The interobserver variability of digital rectal examination in a large randomized trial for the screening of prostate cancer

Abstract

Background: To analyze to what extent the percentage of suspicious digital rectal examination (DRE) findings vary between examiners and to what extent the percentage of prostate cancers (PCs) detected in men with these suspicious findings varies between examiners.

Methods: In the first screening round of the European Randomized study of Screening for PC (ERSPC) Rotterdam, 7,280 men underwent a PSA-determination and DRE of whom 2,102 underwent prostate biopsy (biopsy indication PSA > or = 4.0 ng/ml and/or suspicious DRE and/or TRUS). Descriptive statistics of DRE-outcome per PSA-range were used to determine the observer variability of six examiners. Because this analysis did not correct properly for other predictors of a suspicious DRE (PSA-level, biopsy indication, TRUS-outcome, prostate volume and age), a logistic regression analysis controlling for these explanatory variables was performed as well.

Results: In 2,102 men biopsied, 443 PCs were detected (PPV = 21%). For all PSA levels the percentage suspicious DRE varied between examiners from 4% to 28% and percentage PC detected in men with a suspicious DRE varied from 18% to 36%. Logistic regression analysis showed that three of six examiners considered DRE significantly more often abnormal than others (ORs 3.48, 2.80, 2.47, P < 0.001). For all examiners the odds to have PC was statistically significantly higher in case of a suspicious DRE (ORs 2.21-5.96, P < 0.05). This increased chance to find PC was not significantly observer-dependent.

Conclusions: Three of six examiners considered DRE significantly more often suspicious than the others. However, under equal circumstances a suspicious DRE executed by each examiner increased the chance of the presence of PC similarly.