Application of risk-adapted treatments has greatly improved the survival of children and adolescents with acute myeloid leukemia (AML). However, a significant number of patients still suffer from relapse or refractory disease. Early identification of these patients and development of effective novel therapies are paramount steps for improvement of their survival. In this thesis, we investigated the prognostic significance of various genetic aberrations in AML blasts for the outcomes of treatment with recent Berlin-Frankfurt-Münster (AML-BFM) protocols (2004 - 2017). We identified a subgroup of patients with mutations in Wilms tumor 1, frequently co-occurring with internal tandem duplication of fms related receptor tyrosine kinase 3 gene (FLT3-ITD) and the fusion of nucleoporin 98 with nuclear receptor binding SET domain 1 (NUP98-NSD1), that had very poor survival: patients with AML blasts harboring triple mutations or any combination of double mutations all eventually experienced relapse or refractory AML and only one-third of patients could be rescued. As an example of novel targeted salvage therapies, we evaluated the efficacy and safety of the compassionate use of gemtuzumab ozogamicine (GO) in children and adolescents with relapsed/refractory AML. In a cohort of 76 heavily pre-treated patients with relapsed or refractory AML, using GO facilitated hematopoietic stem cell transplantation in 64% of patients, which then led to a 27% probability of 4-year overall survival. In contrast, all patients in the historic control group died. Conclusively, this thesis confirms the potency of specific combinations of genetic/cytogenetic aberrations in AML blasts to predict poor treatment responses already at diagnosis and underlines the failure of our contemporary salvage therapies to rescue these patients. GO is an example for a novel therapeutic substance that needs to be tested in the front-line of the treatment of pediatric patients at high risk for treatment failure.