Methodologic considerations in comparing imaging methods.

Current methods for evaluating and comparing imaging methods may be inadequate in several important aspects. Prospective investigations often fail to provide uniform conditions for data collection due to variable physician skills in performing the studies being evaluated. The double-blind format, although seemingly objective, is inherently unable to prevent the effects of examiner or observer prejudice when imaging methods are being compared. Commonly used statistical terms are limited in their ability to characterize the clinical efficacy of imaging methods, and are easily misused. Reference examinations, or "gold standards," may be used in a manner preordaining an inferior result for the diagnostic method under evaluation. These problems are discussed and examples of their effects are presented. Suggestions are presented for minimizing existing methodologic limitations.