Using Patient Preferences to Determine Noninferiority Margins in Trials.