A method is presented by which dose distributions in radiotherapy may be judged. This method, based on statistical decision theory, combines the calculated probabilities of radiation induced complications (including failure to control disease) with the therapist's judgment of the morbidity of each complication to yield a single value representing the clinical utility of the dose distribution. Using this figure of merit, alternative dose distributions can be compared on a clinical basis. If the morbidities associated with each injury are satisfactorily estimated, and the dose response parameters are adjusted to match clinical experience, this method can be used to evaluate novel treatment techniques prior to their implementation. Automatic optimization algorithms can be used to maximize the figure of merit as a function of the physical treatment parameters so as to provide a statistically optimal treatment.