Although clinical response to primary chemotherapy in stage II and III breast cancer is associated with a survival advantage, it is the degree of pathological response in the breast and ipsilateral axilla that best identifies patients with a good long-term outcome. A mathematical model of the initial response of 39 locally advanced tumours to anthracycline-based primary chemotherapy has been previously shown to predict subsequent clinical tumour size. This model allows for the possibility of primary resistant disease, the presence of which should therefore be associated with a worse outcome. This study reports the application of this model to an additional five patients with locally advanced breast cancer, as well as to 63 patients with operable breast cancer, and confirms the biological reality of the model parameters for these 100 breast cancers treated with primary anthracycline-based chemotherapy. The tumours that responded to chemotherapy had higher cell-kill (P < 0.0005), lower resistance (P < 0.0001) and slower tumour regrowth (P < 0.002). Furthermore, ER-negative tumours had higher cell-kill (P < 0.05), as compared with ER-positive tumours. All patients with a pathological complete response had zero resistance according to the model. Furthermore, the long-term implication of chemo-resistant disease was demonstrated by survival analysis of these two groups of patients. At a median follow-up of 3.7 years, there was a statistically significantly worse survival for the 37 patients with locally advanced breast cancer identified by the model to have more than 8% primary resistant tumour (P < 0.003). The specificity of this putative prognostic indicator was confirmed in the 63 patients presenting with operable disease where, at a median follow-up of 7.7 years, those women with a resistant fraction of greater than 8% had a significantly worse survival (P < 0.05). Application of this model to patients treated with neoadjuvant chemotherapy may allow earlier identification of clinically significant resistance and permit intervention with alternative non-cross-resistant therapies such as taxoids.