EVIDENCE LIBRARY
Can Repeat Alcohol Tests Give Different Pharmacokinetic Estimates? Breath, Blood and Urine Explained
A controlled repeat-dose study explains why breath and venous blood can yield different ethanol kinetic estimates—and why model precision is not a DUI BAC prediction interval.
Can repeat alcohol experiments yield different pharmacokinetic estimates even when the dose and person are the same? Yes. In a controlled study of 16 healthy adults tested twice, estimates of some ethanol distribution parameters changed between experimental days and differed according to whether the mathematical model used breath or venous-blood measurements. The estimated steady-state distribution volume was comparatively reproducible, but that finding does not turn breath results into directly measured blood alcohol, establish a personal elimination rate, or validate backward reconstruction of alcohol concentration in an individual legal case.
The primary evidence is Norberg, Gabrielsson, Jones and Hahn, British Journal of Clinical Pharmacology (2000), “Within- and between-subject variations in pharmacokinetic parameters of ethanol by analysis of breath, venous blood and urine”. Their experiment examined repeatability of fitted pharmacokinetic parameters following intravenous ethanol—not drivers’ observed performance, an oral-drinking trial, or a validation of a forensic back-calculation method.
What did the researchers actually test?
Eight women and eight men participated. Each received ethanol at a target dose of 0.4 grams per kilogram of body weight as a controlled 30-minute intravenous infusion on two occasions approximately two days apart, after overnight fasting. The investigators studied participants for six hours and measured ethanol repeatedly in expired air and venous blood, as well as ethanol appearing in collected urine. That yielded 32 dosing sessions, although two participants’ atypical venous profiles were excluded from certain formal breath-versus-blood comparisons, leaving 14 for those comparisons.
The intravenous route was deliberately selected to remove variability caused by gastric emptying, intestinal absorption, beverages, and first-pass metabolism. This improves the interpretation of a controlled distribution-and-elimination experiment but makes it less directly transferable to ordinary oral drinking, in which each of those processes may matter. The studied participants were selected healthy volunteers, not injured patients, people with advanced liver disease, or a sample representing drivers in forensic investigations.
Were breath, blood, and urine measured using the same method?
No. Venous blood was obtained from the arm opposite the infusion and analyzed using headspace gas chromatography. The study reports an analytical coefficient of variation of 1.6% at approximately 0.4 g/L. Breath alcohol was tested with a DataMaster infrared instrument; duplicate breath readings were obtained about 60 seconds apart, and the paper reports a breath analytical coefficient of variation of 1.5% at an equivalent concentration near 0.4 g/L. Urine volume and ethanol concentration were measured to estimate cumulative renal elimination.
The authors normalized reported concentrations to the water fraction of blood to account for subject and infusion-related differences in blood water content. That distinction is easy to miss: a water-fraction-normalized ethanol concentration is not numerically identical by definition to an unadjusted whole-blood concentration as conventionally reported in a forensic laboratory. Their breath model also required a blood-to-breath conversion assumption. These conversions should not be hidden when laboratory quantities are compared in litigation.
| Evidence stream | Observation or procedure | What it establishes—and does not |
|---|---|---|
| Intravenous dose | 0.4 g/kg target, infused over 30 minutes; 16 adults tested twice | Controlled systemic ethanol input; does not recreate oral absorption or post-drinking timing. |
| Venous blood | Serial samples, headspace GC; study analytic CV 1.6% near 0.4 g/L | Measured ethanol in sampled venous specimens; not an automatic driving-time or arterial value. |
| Breath | Serial duplicate DataMaster infrared measurements; CV 1.5% near the stated equivalent concentration | Measured expired-air ethanol, subsequently interpreted through a blood–breath relationship; not direct venous-blood analysis. |
| Urine | Collected volumes and analyzed ethanol, with some incomplete emptying | Observed ethanol excretion over intervals; not a synchronized blood concentration or estimate of personal impairment. |
| Mathematical model | Two compartments, capacity-limited metabolic elimination and first-order renal clearance | Fitted parameters conditional on assumptions and input data; not directly measured anatomical volumes or enzyme activities. |
How did the model distinguish measured concentrations from calculated parameters?
Researchers fit breath- or blood-derived concentration–time profiles and urinary excretion to a two-compartment model. The parameters included a maximum modeled metabolic rate (Vmax), the concentration associated with half that modeled rate (Km), intercompartmental distribution clearance, central and tissue compartment volumes, and renal clearance. The steady-state model distribution volume, Vss, was computed as the sum of central and tissue distribution volumes.
