Purpose: We reviewed our institution's experience treating patients with localized prostate cancer with external beam radiation therapy (RT) to determine how differences in the length of follow-up affect the determination of treatment outcome using the American Society for Therapeutic Radiology and Oncology (ASTRO) Consensus Panel Definition of biochemical failure (BF). Methods and Materials: From January 1987 through December 1997, 1109 patients with localized prostate cancer were treated with definitive external beam RT at William Beaumont Hospital, Royal Oak, Michigan. All patients received external beam RT to a median total prostate dose of 66.6 Gy (range: 59.4-70.4 Gy). A total of 1096 patients (99%) had sufficient prostate-specific antigen (PSA) follow-up to determine their biochemical status. To test the impact of differences in follow-up on the calculation of BF, 389 patients with at least 5 years of PSA follow-up were selected as the reference group for the initial analysis. BF was then retrospectively determined using the Consensus Panel definition at yearly intervals, ignoring the remainder of each patient's follow-up. The median follow-up for this group of patients was 6.6 years (range: 5.0-11.6 years). In a second analysis, patient cohorts were randomly selected with varying median PSA follow-up intervals in order to more accurately represent a population whose follow-up is distributed continuously over a defined range. Seven cohorts were randomly selected with 200 patients in each cohort. Cohorts were individually identified such that half of the patients (100) had 2 years or less follow-up than the stated time point for analysis and half (100) had up to 2 years more follow-up than the time point chosen for analysis. For example, in the cohort with a median follow-up of 3 years, 100 patients with a PSA follow-up from 1 to 3 years were randomly selected, and 100 patients with a follow-up from 3 to 5 years were randomly selected, thus generating a median follow-up of 3 years for this cohort (range: 1 to 5 years). This process was repeated five times for five random samples of seven cohorts each. Biochemical failure was calculated according to the Consensus Panel definition. Results: In the first analysis, significantly different rates of biochemical control (varying by 6-21%) were calculated for the same actuarial year chosen for analysis depending only upon the length of follow-up used. For example, the 3-year actuarial rate of biochemical control (BC) varied from 71% when calculated with 3 years of follow-up versus 50.4% with 7 years (p < 0.01). These differences in actuarial rates of BC were observed in all subsets of patients analyzed (e.g., PSA < 10, Gleason less than or equal to 6, n = 132,p < 0.001; PSA < 10, Gleason greater than or equal to 7, n = 33, p = 0.03; PSA greater than or equal to 10, Gleason less than or equal to 6, n = 109, p < 0.001; and PSA greater than or equal to 10, Gleason greater than or equal to 7, n = 72, p = 0.002). The absolute magnitude of the difference in actuarial rates of BC was greatest during years 2 (range 18-30%), 3 (range 16-25%), and 4 (range 15-24%) after treatment. In the second analysis using median PSA follow-ups las defined above), statistically significant differences in actuarial rates of BC were again observed. For example, the 3-year actuarial rate of BC varied from 74.8% with a median follow-up of 2 years versus 49.2% with a median follow-up of 6 years. These dramatic differences in BC were still observed beyond 5 years. Conclusion: When the ASTRO Consensus Panel definition of BF is used to calculate treatment success with external beam RT for prostate cancer, adequate follow-up is critical. Depending upon the length of time after treatment, significantly different rates of BC (varying by 15% to 30%) can be calculated for the same time interval chosen for analysis. These results suggest that data should only be reported if the length of follow-up extends at least beyond the time point at which actuarial results are examined for the majority of patients. (C) 1999 Elsevier Science Inc.
