Using both genomic and clinical data, researchers have developed a new prognostic framework to predict outcomes, including mortality, in patients with acute myeloid leukemia (AML), according to a recent study conducted by Moritz Gerstung, PhD, and colleagues from the European Bioinformatics Institute EMBL-EBI and Wellcome Trust Sanger Institute in Cambridge, United Kingdom. Dr. Gertsung presented the results at the 2015 ASH Annual Meeting.
“Our approach is the first attempt to predict outcomes based on a comprehensive set of prognostic variables,” Dr. Gerstung said, “unlike previous approaches that force patients into a small set of categories, such as high, intermediate, and low-risk based on only a few markers.”
The impact of genomic risk factors on patient outcomes can sometimes be difficult to determine, particularly in the setting of co-occurring genomic lesions. The tool developed and tested by the researchers recognizes mutations in 55 prognostically relevant genes along with 18 of the most common cytogenic legions. It also takes into account blood counts, patient age, and gender.
The authors used detailed diagnostic, treatment, and survival data gathered from 1,540 patients with AML who were enrolled in three trials from the German-Austrian AML Study Group.
Through systematic evaluation of risk-modeling strategies, the researchers identified factors associated either positively or negatively with mortality, using a combination of “many small and few large” variables, including:
- Fusion genes generated by t(15;17), inv(16), and inv(3) rearrangements
- Complex karyotype
- trisomy 21
“Multiple risk factors act mostly additively,” the authors reported, “with the exception of gene-gene interaction terms, including NPM1:FLT3ITD:DNMT3A (n=93; hazard ratio [HR] = 1.50; p<0.03).”
“Additionally, our data show that almost all patients have multiple genomic lesions, and they occur in unique constellations,” Dr. Gertsung said, adding that he believes this finding should prompt clinicians to move toward prognostic schemes that account for multiple co-occurring factors.
Overall, about two-thirds of predicted inter-patient risk variation was related to genomic factors (i.e., balanced rearrangements, copy number changes, and point mutations). The remaining one-third could be attributed primarily to demographic variables, treatment, and diagnostic blood counts. “A large share, but not all, prognostic information seems to be determined by genomic factors,” the authors wrote.
The researchers also tested whether these models could accurately predict the probability of survival and mortality during induction therapy, first complete remission, and after relapse.
“The resulting personalized predictions provide a quantitative risk assessment and allow evaluating the effect of treatment decisions such as allogeneic hematopoietic cell transplantation versus standard chemotherapy in first complete remission,” the authors concluded.
This study acts as proof of principle of this prediction framework but, before the framework could be used as a clinical decision support tool, it needs to be tested in larger cohorts of AML patients prospectively.
Gerstung M, Papaemmanuil E, Martincorena I, et al. Personally tailored risk prediction of AML based on comprehensive genomic and clinical data. Abstract #85. Presented at the American Society of Hematology’s Annual Meeting, December 5, 2015; Orlando, FL.