Validation of a prognostic index in breast cancer

The validation of a prognostic index for patients with primary breast cancer is described. The actual survival data in a group of 383 patients was compared with expected survival predicted by an index based on axillary lymph node status, oestrogen receptor, progesterone receptor and age. There was no significant difference between actual and predicted deaths for five ranges of index value at 2 years and for the three highest ranges of index value at 5 years. However, in the two lowest ranges, the index significantly underpredicted deaths at 5 years. The Cox proportional hazards model was used to determine if there was any significantly better combination of coefficients and covariates to predict survival in the test group. The original index--1 = N + E + P + A, where N = 0 if no nodes are involved, 13 if one to three nodes are involved, and 31 if more than three nodes are involved; E = 15 if oestrogen receptor value is less than 10 fmol/mg cytosol protein and 0 otherwise; P = 12.5 if progesterone receptor value is less than 10 fmol/mg cytosol protein and 0 otherwise; and A = number of years over age 65--was as good as any competing model in ranking survival prospects in the test group. However, it was a less sensitive predictor in this group than in the original set of patients. A second index based on tumour size, hormone receptor status and age, for use when lymph node status was unavailable, was tested by the same method and validated.(ABSTRACT TRUNCATED AT 250 WORDS)