LAUSR

Core-binding factor (CBF) acute myeloid leukemia (AML) is characterized by t(8;21)(q22;q22) and inv(16)(p13;q22)/t (16;16)(p13;q22) abnormalities, and despite mutational and cytogenetic heterogeneity among t(8;21) and inv(16), these entities share similarities in prognosis, are grouped together, and managed similarly [1]. Our historical data demonstrated that fludarabine-based (fludarabine, cytarabine, and G-CSF) frontline regimens may be superior to the traditional anthracycline and cytarabine combination regimen in treating CBF leukemia [2]. FLAG when combined either with gemtuzumab ozogamicin (FLAG-GO) [2] or idarubicin (FLAG-IDA) [3] has resulted in high remission rates and impressive long-term disease-free and overall survival when employed in the frontline setting. The detection of RUNX1/RUNX1T1 (in t(8;21)) and CBFB/MYH11 (in inv(16)) gene-fusion transcripts by quantitative real-time PCR (qPCR) affords a highly sensitive and precise technique for disease monitoring and response to treatment [4]. Relevant cutoff levels predicting relapse vary based on cytogenetic subgroup, timing of the analyzed sample, and source (peripheral blood or bone marrow) [5–13]. Promising results using the FLAG-based treatment prompted us to assess the prognostic utility of depth and slope (rapidity) of reduction of qPCR levels on relapse and survival in this treatment setting. Patients newly diagnosed with CBF-AML from 2007 to 2014 were treated in two consecutive FLAG-based regimens in a phase-II clinical trial (NCT00801489), initially with gemtuzumab (GO), and, when GO was withdrawn from the United States market, with idarubicin (IDA). Induction in both regimens is as previously described [2, 3]. Among patients receiving FLAG-GO, GO was given twice during consolidation, either during cycle (C) 3/4 and then in C 5/6. In regard to FLAG-IDA, idarubicin was given once during one of the consolidation cycles. The primary aim of our study was to determine whether depth and slope (rapidity) of reduction in qPCR levels help predict for relapse and survival. qPCR to detect fusion transcripts was performed in serial bone marrow aspirates and normalized to ABL1 transcript as previously described [3]. Testing was performed at baseline, after induction, and after every 1–2 cycles of consolidation. The qPCR data were logarithmically (log) transformed and depth of response was measured as changes in log qPCR from baseline. Given the size of the cohorts, we decided to combine both cytogenetic subgroups for the analysis. As rapidity of response was one aspect that we were interested in, we used a random regression model [14] to assess the slope of response (defined as the change in log qPCR over time). The random effects regression is an extension of linear regression that adds a “random effect” which would account for multiple measurements made longitudinally on single individuals. A regression was generated for each individual simultaneously, accounting for the distribution of measurements in the entire data set. From the model, regression coefficients for each individual were generated resulting in the slope. Overall survival (OS) was measured from the time of presentation to date of death or censored at the last follow* Gautam Borthakur [email protected]