Assessing the impact of an alternative biochemical failure definition on radiation dose response for high-risk prostate cancer treated with external beam radiotherapy

Purpose The American Society for Therapeutic Radiology and Oncology (ASTRO) biochemical failure definition has recently been compared with various alternative definitions. We assessed the effect of using an alternative failure definition on the dose-response characteristics of high-risk prostate cancer treated with radiotherapy alone. Methods and materials This study included 363 high-risk prostate cancer patients treated with external beam radiotherapy alone from 1987 to 1999. These patients have one or more of the following: 1992 American Joint Committee on Cancer (AJCC) digital rectal examination (DRE) stage ≥ cT3, prostate-specific antigen (PSA) > 20 ng/mL, and/or biopsy Gleason score ≥ 8. We previously reported the dose response based on the ASTRO definition for these patients. In this study, a biochemical failure is defined as a PSA rise ≥ 2 ng/mL above the current nadir PSA (CN + 2). The failure date is defined as the time at which the event occurred (i.e., the call date). Results Using CN + 2, the tumor control probability (TCP) continues to decrease with time as opposed to reaching a plateau as with the ASTRO definition. At 5 years, TCD50 (95% CI), the dose to achieve 50% tumor control, for high-risk prostate cancer, is 70.4 (68.0-72.9) Gy using CN + 2 [ASTRO: 75.5 (70.7-80.2) Gy]. The relative slope, γ50 (95% CI) is 1.8 (0.8-2.8) [ASTRO: 1.7 (0.7-2.7)]. Recursive partitioning again identified two subgroups: PSA < vs. ≥ 13 ng/mL (ASTRO: PSA ≤ vs. > 20 ng/mL). The difference in TCD50 between the two subgroups is about 20 Gy at 5 years (ASTRO: about 15 Gy at 5 years). Conclusion This analysis using the CN + 2 failure definition continues to show a dose response for the high-risk group of patients. However, the dose-response characteristics differ from those estimated using the ASTRO definition. We observed that the position (TCD50) and steepness (γ50) of the dose-response curve changed with time as long as the TCP continued to decrease. This suggests that the dose response characteristics derived from data with longer follow-up may be different from those derived with shorter follow-up using the CN + 2 or similar failure definitions which do not back-date the failure. These changes in dose-response characteristics as well as the time dependence of dose response should be noted when investigators design dose escalation trials for the high-risk prostate cancer patients.