EVIDENCE LIBRARY
Does Total Body Water Predict Breath Alcohol Better Than Body Surface Area? A 77-Person Study
In a 77-person study, total body water predicted modeled breath alcohol better than body surface area. Learn what was measured, estimated, and not proved.
Short answer: In a controlled study of 77 adults given the same 36-gram alcohol dose, measured total body water (TBW) was more closely associated with breath alcohol concentration than body surface area (BSA), a quantity calculated from height and weight. A model using TBW also fit the study’s retrospectively estimated breath-alcohol values better than the comparison BSA models. That finding does not turn a person’s height, weight, or body-water reading into a validated breathalyzer, establish their blood alcohol concentration, or prove impaired driving. It addresses an important question about population-level estimation, not certainty about one person.
The evidence comes from Reiter, Boeckle, Reiter, and Seltenhammer’s 2020 controlled drinking study, “The impact of total body water on breath alcohol calculations”, published in Wiener klinische Wochenschrift. Its contribution is narrower, and more useful, than the familiar claim that people with different body compositions may produce different alcohol readings: it compares specific predictors against repeated breath measurements and a derived time-zero estimate.
What question did the investigators test?
The researchers asked whether an estimate based on an individual’s total body water would explain breath alcohol concentration (BrAC) better than body surface area. BSA is calculated from height and weight; it is not a direct measure of the water-containing compartments into which ethanol distributes. TBW was assessed with multifrequency hand-to-foot bioelectrical impedance analysis (BIA). Alcohol is water-soluble, so the distinction between body size and body composition can matter to concentration estimates.
That is a comparison of predictors. Neither TBW nor BSA is a laboratory measurement of ethanol. A predictor may correlate strongly with an alcohol result and still yield substantial error when used to estimate one person’s reading, especially when the dose, timing, absorption phase, and physiological conditions differ from those in the experiment.
How was the 77-person experiment conducted?
The cohort included 37 men and 40 women, described by the authors as healthy white participants. All drank 375 mL of white wine (12% alcohol by volume), equivalent to approximately 36 grams of ethanol, over 15 minutes. Before the session, participants were instructed to eat a standardized pasta meal more than four hours earlier. The fixed alcohol dose, rather than a dose adjusted to each participant’s body weight, made body-composition differences particularly relevant to the comparison.
Breath alcohol was analyzed with a Dräger Alcotest 7110 MKIII A using infrared and electrochemical systems. Before drinking, participants supplied baseline samples; thereafter, pairs of breath measurements were collected at 30-minute intervals through 150 minutes after drinking began. For the research analysis, the investigators averaged paired readings. They separately used multifrequency BIA to estimate each participant’s TBW.
The design is not interchangeable with a real-world roadside encounter: drinking dose, meal instructions, equipment, observation period, and collection schedule were controlled. No venous blood alcohol measurements, simulated-driving test, crash investigation, or individual impairment classification was reported for the study.
The study’s measurements versus its calculations
| Variable | How obtained | What it supports—and does not |
|---|---|---|
| BrAC (mg/L) | Repeated instrumental breath tests | An actual breath-alcohol measurement at a sampled time; not a measured venous BAC. |
| Total body water (liters) | Multifrequency BIA | An instrument-derived estimate of body water; not a measured alcohol concentration. |
| Body surface area (m²) | Formula using height and weight | A computed body-size proxy; not an individual ethanol distribution-volume measurement. |
| BrAC at theoretical time zero (BrAC0) | Extrapolation based on breath values at 90 and 150 minutes | A modeled intercept dependent on elimination assumptions; not a breath sample actually obtained at time zero. |
| Breath-alcohol elimination slope | Change between the later sampled breath values | An interval-dependent estimate of BrAC decline; not a directly measured blood-alcohol elimination rate. |
What were the main results?
In the full cohort, the calculated BrAC0 showed a stronger inverse association with TBW than with BSA. Because everyone received the same approximate ethanol dose, a larger distribution-related body-water measure tended to accompany a lower estimated breath concentration. The researchers reported Spearman correlations of −0.889 for TBW and −0.865 for BSA (both p ≤ .001). The direction is important: the magnitudes of the correlations were high, but the relationship was negative.
