Background: The main treatments for mCRPC are 2 new ART (ABI, ENZA) and 2 taxanes [doctaxel (DOC) and Cabazitaxel]. The optimal sequence is currently not defined. An increasing number of pts are first treated with one of ART. In this population, data about efficacy of 2nd line (L2) are spare. We designed a retrospective study to evaluate activity of DOC vs ART in mCRPC pts treated with ART as L1 therapy. Methods: In this observational cohort study, we included all consecutive pts with ENZA or ABI as L1 for chemo-naïve mCRPC. We measured PFS and OS during both L1 & L2. To compare the efficacy of DOC versus ART as L2, we measured Growth Modulation Index (GMI=Time To progression with L2/TTP with L1) and we used Cox model to compare PFS/OS in both arm, in both univariate & multivariate analysis using propensity score. Results: We included 175 pts, including 75 treated ENZA and 100 with ABI as L1. 69 (39%) pts received DOC and 30 (17%) pts received ART as L2; 76 (43%) did not receive L2. Median follow-up was 36 months (CI95%: 30.4 - 40). From the starting of L1, PFS was 13.0 months (CI95% 11.0 – 15.0), OS was 34.5 months (CI95%: 28.7 – 38.6). There was no difference between ENZA and ABI in PFS and OS (p=0.684). From the start of L2, the median PFS was 6.0 months (CI95%: 5.0 – 6.6) and the median OS was 18.0 months (CI95%: 13.9 – 21.4). We found a significant benefit in PFS for DOC vs ART in L2 (6.7 months vs 4.0 months, HR 0.60 [CI95%: 0.31 – 0.96], p=0.03). This benefit did not reach the level of significance in OS (19.5 months vs 12.0 months, HR 0.60 [CI95%: 0.35 – 1.03], p=0.1). ECOG-PS and time of castration resistance were associated with OS in multivariable analysis and then used in propensity score. After adjustment to both parameters, we found no difference in PFS (p=0.2) and OS (p=0.1) in pts receiving DOC vs ART as L2. Only 56% of pts received L2. Probability for GMI being >1.33 was not significantly different between DOC (19%) and ART (13%), p=0.099. Conclusions: ABI and ENZA are similarly active in L1 mCRPC. In univariate analysis, L2 with DOC seemed more active than ART. However this benefit was not retained after adjustment to propensity score.