Our aim is to improve prediction models for regional and distant metastatic spread in breast cancer in order to individualise treatment selection for the individual patients. The prediction models integrate patient´s, clinical and tumour related data to achieve high accuracy. Archival tumour tissue from primary tumours and metastasis from patients included in clinical trials is characterised on protein and genomic level and the results linked to clinical outcome. By comparing the characteristics in primary tumours with those in lymph node metastasis and distant recurrences, respectively, change of biological features during tumour progression can be linked to clinically relevant information. One prediction model aims to preoperatively identify node-negative breast cancer patients for whom axillary surgery could be abstained. By integrating biological and radiological features this model can be developed into a clinically useful tool for the treating physician. Our groups is multidisciplinary including clinical PhD students and clinical researchers along with a biostatistician, molecular biologist and technical personnel.