The correlation of serial prostate specific antigen measurements with clinical outcome after external beam radiation therapy of patients for prostate carcinoma

BACKGROUND: The authors analyzed retrospectively their institution's experience in treating patients with localized prostate carcinoma with external beam radiation therapy (EBRT) to determine the correlation of various biochemical failure (BF) definitions with clinical failure and cause specific survival (CSS). METHODS: Between January 1987 and December 1997, 1,094 patients with clinical T1-T3N0M0 prostate carcinoma were treated with definitive EBRT alone at William Beaumont Hospital, Royal Oak, Michigan. All patients received EBRT alone (no adjuvant hormones) to a median total prostate dose of 66.6 grays (Gy) (range, 59.4-70.4 Gy). Multiple BF definitions were tested for their correlation with clinical failure and cause specific death (CSD = 1-CSS). All BF definitions were tested for sensitivity, specificity, and accuracy of predicting subsequent clinical failure and CSD. Positive and negative predictive values were calculated in the form of 10-year actuarial clinical failure and CSD rates. Analyses were performed on all 1,094 patients as well as for those 727 patients who had at least 5 post-RT prostate specific antigen (PSA) level measurements and who did not receive hormonal therapy for post-RT PSA elevations. The median PSA follow-up was 4.0 years for the entire population and 4.5 years for those 727 patients included in the second analysis. RESULTS: In the entire population, 167 patients (15%) experienced clinical failure corresponding to 5- and 10-year actuarial rates of 16% and 34%, respectively. The correlation of various BF definitions with outcome was calculated in those 727 patients who did not receive hormonal therapy. For these patients, BF (as defined by the American Society for Therapeutic Radiology and Oncology Consensus Panel) yielded a 73% sensitivity, 76% specificity, and 75% overall accuracy for predicting clinical failure and a 74% sensitivity, 69% specificity, and 69% overall accuracy for predicting CSD. The 10-year clinical failure rate for those 251 patients demonstrating 3 consecutive PSA rises (BF) was 64% versus 14% for those patients who did not meet these criteria (biochemically controlled [BC]). As expected, definitions requiring only two rises were more sensitive but less specific in predicting clinical failure than those definitions requiring three or four rises. Because there were dramatically more clinically controlled patients (85%) than clinical failures (15%), the overall accuracy for each definition more closely approximated its specificity. The definitions classifying BF as a postnadir increase of > or = 3 or > or = 4 ng/mL above the nadir yielded the highest accuracies of 87% and 88%, respectively. In addition, these definitions also appeared to provide the greatest separation in clinical failure rates between BC and BF patients, an absolute difference of 77% and 76%, respectively. CONCLUSIONS: The correlation between BF and clinical failure and CSD varies markedly depending on the BF definition used. Definitions incorporating a fixed baseline (the nadir level) and the postnadir PSA profile may have better correlation with clinical failure than definitions using the nadir only or a specific number of consecutive rises in which a variable baseline "resets" after a PSA decrease.