The role of imaging in screening for prostate cancer. A decision analysis perspective

Using Bayesian and decision analytical concepts, we can define the ideal characteristics of any screening test for cancer of the prostate: high specificity, reasonably high sensitivity, noninvasiveness, low cost, and low interobserver variability. Computed tomography (CT) fails as a test, since it cannot show the internal structure of the prostate; MRI is too expensive and has an unacceptably poor specificity. Transrectal sonography does have many desirable characteristics, including relatively high sensitivity and, if no biopsy is done, low cost and noninvasiveness. But it has an unacceptably low specificity for early, clinically significant lesions: over half of all patients tested will have a positive result, requiring a confirming biopsy, which means that transrectal sonography will in the end be too invasive and too expensive. Therefore, we cannot recommend transrectal sonography as a primary screening tool for cancer of the prostate at this time.