These parameters were estimated, not directly assayed. Venous GC quantified ethanol in a blood specimen. A breath instrument assessed expired-air ethanol. The model then used repeated measurements, infusion details and its specified kinetic structure to estimate volumes and clearances. A fitted Vmax is not a measured individual criminal-case elimination slope; a fitted volume compartment is not a separately measured organ or physical body-water volume.
That difference explains why two data streams can support good overall curve fitting while still yielding materially different distribution-parameter estimates. Model fit alone cannot establish that the model is uniquely correct or that an individual parameter will transport unchanged to a different route, date, specimen or clinical condition.
What numerical differences mattered in the repeated experiments?
Using venous-blood data, the mean modeled maximum metabolic rate was 95 ± 25 mg/min, and mean Km was 27 ± 19 mg/L. The reported mean modeled steady-state distribution volume was 35.8 ± 4.3 L. Those are study parameter summaries, not constants to assign to every person.
Among the 14 participants retained for the formal paired comparison, the model’s breath-derived central distribution volume was on average 21% lower than the venous-derived estimate, while its tissue distribution volume was 10% higher. Model-derived central volume was also approximately 15% lower on the later test day, and model-derived intercompartmental clearance approximately 9% higher. These differences describe parameter estimates under this experimental model, not proportional changes in actual bodily organs.
| Reported result | Source value | Forensic interpretation |
|---|---|---|
| Mean venous-data Vmax | 95 ± 25 mg/min | Population summary of fitted maximum rate; not a universal BAC decline per hour. |
| Mean venous-data Km | 27 ± 19 mg/L | Fitted kinetic parameter, relatively uncertain and sensitive to late measurements. |
| Mean modeled Vss | 35.8 ± 4.3 L | Model distribution volume; not independently confirmed total body water. |
| Within-subject share of total Vss variation | 2% | Component of the study’s variance decomposition; not “2% error” for every individual BAC estimate. |
| Breath vs venous model volumes | Central 21% lower; tissue 10% higher in breath-derived fit | Sampling site and conversion assumptions affect inferred distribution parameters. |
| Alcohol recovered unchanged in urine | 1.52 ± 0.71% of infused dose | Small renal recovery in this experiment; not a person-specific correction to BAC. |
Does a 2% within-person variation mean ethanol distribution can be predicted within 2%?
No. The 2% figure is the share of the study’s total variance in fitted Vss assigned to within-subject variation under its variance-components analysis. It is not an individual prediction interval, not analytical measurement uncertainty, not a coefficient of variation for a single breath test, and not permission to infer a prior BAC with a ±2% error bar. The study separately reports substantial between-person and other variation in several kinetic parameters.
Similarly, the authors’ suggestion that ethanol dilution might serve as an index of total body water was explicitly provisional. They stated that comparison with an established total-body-water marker was still required. Using Vss as a clinical body-water method is a different question from applying a blood-breath partition ratio or a Widmark factor in a contested individual forensic case.
Why did breath and venous blood produce different modeled volumes?
The authors proposed that regional tissue drainage and local circulation can influence a peripheral venous sample, while breath measurements may track a different aspect of circulation. The data also required a blood–breath partition assumption. The investigators used 2,300 in modeling and reported an empirical partition estimate of 2,272 ± 117 in 28 sessions with 1,445 paired blood-breath observations. Such a group finding does not fix any one person’s instantaneous partition ratio.
The study found two markedly atypical venous profiles that did not display similarly distorted breath curves. The researchers excluded those two volunteers from certain comparative statistics rather than presenting their data as interchangeable with ordinary curves. The proposed circulatory explanation is a hypothesis, not direct proof of a particular cause in those volunteers.
For context on instrumentation, see the site’s separate guides to breath alcohol measurement and blood alcohol analysis by headspace chromatography. Those methods answer different immediate measurement questions, regardless of any later model fitted to the results.
What did urine add—and what could it not tell the investigators?
Urine measurements enabled the researchers to include renal clearance explicitly rather than assigning all loss of ethanol to metabolic elimination. The dose recovered unchanged in urine averaged about 1.5%, with substantial variation; renal clearance estimates were sensitive to differing urine flow. Urine was collected less frequently than breath and blood, and five participants had residual urine that complicated quantitative recovery.
Urine concentration therefore did not supply a direct contemporaneous venous BAC, and urinary recovery could not validate a prior drinking-time concentration. The paper’s observed small renal fraction applies to this intravenous protocol and dose; it does not justify a fixed correction for every clinical or legal case.
Did ethanol elimination stay identical between test days?
