The analytic validation of prostate-specific mRNA detection in whole blood by RT-PCR as a prognostic biomarker for patients with castration-resistant prostate cancer (CRPC).

Abstract

36 Background: To estimate the association between molecular biomarkers and outcomes, robust assays are needed before clinical qualification. Although tumor specific transcripts detected in blood by PCR have been associated with outcome, an analytically validated platform enabling detection of CTC-associated transcripts in clinical settings remains to be identified. Here we explore the efficacy of detecting a panel of prostate specific transcripts in CRPC. Methods: Blood was collected from 97 pts with progressive CRPC in PAXgene tubes (2.5 ml) for total RNA extraction using PAXgene Blood RNA Kit. Five genes expressed in prostate but not in nucleated blood cells were analyzed by RT-PCR. We used Fluidigm Dynamic Array platform to analytically validate RT-PCR assays for KLK3, KLK2, HOXB13, GHRL2, and FOXA1 transcripts. Each PCR-reaction was run in 6 replicates, detection thresholds for each gene were chosen by ROC analysis of an independent set of 56 CRPC patients versus 51 healthy volunteers, and results were reported as transcripts present or absent. CellSearch reported CTC number/7.5 ml of blood drawn into CellSave tube. Results: Prostate specific transcripts (≥ 2 genes) were detected in 52/97 (54%, 95% CI 43-64%) of patients with progressive CRPC studied with a median survival time of 17 months (95% CI: 13.8, 23.5). 40/45 CRPC-patients with ≥ 5 CTC and 15/52 CRPC patients with < 5 CTC had transcripts present in blood. Similar to CTC enumeration, transcript detection predicted overall survival in proportional hazards models that included LDH dichotomized at above and below 300. The predictive accuracy (CPE) of PCR detection + LDH was 0.785 (SE 0.036), comparable to CTC-enumeration + LDH (0.796, SE 0.035). Conclusions: Detecting prostate specific mRNA in whole blood using validated RT-PCR assays, applicable to larger CRPC cohorts, provides prognostic biomarkers that predict overall survival. The clinical utility in monitoring treatment will be prospectively explored. [Table: see text]