Discussion

Analytical Performance

The analyzers exhibited good performance at measuring vaporous ethanol in vitro, especially considering the low cost of the devices. At the 0.080 g/210L concentration, for example, both the C6 and the C8 showed levels of uncertainty similar to those found in more advanced breath alcohol analyzers used for evidential purposes \cite{Hwang_2016,Brockley_Drinkman_2019}.
In the United States, the National Highway Traffic Safety Administration (NHTSA) promulgates performance recommendations for alcohol screening devices (ASDs) \cite{NationalHighwayTrafficSafetyAdministration2008}. The NHTSA's total allowable error for ASDs at the 0.020 g/210L ethanol vapor concentration is ± 0.012 g/210L at the 95% coverage interval. The analyzers examined met this requirement, showing a total error of \(\le\) ± 0.003 g/210L at 0.020 g/210L concentration, although they are not listed on the NHTSA's ASDs conforming products list \cite{2012}.  
For in vivo use, manufacturers of these devices should consider programing the instrument for a duplicate breath test sequence, reporting the mean of the two measurements along with the associated uncertainty. An uncertainty function could be built into the software of the smartphone app to provide users with additional information about the uncertainty of the measurements obtained \cite{Gullberg_2011}. Also, calibration reports could be recorded by the smartphone app and notify the user when recalibration is needed.  

Potential Interfering Substances

Potential interfering substances are volatile organic substances other than ethanol on a person’s breath which has the potential to interfere with the accurate analysis of vaporous ethanol \cite{Gullberg1994,Jones1996,Jones1989,Jones2008a,Logan_1994}. These substances have an impairing effect similar to or greater than ethanol \cite{Caldwell1997-in,Cowan1990-xl}. Normally, potential interfering substances are found in such low concentrations that they are unlikely to interfere with an ethanol breath test \cite{Flores1985a}. However, there are some circumstances in which interfering substances may be present in high enough concentrations that they may falsely elevate an ethanol breath test \cite{Caravati2010,Jones2007,Norfolk1997}. The analyzers examined do not have an interfering substance detection mechanism such as those found in more advanced electrochemical analyzers \cite{Chan_2018}.

Acetone

Acetone has been found in the breath of people with diabetes, during times of fasting, and in very low carbohydrate dieters \cite{Ruzsanyi2017}. Electrochemical fuel cell breath alcohol analyzers are known to be unaffected by acetone \cite{Falkensson1989}. The analyzers examined in this study use a fuel cell and did not respond to acetone.

Isopropanol

Isopropanol may be present in elevated concentrations on a person’s breath after drinking denatured alcohol \cite{Jones1989a,Jones_1992,Logan_1994} or produced endogenously from the biotransformation of acetone to isopropanol \cite{Jones1995a}. In one case study, a self-reported teetotaler obtained a false positive result on an electrochemical fuel cell breath alcohol analyzer after following a very low carbohydrate ketogenic diet \cite{Jones2007}. In another study, isopropanol in the breath was found to be elevated after eating a ketogenic meal \cite{Li2017}. The analyzers showed an apparent ethanol response to isopropanol. Users engaged in very low carbohydrate ketogenic dieting should be aware of the possibility of obtaining elevated BrAC results based on their diet.

Methanol

Methanol may be present in elevated concentrations on a person’s breath after consuming large amounts of fruit \cite{Lindinger1997}, in alcoholics \cite{Wigmore2008,Jones1988}, or through accidental exposure due to the improper production of distilled spirits \cite{Kane1968,Paasma2007,Aghababaeian_2019}. There are two unfortunate cases reported in the literature where a breath alcohol analyzer mistook methanol for ethanol, delaying medical treatment, resulting in the subjects dying from methanol poisoning \cite{Jones1989}. Users of these devices should be aware of the potential, but the unlikely possibility of elevated BrAC results due to methanol.

Limitations

Vaporous ethanol reference material produced by breath simulators cannot account for the complex physiologic gas exchange taking place in the lungs and airways of live subjects \cite{Jones1990,Gullberg1990,Anderson2003,Lubkin1996,Wilson1986,Vosk2014}. The SD of measurements taken in vivo has been shown to be greater than the SD produced by breath simulators \cite{Gullberg1989}. Further research is needed to determine the measurement uncertainty for in vivo results.
The calibration longevity of the analyzers was not examined in this study. An important consideration for those wishing to use these instruments is that the device must be sent back to the manufacturer regularly for recalibration. Individuals or institutions using these instruments may need to keep several on hand while periodic recalibrations are performed. 
Users of these instruments should incorporate quality assurance practices to ensure the accuracy meets the requirements of the intended use  \cite{Dubowski1994}. The use of compressed ethanol-gas reference standards would be a convenient way to perform accuracy checks \cite{Dubowski1996-dw,Silverman1997-zs}. Further investigation with these analyzers using compressed ethanol-gas standards is needed, as the efficacy was not assessed in this study.

Conclusions

The breath alcohol analyzers examined in this study showed the ability to measure vaporous ethanol with confidence in the results, especially at concentrations \(\le\) 0.080 g/210L. At the 0.080 g/210L ethanol vapor concentration, the combined expanded measurement uncertainty was \(\le\) ± 0.013 g/210L at the 95% coverage interval for all instruments. The likelihood of false readings from potential interfering substances appears to be small but may be a concern for those engaged in ketogenic diets with elevated levels of isopropanol. More work needs to be conducted with these instruments in vivo to determine the measurement uncertainty for the results which include a biological component.