End points to establish the efficacy of new agents in the treatment of acute leukemia

During 2006, nearly 16 000 people will be diagnosed with acute leukemia in the United States and despite many important advances in diagnosis and therapy, the majority will die of their disease.¹  Acute myeloid leukemia (AML), which accounts for 75% of acute leukemias, is a disease predominantly of adults with a median age at diagnosis of 68 years and a 75% overall mortality.²  The mortality rate of AML is highly age related with 5-year survival of approximately 50% in children, approximately 30% in adults younger than age 60, and less than 5% in those over age 60 years.^3–5  Acute lymphocytic leukemia (ALL) comprises approximately 25% of acute leukemia cases. Two thirds of ALL cases occur in children who have, on average, cure rates approaching 85% with contemporary therapy.⁶  The prognosis is poorer for adults, with a cure rate of approximately 35% in those aged 20 to 60 years, but only 5% in those older than 60 years.⁷  Thus, acute leukemia in adults and children represents a spectrum of diseases ranging from those with a very poor prognosis, such as AML in older patients, to diseases such as childhood ALL where cure rates are high. However, even for those subtypes with high cure rates, the “cost” of cure may be substantial with short- and long-term side effects. Therefore, new agents are needed to extend survival, improve cure rates, and avoid undesired treatment-related toxicities.

The US Food and Drug Administration (FDA) grants marketing approval for new drugs for a specified indication in a defined population. Drugs must be both safe and effective. The Food, Drug and Cosmetics Act of 1938 (the Act) required that new drugs be safe.

A 1962 amendment to the Act specified that new drugs must demonstrate substantial evidence of efficacy in adequate and well-controlled studies. This requirement is described in the regulations, 21 Code of Federal Regulations (CFC), Part 314.126, and has usually been taken to require 2 trials that demonstrate clinical benefit, although in some cases data from a single trial may be sufficient for approval. Detailed discussion of demonstration of effectiveness is also provided in “A Guidance for Industry, Providing Clinical Evidence of Effectiveness for Human Drug and Biological Products.”⁸ 

Federal regulations provide 2 pathways for new drug approval, regular and accelerated approval. Regular approval requires demonstration of clinical benefit or an effect on an end point established as a surrogate for clinical benefit. Clinical benefit is broadly defined as prolongation of life or improved quality of life (living longer or better). Established surrogate end points are those that, based on the preponderance of evidence, the FDA has accepted as being predictive of clinical benefit. For example, the achievement of durable complete response (CR) in acute leukemia has been accepted as a surrogate for clinical benefit because of the observations in multiple trials that patients who achieve durable CR have improved survival compared with those who do not, and the difference in survival is entirely due to the time spent in CR.⁹  Durable CR could also be considered to be a direct clinical benefit if it were shown to have a significant impact on quality of life by reducing the morbidity associated with acute leukemia. In 1992, a second drug approval pathway, accelerated approval, was added to the New Drug Application regulations under Subpart H (21CFR314.510). This was designed to provide patients with life-threatening diseases more rapid access to therapies that appear to have benefit when compared to other available therapies.¹⁰  Accelerated approval is based on the demonstration of an effect on a surrogate measure “reasonably likely” to predict clinical benefit. Subsequent confirmation of clinical benefit is required, and to that end, the drug sponsor must provide a commitment to complete studies after approval that demonstrate clinical benefit.

A limited number of agents have so far been approved for acute leukemia and related diseases (Table 1). Asparaginase, cytarabine, daunorubicin, idarubicin, 6-mercaptopurine, mitoxantrone and teniposide, thioguanine, vincristine, and recently, dasatinib, have received regular approval for types of acute leukemia. Tretinoin (all-trans-retinoic acid) and arsenic trioxide have received regular approval for acute promyelocytic leukemia (APL; second-line therapy). Gemtuzumab ozogamicin, clofarabine, and nelarabine have received accelerated approval for specific populations of relapsed acute leukemia. Azacitidine, decitabine, and lenalidomide have received regular approval for subtypes of myelodysplasia.

Progress in our understanding of acute leukemia has identified an increasing number of potential therapeutic targets and, accordingly, there has been a recent increase in the number of novel and potentially effective agents in preclinical development and in early clinical trials. There has also been progress in the development of tools to monitor the effects of therapy. Both these novel agents and emerging tools for monitoring therapy require confirmation of their efficacy in appropriately controlled clinical trials.

