Uncertainty estimation of ethanol concentration in simulated breath using enzyme-based biosensors

Abstract

Exhaled breath (EB) contains volatile and nonvolatile compounds that are correlated with physiological processes in the body, and these breath biomarkers hold enormous diagnostic potential when they are adequately measured and monitored. Thus, the development of instrumentation, including enzyme-based biosensors, for breath monitoring applications has been expanding rapidly. In this paper, the process of estimating the overall combined uncertainty in predicting ethanol concentration, u(Cv)pred, using a calibrated alcohol oxidase-based amperometric biosensor is presented. Components that contributed to u(Cv)pred were the standard uncertainties associated with simulation of a breath sample with trace ethanol concentration, sampling temperature, biosensor instrumentation, and regression analysis. In both EB and exhaled breath condensate (EBC) sensing, the largest contributor to overall uncertainty was the random effects captured by the regression model at 38.2 % and 39.8 %, respectively, for EB and EBC. This was followed by biosensor instrumentation (34.5 %) and simulation (25.3 %) in EB sensing. The trend was reversed in EBC sensing with EB simulation having a larger contribution (33.8 %) than biosensor instrumentation (25.5 %) owing to a better repeatability of amperometric measurements with aqueous samples. The remaining 2.0 % and 0.9 % were due to breath sampling temperatures in EB and EBC sensing, respectively. This study provides a framework for how to incorporate uncertainty estimation in both breath monitoring and is applicable to biosensing of other breath biomarkers.