To the editor: Cyclosporine modulation in poor-risk acute myeloid leukemia

The recent report from the Southwest Oncology Group by List et al1 regarding the benefit of cyclosporine modulation of Pglycoprotein drug resistance in poor-risk acute myeloid leukemia (AML) was read with anticipation, as optimal regimens for this difficult patient population have remained elusive. Upon reviewing the paper, however, several concerns became apparent. The use of one-sided analysis to determine superiority of the cyclosporine arm was explained as follows: “the principal objective of the study was to assess whether [cyclosporine A] improves treatment outcomes.” 1(p3214) But the paper acknowledges that the pilot study of cyclosporine modulation showed that “treatment with CsA delayed the hepatic elimination of bilirubin . . . resulting in reversible conjugated hyperbilirubinemia and increased systemic anthracycline exposure” 1(p3212); therefore it seems difficult to justify dismissing the possibility that the cyclosporine arm could do worse than the control arm necessitating 2-sided analysis. This significant flaw in design diminishes the power of the study. This seems particularly important given that the primary end point of the study, increased complete remission (CR) rate with cyclosporine induction, was not reached and that the conclusions reached by the authors are based entirely on secondary and subgroup analysis. The most powerful conclusion from the study is that cyclosporine modulation results in a significant improvement in the duration of remission and in overall survival. But the use of stem cell transplantation for patients in remission is a potent confounding variable for each of these results. The authors accurately point out that, among patients under age 60 who achieved CR, 9 of 25 (36%) of patients in the control arm and 17 of 39 (43%) of patients in the cyclosporine arm received stem cell transplantation. Comparison of the percentages of eligible patients transplanted in each arm appears tantalizingly similar. But an equally accurate manner in which to present this data is that, in absolute numbers, 8 more patients in the cyclosporine arm received transplantation. Although this number is small, this represents nearly 7% of the patients randomized to the cyclosporine arm and is nearly double the number of patients transplanted in the control arm. This small number is pertinent given that the basis of the significance favoring the cyclosporine arm at 2 years for relapse-free survival (RFS) in patients in CR and overall survival for all patients is, in absolute terms, 8 patients and 9 patients, respectively. Whereas the authors provide some statistical reasoning to support their conclusion that “the CsA effect was not attributable to transplantation in remission,” 1(p3215) this significant conclusion warrants more detailed analysis. Who were the patients that received transplants, and by what criteria was transplantation determined? What type of transplants did they receive? Was HLA matching similar between these groups? Was there transplantation-associated mortality? When dealing with such a small number of patients, a factor as random as the availability of matched siblings may dramatically affect the results. Combining the improper use of one-sided analysis and the confounding effect of stem cell transplantation for patients in remission with a small but real imbalance in pretreatment cytogenetics favoring the cyclosporine arm, a clear conclusion that cyclosporine modulation has significant long-term benefit does not seem capable of eliminating type I error (ie, the probability of incorrectly concluding that there is a statistical difference in a data set). Although this study clearly shows that cyclosporine modulation in poor-risk AML results in a significant decrease in residual disease following induction therapy, the long-term significance regarding relapse-free survival and overall survival appears to be a matter of speculation.