Little data exist on whether efficacy benefits or side-effects persist after 5 years of adjuvant treatment with an aromatase inhibitor. We aimed to study long-term outcomes in the Arimidex, Tamoxifen, Alone or in Combination (ATAC) trial that compares anastrozole with tamoxifen after a median follow-up of 100 months. We analysed postmenopausal women with localised invasive breast cancer. The primary endpoint disease-free survival (DFS), and the secondary endpoints time to recurrence (TTR), incidence of new contralateral breast cancer (CLBC), time to distant recurrence (TTDR), overall survival (OS), and death after recurrence were assessed in the total population (intention to treat; ITT: anastrozole, n=3125; tamoxifen, n=3116; total 6241) and the hormone-receptor-positive subpopulation, the clinically important subgroup for which endocrine treatment is now known to be effective (84% of ITT: anastrozole, n=2618; tamoxifen, n=2598; total 5216). After treatment completion, fractures and serious adverse events continued to be collected blindly (safety population: anastrozole, n=3092; tamoxifen, n=3094; total 6186). This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN18233230. At a median follow-up of 100 months (range 0-126), DFS, TTR, TTDR, and CLBC were improved significantly in the ITT and hormone-receptor-positive populations. For hormone-receptor-positive patients: DFS hazard ratio (HR) 0.85 (95% CI 0.76-0.94), p=0.003; TTR HR 0.76 (0.67-0.87), p=0.0001; TTDR HR 0.84 (0.72-0.97), p=0.022; and CLBC HR 0.60 (0.42-0.85), p=0.004. Absolute differences in time to recurrence increased over time (TTR 2.8% [anastrozole 9.7%vs tamoxifen 12.5%] at 5 years and 4.8% [anastrozole 17.0%vs tamoxifen 21.8%] at 9 years) and recurrence rates remained significantly lower on anastrozole compared with tamoxifen after treatment completion (HR 0.75 [0.61-0.94], p=0.01). The fewer deaths after recurrence (anastrozole 245 vs tamoxifen 269) was not significant (HR 0.90 [0.75-1.07], p=0.2), and no effect was noted for OS (anastrozole 472 vs tamoxifen 477) HR 0.97 [0.86-1.11], p=0.7). Fracture rates were higher in patients receiving anastrozole than in those receiving tamoxifen during active treatment (number [annual rate]: 375 [2.93%] vs 234 [1.90%]; incidence rate ratio [IRR] 1.55 [1.31-1.83], p<0.0001), but were not different after treatment was completed (off treatment: 146 [1.56%] vs 143 [1.51%]; IRR 1.03 [0.81-1.31], p=0.79). We did not note any significant difference in risk of cardiovascular morbidity or mortality between anastrozole and tamoxifen treatment groups. These data show long-term safety findings and establish clearly the long-term efficacy of anastrozole compared with tamoxifen as initial adjuvant treatment for postmenopausal women with hormone-sensitive, early breast cancer, and provide statistically significant evidence of a larger carryover effect after 5 years of adjuvant treatment with anastrozole compared with tamoxifen.