An uncertainty budget was constructed for the measurement of ethanol in blood by headspace gas chromatography. The uncertainty budget, covering the analytical range of ethanol concentrations up to 3.00 g/kg, included analytical uncertainty components, traceability uncertainty components, and effects caused by interindividual variation in blood water content. The analytical combined standard uncertainty was estimated from duplicate measurements of real samples and included contributions from headspace recovery, variation between columns, injection, repeatability of analytical signals, and statistical uncertainty of the calibration function. The traceability uncertainty was estimated in a sub-budget based on information about the calibrator and about the preparation of the aqueous standards. Two uncertainty components depended on the interindividual variation in blood water content. First, it caused uncertainty on the density of the blood, and second, it had an effect on the gas phase concentration of ethanol when doing the headspace sampling. These effects as well as their covariance were included in the uncertainty budget. For fresh blood samples, the analytical uncertainty was the dominating uncertainty component, accounting for approximately 90% of the variance. For blood samples collected 100 h postmortem, the interindividual variation in blood water content was the largest uncertainty component. It was demonstrated that subtracting a "safety margin" of 0.1 g/kg from the results was sufficient to keep the risk of committing a type 1 error below 0.1% in ethanol concentrations ranging up to 2 g/kg for fresh blood samples. This risk was higher for postmortem blood samples because of the higher uncertainty of measurement, but still less than approximately 1.4%.