• J M Taylor, K A Griffith, and H M Sandler.
• Department of Biostatistics, University of Michigan, Ann Arbor, MI 48104, USA. jmgt@umich.edu
• Int. J. Radiat. Oncol. Biol. Phys. 2001 Aug 1; 50 (5): 1212-9.

PurposeThe American Society for Therapeutic Radiology and Oncology (ASTRO) published a consensus panel definition of biochemical failure following radiation therapy for prostate cancer. In this paper, we develop a series of alternative definitions of biochemical failure. Using data from 688 patients, we evaluated the sensitivity and specificity of the various definitions, with respect to a defined "clinically meaningful" outcome.Methods And MaterialsThe ASTRO definition of biochemical failure requires 3 consecutive rises in prostate-specific antigen (PSA). We considered several modifications to the standard definition: to require PSA rises of a certain magnitude, to consider 2 instead of 3 rises, to require the final PSA value to be greater than a fixed cutoff level, and to define biochemical failure based on the slope of PSA over 1, 1.5, or 2 years. A clinically meaningful failure is defined as local recurrence, distant metastases, initiation of unplanned hormonal therapy, unplanned radical prostatectomy, or a PSA > 25 later than 6 months after radiation.ResultsRequiring the final PSA in a series of consecutive rises to be larger than 1.5 ng/mL increased the specificity of biochemical failure. For a fixed specificity, defining biochemical failure based on 2 consecutive rises, or the slope over the last year, could increase the sensitivity by up to approximately 20%, compared to the ASTRO definition. Using a rule based on the slope over the previous year or 2 rises leads to a slightly earlier detection of biochemical failure than does the ASTRO definition. Even with the best rule, only approximately 20% of true failures are biochemically detected more than 1 year before the clinically meaningful event time.ConclusionThere is potential for improvement in the ASTRO consensus definition of biochemical failure. Further research is needed, in studies with long follow-up times, to evaluate the relationship between various definitions of biochemical failure and true clinical outcome.

32 keywords...

hide…