The association was also more pronounced for TBW in each analyzed sex group. Among women (n = 40), the coefficients were −0.835 for TBW and −0.801 for BSA; among men (n = 37), they were −0.759 and −0.682, respectively. Those numbers describe this sample. They are not accuracy or misclassification rates for identifying a person’s intoxication or measured BAC.
Model comparison: stronger association does not equal perfect prediction
| Study comparison | Reported finding | Forensic meaning |
|---|---|---|
| TBW versus BSA correlation with derived BrAC0 | Spearman −0.889 versus −0.865; n = 77 | TBW showed the stronger association under this fixed-dose protocol; neither value is an individual error bound. |
| Model using TBW alone | Corrected Akaike information criterion (AICc) = −470.75 | Best-scoring of the displayed candidate models within this dataset; not independent validation. |
| Model using BSA and sex | AICc = −468.55 | Close but less favorable model fit; an approximately 2.2-unit difference does not prove clinical or legal precision. |
| Model using BSA alone | AICc = −459.41 | Less favorable fit than the TBW model in these observations. |
| Mean estimated BrAC decline | Men 0.065 ± 0.011 and women 0.074 ± 0.017 mg/L/hour; p = .006 | Group averages from the study’s interval; not a universal slope for an individual or a blood BAC unit. |
The paper’s full results report the overall mean breath-elimination slope as 0.0699 ± 0.015 mg/L/hour. Other parts of the article report different plus/minus values for these averages; the table above deliberately uses the sex-specific standard deviations reported in the detailed results. The important evidentiary boundary is unaffected: even a carefully estimated average does not make a fixed retrograde-extrapolation rate correct for every person.
Why does a measured BrAC differ from an estimated BrAC0?
A breath instrument provides a concentration reading for the breath sample collected at a particular time, subject to instrument validation and sampling requirements. In contrast, BrAC0 in this study was calculated retrospectively from measurements at 90 and 150 minutes, using an assumption of approximately zero-order ethanol elimination during the chosen interval. The model also depends on the investigators’ attempt to analyze the post-absorptive period.
Consequently, a graph or regression showing good association with BrAC0 is not the same as a validation against a directly observed concentration at the hypothetical earlier moment. The authors did not independently test the formula on a second cohort with different drinking patterns or establish case-level prediction intervals that would support a precise person’s earlier BrAC. An observed association between body water and the modeled intercept cannot remove the uncertainty already present in that intercept.
This distinction is especially important if someone proposes to reconstruct a breath reading before or after the sampled time. Both the assumed absorption state and the elimination slope deserve examination. For general background on that issue, see the site’s pharmacokinetics and forensic interpretation guide.
Does body water change the result of a breathalyzer?
The study does not show that a breathalyzer mechanically adjusts its reading according to the subject’s measured body water. The instrument analyzes ethanol in an exhaled breath sample. TBW was a separate explanatory variable associated with the concentration resulting from a standardized dose. The appropriate distinction is between what affected concentration under controlled conditions and what the device directly measured.
A person could have a valid instrument reading even if a body-size-only dose calculation predicted something else. Conversely, estimating BrAC from dose and TBW would not substitute for the breath result. Breath sampling issues, quality controls, physiological variation, and timing remain separate questions discussed in the site’s breath alcohol measurement guide.
Did the study establish different elimination rates for all men and women?
No. The observed group means differed: 0.065 mg/L/hour for men and 0.074 mg/L/hour for women, with p = .006 for the between-group comparison. That supports a difference in the group averages within this sample and protocol. It does not assign one elimination rate to every man or another to every woman, and it does not establish what would have happened in a particular person’s earlier or later drinking episode.
The distinction is more than semantic. A difference between the averages cannot establish the value of any particular participant’s slope; the study’s summary data are not an individual diagnostic instrument. Sex-linked group patterns may relate to numerous physiological factors and the fixed dosing design. The experiment excluded people with certain conditions, standardized food and timing, and ran for a limited observation period. Applying its averages uncritically to a hospital patient, injured person, unusual drinking sequence, or an unrelated forensic case would require assumptions that this experiment did not test.
