Background: ARSI [enzalutamide (ENZA) and abiraterone (ABI)] demonstrated to be efficacious as a first-line therapy for mCRPC, showing a median progression-free survival of 16 and 20 months (mos) in pivotal trials, respectively. Nevertheless, some pts of daily clinical practice show exceptionally prolonged disease control with very long duration of treatment. The present study aimed at identifying the characteristics of LTR pts who received one ARSI as a first-line for mCRPC. Methods: We retrospectively reviewed the clinical records of a consecutive series of 647 pts treated in the daily clinical practice with one ARSI as first-line mCRPC in four Italian high-volume Institutions. Pts who had previously received docetaxel in castration-sensitive setting were excluded. Pts received standard AA or ENZA doses (1,000 mg po daily plus prednisone 10 mg po daily or 160 mg po daily, respectively). ARSI were administered until progression and were followed according to each Institution policy. For each pt we recorded pre and post-ARSI clinical history, baseline characteristics, treatment details and clinical outcomes. Pts were considered as LTR in presence of an ARSI exposure ≥ 36 mos. LTR were compared to the progressed pts with an ARSI exposure < 36 mos. On-treatment pts with an ARSI exposure < 36 mos were not considered. Results: We identified 81 LTR pts (53 treated with ENZA and 28 treated with ABI). The median age was 76 yrs (range 56-89). The median treatment duration was 46.1 mos (range 36.4-88.6) and 52 pts were still on treatment. LTR were compared to no-LTR 415 pts (233 treated with ENZA and 182 treated with ABI). Compared to no-LTR, LTR pts showed longer PSA doubling time (PSADT) (PSADT) (p = 0.001) and time to mCRPC (p < 0.0001), lower levels of basal PSA (p = 0.01), and alkaline phosphatase (ALP) (p = 0.01), better ECOG PS (p = 0.01), lower Gleason score (p = 0.01), and less pain level (p = 0.04). ECOG PS, ALP, time to mCRPC and PSADT were all identified as independent predictors of time to treatment interruption in multivariate analysis. LTR showed an 89% decreased risk of death compared to no-LTR (HR = 0.11 IC95% 0.06-0.18). Conclusions: In our experience, several variables, which are usually considered as having a prognostic value, demonstrated to be able to predict the probability of prolonged disease control in mCRPC pts treated with first-line ARSI.