Background: Fulvestrant (F) is a selective estrogen degrader for hormone receptor-positive (HR+) locally advanced or metastatic breast cancer (LA/MBC). This network meta-analysis examined the efficacy of F (500 mg) vs alternative endocrine therapies (ETs) for first-line treatment of ET-naïve HR+ LA/MBC. Methods: Randomized controlled trials of first-line F, tamoxifen (Tam), anastrozole (A), exemestane (E), letrozole (L), and toremifene (T) for women (≥18 years) with HR+ LA/MBC and no prior ET were identified in a systematic review of MEDLINE, EMBASE, and Cochrane databases from inception to October 2016. Conference proceedings of the American Society of Clinical Oncology, European Society of Medical Oncology, and San Antonio Breast Cancer Symposium from 2013-2016 were hand searched. Trials of targeted combination therapies were excluded. Studies were checked for heterogeneity. A standard fixed-effect Bayesian network meta-analysis was conducted based on hazard ratios (HRs) and assuming proportional hazards for progression-free and overall survival (PFS/OS). Results: Seven eligible studies (1 Phase [Ph] 2, 5 Ph 3, 1 Ph 2/3) were identified. All had PFS data; five had OS data. Two trials compared F vs A; PFS data were available for both trials; sufficiently mature OS data for F were available from Ph 2 only. The proportional hazards assumption was met for PFS only.F had significantly improved PFS vs Tam (HR 0.57, 95% credibility interval [Crl] 0.44-0.73), A (HR 0.75, 95% Crl 0.62-0.91), E (HR 0.65, 95% Crl 0.47-0.91), and T (HR 0.53, 95% Crl 0.37-0.78). Numerically improved PFS was observed for F vs L (HR 0.81, 95% Crl 0.59-1.11). F had significantly improved OS vs Tam (HR 0.63, 95% Crl 0.40-0.98), A (HR 0.63, 95% Crl 0.42-0.94), and E (HR 0.56, 95% Crl 0.33-0.95). OS was numerically improved with F vs L (HR 0.66, 95% Crl 0.41-1.04). Conclusions: This analysis suggests improved PFS and OS for fulvestrant vs tamoxifen, anastrozole, exemestane, and letrozole, and PFS for fulvestrant vs toremifene. Further analysis should be conducted, using non-proportional hazard methods and more mature OS data, to confirm the OS results.