetics, suggesting that it may be unrealistic to expect that any therapy will produce a general prolongation in OS (39-40). Of course if randomized trials are done in specific but relatively small subsets, many years may be required to complete the trial unless we are prepared to either look for larger differences or accept less power or larger p-values; the latter would be a welcome development in my opinion since p
