An 86-Probe Gene Expression Signature Can Predict Survival in AML with Normal Karyotype Independently of FLT3 ITD and NPM1 Mutation Status - A Collaborative Study from the AMLCG and CALGB Study Groups.

Abstract

Patients with acute myeloid leukemia and a normal karyotype (NK AML) comprise 50% of all AML cases and show heterogeneous treatment outcomes and survival. We used gene expression profiling to develop a prognostic gene signature that predicts survival in this clinically relevant AML subgroup. Our analysis was based on data from 163 patients with newly diagnosed NK AML treated in the German multicenter AMLCG 2000 trial, for whom pretreatment gene expression profiles were obtained using Affymetrix HG-U133 microarrays. We used supervised principal component analysis to identify 86 oligonucleotide probesets (corresponding to 66 different genes and ESTs) that were correlated with overall survival (OS), and to define a prognostic score based on these probesets. When applied to an independent test cohort of 79 NK AML cases from the same AMLCG trial, the continuous prognostic score was predictive of OS (P=0.002, hazard ratio [HR] for a change in prognostic score equal to the difference between the 75th and 25th percentiles of the score = 1.94) and event-free survival (EFS) (P = 0.001, HR=1.70). The score based on our gene signature showed a strong correlation with the presence of the FLT3 internal tandem duplication (ITD), but retained its prognostic value for OS in the test cohort even after adjustment for FLT3 ITD, NPM1 status and age (P=0.037, HR=1.65). When we defined a cut-off value in the training population and used it to dichotomize the gene expression score values in the test cohort, the resulting two subgroups had significantly different OS (median, 259 days vs. not reached, P=0.002) and event-free survival (EFS) (median, 72 vs. 300 days, P = 0.015). We subsequently confirmed our findings in a group of 64 NK AML patients (Blood 2006; 108:1677–83) treated on CALGB 9621. In this validation cohort, our continuous gene expression score was predictive of OS (P < 0.001, HR=4.11) and EFS (P < 0.001, HR=2.90). In multivariate analyses that adjusted for age, NPM1 and FLT3 ITD status, the gene expression score remained significant for OS (P = 0.007, HR=3.40). When we used the prognostic score to split the CALGB validation cohort into two groups, based on the same cut-off value as in the AMLCG test population, the two resulting subgroups differed in their OS (median, 375 days vs. not reached, P < 0.001) and EFS (median, 258 vs. 728 days, P = 0.027). In summary, we present a novel and robust gene expression signature that offers independent prognostic information for patients with normal karyotype AML.