A number of trials have studied the value of aromatase inhibitors (AIs) for the adjuvant treatment of early hormone-responsive postmenopausal breast cancer. Three different AIs have been used and they have been compared as initial treatment (two trials) or after 2-3 years of tamoxifen (four trials), in both cases against a standard arm of 5 years of tamoxifen. In addition, two trials have evaluated AIs against no treatment after 5 years of tamoxifen. In all circumstances, the AIs have demonstrated superior efficacy. However, no results are currently available for the key question, that is - is it better to start initially with an AI or use it sequentially after 2 years of tamoxifen? Here, we review the trial results and present two models, which address this issue. The models clearly show that early treatment with an AI is superior to using it after 5 years of tamoxifen. They also favour an upfront strategy to sequencing after 2 years of tamoxifen, but in this case the differences are small and model-dependent. A key question is whether AIs have substantially better efficacy than tamoxifen for ER-positive-PgR-negative tumours, where the data are currently contradictory. A mechanism explaining why greater efficacy might be so is proposed. Further results from ongoing trials will be needed to resolve this issue.