• © 2002 by American Society of Clinical Oncology

Treatment of Myelodysplastic Syndrome: Questions Raised by the Azacitidine Experience

IN THIS ISSUE of the Journal of Clinical Oncology, Silverman et al¹ report the results of a randomized trial in which 191 patients with myelodysplastic syndrome (MDS) received either subcutaneous azacitidine or supportive care. Use of azacitidine was associated with statistically significantly higher complete response (CR) plus partial response (PR) rates, overall response rate, longer time to leukemic transformation or death, and better quality of life, but no significant difference in overall survival. This is the first randomized trial to report that treatment for MDS was better than supportive care, as assessed by multiple end points.^2-5 Silverman et al should be commended for their perseverance in developing azacitidine in MDS through a series of phase II studies of varying intravenous and subcutaneous schedules, which have culminated in this successful phase III randomized trial.^6,7

This study raises several issues that affect the design, conduct, and interpretation of trials in MDS and other hematologic cancers. The first issue concerns response criteria in MDS. The current study used strict conventional criteria for CR and PR, but there was a less strictly defined mono- or bilineage response category of hematologic improvement or “improved.” This category constituted the majority of responses in the azacitidine arm (37%) compared with 23% for CR + PR. The response criteria were validated by their association with longer time to transformation or death, improved quality of life, and reduced transfusion requirements, particularly in the improved response category. A separate analysis of CR + PR and improved categories might help determine whether the longer time to transformation is due solely to the CR + PR categories or is also contributed to by the improved response subgroup. The United States Food and Drug Administration (FDA) has asked investigators to define response criteria in MDS and other tumors that are associated with improved quality of life, such as significant reductions in infections, hospitalizations, and packed RBC and platelet transfusions. This has led a group of investigators to propose such response criteria for MDS, which should be considered in the evaluation of future trials and modified as indicated by future results.⁸

A second issue concerns the randomization design used in this trial, with subsequent cross-over to azacitidine therapy at the time of disease progression for patients assigned to supportive care. When phase II data with a new agent appear positive in a disease associated with a particularly poor prognosis, it becomes difficult to convince some investigators and most patients to accept randomization. This is particularly true when the control arm involves observation, placebo, or supportive care alone, and when the standard of care has been associated with poor outcome. Randomized trials are, therefore, more acceptable in relatively benign disorders, in those with indolent courses, when the standard of care is relatively effective, when a standard of care is offered with an added investigational component, or when the relative merits of the control and investigational arms cannot be judged a priori based on past experience. However, when a randomized trial offers therapy with promising phase II experience in a poor-prognosis disease, innovative designs may be needed to encourage participation in the trial. One possibility is a cross-over design for patients whose disease progresses on a treatment arm, as in this study. The cross-over design is attractive and ethical, but it may reduce our capacity to demonstrate significant differences in survival outcome because the beneficial treatment would be offered sequentially.⁹ Alternative designs have been proposed that offer patients a potentially effective therapy earlier, including the “randomized discontinuation” design.¹⁰ In this design, all patients are treated initially with the investigational drug. After a period of time, patients who respond continue to receive therapy, those who progress are taken off study, and those with stable disease are randomized to the standard of care or continuation of the investigational treatment. This design allows for enrichment of the randomized portion of the study, but it may also require a large number of patients. Other randomization designs may include randomization with fixed unequal allocation (higher ratio of patients assigned to the investigational arm), adaptive randomization (higher proportion of patients assigned to the arms showing superior outcome), or others.^11,12

This issue brings up an important question pertinent to the current azacitidine study in MDS. Whereas survival was not significantly different between the azacitidine and supportive care arms, the trend favored azacitidine therapy (median survival, 20 months v 14 months; P = .10). Could the survival difference have become statistically significant if the study had accrued a larger number of patients? A total of approximately 350 to 530 assessable patients would have to have been entered, a difficult task in future studies of MDS if the control arm remains supportive care.

The third issue highlighted by this trial and others is whether improvement in survival should always be the ultimate treatment end point, or whether other surrogate end points for patient benefit can be used. These may include time to transformation; high-quality responses which are consistently associated with improved survival, such as major cytogenetic response in chronic myelogenous leukemia or molecular complete responses in acute promyelocytic leukemia or other tumors; significant and durable responses from treatments with negligible side effects or extremely low mortality rates; or significant objective improvements in quality-of-life measures. These surrogate end points have been accepted by the FDA for the approval of several drugs. Future surrogate end points may include modulation of cancer target signals for particular target-specific strategies.

Finally, the ultimate question is whether the results of this trial are sufficient to declare azacitidine the new standard of care for the treatment of MDS. I believe they are; others may disagree. They may raise questions about the response criteria, including the fact that both investigators and patients were aware of the treatment assignment, which could affect their assessment of quality of life. In addition, central pathology review and independent auditing of data quality and protocol compliance were undertaken in a minority of patients, supportive care measures may not have been standardized, and blood and marrow studies may not have been obtained at similar intervals in the azacitidine and supportive care groups, with implications for analyses of time to treatment failure and acute myelogenous leukemia transformation. Most important, survival was not statistically improved with azacitidine therapy compared with supportive care. FDA approval of azacitidine as the first treatment for MDS may not only help many patients but will also encourage future investigators to compare or add new agents to azacitidine for the treatment of MDS. These may include potentially better hypomethylating agents (decitabine), farnesyl transferase inhibitors (R115777), chemotherapeutic agents with novel mechanisms of action (clofarabine, homoharringtonine), new cytokines (thrombopoietins), and many others.^13-18 Certainly, the positive experience with azacitidine, made possible through the endeavors of Dr Silverman and others, brings new hopes to thousands of patients with MDS and to investigators who dedicate their research to improving prognosis in this relatively common and lethal malignancy.