Validation of a Tissue Proteomic Classifier for Prediction of Metastatic Prostate Cancer

Background and objective: Improved clinical management of prostate cancer (PCa) is based not only on early detection of neoplastic lesions in the prostate but also very significantly on the early discrimination of indolent from aggressive PCa. This study aimed to validate the performance of our previously developed 5-protein classifier (SPARC, TGFB1, FOLH1, KLK3, and CAMKK2) for predicting metastatic PCa.
Methods: The independent case-cohort sample cohort consisted of 193 PCa patients, of whom 38 had metastases. The radical prostatectomy (RP) samples were analyzed using the same targeted mass spectrometry assays described in our previous study. Biomarkers were evaluated in models adjusted for either NCCN risk strata (as in the original study) or CAPRA-S.
Key findings and limitations: Univariate analyses of individual biomarkers versus metastasis-free survival showed statistically significant results for CAMKK2, SPARC and TGFB1 (p <0.05). Multivariable analyses of individual biomarkers, each adjusted for NCCN risk group, versus metastasis-free survival showed statistically significant results for SPARC and TGFB1 (p <0.05); all but FOLH1 had concordance indexes (c-index) higher than that for NCCN risk group alone. This largely validated the results in our original study. Although in the multivariable analyses of individual biomarkers, each adjusted for CAPRA-S, versus metastasis-free survival, none of the biomarkers were statistically significant, both SPRAC and TGFB1 had c-index higher than that for CAPRA-S alone.
Conclusions and clinical implications: Considering both the large hazard ratios and increased c-indexes over those of NCCN and CAPRA-S alone, SPARC and TGFB1 may provide improved prognostication of metastatic PCa in clinical application.