Although a number of susceptibility loci for neuroblastoma (NB) have been identified by genome‐wide association studies, it is still unclear whether variants in the HLA region contribute to NB susceptibility. In this study, we conducted a comprehensive genetic analysis of variants in the HLA region among 724 NB patients and 2863 matched controls from different cohorts. We exploited whole‐exome sequencing data to accurately type HLA alleles with an ensemble approach on the results from three different typing tools, and carried out rigorous sample quality control to ensure a fine‐scale ancestry matching. The frequencies of common HLA alleles were compared between cases and controls by logistic regression under additive and non‐additive models. Population stratification was taken into account adjusting for ancestry‐informative principal components. We detected significant HLA associations with NB. In particular, HLA‐DQB1*05:02 (OR = 1.61; padj = 5.4 × 10−3) and HLA‐DRB1*16:01 (OR = 1.60; padj = 2.3 × 10−2) alleles were associated to higher risk of developing NB. Conditional analysis highlighted the HLA‐DQB1*05:02 allele and its residue Ser57 as key to this association. DQB1*05:02 allele was not associated to clinical features worse outcomes in the NB cohort. Nevertheless, a risk score derived from the allelic combinations of five HLA variants showed a substantial predictive value for patient survival (HR = 1.53; p = 0.032) that was independent from established NB prognostic factors. Our study leveraged powerful computational methods to explore WES data and HLA variants and to reveal complex genetic associations. Further studies are needed to validate the mechanisms of these interactions that contribute to the multifaceted pattern of factors underlying the disease initiation and progression.