Survival impact of initial local therapy selection for men under 60 with high risk prostate cancer.

Abstract

5082 Background: The impact of initial local therapy selection on survival for high risk prostate cancer (PCa) patients remains uncertain. We sought to assess this effect, while limiting competing causes of death, through the examination of a younger PCa patient cohort within the National Cancer Database. Methods: We evaluated the overall survival (OS) of men under 60 with high risk PCa receiving either radiation therapy (RT) or radical prostatectomy (RP). All men in this age group were treated between 2004 and 2013, harbored cN0M0 disease, and presented with Gleason Scores (GS) of 8 to 10. The RT group included patients who received external beam radiation (EBRT) alone or EBRT in combination with brachytherapy (BT). Overall survival and covariates were evaluated using multivariable Cox regression analysis. Results: A total of 16,944 patients met inclusion criteria of which 12,155 underwent RP and 4,789 received RT as initial therapy. 82.5% of RT patients received hormonal therapy, and the median dose was 77.4 Gy. In the RP group, 17.2% of patients received postoperative radiation, and 87% of these cases received a dose exceeding 64.80 Gy. The RP group had a higher proportion of cases with Charlson-Deyo comorbidity score > 0 (15.2% v. 11.2%, p < 0.000001). At a median follow-up of 50 months (0 - 131 months), RP was associated with improved OS in comparison to RT (hazard ratio = 0.52; 95% CI (0.47, 0.58); p < 0.000001). The estimated 8-year OS (±1 standard error of the estimate) was 85.1±0.7% and 74.9±0.7%, after RP and RT, respectively. This benefit remained present when adjusting for age, year of treatment, race, comorbidity score, Gleason score, T stage, hormonal therapy, chemotherapy, form of radiation, PSA, or insurance status. Conclusions: Compared to RT, initial treatment of men under 60 with high risk PCa with RP results in a large, statistically significant improvement in overall survival that remains consistent over time and remains significant in a multivariable model adjusting for known prognostic variables. These results are limited by the retrospective nature of the database analysis, and the lack of cancer specific survival information.