Purpose: To compare long-term outcome using alternative failure definitions after external beam radiation for localized prostate cancer.
Methods and materials: Data from 4839 patients with stage T1b, T1c, and T2 adenocarcinoma of the prostate who were treated solely with external beam radiation between 1986 and 1995 at nine U.S. institutions were analyzed. Outcome using the following prostate-specific antigen (PSA) failure definitions was compared: (1) three consecutive PSA rises backdated (American Society for Therapeutic Radiology and Oncology [ASTRO]), (2) two PSA rises of at least 0.5 ng/mL each, backdated (0.5 x 2), (3) three consecutive PSA rises with failure recorded at the call date (ASTRO call date), (4) PSA > or =current PSA nadir + 2 ng/mL (Houston + 2), (5) PSA > or =current PSA nadir + 3 ng/mL (Houston + 3), (6) PSA >0.2 ng/mL, or (7) PSA >0.5 ng/mL. For definitions 3-7, the failure date was recorded as the date the criterion was met, without backdating.
Results: PSA disease-free survival (PSA-DFS) varied according to the failure definition used with differences of up to 13% with PSA rise definitions and up to 44% with absolute nadir value surgical-type definitions within the first 5 years post-therapy as compared with the ASTRO definition. PSA-DFS was 66%, 66%, 68%, 72%, 15%, and 25% at 5 years postradiation for definitions 2-7, respectively, vs. 59% for the ASTRO definition. Sensitivity and specificity of definitions 2, 4, and 5 were better than for the ASTRO definition, whereas, for definitions 6 and 7, the sensitivity was at least 90% but the specificity was only 9% and 26%, respectively. This analysis shows that the ASTRO definition does not overestimate outcome, particularly in the first 5 years after therapy, as compared with other definitions appropriate to irradiated patients.
Conclusion: There are notable differences in both short- and long-term outcomes after definitive radiation for prostate cancer depending on the failure definition applied. Failure definitions must be tested objectively for sensitivity and specificity in predicting clinical outcome, and it is only in this manner that reasonable choices can be made. Although traditional surgical-type failure definitions do not seem applicable to patients treated with external beam radiation, further analysis of definitions across multiple therapeutic modalities is necessary to determine whether a universal failure definition might be feasible, at least for research and comparative purposes.