Results of adjuvant dose intensification studies in patients with localised breast cancer have raised questions regarding the clinical usefulness of this treatment strategy. Here, we develop and fit a natural history model for the time to clinical tumour recurrence as a function of the number of involved lymph nodes, and derive plausible predictions of the effects of dose intensification under various conditions. The time to tumour recurrence is assumed to depend on the residual postoperative micrometastatic burden of tumour, the fractional reduction of residual tumour burden (RTB) by treatment, and the rate of regrowth of the RTB to a clinically detectable size. It is assumed that a proportion of micrometastatic tumours are unresponsive to adjuvant chemotherapy even at maximal dose intensity. Data fitted included the San Antonio Cancer Institute (SACI) database of untreated patients, and CALGB #9082, a study comparing a highly intensive and moderately intensity adjuvant regimen in patients with 10+ positive axillary nodes. The proportion of tumours unresponsive to maximally intensive adjuvant treatment is estimated to be 48% (29-67%). The estimated log kill for intermediate-dose therapy from CALGB #9082 was 6.5 logs, compared with 9 logs or greater for high-dose therapy. The model is consistent with a modest but nonnegligible advantage of dose intensification compared with standard therapies in patients with sensitive tumours who have 10+ positive axillary nodes, and suggests that much of this clinical benefit could be achieved using intermediate levels of treatment intensification. The model further suggests that, in patients with fewer than 10 involved axillary nodes, any advantage of treatment intensification over standard therapy would be much reduced, because in patients with smaller tumour burdens of sensitive tumour, a larger proportion of cures achievable with intensified therapy could be achieved as well with standard therapy.