Allogeneic blood transfusion has been implicated as an independent risk factor for postoperative bacterial infection in clinical and animal studies. The association among transfusion, quantitative immunologic factors, and infection was examined in 102 patients undergoing 109 spinal fusion procedures. In 60 procedures, patients received autologous blood only; in 24 procedures, they received at least 1 unit of allogeneic blood, and in 25 procedures, they received no transfusions. Twenty-two patients developed bacterial infections, in 8 cases while in hospital and in 14 cases after discharge. Univariate analysis revealed that patients who received any allogeneic blood and those who received no allogeneic blood differed significantly in the rate of hospital-acquired infection (20.8 vs. 3.5%), length of stay (12.3 vs. 9.7 days), days of fever greater than or equal to 38 degrees C (4.0 vs. 2.9), days on antibiotics (3.9 vs. 2.5), duration of surgery (309 vs. 231 min), blood loss (1343 vs. 887 mL), surgeon, and postoperative drop in natural killer (NK) cells (-174 vs. -42/microL). Multivariate logistic and linear regressions revealed that the number of allogeneic units transfused was the only significant predictor of in-hospital infection (p = 0.016) or days on antibiotics and length of stay. None of the clinical, surgical, or transfusion variables was significantly associated with posthospital infection, although a significantly greater drop in NK cells had occurred in patients who developed infection (p = 0.0035). These data strongly implicate allogeneic transfusion as a risk factor for in-hospital postoperative bacterial infection.(ABSTRACT TRUNCATED AT 250 WORDS)