In the JournalsPerspective

Novel 24-gene signature may predict outcomes in AML

A novel risk classification method that involves a 24-gene signature may predict survival outcomes in acute myeloid leukemia, according to study results.

The researchers aimed to describe a prognostic gene expression signature with the capacity to independently predict survival in patients with AML. Such a signature may improve current risk classification methods.

Four cohorts that included 499 AML patients with a variety of cytogenetic and molecular abnormalities served as the training set. Two independent validation sets included a combined 825 patients.

Different protocols were used to treat patients from different cohorts. Also, various microarray platforms described the gene expression profiles of patients.

Cox regression analysis of the four training sets yielded a prognostic signature composed of 24 genes. Multivariate model results indicated that a higher sum value of the signature independently predicted shorter OS and EFS in the training and validation patient sets (P<.01).

Researchers observed significantly distinct OS and EFS (P<.001) in three new risk groups classified by the integrated risk classification model.

“The integrated risk classification incorporating this gene signature provides a better framework for risk stratification and outcome prediction than the [European LeukemiaNet] classification,” the researchers concluded.


  • The disease label ‘acute myeloid leukemia’ (AML) — though based on clear morphologic criteria — captures a diverse group of biological entities, driven by a variety of genetic and epigenetic abnormalities, and causing a spectrum of responses to standard therapies. Accordingly, an enduring goal has been to sub-classify AML so that specific biological entities can be matched with the best corresponding treatment.

    Toward this end, the WHO classification of AML incorporates biological information such as the presence of recurrent chromosome abnormalities. Gene expression profiles are not part of the WHO classification. Should they be a standard component of the clinicopathologic evaluation of AML? One problem is that gene expression is a continuous variable, unlike translocations or mutations, which are inherently either absent or present.

    Another problem is that RNA, used for gene expression analyses, is not as stable as DNA. This issue, together with the choice of microarray and bioinformatics used for analyses, can produce technical variation in results.

    To address the problem of variability between analyses, Li and colleagues evaluated four independent datasets (with a total of 499 patients) from different groups in the United States and Europe. They identified 24 genes with expression levels that had consistent, robust associations with OS. They then validated this 24-gene expression signature in multivariate analyses.

    This 24-gene expression signature improved prognostic discrimination when added to the European LeukemiaNet score that incorporates age, cytogenetics, antecedent myelodysplastic syndrome/therapy, FLT3 mutation and proportion of bone marrow CD34+ blasts.

    So back to the question: Should gene expression analyses be a standard component in AML evaluation? Unfortunately, not yet. More work needs to be done, beginning with basic technical issues, such as choice of methods used to measure gene expression. Furthermore, prognosis is a function of the particular treatments (or lack thereof) administered.

    Thus, as novel treatments are developed and applied in AML — as opposed to the generally uniform cytarabine-based chemotherapy used today — use of differences in prognoses as clues to biological subtypes of AML could become complicated. Nonetheless, it could very well be that in the future, gene expression analyses will complement DNA-based analyses in the effort to match patients to the best possible treatment for their individual diseases.

    • Yogen Saunthararajah, MD
    • Staff Physician
      Department of Hematologic-Oncology and Blood Disorders
      Taussig Cancer Institute
      Cleveland Clinic