Not every kinetic quantity changed in the same way. In this study, estimates of the maximum metabolic rate were relatively reproducible over the two closely spaced sessions, while several distribution estimates changed with test day and sampling method. But the fitted parameters Vmax and Km were highly correlated, and late, low-concentration measurements were particularly influential for Km. Such statistical coupling complicates any attempt to treat each fitted parameter as an independently known physical constant.
Across the 32 sessions, the coefficient of variation across breath concentration measurements at a particular postinfusion time increased from approximately 14% at 60 minutes to 49% at 240 minutes; the analogous venous values were 14% and 45%. These larger late-time figures describe variability in the concentration profiles across study sessions and subjects, not a deterioration of the gas chromatograph’s analytical CV to 45% or of the breath instrument’s analytic CV to 49%. Distinguishing biological/time-profile variability from method precision is essential when reading the paper.
Can the study support retrograde BAC extrapolation in a DUI case?
It supplies important cautionary evidence, not an individually validated reconstruction algorithm. The experiment used carefully measured intravenous dosing, frequent samples over six hours, standardized fasting conditions and fitted nonlinear kinetics. Typical forensic disputes may instead involve unknown oral doses, staggered consumption, uncertain absorption completion, variable sampling times, potential medical or postcollection confounders, and no repeated concentration–time curve for the person at issue.
A proper forensic evaluation should separately identify the actual specimen and collection time, the direct analytical method and result, whether units describe whole blood or a normalized water fraction, any breath partition assumption, the assumed absorption phase, uncertainty in model parameters, and corroborating behavioral or event evidence. Replacing an observed blood or breath value with a fitted parameter estimate does not convert that estimate into a direct sample from the earlier event.
The broader scientific framework is explained in the site’s pharmacokinetics and forensic interpretation guide. That framework is not a substitute for an individualized reconstruction with actual case records and disclosed assumptions.
What are the most consequential limitations?
The sample was small (16 healthy volunteers), and only 14 were included in specific model comparisons after exclusion of two atypical venous profiles. The study involved intravenous ethanol, not drinking. Sessions were only about two days apart, so the results do not characterize longer-term variability or disease-related effects. Repeatable fitted volumes depend on the chosen two-compartment structure, the weighting of low concentrations and blood-water normalization. Expired-air measurements depend on instrument performance and a blood–breath ratio assumption. Urine recovery was incomplete in some participants.
The paper did not validate individual forensic impairment thresholds, estimate crash causation, test actual driving, measure cognitive or motor function, or demonstrate a legally applicable blood-alcohol back-calculation margin. Its conclusions about a possible total-body-water index were expressly subject to further external validation. These are limits on the scope of inference, not reasons to dismiss the underlying controlled measurements.
Frequently asked questions
Did the study directly measure blood alcohol?
Yes. Serial venous-blood ethanol concentrations were measured using headspace gas chromatography. For comparison and model fitting, the authors also reported water-fraction-adjusted concentrations, which must not be silently equated with unadjusted whole-blood units.
Was breath alcohol the same thing as venous BAC?
No. Expired-air ethanol was separately measured with an infrared instrument and interpreted through a blood–breath conversion assumption. The two data streams did not always yield the same fitted distribution parameters.
Did the researchers find a universal ethanol elimination rate?
No. They estimated model parameters in a 16-person controlled experiment. The mean fitted Vmax is not an individual forensic elimination slope and cannot simply be assigned to an unrelated person.
Can 2% model variation be used as the uncertainty of a legal BAC?
No. It describes a portion of the fitted distribution-volume variance within the study, not the uncertainty of a direct legal BAC measurement or a reconstructed earlier BAC.
How much infused ethanol was recovered in urine?
Approximately 1.52% on average, with a standard deviation of 0.71 percentage points. The amount is protocol-specific and does not make urine concentration interchangeable with blood ethanol.
Why did venous and breath models disagree?
Potential contributing factors include differences in the physiological compartments reflected by measurements, time-dependent distribution and the assumed blood–breath ratio. The study could not uniquely prove one mechanism for every discrepancy.
Does this study establish that a particular person was impaired or caused a collision?
No. It examined ethanol measurement and mathematical pharmacokinetics, not road performance or case-level causation. Those claims require independent evidence and cannot be drawn from the reported model means alone.
Primary scientific source
Norberg Å, Gabrielsson J, Jones AW, Hahn RG. Within- and between-subject variations in pharmacokinetic parameters of ethanol by analysis of breath, venous blood and urine. British Journal of Clinical Pharmacology. 2000;49(5):399–408. DOI: 10.1046/j.1365-2125.2000.00194.x. The conclusions above distinguish the authors’ directly measured venous, breath and urinary ethanol from fitted compartmental parameters and from legal or behavioral determinations the experiment did not test.