An analysis of a multiple biomarker panel to better predict prostate cancer metastasis after radical prostatectomy

Abstract

A plethora of individual candidate biomarkers for predicting PSA relapse in localized prostate cancer (PCa) have been proposed. Combined biomarkers may improve prognostication, and ensuring validation against more clinically relevant endpoints are required. The Australian PCa Research Centre NSW has contributed to numerous studies of molecular biomarkers associated with PSA relapse. In the current study, these biomarkers were re-analysed for PSA relapse, metastatic relapse and PCa death with extended follow-up. Biomarkers of significance were then used to develop a combined prognostic model for clinical outcomes and validated in a large independent cohort. The discovery cohort (n=324) was based on 12 biomarkers with a median follow-up of 16 years. Seven biomarkers were significantly associated with PSA relapse. Three biomarkers were associated with metastases: AZGP1, Ki67 and PML. Only AZGP1 was associated with PCa death. In their individual and combinational forms, AZGP1 and Ki67 as a dual BM signature was the most robust predictor of metastatic relapse (AUC 0.762). The AZPG1 and Ki67 signature was validated in an independent cohort of 347 PCa patients. The dual BM signature of AZGP1 and Ki67 predicted metastasis in the univariable (HR 7.2, 95% CI, 1.6-32; P=0.01) and multivariable analysis (HR 5.4, 95% CI, 1.2-25; P=0.03). The dual biomarker signature marginally improved risk prediction compared to AZGP1 alone (AUC 0.758 versus 0.738, P<0.001). Our findings indicate that PSA relapse is not an adequate surrogate for metastasis or PCa death. The dual biomarker signature of AZGP1 and Ki67 offers a small benefit in predicting metastasis over AZGP1 alone.