OBJECTIVES: This study aims to evaluate the cost-effectiveness of two aromatase inhibitors for adjuvant treatment of postmenopausal hormone receptor positive early breast cancer, and to address the most reasonable treatment option when population is stratified by nodal status. METHODS: A Markov model was developed defining six Markov states based on breast cancer progression. Annual probabilities of recurrence by adjuvant treatment, anastrozole, letrozole, and tamoxifen were estimated from published studies in overall population, node negative, and node positive group. Costs of defined breast cancer events were measured by micro-costing methods based on 2009 National Health Insurance Fee Schedule and the third Clinical Guideline of Breast Cancer Treatment. Anastrozole and letrozole were compared with tamoxifen respectively, using the same Markov model. Incremental cost-effectiveness ratios (ICERs) in overall population and each subgroup were estimated. RESULTS: When anastrozole was compared with tamoxifen, anastrozole was more effective and costly than tamoxifen costing additional KRW 22,461,689 per QALY. Letrozole showed similar incremental cost of KRW 21,004,142 per QALY. In the node negative group, anastrozole was the most cost-effective with incremental cost of KRW 19,717,770 per QALY, while letrozole was the most cost-effective with incremental cost of KRW 8,150,512 per QALY in node positive group. Sensitive analysis showed that these results were robust. CONCLUSIONS: Subgroup analysis clearly demonstrated which treatment was superior among aromatase inhibitors. Such a demonstration was not confirmative in the cases of overall population. The implication of this study is that the decision maker should be careful when generalizing the cost-effectiveness results. The stratified analysis in this context may help reach a reasonable decision on resource allocation.