Purpose: To study the use of the year of therapy as an independent predictor of outcomes, serving as a proxy for time-related changes in therapy and tumor factors in the treatment of prostate cancer. Accounting for these changes would facilitate the retrospective comparison of outcomes for patients treated in different periods.
Methods and materials: Nine institutions combined data on 4,537 patients with Stages T1 and T2 adenocarcinoma of the prostate who had a pretherapy prostate-specific antigen (PSA) level and biopsy Gleason score, and who had received > or = 60 Gy external beam radiotherapy without neoadjuvant androgen deprivation or planned adjuvant androgen deprivation. All patients were treated between 1986 and 1995. Two groups were defined: those treated before 1993 (Yr < or = 92) vs. 1993 and after (Yr > or = 93). Patients treated before 1993 had their follow-up truncated to make the follow-up time similar to that for patients treated in 1993 and after. Therefore, the median follow-up time was 6.0 years for both groups (Yr < or = 92 and Yr > or = 93). Two separate biochemical failure endpoints were used. Definition A consisted of the American Society for Therapeutic Radiology Oncology endpoint (three PSA rises backdated, local failure, distant failure, or hormonal therapy). Definition B consisted of PSA level greater than the current nadir plus two, local failure, distant failure, or hormonal therapy administered. Multivariate analyses for factors affecting PSA disease-free survival (PSA-DFS) rates using both endpoints were performed for all cases using the following variables: T stage (T1b, T1c, T2a vs. T2b, T2c), pretreatment PSA (continuous variable), biopsy Gleason score (continuous variable), radiation dose (continuous variable), and year of treatment (continuous variable). The year variable (defined as the current year minus 1960) ranged from 26 to 35. To evaluate the effect of radiation dose, the multivariate analyses were repeated with the 3,897 cases who had received < 72 Gy using the same variables except for radiation dose.
Results: For all 4,537 patients, the 5- and 8-year PSA-DFS estimate using definition A (ASTRO consensus definition) was 60% and 55%, respectively. The 8-year PSA-DFS estimate for Yr < 93 vs. Yr > or = 93 was 52% vs. 57%, respectively (p < 0.001). In the subgroup of patients receiving < 72 Gy, the 8-year PSA-DFS estimate for Yr < 93 vs. Yr > or = 93 was 52% and 55%, respectively (p = 0.004). The differences in PSA-DFS rates in the different subgroups were similar when definition B was used. The multivariate analyses for all 4,537 cases with either PSA-DFS definition revealed T stage (p < 0.001), pretherapy PSA level (p < 0.001), Gleason score (p < 0.001), radiation dose (p < 0.001), and year of treatment (p < 0.001) to be independent predictors of outcomes. The multivariate analyses restricted to the 3,897 cases receiving < 72 Gy still revealed year of treatment to be an independent predictor of outcomes (p < 0.001), in addition to T stage (p < 0.001), pretherapy PSA level (p < 0.001), and Gleason score (p < 0.001).
Conclusions: Independent of tumor stage, radiation dose, failure definition, and follow-up parameters, the year in which RT was performed was an independent predictor of outcomes. These findings indicate a more favorable presentation of localized prostate cancer in current years that is not necessarily reflected in the patients' PSA levels or Gleason scores. This phenomenon is probably related to a combination of factors, such as screening, increased patient awareness leading to earlier biopsies and earlier diagnosis, more aggressive pretherapy staging, and unrecognized improvements in therapy, but perhaps also to changing tumor biology. Outcomes predictions should be based on contemporaneous series. Alternatively, the year of therapy could be incorporated as a variable in outcomes analyses of localized prostate cancer patients treated in different periods within the PSA era.