Acute leukemias are comprised of many molecularly distinct subgroups, each of which may require very distinct forms of therapy. There is increased appreciation for the heterogeneity of patients within conventional disease categories. The challenge is to evaluate diagnostic tools and therapeutic agents targeted to very specific and relatively uncommon patient subsets. An additional challenge is that some of the developing agents differ from conventional chemotherapy in their mode of action, for example, promoting differentiation rather than cytotoxicity, suggesting that novel surrogate end points may be needed to predict their clinical utility. Further complicating the evaluation of novel agents is the availability of hematopoietic cell transplantation (HCT), a potentially life-saving intervention that, if used after patients have been treated with experimental therapies, can make it impossible to judge either the response duration or the impact on overall survival of the investigational approach.

To explore some of the issues pertinent to the choice of end points for drug approval in acute leukemia, the FDA invited the American Society of Hematology (ASH) to participate in the organization and conduct of a joint workshop. The workshop would consider possible end points other than durable CR that could serve as surrogates for clinical benefit in the evaluation of new agents. A steering committee established 5 subcommittees, each focused on a specific category of acute leukemia, including pediatric ALL, adult ALL, pediatric AML, adult AML, and leukemia in the elderly. Subcommittee members were selected based on their expertise in acute leukemia and related topics, including clinical trials, molecular markers, hematopathology, statistics, and quality-of-life (QOL) assessment. An FDA reviewer/hematologist was also assigned to each subcommittee, which held a series of discussions and produced a brief working document. A subsequent workshop was organized around the major themes that emerged from the subcommittee discussions, namely, patient heterogeneity, hematologic responses other than CR, elimination of minimal residual disease, “bridge to transplantation,” and QOL. This article summarizes the subcommittee reports and subsequent public discussions that occurred at the workshop.

Patients with acute leukemia vary greatly based on age, health status, and the nature of their leukemia. The biology of both AML and ALL varies greatly with age, with older patients with AML far more likely to have unfavorable cytogenetics and older ALL patients more likely to have Ph⁺ disease.^11–13  Age and performance status greatly affect treatment outcome. In a recent review from the Southwest Oncology Group (SWOG), the likelihood of dying during the first month after receiving induction therapy for newly diagnosed AML was less than 5% for patients younger than 56 years irrespective of performance status, and was less than 15% for patients of any age, even older than 75, with a good performance status. However, for patients who were both older and had a poor performance status, the risk of early death increased dramatically, reaching 82% for those older than 75 with a performance status of 3.¹⁴ 

Disease-related factors that consistently have been shown to influence outcome of adult AML include cytogenetics, white blood cell (WBC) count, and whether the disease is primary (de novo) or preceded by a prior hematologic malignancy.¹⁵  Other disease-specific factors reported as having prognostic importance include MDR1 expression, FLT3 mutational status, NPM1 mutational status, and gene array expression among others.^16,17  The power of these factors was illustrated in a report from SWOG showing that among patients older than 55 years, a CR rate of only 11% could be anticipated in patients with unfavorable cytogenetics, a prior hematologic disorder, and MDR1 expression compared to an 81% CR rate in patients without any of the 3.¹⁸ 

The major prognostic factors in adult ALL generally have included age, cytogenetics, WBC count at diagnosis, and immunophenotype. As one example of the potential impact of prognostic factors on outcome in newly diagnosed adult ALL, in a study from Cancer and Leukemia Group B (CALGB), Larson et al found that age 60 or older, presence of t(9;22) or t(4;11), WBC count of 30 × 10⁹/L, FAB L3 morphology, and absence of a mediastinal mass were each associated with a poorer outcome, and if all 5 factors were present, no patient survived 3 years, whereas if all 5 were absent, survival at 3 years was 100%.¹⁹  In pediatric ALL, age, sex, WBC count at diagnosis, immunophenotype, cytogenetics, and early response to therapy have all been identified as important risk factors.⁶  However, the impact of any risk factor on prognosis is subject to a variety of influences and especially to the specific type of therapeutic regimen used. As an example, although T-cell ALL in children and adults was previously thought to have a poorer outcome than precursor B-cell disease, with changes in therapy this difference has markedly decreased.

