Introduction: Hodgkin lymphoma (HL) is a common pediatric malignancy, and although considered highly curable, treatment success comes at a high cost in the form of long-term toxicity and morbidity. To overcome this challenge clinical trials have evaluated risk-adapted treatment regimens aiming to maintain high survival rates and reduce treatment-related toxicity. However, risk stratification is currently limited to the use of clinical factors as there are no validated molecular biomarkers that can be employed as predictors for treatment outcome in pediatric HL. Therefore, we aimed to perform gene expression profiling (GEP) to uncover disease biology underlying treatment response and develop a prognostic model to tailor first-line therapy in pediatric HL. Methods: We selected specimens from patients enrolled in a phase 3 clinical trial (AHOD0031) of the Children’s Oncology Group (COG) randomizing patients dependent on CT imaging-based early response criteria (slow early response, SER; rapid early response, RER). We performed intermediate density GEP (784 genes) using NanoString on RNA extracted from pre-treatment formalin-fixed, paraffin-embedded tissue biopsies. Results: Of the 206 tissue samples obtained, 185 (89.8%) passed quality assurance testing. We applied our previously published 23-gene predictor - developed to predict OS in adult HL patients - to the pediatric cohort. This assay failed to predict outcomes, with patients in the “high-risk” group, as assigned by the assay, trending to have superior outcomes than the “low risk” patients. Therefore, we sought to develop a novel EFS predictive model for pediatric patients. Using penalized Cox regression, we developed a 16-gene model to predict EFS in a training cohort. This model was applied to an independent cohort of 117 specimens from the same clinical trial which were enriched for treatment failure (no EFS event to EFS event ratio = 1:1). Using this validation cohort, the 16-gene model separated high-risk and low-risk groups with significantly different EFS in the SER (p=0.034), but not in the RER group (p=0.81). Conclusions: Failure of the GEP-based model developed in adult HL suggests distinct biology underlies treatment failure in the pediatric age group, although differences in therapy may also be a contributing factor. We describe the development of a novel predictive model for EFS in intermediate-risk pediatric HL patients which successfully risk-stratifies patients in the SER subgroup.