Interobserver variability in histologic evaluation of radical prostatectomy between central and local pathologists: findings of TAX 3501 multinational clinical trial

Abstract

Objectives: To determine the agreement between the local pathologist findings and central pathologist findings using data from the TAX 3501 trial. TAX 3501 was a randomized, multinational trial comparing the outcomes of patients with high-risk prostate cancer treated with androgen deprivation with or without docetaxel after radical prostatectomy (RP). Patient eligibility was determined by a minimal 5-year progression-free survival estimate of 60% using Kattan's nomogram.

Methods: The pathologic findings were reassessed in 257 consecutive RP specimens by 2 central pathologists and compared with the local pathologist data.

Results: For the Gleason score, agreement was found in 181 (70%) of 257 cases, upgrading in 57 (75%), and downgrading in 25% of the RP specimens The most frequent upgrade was from Gleason score 7 to 8 or 9 and downgrading from Gleason score 8 to 7. Of the upgrades and downgrades, 37% and 21% were of 2 Gleason score points, respectively. For the tumor extent, agreement was found in 179 (70%) of 256 specimens, with upstaging in 70 (91%) and downstaging in 9%. The most frequent upstage was from focal to extensive extraprostatic extension (45%). For seminal vesicle invasion, agreement was found for 238 (93%) of 256 RP specimens Almost equal rates of underdiagnosing and overdiagnosing seminal vesicle invasion was observed. For margin status, agreement was present for 229 (89%) of 256 cases. The central pathologist review led to reclassification as a positive margin in 17 cases and a negative margin in 10. For lymph node status, 2 (1%) of 210 RP specimens had positive nodes identified only by the central pathologist. Agreement was observed in 154 negative and 54 positive cases.

Conclusions: Significant interobserver variations were found between the central and local pathologists. From the central pathologist review, the progression-free survival estimates were altered in 31 patients (13%), including 22 who were reassigned a greater risk estimate, rendering them study eligible. Thus, interobserver variability affected prognostication and trial accrual.