Aim: The study aim was to determine the frequency with which women decline clinicians' treatment recommendations and variations in this frequency by age, cancer and service descriptors. Design: The study included 36,775 women diagnosed with early invasive breast cancer in 1998-2005 and attending Australian and New Zealand breast surgeons. Rate ratios for declining treatment were examined by descriptor, using bilateral and multiple logistic regression analyses. Proportional hazards regression was used in exploratory analyses of associations with breast cancer death. Results: 3.4% of women declined a recommended treatment of some type, ranging from 2.6% for women under 40 years to 5.8% for those aged 80 years or more, and with parallel increases by age presenting for declining radiotherapy (p<0.001) and axillary surgery (p=0.006). Multiple regression confirmed that common predictors of declining various treatments included low surgeon case load, treatment outside major city centres, and older age. Histological features suggesting a favourable prognosis were often predictive of declining various treatments, although reverse findings also applied with women with positive nodal status being more likely to decline a mastectomy and those with larger tumours more likely to decline chemotherapy. While survival analyses lacked statistical power due to small numbers, higher risks of breast cancer death were suggested, after adjusting for age and conventional clinical risk factors, (1) for women not receiving breast surgery for unstated reasons (RR=2.29; p<0.001); and (2) although not approaching statistical significance p≥ 0.200), for women declining radiotherapy (RR=1.22), a systemic therapy (RR1.11), and more specifically, chemotherapy (RR=1.41). Conclusions: Women have the right to choose their treatments but reasons for declining recommendations require further study to ensure that choices are well informed and clinical outcomes are optimized.