Adolescent and young adult survivors of pediatric brain tumors often live with long-term neuropsychological deficits, found to be related to functional and structural brain changes associated with the tumor itself as well as treatments such as radiation therapy. The importance of brain networks has become a central focus of research over recent decades across neurological populations. Graph theory is one way of analyzing network properties that can describe the integration, segregation, and other aspects of network organization. The existing literature using graph theory with survivors of brain tumor is small and inconsistent; therefore, more work is needed, particularly in survivors of pediatric brain tumors. The present study used graph theory to determine whether functional network properties in this population differ from healthy controls; whether graph metrics relate to core cognitive skills: attention, working memory, and processing speed; and whether they relate to a cumulative measure of neurological risk. 31 survivors and 31 matched controls completed neuropsychological testing and functional magnetic resonance imaging. Neuroimaging was preprocessed and spatially constrained ICA was completed, followed by calculation of area under the curve values of graph metrics. Results revealed a significant difference such that brain tumor survivors exhibited less small-world properties in their functional brain networks. This was found to be related to working memory, such that less small-worldness in the network was related to poorer performance. There were no significant relationships with neurological risk, but there were nonsignificant correlations of small-moderate effect size such that lower global efficiency and clustering coefficient were associated with greater neurological risk. Comparisons to structural network analysis from a similar sample and additional post-hoc analyses are also discussed. These findings reveal that survivors of pediatric brain tumor indeed display significant differences in functional brain networks that are quantifiable by graph theory. It is also possible that with further work, we might better understand how metrics such as small-worldness can be used to predict long-term cognitive outcomes and functional independence in adulthood. This would ideally allow neuroimaging to play a part in determining the most efficient and impactful allocation of intervention resources among pediatric brain tumor patients.