To investigate the process of metastasis, primary clinical data and disease events such as metastases, local recurrence and survival (median follow-up 9.4 years) from the Munich Cancer Registry from 1978 to 1996 were analysed. Since metastases, even from small tumours, may be initiated before the diagnosis of the primary tumour, the growth of the primary tumour and metastasisation may be two autonomous processes. In our data, survival following metastases was almost unrelated to primary tumour size. However, the number of M1 cases and the time to metastasisation depended on the tumour diameter at diagnosis. The time from initiation of metastases to its diagnosis was estimated as 5.8 years. The growth of metastases was almost homogeneous. However, the growth time following metastasisation-depending on the metastases-free time, receptor status and histological grade-only varied by approximately a factor of 2. Local recurrence, above all, was an indicator of metastases. Furthermore, local recurrence may also have the potential to metastasise. Excess mortality due to local recurrence was estimated up to 9.3 years after diagnosis. Our hypothesised metastases model illustrates the importance of early detection, the concept of breast-conserving therapy and additional metastases from local recurrence. It highlights the benefits of optimal local therapy of the primary tumour and the limitations of systemic therapy. It also questions the use of axilla dissection and lymph node irradiation. Its generalisation to solid tumours may help to clarify many of the current controversial debates.