Early peripheral clearance of leukemia-associated immunophenotypes in AML: centralized analysis of a randomized trial.

Although genetics is a relevant risk factor in acute myeloid leukemia (AML), it can be minimally informative and/or not readily available for the early identification of patients at risk for treatment failure. In a randomized trial comparing standard vs high-dose induction (ClinicalTrials.gov #NCT00495287), we studied early peripheral blast cell clearance (PBC) as a rapid predictive assay of chemotherapy response to determine whether it correlates with the achievement of complete remission (CR), as well as postremission outcome, according to induction intensity. Individual leukemia-associated immunophenotypes (LAIPs) identified pretherapy by flow cytometry were validated and quantified centrally after 3 days of treatment, expressing PBC on a logarithmic scale as the ratio of absolute LAIP+ cells on day 1 and day 4. Of 178 patients, 151 (84.8%) were evaluable. Patients in CR exhibited significantly higher median PBC (2.3 log) compared with chemoresistant patients (1.0 log; P < .0001). PBC < 1.0 predicted the worst outcome (CR, 28%). With 1.5 log established as the most accurate cutoff predicting CR, 87.5% of patients with PBC >1.5 (PBChigh, n = 96) and 43.6% of patients with PBC ≤1.5 (PBClow, n = 55) achieved CR after single-course induction (P < .0001). CR and PBChigh rates were increased in patients randomized to the high-dose induction arm (P = .04) and correlated strongly with genetic/cytogenetic risk. In multivariate analysis, PBC retained significant predictive power for CR, relapse risk, and survival. Thus, PBC analysis can provide a very early prediction of outcome, correlates with treatment intensity and disease subset, and may support studies of customized AML therapy.