BACKGROUND. Prostate cancer screening is studied in a randomized trial
in Antwerp, Belgium. The case group receives three screening tests (D
RE, TRUS, and PSA). Intermediate evaluation shows that only 1/3 of the
biopsy results is positive (35/125). The proposed analysis identifies
variables that determine the biopsy outcome. METHODS. Multiple logist
ic regression analysis is used to regress biopsy results (n = 125) by
age (60-74), PSA, PSA-D, prostate volume, TRUS, and DRE. Continuous va
riables are transformed into quartile values. Robustness of the outcom
e is tested with ROC and sensitivity analysis on age. RESULTS. Biopsy
outcomes are best explained (82.3%) by PSA, DRE, and DRE related to vo
lume. Volume is more sensitive than age to explain the biopsy result.
PSA-D, instead of PSA, does not procure more precise information when
a high PSA cut-off level is used. Restricting the analysis to the 60-7
0-year-old age group shows that volume is more sensitive. ROC-analysis
confirms the findings. CONCLUSIONS. When performing prostate cancer m
ultitest screening among a wide age range, the use of uniform screenin
g criteria is difficult to accept due to differences in prostate volum
e. Logistic regression analysis is an appropriate method to identify c
ut-off levels for prostate volume. (C) 1997 Wiley-Liss, Inc.