Simulated dialysis of whole blood was used to determine whether membrane factors (biocompatibility), endotoxin (ET) membrane diffusion, or transmembrane monocyte-ET interactions would stimulate tumor necrosis factor (TNF alpha) release. Whole blood containing EDTA and aprotinin was recirculated in the blood compartment of hollow fiber dialyzers containing either regenerated cellulose or polysulfone membranes. ET-free and ET-spiked dialysate were recirculated consecutively in the dialysate compartment for 30 min each. Blood and dialysate samples were collected at to and after each 30 min of simulated dialysis for determination of TNF alpha and ET concentrations. TNF alpha was not detected in any blood samples collected after simulated dialysis with regenerated cellulose (RC) membranes and ET-free or ET-spiked dialysate. However, blood ET concentrations, as determined by the Limulus amebocyte lysate (LAL) assay, increased in RC dialyzers after each 30 min of simulated dialysis even with ET-free dialysate. Since TNF alpha was not detected in these blood samples, the material detected by the LAL assay probably was not ET but an LAL-reactive material. After simulated dialysis with polysulfone dialyzers and ET-free dialysate, TNF alpha and ET were not detected in blood samples. ET also was not detected in blood samples after dialysis with ET-spiked dialysate. However, TNF alpha was detected in 7 of 13 (54%) of the blood samples following the 500 ng/ml of ET dialysate spike. TNF alpha release during simulated dialysis with polysulfone membranes and ET-contaminated dialysate may be due to transmembrane stimulation of circulating mononuclear cells and not diffusion of ET across the membrane.