Optimal treatment levels in cancer therapy

A two‐part method is proposed for optimizing treatment levels in cancer therapy. The first part makes use of a treatment characteristic (TC) curve as a graphical method for summarizing the responses of normal and tumor tissue to therapy. For treatment responses that are cumulative normal, the TC curve, formed by plotting the probit probability of ablation against the probit probability of complication, is a straight line whose slope is the ratio of standard deviations of the two responses and whose position measures the relative sensitivity of the cancer and normal tissue. TC lines are completely defined by studying a homogeneous group of patients at two treatment levels within the same protocol. In the second part of the method, all possible outcomes of therapy are assigned costs or weights. This assignment converts the probit TC line into an expected loss function which estimates the effectiveness of therapy. The minimum of this function locates the optimum probabilities of ablation and complication and is the best therapy, given the therapist's cost structure, the protocol, and the class of patients being studied.