Defining biochemical failure after radiotherapy with and without androgen deprivation for prostate cancer

Purpose: To compare several characteristics of alternative definitions of biochemical failure (BF) in men with extended follow-up after radiotherapy (RT) with or with androgen deprivation therapy (ADT) for prostate cancer. Methods and Materials: From December 1, 1991, to April 30, 1998, 688 men with Stage T1c-T3NX-N0M0 prostate cancer received RT alone (n = 586) or RT plus ADT (n = 102) with a minimal follow-up of 4 years and five or more "ADT-free" posttreatment prostate-specific antigen levels. BF was defined by three methods: (1) the ASTRO definition (three consecutive rises in prostate-specific antigen level); (2) a modified American Society for Therapeutic Radiology Oncology (ASTRO) definition requiring two additional consecutive rises when a decline immediately subsequent to three consecutive rises occurred; and (3) the "Houston" or nadir plus 2-ng/mL definition (a rise of at least 2 ng/mL greater than the nadir). The sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy were determined for each using clinical progression as the endpoint. Furthermore, the misclassification rates for a steadily rising prostate-specific antigen level, ability to satisfy the proportional hazards (RT with or without ADT), effects of short follow-up, and intervals to the diagnosis of BF were compared. Results: The misclassification rate for BF using the nadir plus 2-ng/mL definition was 2% for RT alone and 0% for RT plus ADT compared with 0% and 0% for the modified ASTRO definition, and 5% and 23% for the ASTRO definition, respectively. The hazard rates for RT alone and RT plus ADT were proportional only for the nadir plus 2 ng/mL definition and seemingly unaffected by the length of follow-up. For RT with or without ADT, the nadir plus 2 ng/mL definition was the most specific (RT, 80% vs. RT plus ADT, 75%) with the greatest positive predictive value (RT, 36% vs. RT plus ADT, 25%) and overall accuracy (RT, 81% vs. RT plus ADT, 77%). A greater proportion of BF was diagnosed in the first 2 years of follow-up with the nadir plus 2 ng/mL definition compared with the ASTRO definition (13% vs. 5%, p = 0.0138, chi-square test). Conclusion: The nadir plus 2 ng/mL definition was the best predictor of sustained, true, biochemical, and clinical failure, and was not affected by the use of ADT or follow-up length.