There is no satisfactory assay procedure of PAF in human whole blood in terms of sensitivity, reproducibility and simplicity. This is due to coexisting lipids from plasma and cellular membranes which inhibit measurement of PAF in various assay procedures, including bioassay. In the present study, an attempt was made to eliminate these interfering lipid inhibitors from blood samples. Lipids in human whole blood were extracted according to the method of Bligh & Dyer and the organic layer was dried under a stream of nitrogen. Then, the dried organic layer was dissolved in diethyl-ether and the solution was kept at -20 degrees C which was then centrifuged. The resulting supernatant was then applied to an anion-exchange column and the PAF fraction was obtained by step-wise gradient elution. The fraction was further purified by normal phase HPLC. Then PAF in the final sample was determined by sensitive bioassay using rabbit platelets containing fibrinogen and epinephrine. The recovery rate of PAF throughout this procedure was constant and satisfactory (37.4 +/- 9.7%), which was confirmed using [3H]-PAF. The lower limit of the present assay was estimated to be 5pg in 1 ml of blood and it was sensitive enough to detect PAF in blood samples from healthy volunteers and patients with sepsis or liver cirrhosis. Furthermore, attempts were made to compare the sensitivity and the recovery of our method with these of a commercially available radioimmunoassay (RIA) kit for PAF. However, it was not possible to detect any amount of authentic PAF added to whole blood.