Modifying the American Society for Therapeutic Radiology and Oncology definition of biochemical failure to minimize the influence of backdating in patients with prostate cancer treated with 3-dimensional conformal radiation therapy alone

Purpose: Adoption of the American Society for Therapeutic Radiology and Oncology (ASTRO) consensus definition has been critical for evaluating and comparing outcome following treatment with radiation. However, since its almost universal adoption, several points have remained controversial, notably backdating the date of failure to the point midway between the posttreatment prostate specific antigen (PSA) nadir and the first increase. We evaluated the impact of backdating on no biochemical evidence of disease (bNED) control and suggest changes in the definition. Materials and Methods: Between April 1, 1989 and November 30, 1998, 1,017 patients with nonmetastatic prostate cancer were treated with 3-dimensional conformal radiation therapy alone. bNED control was defined using the ASTRO consensus definition. bNED failure was calculated from the time midway between the posttreatment PSA nadir and the first of the 3 consecutive increases in PSA (date of failure A). Four alternate failure time points were chosen, including backdating to the date of the first increase in PSA after the nadir, the date between the first and second consecutive PSA increases, the date between the second and third consecutive PSA increases, and the date of the third increase in PSA after the nadir (dates of failure 1 to 4). Kaplan-Meier estimates were calculated for all definitions of failure as well as hazard functions with time. Subset analyses based on prognostic group and followup time were also performed. Results: The 10-year Kaplan-Meier bNED control rates were 64%, 52%, 47%, 42% and 39% using dates of failure A and 1 to 4, respectively. These differences persisted when patients were stratified by prognostic group. These same differences in bNED control were observed for the long-term followup subset, in which 10-year bNED control rates were 48%, 47%, 44%, 41% and 39% using dates of failure A and 1 to 4, respectively. Conclusions: Adoption of the ASTRO consensus definition has been crucial for evaluating outcome in the radiation oncology community. However, the date of failure should be moved from the current point to one closer to the point at which failure is declared. Additional analysis with large numbers of patients from multiple institutions is necessary to determine the point.