RESUMO
PURPOSE: Patients with clinically localized, high-risk prostate cancer are often treated with surgery, but exhibit variable prognosis requiring long-term monitoring. An ongoing challenge for such patients is developing optimal strategies and biomarkers capable of differentiating between men at risk of early recurrence (<3 years) that will benefit from adjuvant therapies and men at risk of late recurrence (>5 years) who will benefit from long-term monitoring and/or salvage therapies. PATIENTS AND METHODS: DNA methylation changes for 12 genes associated with disease progression were analyzed in 453 prostate tumors. A 4-gene prognostic model (4-G model) for biochemical recurrence (BCR) was derived utilizing LASSO from Cohort 1 (nâ¯=â¯254) and validated in Cohort 2 (nâ¯=â¯199). Subsequently, the 4-G model was evaluated for its association with salvage radiotherapy (RT) and/or hormone therapy, and the additive potential to CAPRA-S to develop an integrative gene model was assessed. RESULTS: The 4-G model was significantly associated with BCR in both cohorts (chi-squared analysis P≤ 0.004) and specifically, with late recurrence at 5+ years (P < 0.001, Cohort 1; P= 0.028, Cohort 2). Multivariable Cox proportional regression analysis identified the 4-G model as significantly associated with salvage RT or hormone therapy in Cohort 1 (hazard ratio (HR) 1.64, 95% confidence interval (CI) 1.29-2.10, P< 0.001) and further validated in Cohort 2 (HR 1.63, 95% CI 1.18-2.25, P< 0.001). The integrative model outperformed prostate-specific antigen and the 4-G model alone for predicting BCR and was associated with patients who received hormone therapy 3+ years postsurgery. CONCLUSIONS: We have identified and validated a novel integrative gene model as an independent prognosticator of BCR and demonstrated its association with late BCR. These patients require more long-term postsurgical monitoring and could be spared the comorbidities of adjuvant therapies.
