Background: Playing an important role in prostate cancer, androgen receptor (AR) signaling is a common therapeutic target. Novel hormonal treatment (NHT) using enza or abi prolongs overall survival in men with metastatic castration-resistant prostate cancer (mCRPC). However, biomarkers predicting therapy response are limited. AR-V7, as the most abundant AR splice variant, has gained clinical interest. Nonetheless, current discussions on its predictive power are diverse. Given that AR-V7 as a sole biomarker is not efficient in predicting response to NHT, we aimed to increase the predictive potential by analysis of combinatorial AR splice variant (AR-V) expression in mCRPC patients undergoing NHT. Methods: We prospectively enrolled 60 patients who started on either abi or enza. Presence of circulating tumor cells (CTC) as well as expression of AR-V3, -7 and -9 were assessed. Outcomes in CTC-, CTC+/AR-V- and CTC+/AR-V+ patients were analyzed considering PSA reduction, PSA-PFS, PFS and OS. Results: PSA reduction of 50% was predominantly found in CTC- patients (78.5%) compared to CTC+/AR-V- (55.5%) and CTC+/AR-V+ (39.3%) without statistical significance (P = 0.059). When taking co-expression of two or more AR-V into account there was no difference in PSA response either (one AR-V 42.9%, two AR-V 33.3%, three AR-V 41.6%, P = 0.154). Median PSA-PFS was 17 months (95%CI 15.7 – 18.3), 13 months (95%CI 6.8 – 19.2) and 5 months (95%CI 3.6 – 6.4) for CTC- pts, CTC+/AR-V- pts and CTC+/AR-V+ pts, respectively (P = 0.005). However, comparing CTC- and CTC+ pts, differences become even more apparent (P = 0.004), CTC+/AR-V- and AR-V+ pts showed less statistically significant differences (P = 0.029). Median PFS and OS were not reached for CTC- pts. PFS was 10 months (95%CI 6.2 – 13.8) for CTC+/AR-V- pts and 9 months (95%CI 1.1 – 16.9) for CTC+/AR-V+ pts (P = 0.004, only CTC- vs. CTC+ P = 0.002). OS was 28 months (95%CI 16.8 – 39.2) for CTC+/AR-V- pts and 15 months (95%CI 7.9 – 22.1) for CTC+/AR-V+ pts (P = 0.014, only CTC- vs. CTC+ P = 0.006). Regarding PFS and OS, there was no difference comparing only CTC+/AR-V- and AR-V+ pts (P = 0.356 and P = 0.244). Conclusions: AR splice variants have prognostic power in stratifying mCRPC patients suffering from a more advanced stage of disease. Nonetheless, our study clearly demonstrates the lack of predictive power of AR splice variants for response to NHT. Additionally, we prove the importance of CTC analysis rather than AR-V expression being more valuable in mCRPC.