OBJECTIVES: To establish the clinical and cost-effectiveness of aromatase inhibitors (AIs) anastrozole, letrozole and exemestane compared with tamoxifen in the adjuvant treatment of early oestrogen receptor-positive breast cancer in postmenopausal women. DATA SOURCES: Major electronic databases and three trials registers were searched from May to June 2005. Three conference abstract databases were searched in December 2005. Industry submissions. REVIEW METHODS: Studies evaluating the clinical effectiveness of AIs against 5 years' tamoxifen treatment were included and critically appraised. The review of the health economics of AIs in early breast cancer in comparison with standard therapies included a review of existing economic evaluations of the relevant therapies, a critique of each of the economic evaluations submitted to the National Institute for Health and Clinical Excellence (NICE) by pharmaceutical manufacturers and a detailed explanation of the methodologies and results of the authors' economic model. The three treatment strategies (primary adjuvant therapy, unplanned switch therapy and extended adjuvant therapy) were considered separately within the authors' economic analysis. RESULTS: A meta-analysis of three trials found a significant difference in overall survival when an unplanned anastrozole switching strategy was compared with 5 years' tamoxifen. Significant improvements in overall survival are yet to be demonstrated in other strategies. Compared with 5 years' tamoxifen, disease-free survival (disease recurrence or death from any cause) was significantly improved in the primary adjuvant setting with anastrozole and letrozole, and with an exemestane switching strategy. Other trials did not report this outcome. Breast cancer recurrence (censoring death as an event) was significantly improved with primary adjuvant anastrozole and letrozole, anastrozole switching, extended adjuvant anastrozole or letrozole. The AIs and tamoxifen have different side-effect profiles, with tamoxifen responsible for small but statistically significant increases in endometrial cancer and, sometimes, thromboembolic events and stroke. AIs show a trend towards increases in osteoporosis, the statistical significance of which increases with follow-up time. The absence of tamoxifen treatment also increases the risk of hypercholesterolaemia and cardiac events in postmenopausal women. There was no significant difference in overall health-related quality of life between standard treatment and either primary adjuvant anastrozole and extended adjuvant letrozole strategies. The cost-effectiveness results for AIs compared with tamoxifen in the primary adjuvant setting, are estimated to be between 21,000 pounds and 32,000 pounds per quality-adjusted life-year (QALY) based on an analysis over 35 years. There is currently no trial evidence for exemestane in this setting. The cost-effectiveness results for anastrozole and exemestane, compared with tamoxifen in the unplanned switching setting, are estimated to be 23,200 pounds and 19,200 pounds per QALY, respectively, based on an analysis over 35 years. There is currently no trial evidence for letrozole in this setting. In the extended adjuvant setting, the cost per QALY for letrozole compared with placebo is estimated to be 9800 pounds, based on an analysis over 35 years. All these results are considered to be conservative. In the base case it is assumed that the benefits of AIs over tamoxifen or placebo seen during the therapy period are gradually lost during the following 10 years. An alternative scenario, the 'benefits maintained' scenario, is tested in sensitivity analysis. Here it is assumed that following the treatment period the annual rate of recurrence in both arms is the same. This reduces the cost-effectiveness ratio by over 50%, to around 10,000-12,000 pounds, 5000 pounds and 3000 pounds in the primary adjuvant, unplanned switching and extended adjuvant setting, respectively. The limited evidence to date of benefits after the therapy period suggests that the 'benefits maintained' scenario may be realistic. The results from the economic analyses within the industry submissions are generally lower than the results from the authors' model and are close to or below 12,000 pounds in all three settings. The authors' analyses generally produce a lower estimate of QALY gain for the aromatase inhibitors, due to the more conservative assumption regarding benefits, along with differences in the utility values used in the their analysis. CONCLUSIONS: On the basis of the current data and within their licensed indications, AIs can be considered clinically effective compared with standard tamoxifen treatment. However, their long-term effects, in terms of both benefits and harms, remain unclear. Under the conservative assumption that benefits gained by AIs during the treatment period are gradually lost over the following 10 years, the cost per QALY for AIs compared with tamoxifen is estimated to be between 21,000 pounds and 32,000 pounds in the primary adjuvant setting and around 20,000 pounds in the unplanned switch setting. The cost per QALY for AIs compared with placebo in the extended adjuvant setting is estimated to be around 10,000 pounds. Under the less conservative assumption that rates of recurrence are the same in both arms after the therapy period is complete, the incremental cost-effectiveness ratios are typically at least 50% lower, suggesting that AIs are likely to be considered cost-effective in all three settings. Understanding of the long-term treatment effects on cost-effectiveness is, however, incomplete. Data on the impact of AIs on survival are awaited from the majority of the trials to confirm whether or not the benefits seen in disease-free survival and recurrence rates are translated into overall survival benefit in the medium to long-term.