Skip to main content
. 2012 Jun;180(6):2240–2248. doi: 10.1016/j.ajpath.2012.03.008

Figure 4.

Figure 4

Median size variation of CNV of blood and tumor samples predicts prostate cancer relapse and short PSADT. A: ROC curves of CNV median size of blood samples predicting prostate cancer relapse. The blood samples were separated as described in A. The optimal prediction rates for CNV median size of blood model are 86% (57/66) sensitivity and 61% (11/18) specificity. B: ROC curves of CNV median size of blood predicting prostate cancer short PSADT. The blood samples were separated into a group that had PSADT within 4 months of prostatectomy (n = 31) and a group that did not (n = 53). The optimal prediction rates for CNV median size of blood model are 68% (21/31) sensitivity and 70% (37/53) specificity. C: ROC curves of predicting prostate cancer relapse using median sizes of CNV from tumor samples. The prostate cancer was separated into a group that relapsed within 5 years of prostatectomy (n = 75) and a group that did not relapse (n = 27). The optimal prediction rates for CNV median size of tumor model are 71% (53/75) sensitivity and 89% (24/27) specificity. D: ROC curves of predicting prostate cancer short PSADT using CNV median sizes from tumor samples. The prostate cancer was separated into a group that had PSADT within 4 months of prostatectomy (n = 33) and a group that did not (n = 69). The optimal prediction rates for CNV median size of tumor model are 61% (20/33) sensitivity and 90% (62/69) specificity. Prediction using various CNV median sizes was performed to produce the ROC charts. Dotted line, random prediction baseline; broken line, prediction generated from Gleason grading.