Why breath-versus-blood units matter in litigation
The authors reported BrAC in milligrams per liter of breath (mg/L). Blood alcohol concentration may instead appear as grams per deciliter, grams per kilogram, or another unit depending on the jurisdiction and specimen. These are different measured matrices and reporting conventions. An elimination rate in mg/L of breath per hour must not silently become a blood-alcohol elimination rate in g/dL/hour.
The paper discusses Austrian practice and historical blood-to-breath conversion factors. That background should not be repackaged as proof of a fixed conversion factor in another legal system. The 77-person protocol did not obtain paired venous blood samples to measure an individual blood-to-breath ratio. It therefore cannot validate a personal BAC calculated from this study’s BrAC or a specific retrograde blood-alcohol estimate. See the alcohol and volatiles evidence collection for broader matrix-specific considerations.
What records would matter when someone relies on this study in a case?
First identify the proposition being advanced: an actual breath result, an estimated breath result at another time, an estimated blood concentration, or an impairment opinion. These are separate scientific propositions. A careful case review would examine the instrument’s test records and calibration, collection times and duplicate readings, any independent blood tests, the actual alcohol-consumption evidence, food and absorption history, and the calculation procedure—including which body-water method and which elimination assumption were used.
If the claimed estimate depends on TBW, ask how and when TBW was determined, whether BIA conditions were comparable to the study, whether the subject’s clinical state might influence body-water measurement, and whether there are validated individual prediction errors for the intended application. The source itself is a controlled association and modeling study, not a validated forensic calculator. The defensible expert statement is therefore limited to what these methods and observations support.
Frequently asked questions
Can two people who drink the same amount have different breath alcohol concentrations?
Yes. In this controlled fixed-dose study, body water was strongly associated with estimated breath concentration. That does not mean total body water is the only cause of variation or that the study predicts a precise difference for any two people.
Is total body water simply another name for body weight?
No. TBW describes water-containing body compartments. Weight includes tissues with different water fractions. In the experiment TBW was estimated with BIA; BSA was calculated from height and weight.
Did BIA directly measure alcohol?
No. BIA estimated body water. The breath instrument measured alcohol in breath. Combining them in a statistical analysis did not turn the BIA device into an ethanol analyzer.
Does a correlation of −0.889 mean 88.9% accurate prediction?
No. A correlation coefficient describes the direction and strength of association in the study sample. It is not a percentage of correctly predicted individual results, a measurement-uncertainty interval, or an impairment-classification accuracy.
Can the study prove someone’s BAC at the time of driving?
No. The researchers did not collect paired venous BAC samples or validate an individual blood reconstruction. Their BrAC0 was modeled using later breath readings, not measured at time zero.
Do the male and female average breath-elimination slopes apply automatically to a defendant?
No. They are sample averages observed during a specified interval under controlled conditions. Extrapolating a particular person’s BrAC requires an appropriately supported person-specific and timing-specific analysis.
Does the study establish that the participants were impaired drivers?
No. It measured breath alcohol and body-composition variables in a research setting. It did not study driving, field sobriety performance, collision causation, or a validated impairment threshold for an individual.
Bottom line
Reiter and colleagues showed that directly estimated total body water was a better explanatory variable than body surface area for their derived breath-alcohol values after a fixed dose, and they documented different average breath-alcohol slopes in two study groups. Those observations argue against treating body size, one population formula, or a uniform elimination assumption as an error-free account of an individual’s alcohol concentration. The experiment did not establish an individual’s blood BAC, impairment, or legal causation. Measured breath, calculated earlier breath, estimated blood concentration, and demonstrated functional impairment must remain separate conclusions.
Primary scientific reference
Reiter GS, Boeckle M, Reiter C, Seltenhammer MH. The impact of total body water on breath alcohol calculations. Wiener klinische Wochenschrift. 2020;132:535–541. https://doi.org/10.1007/s00508-020-01663-4. The article above is an evidence-based interpretation of this original primary study, not a separate experimental replication.
Editorial note: This research briefing was prepared with AI assistance using the cited primary paper. It provides scientific education, not a case-specific expert opinion, an individually reviewed report, or legal or medical advice.