Development and validation of a novel circulating tumor cell response stratification criteria for metastatic castration-resistant prostate cancer clinical trials.

Abstract

e17072 Background: While circulating tumor cells (CTC) have been used in metastatic castration-resistant prostate (mCRPC) clinical trials, established response measures are limited by their fixed criteria (decline to 0 or <5). We developed and validated a CTC response stratification (CTC-RS) criteria (table 1) to classify patients as responders (CTC-CR & -PR) and non-responders (CTC-SD & -PD) by refining elements of existing criteria. Methods: We analyzed two phase 3 TAK-700 clinical trials in mCRPC (ELM-PC4, NCT01193244 ; ELM-PC5; NCT01193257 ). Patients with detectable CTCs at baseline (BL) and/or the 12wk timepoint were evaluable for CTC-RS. Response rate as a function of BL CTC burden was evaluated for CTC0, CTC conversion, and CTC-RS. Associations with outcomes were assessed using landmarked Kaplan-Meier estimates and the separation evaluated using the predictive concordance probability (pCP). CTC-RS was compared to RECIST best overall response (BOR), and the association between CTC-RS and survival was assessed in the RECIST stable disease (RECIST-SD) subgroup. Results: CTC data was available at BL and the 12wk timepoint from 1107 (PC4) and 666 patients (PC5), and 780 and 560 were evaluable for CTC-RS. Response rates were dramatically lower in patients in the upper quartile of BL CTC counts versus those in the lower quartile using CTC0 (PC4: 10% vs 65%; PC5: 8% vs 49%) and CTC conversion (PC4: 12% vs 76%; PC5: 15% vs 56%), while CTC-RS response rates remained relatively high (PC4: 48% vs 65%; PC5: 43% vs 49%). CTC-RS responders had significantly longer OS (PC4: 30.5 vs 19.8mo; PC5: not reached vs 13.1mo) and PFS (PC4: 13.7 vs 8.3mo; PC5: 11.0 vs 7.0mo) compared to non-responders. In PC4 and PC5, CTC0 (pCP: 0.70 & 0.75) and CTC conversion (pCP: 0.73 & 0.74) demonstrated better separation in the OS curves than CTC-RS (pCP: 0.60 & 0.64), but this finding is partly due to the substantially better BL prognostic characteristics of patients who achieved CTC0 and/or CTC conversion. In PC5 patients evaluable for BOR (n=187), CTC-RS demonstrated modest accuracy (0.66) and high NPV (0.91) using BOR as ground truth. The PPV (0.27) was low, driven by CTC-RS “false positives” in RECIST-SD patients. In the RECIST-SD subgroup, CTC-RS responders (n=41) had significantly longer OS than non-responders (n=70) (not reached vs 13.1mo). Conclusions: CTC-RS is associated with clinical outcome and untethers response from BL CTC burden by using percentage-based criteria, enabling improved evaluation of high CTC burden patients. CTC-RS has modest agreement with BOR and stratifies RECIST-SD patients into responders and non-responders, reflecting the ability to assess the totality of disease. CTC-RS criteria. CTC-RS Group CTC-RS Stratification BL Criteria 12wk Criteria CTC-CR Responder >0 CTC 0 CTC CTC-PR Responder ≥5 CTC ≥70% CTC decline CTC-SD Non-Responder None Absence of CTC-CR/PR/PD (CTC >0 at BL and/or 12wk) CTC-PD Non-Responder None ≥5 CTC & ≥30% CTC increase