Molecular Imaging of Androgen Receptor Expression and Glycolysis as Biomarkers for Clinical Outcome for Metastatic Castration-Resistant Prostate Cancer

PurposePatients with metastatic castration-resistant prostate cancer (mCRPC) with tumors that display disease heterogeneity, including features of androgen receptor (AR)-independence and/or lineage plasticity, have particularly poor outcomes. Non-invasive methods to prognosticate on such poor-risk features continue to pose a major challenge. We previously showed that imaging could be used to quantitatively identify patients by virtue of AR-expression and glycolysis, and that those with AR-negative and glycolytic-positive disease were associated with a poorer prognosis. Here, we extend our analysis by creating a model integrating imaging parameters with clinical determinants to comprehensively prognosticate progression free (PFS) and overall survival (OS). Patients and Methods124 CRPC patients underwent dual FDG/FDHT PET scans in a clinical trial spanning 7-years. For each patient, clinical markers and five index lesions showing abnormal tracer uptake and anatomical tumor spread were analyzed. Univariate analysis was performed on PET and clinical variables for OS and PFS. Cox proportional-hazard models were developed for OS and PFS with an analysis of covariance determining association between FDG/FDHT and OS/PFS, after adjustment for serum markers. Concordance probability estimates (CPE) quantified the discriminatory power of our model. ResultsFDG SUVmaxavg uptake was associated with OS (HR=1.55, P=0.027) with the 4th quartile of FDG SUVmaxavg ([≥]7.8) showing a 14-month reduction in OS relative to the 1st quartile ([≤]4.22). Moreover, through multivariate modeling, the incorporation of FDG PET with clinical markers yielded a model with moderate discriminatory power (CPE=0.74). FDHT SUVmaxavg uptake was however not associated with OS (HR=1.14, P=0.5). Furthermore, FDHT SUVmax as a covariate lacked robust association with PFS (HR=0.78, P=0.08). ConclusionsIn this prognostic model that integrates clinical and molecular imaging data, FDG PET is an important element of identifying poor-risk patients by OS. FDHT, however, does not prognosticate for OS, and further lacks robust association with PFS, likely due to the complexities of imaging AR in mCRPC patients.