Scrutiny of the ASTRO consensus definition of biochemical failure in irradiated prostate cancer patients demonstrates its usefulness and robustness. American Society for Therapeutic Radiology and Oncology.

PURPOSE

The goals of this study are: (1) to establish the robustness of the Fox Chase Cancer Center (FCCC) and the American Society for Therapeutic Radiology and Oncology (ASTRO) consensus definitions of failure by comparing biochemical estimates under various modifications of the censoring and failure time components to their respective unaltered definitions; (2) to isolate the source of variation between the two definitions of failure; and (3) to describe the hazard of failure over time for each definition.

METHODS

Between May 1989 and May 1997, 670 men were treated at Fox Chase Cancer Center for localized prostate cancer using three-dimensional conformal radiation therapy (3DCRT). These men were stratified into three groups for analysis: 111 men treated with adjuvant hormones; 204 men treated with radiation therapy alone and presenting with more favorable prognosis tumor characteristics; 255 men treated with radiation therapy alone and presenting with less favorable prognosis tumor characteristics. For each group, biochemical failure was estimated and compared using the FCCC and ASTRO definitions of failure. The robustness of each definition was evaluated by comparing estimates under the definition as stated to those under various modifications of the censoring and failure components. Analyses were also performed while excluding slow-progressing patients. To isolate the source of variation between the two failure definitions, estimates were compared for patients with agreement in failure status. Estimates of biochemical failure, and thus hazard rates, were made using Kaplan-Meier methodology.

RESULTS

ASTRO biochemical failure estimates were higher than the FCCC failure estimates in the first 5 years post-treatment. Beyond 5 years, ASTRO estimates level off, while the FCCC failure estimates continued to increase. These failure patterns were similar in all patient groups; however, patients treated with adjuvant hormones had a much higher risk of failure immediately following treatment under the ASTRO definition. Modifying the censoring pattern had little effect on failure estimates in any patient group, regardless of definition used. The exclusion of patients with slow prostate-specific antigen (PSA) doubling time did not result in biochemical estimates that differed significantly from those for all patients. The analysis of patients with agreement in failure status continued to demonstrate significant differences in estimates between the two definitions, and thus differences may be attributed to the specification of time to failure. For all patient groups, hazard rates were dependent upon failure definition: under the FCCC failure definition, patients were at constant risk of failure over the observation period; under the ASTRO failure definition, patients were at risk of failure during the first 4 years following treatment, and then at low risk of failure beyond 5 years.

CONCLUSIONS

Both FCCC and ASTRO failure definitions were robust to modifications in censoring and the inclusion of patients with long doubling times. The ASTRO failure definition was robust to specifying the time to failure at first rise, as opposed to midway between nadir and first rise. Similarities in estimates for all patients versus patients with agreeing failure status suggest that differences in failure definition lie in the specification of time to failure. The ASTRO definition of failure is more appropriate because it does not impose an empirical failure marker but is based on the initiation of biochemical rise. The use of the ASTRO consensus definition demonstrated little risk of biochemical failure 4 years beyond treatment. The ASTRO failure definition should be adopted in all research involving biochemical failure analysis of men treated with radiation therapy.