Glioblastomas make up 50% of all primary brain tumors and, despite aggressive treatment, result in the worst prognosis with a median survival time of 14 months. The disease is characterized by diffuse invasion and heterogeneity, presenting a difficult clinical challenge. The standard of care includes spatially uniform doses with wide margins due to uncertainty in subclinical disease location, potentially underdosing volumes of high tumor cellularity and overdosing regions with minimal disease. Patient-specific proliferation, invasion and response to radiation therapy can be predicted by a reaction-diffusion mathematical model. We optimized dose plans subject to clinically relevant decision criteria by integrating the mathematical model with a multi-objective evolutionary algorithm (MOEA) for intensity-modulated radiation treatment (IMRT). The integrated model generates adaptive, customized plans that take into account tumor-specific growth and response kinetics, improving treatment efficacy and reducing unnecessary dose to normal tissue. Simulations of tumor growth and radiation therapy with both patient-specific optimized plans and standard-of-care plans were run for a cohort of 11 actual patients. Results were evaluated in terms of volume weighted equivalent uniform dose (EUD) to critical areas, therapeutic ratio: the ratio of tumor EUD to normal brain EUD, and virtual evaluation of cancer treatment response (VECTR) scores. The optimized plans resulted in significantly decreased normal tissue EUD and increased therapeutic ratio and VECTR scores across all patients. The normal tissue EUD and therapeutic ratio, as well as the improvements due to the optimization were linearly correlated with patient-specific tumor invasiveness, a relationship that is robust to uncertainty in measured tumor volume and model parameters with p values < 0.005. Biologically informed, patient-specific radiation therapy plan design can improve the efficacy of radiation therapy and potentially reduce complications across a variety of patients, supporting integration of such approaches into clinical trials.