Info-gap decision theory and its potential applications in the clinic

Personalized medicine seeks to match innate characteristics of patients with potential interventions in order to secure the most favorable outcome while mitigating risk. Personalized medicine does not supplant the decisions made by health care professionals in conjunction with patients. In fact, ideally, the application of personalized medicine greatly augments the quality of that interaction by adding an objective measure of disease and response to the physician's professional judgment. In that context, the practitioner faces a disparity between what is known and what needs to be known in order to realize the desired goal of a favorable outcome of treatment or an effective avoidance of disease or complication. Info-gap theory can help manage this challenge. Physician and patient both wish to choose a course of action with confidence that the consequence will be satisfactory to the patient. However, these two attributes of a treatment - success and reliability - are different from each other and conflict with each other. Ambitious goals may be risky. The crux of the challenge is uncertainty: fragmentary data and imperfect understanding. Choices that exhaustively exploit uncertain knowledge may tend to fall short of the quality of outcome which is predicted by that knowledge, precisely because some of that knowledge is imprecise. On the other hand, better-than-anticipated outcomes are also possible when our knowledge is imperfect. Treatments must be selected not only according to their predicted outcomes, but also according to the immunity of those outcomes to errors in the knowledge underlying the predictions, and according to the opportunities inherent in the uncertainty. Info-gap decision theory [1] provides a tool for evaluating both the robustness against pernicious uncertainty as well as the opportuneness from propitious uncertainty. Info-gap theory is a methodology for decision under severe uncertainty which has been applied in many disciplines [101].