Background: In mCRPC, AR T878 and/or H875 and L702H mutations have been associated with disease progression and poor prognosis. This study characterized real-world overall survival (rwOS), testing practices, mutation prevalence, and treatment patterns in patients with mCRPC with and without tumors with missense mutations in the AR LBD (amino acids 671–920). Methods: In this retrospective analysis of the Guardant Inform Database (Mar 11, 2014–Jun 30, 2022), men (aged ≥18 y) with mCRPC who were tested with Guardant360 (G360) next-generation sequencing and diagnosed/treated at a clinical site in the US with a first-line (1L) treatment were included. Matched rwOS was assessed in 2 subpopulations: patients with any missense mutation in AR LBD, excluding those with AR L702H alone (AR LBD group), and patients with AR T878 and/or H875 without L702H (AR 878/875 group) and compared with patients without any AR LBD missense mutations (control group). Patients were matched (1:5) on key prognostic variables, including age, Elixhauser Comorbidity Index weighted score, smoking status, prior novel hormonal agent (NHA) use, and 1L treatment year. Mutation prevalence, patient demographics, frequency and timing of G360 tests, and treatment patterns were assessed in patients treated from 2018 to 2022 to reflect more recent trends. Results: In the matched rwOS analysis, patients in the AR LBD group (n=275) had statistically significantly shorter 1L median rwOS compared with the control group (n=1375): 27.3 vs 47.8 months; P<0.0001. Similarly, patients in the AR 878/875 group (n=154) had shortened median rwOS compared with the control group (n=770): 28.9 vs 44.4 months; P=0.0004. Of 4833 patients treated for mCRPC from 2018 to 2022, 20% had AR LBD mutations, 4% had AR 878/875 with L702H, and 6% had AR 878/875 without L702H. Most patients (59%) were first tested at some point after 1L therapy, with only 15% of patients receiving >1 G360 test. Mutation prevalence was higher in patients tested in fourth-line (4L) compared with 1L, particularly for AR L702H mutations (17% vs 10%) and all AR LBD mutations (27% vs 20%). Use of NHAs was lower in 4L compared with 1L (20% vs 33%), whereas use of taxanes was higher (31% vs 22%). Conclusions: Patients with AR LBD–mutated mCRPC had notably shorter rwOS than those whose tumors did not harbor AR LBD mutations, indicating an unmet need for this population. The prevalence of AR LBD mutations was higher in later lines of therapy, but repeat testing rates were low, suggesting that AR LBD mutations may go undetected.