To determine whether perioperative blood transfusion affected the recurrence rate of squamous cell cancer of the head and neck, we performed a retrospective study of all patients with stage III and IV disease treated surgically at the University of North Carolina, Chapel Hill, between 1983 and 1986. Those who recurred were compared with those who did not in regard to 16 prognostic variables, including whether or not they had received a perioperative blood transfusion. Analyzing each variable separately, five were significantly related to recurrence. These were (1) surgical margin status, (2) stage, (3) presence of pathologically positive nodes, (4) blood transfusion status, and (5) type of treatment. However, because several of these variables were clearly interrelated, the same data were subjected to a multivariate regression analysis specifically designed to identify significant prognostic variables independent of their association with other variables. By this analysis, only margin status and the presence or absence of a blood transfusion were statistically significant predictors of recurrence.