Purpose: Gene fusions leading to androgen receptor-modulated ERG overexpression occur in up to 70% of metastatic castration-resistant prostate cancers (mCRPC). We assessed the association between ERG rearrangement status and clinical benefit from abiraterone acetate. Experimental Design: COU-AA-302 is a phase III trial comparing abiraterone acetate and prednisone versus prednisone in chemotherapy-naïve mCRPC. ERG status was evaluated by FISH on archival tumors. End points included radiographic progression-free survival (rPFS), time to PSA progression (TTPP), rate of ≥50% PSA decline from baseline, and overall survival (OS). Cox regression was used to evaluate association with time-to-event measures and Cochran-Mantel-Haenszel for PSA response. Results: ERG status was defined for 348 of 1,088 intention-to-treat patients. ERG was rearranged in 121 of 348 patients with confirmed ERG status (35%). Cancers with an ERG fusion secondary to deletion of 21q22 and increased copy number of fusion sequences (class 2+ Edel) had a greater improvement in rPFS after abiraterone acetate and prednisone [22 vs. 5.4 months; HR (95% confidence interval, CI), 0.31 (0.15-0.68); P = 0.0033] than cancers with no ERG fusion [16.7 vs. 8.3 months; 0.53 (0.38-0.74); P = 0.0002] or other classes of ERG rearrangement. There was also greater benefit in this subgroup for TTPP. Conclusions: Both ERG-rearranged and wild-type cancers had a significant improvement in rPFS with abiraterone acetate and prednisone in the COU-AA-302 trial. However, our data suggest that 2+ Edel cancers, accounting for 15% of all mCRPC patients and previously associated with a worse outcome, derived the greatest benefit.