Ninety women with recently diagnosed stage I or stage II breast cancer who had been admitted to the NIH Clinical Center and were participating in a randomized trial were entered onto this behavioral immunology protocol. Patients were immunologically and psychosocially assessed at baseline (approximately 5 days after surgery) and again at 3 and 15 months post surgery. All of the patients were followed up for a minimum of 5 years, and 60% of the patients were followed for 7 years or longer. Twenty-nine women in the study group reported disease recurrences over the entire follow-up period. Causal path modeling statistical techniques showed that natural killer (NK) cell activity was a strong predictor of disease outcome when the outcome variable was defined as recurrence v nonrecurrence of disease (chi 2 = 6.9, p less than .001). Higher NK cell activity at follow-up predicted disease-free survival over the follow-up period. When the disease outcome variable was operationally defined as time to recurrent disease, the psychosocial factors were more strongly predictive of the rate of disease progression for those who had a recurrence (chi 2 = -4.1, p less than .01), but NK cell activity was seemingly less relevant in this latter case. Overall, these findings suggest that including mood and potentially relevant immunological variables, along with important biological prognostic variables, in multivariate and prospective models such as those examined in this study, potentially contributes more to the explanation of greater outcome variance of early-stage breast cancer than has been believed in the past.