Rule-based versus probabilistic selection for active surveillance using three definitions of insignificant prostate cancer

To study whether probabilistic selection by the use of a nomogram could improve patient selection for active surveillance (AS) compared to the various sets of rule-based AS inclusion criteria currently used. We studied Dutch and Swedish patients participating in the European Randomized study of Screening for Prostate Cancer (ERSPC). We explored which men who were initially diagnosed with cT1-2, Gleason 6 (Gleason pattern a parts per thousand currency sign3 + 3) had histopathological indolent PCa at RP [defined as pT2, Gleason pattern a parts per thousand currency sign3 and tumour volume (TV) a parts per thousand currency sign0.5 or TV a parts per thousand currency sign 1.3 ml, and TV no part of criteria (NoTV)]. Rule-based selection was according to the Prostate cancer Research International: Active Surveillance (PRIAS), Klotz, and Johns Hopkins criteria. An existing nomogram to define probability-based selection for AS was refitted for the TV1.3 and NoTV indolent PCa definitions. 619 of 864 men undergoing RP had cT1-2, Gleason 6 disease at diagnosis and were analysed. Median follow-up was 8.9 years. 229 (37 %), 356 (58 %), and 410 (66 %) fulfilled the TV0.5, TV1.3, and NoTV indolent PCa criteria at RP. Discriminating between indolent and significant disease according to area under the curve (AUC) was: TV0.5: 0.658 (PRIAS), 0.523 (Klotz), 0.642 (Hopkins), 0.685 (nomogram). TV1.3: 0.630 (PRIAS), 0.550 (Klotz), 0.615 (Hopkins), 0.646 (nomogram). NoTV: 0.603 (PRIAS), 0.530 (Klotz), 0.589 (Hopkins), 0.608 (nomogram). The performance of a nomogram, the Johns Hopkins, and PRIAS rule-based criteria are comparable. Because the nomogram allows individual trade-offs, it could be a good alternative to rigid rule-based criteria.

• View at Publisher
• View at Pubmed

Free Full Text ( Final Version , 409kb )