In the MRC AML-HR trial,1 Milligan et al describe inferior overall survival with fludarabine and high-dose cytosine (FLA) compared with conventional cytosine, daunorubicin, and etoposide (ADE) reinduction chemotherapy for high-risk acute myeloid leukemia (AML). As type 1 error is a potential explanation of these results, further information to clarify the mechanism underlying the reduced overall survival with FLA would be helpful before a potentially useful regimen is abandoned.
The difference in overall survival appears attributable to increased death in remission (34% vs 12%; P = .1) in the FLA compared with ADE group. Although not statistically significant, the trial was not powered to explore this parameter, and other adverse predictors (resistant disease, induction death, and relapse rate) were almost identical. An analysis of deaths in remission comparing the 2 treatment arms may clarify the inferior overall survival with FLA. For example, fludarabine predisposes to opportunistic infections,2 may influence choice of subsequent consolidation therapy, and may also lead to an increased risk of second malignancy.3 No information on the use of prophylactic antimicrobials was given.
The MRC AML-HR trial1 has made an important contribution to our understanding of the optimal therapeutic approach in high-risk AML. Further information would be valuable to continue development or modification of current treatment algorithms in this challenging disease.
The authors declare no competing financial interests.
Increased mortality with FLA compared with ADE
We agree with Lane et al that the difference in overall survival between fludarabine and high-dose cytosine (FLA) and cytosine, daunorubicin, and etoposide (ADE) could be a chance effect. The hazard ratio, as reported in our paper, has a 95% confidence interval (CI) that ranges from 1.01 to 1.77. Thus, our result is compatible with there being no adverse impact of FLA on survival; on the other hand, it is also compatible with a 75% increase in mortality. Because of this uncertainty, we were cautious in our interpretation: in the abstract we concluded that “FLA may be inferior,”1(p4614) whereas in the discussion we said that “we found no evidence that fludarabine with high-dose cytosine was a better treatment than the standard MRC schedule of ADE.”1(p4620) We believe that the latter statement is justified since, based on our results, it seems unlikely that FLA is actually better than ADE.
With regard to deaths in complete remission (CR), the Kaplan-Meier estimates at 4 years are a bit misleading and make it appear as if the death rate is nearly 3 times greater in the FLA arm (34% versus 12%). This is because there was a small number of late events in the FLA arm (4 deaths beyond 1 year—with only a small number of patients at risk, each event adds 3%-4% to the Kaplan-Meier estimate), but none beyond 1 year in the ADE arm. In terms of crude death rate, 20% (17/83) of patients in the FLA arm died in CR compared with 12% (10/84) in the ADE arm. There is no clear explanation for this moderate, and nonsignificant, difference. More patients died following transplantation in the FLA arm than in the ADE arm (10 versus 5, with 36 versus 37 transplantations performed, respectively), though this difference is not significant (P = .13). Of the remaining 12 deaths in CR, the majority was from infection as would be expected, but there was no excess with fludarabine (4 in each arm). None of the deaths in CR was related to second malignancy. Although there was no difference between the arms in relapses (83 FLA, 84 ADE), slightly more of the relapsing FLA patients have subsequently died (45 FLA, 39 ADE). Thus, a number of factors combine—3 more deaths without CR, 7 more in CR, 6 more after relapse—to give the cumulative excess of 14 deaths in the FLA arm, and the borderline significant adverse effect of FLA on survival compared with ADE, as reported in our paper.
Although we have not identified any clear reason for the apparent adverse impact of FLA on survival, and on the basis of the CI in our study (lower limit of hazard ratio CI = 1.01), it may not actually be any worse than ADE; it could, however, be a lot worse (upper limit of CI = 1.77), so it would be reasonable on the current evidence to avoid using FLA routinely.
The authors declare no competing financial interests.
Correspondence: Donald W. Milligan, Department of Hematology, Birmingham Heartlands Hospital, Birmingham B9 5SS, United Kingdom; e-mail: d.w.milligan@bham.ac.uk.
