I’ve written in the past about how most so-called "breathalyzers" do not measure alcohol: they actually measure the presence of the methyl group in chemical compounds. One of those compounds is ethyl alcohol (aka ethanol), and the machine simply assumes that the detected compound is ethyl alcohol.
Problem: there are thousands of compounds containing the methyl group — of which over one hundred have been found on the human breath. Breathing gasoline or paint fumes, for example, or merely absorbing the fumes through the skin, can create false breath test results for days afterwards. And I’ve posted that the problem is particularly acute when the suspect happens to be a diabetic, as diabetics often have high levels of acetone in their breath — a compound which contains the methyl group.
However, you don’t have to be a diabetic to have high levels of acetone: scientific research has established that acetone can exist in perfectly normal individuals at levels sufficient to cause false high breath-alcohol test readings. See "Excretion of Low-Molecular Weight Volatile Substances in Human Breath: Focus on Endogenous Ethanol", 9 Journal of Analytical Toxicology 246 (1985).
Fasting or radical dieting, such as with the Atkins diet, can also cause significantly elevated acetone. Studies have concluded that fasting can increase acetone in the body sufficient to obtain breathalyzer readings of .06% (this is cumulative — that is, the .06% will be added by the machine to any levels actually caused by alcohol or other compounds). See "The Likelihood of Acetone Interference in Breath Alcohol Measurement", 3 Alcohol, Drugs and Driving 1 (1987). And low-carbohydrate diets, such as Atkins, have long been associated with high levels of acetone production.
Of course, for many years law enforcement denied that any such problem existed, just as they denied that mouth alcohol and radio frequency interference caused false test results — until manufacturers started adding acetone detectors, mouth alcohol detectors and RFI detectors to their machines (none of which, unfortunately, have proven effective.)
How reliable are breathalyzers? "Close enough for government work". As I’ve posted, there seems to be a growing trend toward using blood rather breath analysis, including in some states letting officers draw the blood themselves (in at least two states, at the scene of arrest). Given the reassurances about these machines so often expressed publicly by law enforcement, one has to wonder why they are increasingly turning to the involved process of hypodermic needles, preservatives, anticoagulents, refrigeration and delayed laboratory analysis….