The criteria for active surveillance and focal therapy in Gleason score (GS) 3 + 4 disease remain unclear. We aimed to refine them by evaluating predictors of adverse pathology and low-volume disease at radical prostatectomy (RP).
We retrospectively analyzed 1,665 men with biopsy GS 3 + 4 disease who underwent RP. Associations of biochemical recurrence (>0.2 ng/ml) and three definitions of adverse pathology were evaluated using Cox models: (1) GS ≥4 + 3 or pT3a-4, (2) GS ≥4 + 3 or pT3b-4, and (3) GS ≥4 + 4 or pT3a-4. The definition with the highest C-index was selected as the primary endpoint. Preoperative clinicopathological and MRI predictors of the selected adverse pathology were identified using multivariable logistic regression. With the same predictors, tumor diameter and volume at RP were assessed. The low-risk criteria were defined as a ≤ 20% predicted risk of the selected adverse pathology.
Definition 2 yielded the highest C-index (0.76; 95% CI, 0.71-0.81). Independent predictors were age, Black race, PSA, cT stage, maximum involvement of biopsy core (%), and MRI-detected extracapsular extension or seminal vesicle invasion. They were consistently associated with larger tumor diameter and volume, except for age and race. The low-risk criteria were defined as PSA ≤5 ng/ml, cT1 stage, absence of extracapsular extension or seminal vesicle invasion, and cancer involvement ≤30% of any core. The limitation was surgical selection bias.
Four preoperative predictors, supported by pathologically confirmed low-volume disease, provide a practical framework to refine selection for active surveillance and focal therapy in GS 3 + 4 disease. External prospective validation is warranted.
Urologic oncology. 2026 May 22 [Epub ahead of print]
Yu Ozawa, Marcio Covas Moschovas, Marco Sandri, Rohan Sharma, Shady Saikali, Ari Diamond, Travis Rogers, Vipul Patel
AdventHealth Global Robotics Institute, Celebration, FL. Electronic address: ., AdventHealth Global Robotics Institute, Celebration, FL; Urology Department, University of Central Florida (UCF), Orlando, FL., Big and Open Data Innovation Laboratory (BODaI-Lab) and Data Methods and Systems Statistical, University of Brescia, Brescia, Italy., AdventHealth Global Robotics Institute, Celebration, FL.