Calculation and Verification of Blood Ethanol Measurement Uncertainty for Headspace Gas Chromatography

Abstract

An estimate was made of the measurement uncertainty for blood ethanol testing by headspace gas chromatography. While uncertainty often focuses on compliance to a single threshold level (0.08 g/100 mL), the existence of multiple thresholds, related to enhanced sentencing, subject age, or commercial vehicle licensure, necessitate the use of an estimate with validity across multiple specification levels. The uncertainty sources, in order of decreasing magnitude, were method reproducibility, linear calibration, recovery, calibrator preparation, reference material, and sample preparation. A large set of reproducibility data was evaluated (n = 15,433) in order to encompass measurement variability across multiple conditions, operators, instruments, concentrations and timeframes. The relative, combined standard uncertainty was calculated as ±2.7%, with an expanded uncertainty of ±8.2% (99.7% level of confidence, k = 3). Bias was separately evaluated through a recovery study using standard reference material from a national metrology institute. The uncertainty estimate was verified through the use of proficiency test (PT) results. Assigned values for PT results and their associated uncertainties were calculated as robust means (x*) and standard deviations (s*) of participant values. Performance scores demonstrated that the uncertainty estimate was appropriate across the full range of PT concentrations (0.010-0.370 g/100 mL). The use of PT data as an empirical estimate of uncertainty was not examined. Until providers of blood ethanol PT samples include details on how an assigned value is obtained along with its uncertainty and traceability, the use of PT data should be restricted to the role of verification of uncertainty estimates.