A number of different factors are known to be correlated with survival of patients with breast cancer. Among these are lymph node status, tumour size, oestrogen receptor (ER), progesterone receptor (PR) and androgen receptor (AR) status. The purpose of this study was to investigate the relative significance of these factors and use this information to construct a prognostic index capable of predicting survival. These factors together with age and menopausal status were studied and correlated with prognosis in 796 women with primary breast cancer. The data were analysed in a stepwise manner by the Cox proportional hazards regression technique. Statistically, greater than 3 nodes involved gave the worst prognosis (P less than 0.001). This was followed by ER if less than 10 fmol/mg cytosol protein (P less than 0.001), PR if less than 10 fmol (P less than 0.01), greater than 0 lymph nodes involved (P less than 0.01) and the number of years over age 65 (P less than 0.01). When these factors were accounted for, tumour size, menopausal status and AR did not significantly improve prediction of survival. The significant factors were incorporated into a prognostic index: I = N + E + P + A, where N = 0 if no nodes involved, 13 (if 1-3 nodes involved and 31 if greater than 3 nodes involved, E = 15 if ER less than 10 fmol, P = 12.5 if PR less than 10 fmol and A = number of years over 65. Using this index five year survival curves were constructed corresponding to groups of patients with widely differing prognoses. Predicted five year survival ranged from 96 to 12 per cent.