OBJECTIVES: To compare the efficacy of third-generation aromatase inhibitors (AIs) with tamoxifen in the treatment of early and advanced breast cancer (EBC and ABC) in randomised controlled trials (RCTs). METHODS: Systematic review of CENTRAL, EMBASE, MEDLINE, for RCTs comparing tamoxifen with anastrozole, exemestane, or letrozole. Searching was restricted to English-language publications and was completed in March 2008. Mortality and relapse data were extracted for EBC studies at 2 and 5 years post randomisation.  For ABC, primary outcomes were disease progression and response. Meta-analysis was conducted with a fixed effects model using the Mantel-Haenszel method. Summary effect estimates (Relative Risk [RR]) with 95% confidence intervals (95%CI) were calculated (negative outcomes, e.g. mortality, favour AIs when RR<1.0; positive outcomes, e.g. clinical benefit, favour AIs when RR>1.0). RESULTS: The literature search identified 2417 papers, 15 provided data on 9 RCTs comparing third-generation AIs with tamoxifen. The results for EBC presented a range of findings in favour of AIs at 2 and 5 years with statistically significant outcomes at 2 years: all cause mortality (RR 0.86, 95%CI: 0.77, 0.97); contralateral recurrence (RR 0.53, 95%CI: 0.38, 0.74); distant recurrence (RR 0.75, 95%CI: 0.67, 0.84); local recurrence (RR 0.80, 95%CI: 0.66, 0.96); and 5 years: contralateral recurrence (RR 0.57, 95%CI: 0.42, 0.76); distant recurrence (RR 0.86, 95%CI: 0.78, 0.94); local recurrence (RR 0.64; 95%CI: 0.48, 0.87).  For ABC there were significant differences between AIs and tamoxifen for: clinical benefit (RR1.19, 95%CI:1.09, 1.3); complete response (RR 1.82, 95%CI:1.22, 2.73); objective response (RR 1.58, 95%CI: 1.28, 1.97); partial response (RR1.2, 95%CI 1.02, 1.41) and disease progression (RR0.83, 95%CI:0.76, 0.92). Adjusted indirect comparison of individual AIs were conducted with no significant differences identified. CONCLUSIONS: Treatment with third-generation AIs rather than tamoxifen provides significantly better outcomes for ABC and for EBC at two and five years follow up.