Acknowledging the enormous heterogeneity in acute leukemia based on host factors such as age and performance status, as well as disease characteristics such as genetic subtype, is essential when attempting to judge the clinical benefit provided by any particular intervention. Given the potential for this heterogeneity to influence clinical outcomes in predictable and unpredictable ways, single-arm studies must always be viewed with great caution, even if investigators make heroic efforts to identify historical controls. Single-arm trials might be possible for patients who are not candidates for chemotherapy, but the definition of candidacy for chemotherapy should not be based on age alone, but rather also consider performance status, and perhaps, comorbidity scores as well. An alternative to a single-arm study for patients thought not to be candidates for conventional chemotherapy might be a randomized comparison of the experimental agent to supportive care as chosen by the treating physician.

One well-established approach to reducing the impact of patient heterogeneity in randomized trials is to adequately stratify patients according to the most important prognostic factors. An alternative approach is to narrow the entry criteria and exclude certain categories of patients, particularly those with comorbidities or poor performance status. Although the latter approach has the advantage of allowing for a purer, more homogenous trial, it also should be acknowledged that, if approved, the agent will likely be used in a broader population of patients, and potentially important toxicities might only emerge in this more heterogeneous setting.

Not infrequently, a randomized trial may fail to show an overall benefit of a new agent, but a retrospective analysis may suggest benefit in a particular subset of patients. Such findings are particularly alluring if a biologic argument exists potentially explaining the finding. Retrospective subset analyses can be useful tools for creating new hypotheses and for the design of subsequent studies but are generally not useful for obtaining FDA approval. It should be remembered that with enough inspection, the likelihood of finding a subset of patients that appears to benefit from a specific treatment is high, and that even if the effect is real, because in retrospective analyses one is searching for outliers, the magnitude of the effect will likely diminish in subsequent prospective trials.

CR is generally defined as the achievement of both normal marrow morphology on light microscopy with fewer than 5% blasts and recovery of peripheral blood counts with neutrophils greater than 1 × 10⁹/L and platelets counts greater than 100 × 10⁹/L.²⁰  As noted, the FDA has accepted durable CR as an established surrogate for clinical benefit because those patients who achieve durable CR have been shown to survive longer than those who do not.⁹  CR as a surrogate end point provides more timely conclusions about drug's efficacy without waiting for maturation of survival data. Further, the confounding effects of HCT for patients in CR can be avoided.

Following chemotherapy, patients may fulfill all the requirements of a CR but may not recover peripheral blood counts to the required level, a response category termed morphologic complete remission with incomplete blood count recovery (CRi). ²⁰  CRp describes a subcategory of CRi, wherein patients fulfill all criteria for CR except that platelet counts are less than 100 × 10⁹/L.

Gemtuzumab ozogamicin is currently the only drug approved by the FDA for relapsed AML. Accelerated approval was granted in 2000 largely on the basis of 3 single-arm phase 2 trials in patients with AML in untreated first relapse, which reported a 16% CR rate and a 13% CRp rate (overall response rate of 30%).²¹  Median overall survival for the entire study population (142 patients) was 5.9 months. These studies were not adequately powered to substantiate the use of CRp as an established surrogate, but there was no apparent difference in overall survival between CR and CRp patients, with both categories surviving longer than patients who did not exhibit either response.^21,22  The significance of CRi or CRp might differ among products, regimens, or patient populations. Therefore, even if CRp were established to be a surrogate for clinical benefit in one setting, CRp might not be a useful surrogate in all clinical situations. In an examination of the utility of CRp as an end point, investigators at the MD Anderson Cancer Center analyzed results in 176 patients with AML in first relapse treated with a variety of regimens that did not include gemtuzumab ozogamicin. CRp was less common than after therapy with gemtuzumab ozogamicin but had similar implications. Median survival was similar for CR and CRp patients, and patients in both response categories lived longer than nonresponders. When previously untreated AML patients at the MD Anderson Cancer Center were analyzed, only 6.3% had a response that was categorized as CRp, and although these patients survived significantly longer than nonresponders, they did not survive as long as true CR patients, a result consistent with that seen in a much larger study from the Dutch-Belgian and Swiss group.^23,24 

CR and CRp were both applied as surrogate end points in a small study used as a basis for the accelerated approval of clofarabine for the treatment of patients aged 1 to 21 years with ALL that had relapsed after or was refractory to at least 2 prior regimens. Among 49 patients, 12% CR and 8% CRp were reported. Although the study was not powered to detect a difference between CR and CRp, the overall survival of CR and CRp patients appeared similar. These response categories were associated with superior survival rates compared to partial respone (PR) and nonresponding patients, a finding similar to that reported in a phase 2 trial of clofarabine in adults.²⁵  The data from patients being treated for recurrent leukemia seems to favor the use of CRp as a surrogate for clinical benefit in this clinical setting. Current data do not conclusively allow the extrapolation of CRp to patients receiving initial chemotherapy, where it is possible that incomplete platelet recovery could interfere with the administration of postremission consolidation chemotherapy. Insufficient data were presented at the workshop to evaluate the significance of other categories of CRi.

