Prostate Cancer Test Network

PHI at threshold 35 showed high sensitivity but modest specificity in the prostate cancer estimates. The meta-regression model had the lowest DIC in the prostate cancer dataset, consistent with the sparse network. SelectMDx was treated as binary because too few thresholds were reported to estimate a threshold-accuracy relationship reliably. The standard Nyaga ANOVA model could not be fitted for the prostate comparison after common-threshold selection because no connected evidence network remained. The prostate cancer application included four tests across 37 studies in a connected network.