Method: As a first validation set, we studied 615 men with 353 having prostate cancer and no previous biopsy and 272 with previous biopsy (113 with cancer). Data on total PSA, free PSA, -2proPSA (2pPSA), and other characteristics were collected from four European sites. We used ROC analysis to assess the value of markers in addition to the previously developed risk calculators. We used a novel approach based on likelihood ratios in logistic regression models.
The extended prediction tools were externally validated in an independent second validation dataset: 950 men with 404 cancer cases detected. As outcomes we considered both any biopsy detectable cancer as well as likely advanced or aggressive cancer.
Result: Applying the previously developed risk calculators to the first validation set resulted in AUCs of 0.68 for predicting cancer, and an AUC for predicting potentially aggressive cancer of 0.65. Inclusion of %2pPSA, 2pPSA and PHI markers resulted in an increase in predictive capability where the inclusion of PHI gave the largest increase in AUC of 0.06 in the first and 0.04 in the second validation set.
Conclusion: Inclusion of markers such as PHI increases the predictive ability of previously developed risk calculators. This may lead to important further reductions of unnecessary prostate biopsies, one of the major drawbacks of current screening for prostate cancer.