We have established an ultra-sensitive method for determination of ethanol in whole blood by headspace capillary gas chromatography (GC) with cryogenic oven trapping. After heating a blood sample containing ethanol and isobutyl alcohol (internal standard, IS) in a 7.0-ml vial at 55 degrees C for 15 min, 5 ml of the headspace vapor was drawn into a glass syringe and injected into a GC port. All vapor was introduced into an Rtx-BAC2 wide-bore capillary column in the splitless mode at -60 degrees C oven temperature to trap entire analytes, and then the oven temperature was programmed up to 240 degrees C for GC measurements with flame ionization detection. The present method gave sharp peaks of ethanol and IS, and low background noise for whole blood samples. The mean partition into the gaseous phase for ethanol and IS was 3.06+/-0.733 and 8.33+/-2.19%, respectively. The calibration curves showed linearity in the range 0.02-5.0 microg/ml whole blood. The detection limit was estimated to be 0.01 microg/ml. The coefficients of intra-day and inter-day variation for spiked ethanol were 8.72 and 9.47%, respectively. Because of the extremely high sensitivity, we could measure low levels of endogenous ethanol in whole blood of subjects without drinking. The concentration of endogenous ethanol measured for 10 subjects under uncontrolled conditions varied from 0 to 0.377 microg/ml (mean, 0.180 microg/ml). Data on the diurnal changes of endogenous ethanol in whole blood of five subjects under strict food control are also presented; they are in accordance with the idea that endogenous blood ethanol is of enteric bacterial origin.