Patients who fail to achieve CR or CRi may experience a PR (defined as a ≥ 50% decrease in marrow blasts with normalization of peripheral counts) or some other measure of hematologic improvement. A PR in acute leukemia is generally expected to be of short duration and thus in most circumstances is unlikely to serve as a surrogate reasonably likely to predict for clinical benefit. There may be some categories of patients with less proliferative AMLs who may benefit from achievement of PR such as AML arising from myelodysplasia. In addition, as discussed later (see “Bridge to transplantation”), a PR could contribute to the clinical benefit conveyed by a subsequent intervention, such as HCT, provided the conditions were adequately defined.

Although durable morphologic CR is generally accepted as an established surrogate of clinical benefit, standard light microscopic examination of marrow has a sensitivity of, at best, 1%. More sensitive measures for detection of leukemia cells than morphology exist, including multiparametric flow cytometry (MFC) and polymerase chain reaction (PCR) assays of minimal residual disease (MRD) persisting in remission marrow. MFC generally depends on the expression by malignant cells of a pattern of cell-surface antigens subtly different from those expressed by normal hematopoietic progenitors. Abnormal MFC patterns can be identified on at least 80% of human AML cases and an even greater proportion of ALL cases. The sensitivity of MFC is limited to detection of about 1 cell/10³ to 10⁴, depending on the antigenic marker. PCR is substantially more sensitive than MFC, able to detect 1 cell/10⁵ to 10⁶. However, to be applicable, a specific genetic marker must be known, such as a leukemia-specific chromosomal translocation or rearrangement in the immunoglobulin or T-cell receptor (TCR) genes. Thus, though more sensitive, PCR is less applicable for AML, because known mutational translocations exist in only 30% to 40% of cases. In ALL, however, PCR assays directed at immunoglobulin or TCR rearrangements are applicable in at least 80% of cases. Compared to MFC, PCR-based assays are more labor intensive and expensive.

Measurement of MRD has attracted interest as a potential surrogate for clinical benefit. As with CR, if MRD at the end of induction were the end point, one would not need to wait for subsequent relapse or death to evaluate a new induction regimen. Further, one could evaluate the efficacy of the addition of new induction agents in situations where current regimens already approximate 100% CR rates, such as APL and childhood ALL. It is also possible that the relative efficacy of different consolidation or maintenance approaches could be compared by measuring their impact on MRD without needing to wait until relapse or death.

Most studies of MRD have examined its predictive power when measured at the end of induction. In AML, several investigators have published results demonstrating that the extent of MRD measured by MFC on the first bone marrow in morphologic CR following induction therapy strongly predicts both disease-free and overall survival. For example, San Miguel reported that survival at 5 years after induction varied from 90% in those with low levels of MRD after induction to a low of 29% in those with high levels.²⁶  Similarly, studies in ALL report event-free survival (EFS) at 3 years after induction ranging from 90% in those with no measurable MRD to 27% in those with significant amounts.^27–31 

A second setting where MRD levels appear to predict clinical outcome is prior to bone marrow transplantation. In patients undergoing transplantation for ALL in morphologic remission, MRD level prior to transplantation appears to have a marked impact on posttransplantation outcome.³²  A similar finding has been reported in the treatment of APL in second remission using autologous transplantation where the best results are seen in those who are PCR negative for the PML/RARα fusion gene product prior to the procedure.³³ 

The consistency of observations correlating levels of MRD with ultimate survival are probably strongest with the measurement of postinduction MRD by MFC in childhood ALL and provide a good argument that reduction in MRD at the end of induction might be a surrogate for clinical benefit in that setting. A similar argument can be made for AML. However, even though a particular method of measurement of MRD for a specific disease may serve as a useful surrogate for clinical benefit in that setting, it does not follow that all measures of MRD will be similarly predictive. One striking example is the finding that many patients with t(8;21) AML in long-term (> 10-year) unmaintained remissions continue to test positively by PCR for the presence of the translocation.³⁴  Given this observation, it would be hard to argue that elimination of a positive PCR signal in this setting is evidence of clinical benefit. Thus, although evidence indicates that measurements of MRD might serve as surrogates for clinical benefit in certain subtypes of acute leukemia, each form of measurement and each clinical circumstance should be individually validated.

