Radiation Therapy treatment requires selection of a dose that balances the increased efficacy (tumor control) with the increased toxicity typically associated with higher dose. Historically, a single dose has been selected for a population of patients. However, this dose is not universally optimal as evidenced by patients who have tumor progression without toxicity (potentially underdosed) or toxicity with no tumor progression (potentially overdosed). Biomarkers offer the possibility of individualized dose selection. We propose a utility based approach to identifying an optimal dose for an individual patient based on biomarkers and clinical factors. First, the numeric utility or “desirability” of each possible bivariate (toxicity and efficacy) outcome is elicited. The optimal dose is defined as the dose which maximizes the expected utility for an individual patient. For binary outcomes, the expected utility can be expressed as the probability of efficacy minus a weighted sum of toxicity probabilities which can each be estimated from separate statistical models as a function of dose, clinical factors and biomarkers. This approach is shown to provide optimal efficacy for any fixed rate of toxicity. Intuitively, this is because it “spends” its toxicity risk (from higher dose) in patients who derive the largest benefit in efficacy. We illustrate this approach using an 226example from radiation therapy treatment of lung cancer and also present results for a virtual “in-silico” trial comparing standard to utility based treatment.