Accuracy of biopsy Gleason scores from a large uropathology laboratory: use of a diagnostic protocol to minimize observer variability.

OBJECTIVES To examine the correlation of biopsy Gleason scores with radical prostatectomy specimens from a laboratory that uses protocols designed to minimize observer variability. This protocol mandates consensus case review of all nonbenign cases. METHODS Between August 24, 1993 and June 26, 1997, 106 patients who underwent radical prostatectomy at Johns Hopkins Hospital, Baltimore, Maryland had their prostate cancer diagnosed and graded at one laboratory (DIANON Systems). We analyzed the Gleason scores from the biopsy and radical prostatectomy specimens. RESULTS Exact correlation existed between biopsy and radical prostatectomy Gleason scores for 72 (68%) cases; 103 (97%) correlated within 1 grade, all cases correlated within 2 grades; 26 (25%) biopsies were undergraded and 8 (8%) were overgraded. Positive predictive values for biopsy Gleason scores 5, 6, and 7 were 66%, 67%, and 71%, respectively. Grouped Gleason scores (well differentiated [2 to 4], moderately differentiated [5, 6], moderately to poorly differentiated [7], and poorly differentiated [8 to 10]) correlated exactly for 74 (70%) cases and within 1 group for all cases. Patient age, digital rectal examination results, total number of positive cores, and maximum percentage of tumor on biopsy did not affect correlation, but prostate-specific antigen (PSA) levels did affect correlation (exact correlation 96% when the PSA level was less than 5 ng/mL; 50% when the PSA level was 11 ng/mL or greater, P <0.01). CONCLUSIONS The combination of experience and the protocol described minimizes intra- and interobserver variability, thereby improving the predictive value of biopsy Gleason grading. Biopsy and radical prostatectomy Gleason scores correlate more poorly when the PSA level is high (11 ng/mL or greater) than when the PSA level is low (less than 5 ng/mL).