Strikingly different molecular relapse kinetics in NPM1c, PML-RARA, RUNX1-RUNX1T1, and CBFB-MYH11 acute myeloid leukemias.

Early relapse detection in acute myeloid leukemia is possible using standardized real-time quantitative polymerase chain reaction (RQ-PCR) protocols. However, optimal sampling intervals have not been defined and are likely to vary according to the underlying molecular lesion. In 74 patients experiencing hematologic relapse and harboring aberrations amenable to RQ-PCR (mutated NPM1 [designated NPM1c], PML-RARA, RUNX1-RUNX1T1, and CBFB-MYH11), we observed strikingly different relapse kinetics. The median doubling time of the CBFB-MYH11 leukemic clone was significantly longer (36 days) than that of clones harboring other markers (RUNX1-RUNX1T1, 14 days; PML-RARA, 12 days; and NPM1c, 11 days; P < .001). Furthermore, we used a mathematical model to determine frequency of relapse detection and median time from detection of minimal residual disease to hematologic relapse as a function of sampling interval length. For example, to obtain a relapse detection fraction of 90% and a median time of 60 days, blood sampling every sixth month should be performed for CBFB-MYH11 leukemias. By contrast, in NPM1c(+)/FLT3-ITD(-), NPM1c(+)/FLT3-ITD(+), RUNX1-RUNX1T1, and PML-RARA leukemias, bone marrow sampling is necessary every sixth, fourth, and fourth and second month, respectively. These data carry important implications for the development of optimal RQ-PCR monitoring schedules suitable for evaluation of minimal residual disease-directed therapies in future clinical trials.