Development of prognostic molecular markers in pediatric rhabdomyosarcoma based on gene expression and copy number variations.
9510 Background: Rhabdomyosarcoma (RMS) is the most common pediatric soft tissue sarcoma and comprises two major histological subtypes: alveolar and embryonal. The majority of alveolar tumors harbor PAX/FOXO1 fusion genes. Current patient risk stratification, unlike other pediatric embryonal tumors, does not utilize any molecular data. Therefore, we aimed to improve the risk stratification of RMS patients through the use of molecular biological data. Methods: Two independent data sets of gene expression profiling for 124 and 101 RMS were used to derive prognostic gene signatures by meta-analysis. Genomic array CGH data for 109 RMS was also evaluated to develop a prognostic marker based on copy number variations (CNVs). The performance and usefulness of these derived metagenes and CNVs as well as a previously published metagene signature were evaluated using rigorous leave-one-out cross-validation analyses. Results: The new prognostic gene expression signature, MG15, and one previously published (MG34) (Davicioni. JCO. 2010) performed well with reproducible and significant effects (HR 3.2 [1.7-5.9] p < 0.001 and HR 2.5 [1.5-4.3] p < 0.001, respectively). However, they did not significantly add new prognostic information over the fusion gene status (PAX3/FOXO1, PAX7/FOXO1 and Negative). Similarly, a prognostic CNV marker, although showing HR 2.9 [1.5-5.6] p < 0.01, was also not improving models with fusion gene status. Within fusion negative RMS, the analysis identified prognostic markers based on either gene expression or CNVs and showed significant association with patients outcome (HR 6.3 [1.5-26.3] p ≤ 0.016 and HR 11.2 [2.5-50.7] p < 0.010, respectively). Moreover, these were able to identify distinct risk groups within the COG (Children's Oncology Group) risk categories, which is currently used to guide treatment. Conclusions: Molecular signatures derived using all RMS effectively stratify patients by their risk, but most of their prognostic information is contained in the PAX/FOXO1 fusion gene status which is simpler to assay. New markers developed within the fusion negative population seem improving current RMS risk classifier and should be tested in follow-up studies.