Although large prospective randomized trials are lacking, retrospective analyses suggest that HCT is the treatment of choice for many patients with recurrent acute leukemia. Occasional patients who recur with acute leukemia following a long first remission can be cured with second-line chemotherapy. However, for the majority of patients who recur with their disease within the first 2 years of initial induction, HCT appears to offer the best chance for cure. In the large majority of studies published to date, disease status at the time of transplantation appears to greatly influence treatment outcome. For example, in a report analyzing the results of allogeneic transplantation for recurrent AML in children, disease-free survival at 5 years was 24% overall, but 58% for patients undergoing transplantation in second remission versus 36% for those undergoing transplantation in untreated first relapse versus 9% for those having transplantation in refractory relapse (defined as disease resistant to the last reinduction attempt).³⁵  Similarly, in a study of 174 patients with recurrent AML or ALL undergoing unrelated donor transplantation, the 5-year leukemia-free survival was 38% in those with less than 30% marrow blasts and no peripheral blasts, 15% for those with more than 30% marrow blasts but no circulating blasts, and 0% in those with blasts in the periphery (this last group consisting almost entirely of patients with refractory disease).³⁶  Whether or not these results are due to an impact of tumor burden on outcome or instead are a reflection of inherent drug sensitivity, the practical result is that many transplantation centers are reluctant or refuse to offer standard ablative transplantation to patients with refractory disease with circulating blasts. The impact of disease burden seems to be even greater with nonmyeloablative transplants where the entire antileukemia effect is due to the immunological graft-versus-leukemia reaction.³⁷ 

Durable CR has already been established as a validated surrogate of clinical benefit in acute leukemia. Thus, an agent that is effective in producing second or subsequent remissions in acute leukemia prior to transplantation would be accepted as beneficial regardless of its impact on subsequent transplantation. However, a question of interest, and the one being discussed here, is whether an agent could be viewed as providing significant clinical benefit on the basis that without significantly increasing CR rates, it reduces or controls tumor burden sufficiently with acceptable toxicity to allow patients to proceed to transplantation. And if so, is it sufficient that the agent simply allows more patients to get to transplantation or would the outcome of subsequent transplantation be of importance in determining the benefit provided by the agent?

It would, of course, be preferable to have evidence from prospective randomized trials that an agent that allows more patients to undergo transplantation also results in improved long-term survival. However, randomized trials testing such an outcome would be extremely difficult to perform, given the small numbers and broad heterogeneity of patients, their diseases, and their donors. Further, agreeing on a standard comparison arm would be very difficult given the heterogeneity in agents previously used in such patients. Finally, such a study would take a very long time to complete, considering the time needed to accrue patients and the length of time required to follow until relapse or death.

Why not then accept an agent's ability to allow a patient to get to transplantation as a surrogate for clinical benefit? First, although the evidence is compelling that the presence of a high leukemic burden compared to a lower leukemic burden at the time of transplantation is a powerful negative prognostic factor, that does not necessarily mean that any agent that can reduce that burden sufficiently to allow the patient to undergo transplantation will necessarily improve the outcome of transplantation. Second, without actually observing the outcome of transplantation after the experimental agent, it is possible that any advantage in getting additional patients to transplantation could be lost if there were unpredicted posttransplantation toxicities caused by prior exposure to the agent in question. Although it is tempting to base the assessment of clinical benefit solely on data that correlate disease burden at the time of HCT and immediate outcome after HCT, additional long-term observations are needed for the assessment to be supported.

