To the Editor.
—In their study of the validity of PSA in detecting prostate cancer, Dr Gann and colleagues1 provide valuable information reducing spectrum bias by doing a nested case-control study from a cohort of men followed for at least 10 years and stratifying the results by aggressive and nonaggressive tumors. Unfortunately, the authors report their results as sensitivity and specificity, necessitating dichotomous results, or odds ratios at several levels of PSA. The validity of diagnostic or screening tests are better determined by likelihood ratios (LRs), which allow clinicians to ascertain whether the probability of prostate cancer is revised downward (if the LR is <1.0) or upward (if the LR is >1.0), with the magnitude of the revision determined by the numerical value of the LR.2Using the method of Peirce and Cornell,3 I calculated the LRs for PSA in detecting aggressive cancers based on the seven
Peirce JC. The Utility of Prostate-Specific Antigen for Detecting Prostate Cancer. JAMA. 1995;274(8):608. doi:10.1001/jama.1995.03530080023023