Acute myeloid leukemia (AML) encompasses a large number of morphologically similar but molecularly distinct variants. Recurrent cytogenetic aberrations have been shown to constitute markers of diagnostic and prognostic value. However, despite recent successes in detecting novel molecular markers like FLT3 (fms-related tyrosine kinase 3) mutation, treatment stratification is still difficult, especially for the 40–45% of patients with intermediate-risk, normal karyotype disease. To better characterize AML at the molecular level, and to address the need for improved risk stratification, we recently profiled gene expression in a large series of adult AML patients (

Bullinger et al.,
N Engl J Med
350
:
1605
,
2004
). By unsupervised analysis we identified new prognostically-relevant AML subgroups, and using a supervised learning algorithm we constructed a gene-expression based outcome predictor, which accurately predicted overall survival across all patients, including for the subset of AML cases normal karyotype. Having demonstrated the presence at diagnosis of normal karyotype signatures correlating with clinical outcome, we have now sought to refine a prognostic signature specific for normal karyotype disease. Towards this goal, we have now profiled 119 samples of adult AML patients with normal karyotype using 42k cDNA microarrays from the Stanford Functional Genomics Facility. By semi-supervised analysis using the supervised principal component method (
Bair et al.,
PLoS Biology
2
:
511
,
2004
), we built a cross-validated gene-expression based outcome predictor in a randomly partitioned training set (n=60 samples). This outcome signature, comprising only 16 genes, significantly predicted outcome class for normal karyotype samples in the independent test set (n=59 samples; P=0.001). In multivariate analysis, the 16-gene signature was a strong [odds ratio=0.35 (0.13 to 0.91); P=0.01] factor in predicting overall survival, independent of known prognostic factors including FLT3 mutations and preceeding malignancy. Our findings support the utility of expression profiling for improved risk stratification and clinical management of the clinically important subclass of AML patients with normal karyotype disease.

Topics:
gene expression, karyotype determination procedure, leukemia, myelocytic, acute, ms-like tyrosine kinase 3, biological markers, cancer, dna, complementary, gene expression profiling, molecular profiling, prognostic factors

Author notes

Corresponding author

2005, The American Society of Hematology
2005
Sign in via your Institution