Background: Response to androgen signaling inhibitors is highly variable and dictates prognosis in castration resistant prostate cancer (CRPC). The development of effective combinations for CRPC has been hampered by the absence of biomarkers to identify its distinct biological subsets, obligating the homogeneous application of therapies to a heterogeneous disease in clinical trials. We previously defined the aggressive variant prostate cancers as a framework for the study of those with poor responses to androgen signaling inhibitors (dubbed ‘androgen indifferent’) and observed improved outcomes in this subset when carboplatin was added to cabazitaxel (CABCARB). In contrast, results of prospective trials suggested that anti-CTLA-4 therapy may benefit patients with androgen responsive disease. We hypothesized that said combinations applied to subsets defined by early marker declines to androgen ablation would improve overall survival (OS) over historical data in unselected CRPC. Methods: In an open-label, phase II trial (NCT02703623), men with mCRPC received abiraterone and apalutamide (ABIpAPA; Mod 1). At week 8, patients had either a ‘satisfactory’ (S) decline in PSA (≥ 50% from baseline) and CTC (≤5/7.5mL) or ‘unsatisfactory’ (US). S patients (Mod 2) were randomized to continue ABIpAPA alone (2A) or with up to 4 cycles of ipilimumab (IPI, 2B). US patients (Mod 3) continued ABIpAPA with up to 10 cycles of CABCARB. Thereafter patients continued ABIpAPA until progression. OS was calculated from entry into Mod 2 or 3. Based on historical data, the estimated median OS for men in Mod 2 and Mod 3 were 36 and 16 months respectively. Results: 195 men with CRPC were enrolled between May 2016 and August 2019. 128 (67%) were allocated to S of which 64 each were randomized to receive IPI vs. not. 3 went off study during Mod 1 without clinical decline. Of 64 patients allocated to US, 59 were treated with CABCARB on study. Men in US were more likely to have had de novo metastasis (33% vs 59%, p < 0.001), RECIST measurable disease (23% vs 42%, p = 0.01), and liver metastases (3% vs 13%, p = 0.01). Differences in baseline circulating markers between the groups are shown in Table 1. Serious adverse events were reported for 3 (4.7%), 21 (33%), and 6 (10.2%) patients for Mod 2A, 2B, and 3 respectively. One death due to neutropenic sepsis occurred in Mod 3. At a median follow-up of 48 months, the median (95% CI) OS in Mod2A was 44.3 (38.1, 47.7) months, 41.4 (33.3 months, not reached) in Mod 2B, and 18.7 (14.3, 36.3) months in Mod 3.Conclusions: These data substantiate the impact of underlying biology on CRPC patient outcomes and the urgent need for biomarkers that can enable the development of therapies specific to each subset. Correlative studies to identify said biomarkers will be presented. Clinical trial information: NCT02703623.