A multicenter study demonstrating discordant results from electronic prostate-specific antigen biochemical failure...

Area of Science:

• Oncology
• Medical Informatics
• Biostatistics

Background:

• Biochemical failure (bF) is a critical endpoint in prostate cancer treatment.
• Standardized electronic calculators are used to assess bF, but interobserver variation exists.
• Understanding this variation is key to accurate treatment response evaluation.

Purpose of the Study:

• To evaluate interobserver variability among four electronic biochemical failure (bF) calculators.
• To compare the concordance of three distinct bF definitions (ACD, L2, T3) across these calculators.
• To identify sources of variation in electronic bF reporting.

Main Methods:

• Analysis of data from 1200 men using four institutional electronic bF calculators.
• Comparison of bF identification using the American Society for Therapeutic Radiology and Oncology (ACD) consensus definition, the lowest PSA level to date plus 2 ng/mL (L2), and a threshold of 3 ng/mL (T3).
• Assessment of concordance and variation in bF status and timing across centers.

Main Results:

• Unanimous agreement on bF status was high: 87.3% (ACD), 96.4% (L2), and 92.7% (T3).
• Significant variation in ACD bF timing (>2 months) occurred in 270 men, with one institution contributing 63% of the variability.
• The L2 definition showed considerable timing variation (20.5%), while T3 had minimal variation (2.0%). Fifteen interpretation variations were noted.

Conclusions:

• Reported bF results are influenced by both definition differences and their implementation in electronic calculators.
• Variations in interpretation were most common for the ACD definition.
• A proposed algorithm for L2 and a verification system are needed for consistent electronic bF reporting.