FDA approval of a new drug requires demonstration of clinical benefit. One acceptable definition of clinical benefit is “living better.” Thus, it would seem that an agent that provides an improved QOL should be considered to be of clinical benefit. Yet, lack of validated measures of QOL for acute leukemia and methodologic limitations have hampered use of this end point for drug approval. A number of clinical events associated with acute leukemia and its treatment are objective, easy to measure, and likely affect QOL. These include, but are not necessarily limited to, number of blood transfusions, days with fever, days on antibiotics, and days in hospital. Investigators have suggested that it should be possible to construct a clinical events composite score that incorporates these measures, but one is not currently available. Demonstrated differences in morbidity are most compelling when response rates and survival are similar. If response rates and survival are not similar, in most situations it would be difficult to justify a tradeoff of lower response rates or shorter survival for improved QOL. Particularly for younger patients with acute leukemia where cure is the therapeutic goal, the period during which patients are actually receiving therapy is only a small proportion of their total survival after diagnosis and thus the emphasis in evaluating a drug must be on response rates and survival. The situation might be somewhat different for older patients where response rates are lower and survival shorter. For these patients, given the toxicity of currently available chemotherapies, a drug that provides improved QOL, but at a cost of somewhat shorter survival, might provide clinical benefit. However, in making these comparisons, one would need to be very cautious, because if one is willing to accept lower response rates, then a reduction in toxicity might be achieved with available drugs simply by lowering doses.

Efforts have been made to go beyond simple measurements of clinical events. These efforts have included patient-reported outcomes (PROs), such as symptom relief or health-related quality of life (HRQOL). HRQOL is multidimensional and includes physical, functional, social, emotional, and spiritual QOL, as measured by QOL instruments. Assessments of symptom relief and QOL are hampered by the lack of instruments that have demonstrated validity, reliability, and sensitivity to clinically important changes. Particular challenges are posed by pediatric patients where parents must serve as proxies, and minority populations where cultural differences in perceived QOL may exist.³⁸  Moreover, few instruments are available for measuring HRQOL in acute leukemia. Although Functional Assessment of Cancer Therapy for Leukemia (FACT-LEUK) has been developed, it is undergoing validation and no publications have yet reported its use. European Organization for Research and Treatment of Cancer-Leukemia (EORTC-LEUK) has been used several times, but this module is not listed on the EORTC Quality of Life Questionnaire (QLCQ-30) website.^39,40 

Even if a well-validated instrument were available, the data are difficult to collect, particularly in an ill population with high mortality rates. Because of the time required to fill out surveys, patients frequently do not complete forms or drop out of studies, necessitating complicated analytic approaches to study evaluation. Larger study samples are often needed for QOL assessments than for other clinical end points to reach statistical significance. Because HRQOL is affected by so many inputs, the end point being measured suffers from variability. It is difficult enough to understand minimal clinical differences, but it is even harder to understand the meaning of a given change in an HRQOL scale. In general, the lack of validated instruments, the relatively small numbers of patients available for trials, and the need for large numbers of patients for HRQOL studies, makes the use of these tools as an end point for acute leukemia trials problematic. These facts do not mean that QOL is not important; in fact, along with life expectancy, it is probably the most important issue in treating patients. The lack of currently available objective, reproducible, and understandable HRQOL tools for studying acute leukemia points out an area of research need. In the meantime, individual or composite scores of clinical events might help serve a similar purpose.

FDA evaluation of new agents for the treatment of acute leukemia in children and adults may result in full approval or accelerated approval, the latter requiring demonstration of clinical benefit after approval. Regular approval requires demonstration of clinical benefit, either prolongation of life or improved QOL, or an effect on an end point established as a surrogate for clinical benefit. The design and execution of prospective randomized clinical trials that will lead to definitive evidence supporting a longer or better life for patients with these disorders can be difficult and costly and require lengthy follow-up. Thus, the development of novel trial designs and inclusion of validated surrogate markers for clinical benefit would be welcome. At the workshop, participants discussed several possible surrogate end points, each of which will require further evaluation for validation. These end points include CRp and other variants of CR, various measures of MRD, so-called “bridge to transplantation,” and measures of QOL. At the current time, CR remains the most robust surrogate marker predicting a longer life. The findings of the workshop will be presented at the Oncologic Drugs Advisory Committee (ODAC) in the near future, so that ODAC can provide advice to the FDA regarding the end points and related issues discussed at the workshop.

Contribution: All authors contributed to the organization of the workshop, the selection and presentation of data, and the writing of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Frederick R. Appelbaum, Fred Hutchinson Cancer Research Center, Clinical Research Division, 1100 Fairview Ave N, D5-310, PO Box 19024, Seattle, WA 98109; e-mail: fappelba@fhcrc.org.

The views expressed are the result of independent work (D.R. and N.S.S.) and do not necessarily represent the views and findings of the US Food and Drug Administration or the National Institutes of Health.