Survival modelling of relapse-free survival and competing-risk analysis in patients with high-risk localized prostate cancer treated in GETUG-12

Abstract

Background: Androgen-deprivation therapy (ADT) plus radiotherapy is a standard of care for men with high-risk localized prostate cancer (PC). Adding docetaxel in this setting demonstrated better relapse-free survival (RFS) but not survival. Few data are available on relapse patterns. Our aim was to investigate whether different patterns of relapses exist in men with high-risk localized PC.

Patients and methods: Prospective data from the GETUG12 phase 3 randomized controlled trial comparing ADT alone vs ADT + docetaxel and estramustine (DE) was examined. RFS (main endpoint) was analysed using parametric survival models and second event-free survival (SEFS) defined from biochemical progression (BP) was analysed using a competing-risk approach.

Results: Overall, 413 patients were randomized from 2002 to 2006, 206 treated with ADT and 207 with ADT+DE, in addition to local treatment. First analysis showed that the piecewise-exponential model with two time-intervals ([0-3[; ≥3 years) was the model that best characterized RFS with no time-dependent effect of risk factors and treatment. Second analyses limited to patients with a BP showed that the risk of a second event was significantly lower in patients with a longer time-interval from randomization to BP (hazard ratio (HR_{≥ 3 versus < 3 years}): 0.49 [95% CI 0.30-0.79]). In competing-risk analysis, a significant and protective effect of this time-interval was observed for metastases (sub-HR_{≥ 3 versus < 3 years}: 0.41 [0.23-0.73] accounting for salvage treatment.

Conclusion: Time-interval from randomization to BP is a major prognostic factor for developing local or distant recurrences for patients with high-risk localized PC.

Keywords: Biochemical progression; GETUG-12; High-risk localized prostate cancer; Relapse-free survival; Survival modelling.