The Impact of Endpoint Definitions on Predictors of Progression in Active Surveillance for Early Prostate Cancer

Background/Objectives: There is conflicting data on which factors predict progression events in active surveillance for early prostate cancer. Here we explored the value of different clinicopathological variables and if progression endpoint definitions impacted predictive utility. Methods: Clinicopathological variables were extracted from the STRATified CANcer Surveillance (STRATCANs) prospective AS database, and included biopsy features (core positivity, cancer core length and percentage core involvement) and MRI features (Likert score, lesion size and location) as well as baseline PSA density [PSAd] and Cambridge Prognostic Group (CPG). These were tested against AS endpoint defi-nitions of (1) progression to ≥CPG3 (2) any pathological progression and two definitions from the literature: (3) ≥GG3 or change to treatment, and (4) ≥GG4, metastasis or can-cer-related mortality. Predictors were assessed using regression analysis. Results: Data from 296 men were included (median age 66; follow-up 5 years). Progression per defi-nition (1 – 4) occurred in 46 (15.5%), 54 (18.2%) 84 (28.4%) and 10 (3.4%) men, respectively. In univariate analysis using definition 1, no biopsy parameter was independently pre-dictive of progression while MRI Likert score (p=0.02) was the only significant imaging parameter. For definition 2 core positivity (p=0.003) and MRI Likert score (p=0.01) were significant predictors in univariate analyses. While for definition 3, tumour core length (p=0.005), core positivity (p=0.002) and MRI Likert score (p=0.003) were all predictive in univariate analyses. In multivariate analysis however the only consistent independent predictor was PSAd, regardless of endpoint definition. No variables predicted definition 4 progression. Conclusions: AS endpoint selection appears to define which variables predict progression. Using progression to ≥CPG 3 as an unambiguous AS endpoint neither biopsy nor MRI variables added incremental value in predicting progression. PSAd however appears to be a robust and independent generalisable progression pre-dictor.