Characterization of the behavior of three definitions of prostate-specific antigen-based biochemical failure in relation to detection and follow-up biases: comparison with the American Society for Therapeutic Radiology and Oncology consensus definition.

PURPOSE: To examine the impact of detection biases on three prostate cancer biochemical failure (bF) definitions in comparison with the existing American Society for Therapeutic Radiology and Oncology Consensus Definition (ACD). METHODS AND MATERIALS: Three alternative bF definitions were tested against the ACD: three rises in prostate-specific antigen (PSA) level without backdating, nadir plus 2 ng/mL, and a threshold PSA level of >3 ng/mL, according to data from 1050 men. The mean time between PSA tests (MTBT), regularity of collection, and calendar year of analysis were examined in each bF definition.
RESULTS: The MTBT produced a statistically significant difference in the derived hazard ratio for identification of bF in all definitions. The influence of test regularity was statistically significant beyond the median level of regularity in all definitions. The year of analysis impacted greatly on the ACD, whereas the three alternative definitions exhibited minor follow-up duration variations by comparison. The alternative definitions had reliable follow-up when the crude median time to censoring was at least 1.6 times greater than that of failure.
CONCLUSIONS: Detection biases will always be a significant issue in defining bF. A number of alternative failure definitions have more predictable interactions with these biases than the existing ACD.