EVIDENCE LIBRARY
Can a BAC Below 0.15‰ Be Back-Calculated? Limits of a New 43-Minute Rule
A 2026 study proposes a 43-minute correction for BAC below 0.15‰. See its evidence, sample sizes, analytical limits, and forensic uncertainty.
Quick answer: A very low, analytically confirmed blood-alcohol concentration does not necessarily make retrospective BAC calculation scientifically impossible. In a 2026 reanalysis of controlled-drinking data, Mihan and colleagues found that ethanol concentrations generally continued to fall in an approximately linear pattern near zero. They proposed a research-derived conservative approach for measured BAC values of roughly 0.06–0.14 per mille (‰): shorten the elapsed back-calculation period once by 43 minutes, then apply a decline of 0.1‰ per hour. That proposal is not a universal legal rule, an individually validated elimination constant, or direct proof of someone’s BAC while driving. Its assumptions and uncertainties must be assessed before use in a case.
What does a BAC below 0.15‰ mean for retrograde extrapolation?
It is an evidentiary and modeling problem, not automatically a zero-evidence result. Retrograde extrapolation begins with a specimen collected after an event and mathematically estimates an earlier ethanol concentration. Because it adds an assumed amount of ethanol eliminated between the event and collection, its accuracy depends on the actual timing, the measured specimen, analytical uncertainty, and whether the person was in a relevant post-absorptive phase.
The new study examines an established German forensic practice of generally avoiding back-calculation when the starting measured BAC is under 0.15‰. The authors ask whether data collected as alcohol approaches complete elimination support a more discriminating rule. Their article does not repeal a judicial rule or supply an admissibility ruling in California, Arizona, or any other jurisdiction. A scientifically plausible population algorithm and the legal admissibility of an opinion are separate questions.
Units matter: the paper expresses BAC in per mille (‰), commonly associated with mass of ethanol per mass of blood. This is not interchangeable by typography alone with percent, g/dL, breath-alcohol units, or mg/L in plasma. Any conversion must identify the actual specimen, measurement convention, and method.
What original research did Mihan and colleagues analyze?
Mihan, Toennes, Paulke, and Verhoff reanalyzed raw concentration-time data from four previously conducted controlled-drinking experiments. The combined research set involved 87 adult volunteers drinking wheat beer, apple wine, grappa, or vodka under supervised conditions. The investigators identified 459 post-peak concentration observations at or below 0.15‰; the specific terminal-segment regression analysis used 75 usable individual curves. Those are three different denominators, not three different subject counts.
The researchers defined a person’s elimination phase from the last measured peak concentration to the end of observation, then fitted straight lines to the entire declining curve and several narrower concentration intervals. Their analysis used regression and simulations to explore the consequences of applying a conventional low elimination rate near the end of the BAC curve. This was a secondary analysis of controlled experimental observations, not a prospective validation trial of the proposed rule in unknown real-world DUI defendants.
The original study also describes laboratory blood sampling, alcohol testing, quality controls, and German forensic analytical practice. Breath testing was used in the original experimental protocols, including to track the disappearance of measurable alcohol. Accordingly, a courtroom expert should ascertain exactly which original values were blood measurements, which were converted or otherwise linked to breath measurements, and what records substantiate the reported individual curves. A curve reconstructed statistically is different from a new blood specimen drawn at the earlier incident time.
Did the measured decline really slow near zero?
Yes, in one defined regression comparison—but not every interval produced the same slope. The full-curve mean post-peak decline was reported as approximately 0.156‰/h. Across the 75 analyzed terminal curves under 0.15‰, the mean linear decline was about 0.083‰/h, below the 0.1‰/h convention used by the authors for conservative minimum-BAC calculations. The reported difference from 0.1‰/h was statistically significant in their one-sample test.
Yet when the authors pooled a broader 0.3-to-zero interval, their mean rate was approximately 0.13‰/h. Their Table 1 also shows differences between beverage-specific interval regressions. These are not competing laboratory measurements of one constant; they are estimates formed from different portions of a nonlinear and variable physiological trajectory. The selection of start and end points affects a slope, especially when the measured concentrations are small.
| Study finding or interval | Reported value | What it can—and cannot—tell us |
|---|---|---|
| All original controlled-drinking participants | 87 adults | Source population for reanalysis, not 87 independently validated criminal cases. |
| Post-peak observations at or below 0.15‰ | 459 measurements | Repeated observations; not 459 statistically independent people. |
| Terminal regression subset | 75 individual curves | Basis of the reported mean low-end rate; not the entire n=87 pool. |
| Full elimination phase from measured peak to near zero | Mean 0.156‰/h | Describes an extended post-peak interval. |
| Specific terminal segment below 0.15‰ | Mean approximately 0.083‰/h | Group regression estimate; not the elimination rate of an unknown defendant. |
| Broader 0.3-to-zero concentration interval | Mean approximately 0.13‰/h | Different interval, so it should not silently replace the terminal-segment result. |
The central lesson is methodological. A measured ethanol concentration is an analytical result at a known collection time. A regression slope is a mathematical summary of measurements over time. An earlier BAC reconstructed from that slope is another inferred value, with added assumptions and uncertainty.
How does the proposed 43-minute correction work?
The authors compared an unmodified conservative back-calculation with a “realistic” time-adjusted version and a more protective “extreme” version. The starting formula is estimated earlier BAC = measured later BAC + (0.1‰/h × elapsed hours). For a very low starting concentration, adding the full elapsed interval could overstate the presumed earlier value if the final part of the curve actually falls more slowly than 0.1‰/h.
The proposed extreme model therefore subtracts 43 minutes once from the elapsed event-to-collection interval before multiplying by 0.1‰/h. It does not subtract 43 minutes for each hour; nor does it subtract 43 minutes from the laboratory collection time or change the original measured concentration. Its uniform deduction is derived from the lowest individual terminal regression slope observed in the authors’ selected 75-curve data set, 0.02859‰/h. That empirical extreme should not be misrepresented as a physiologically universal slowest possible rate.
The paper recommends a starting range of 0.06–0.14‰ for this special treatment and omitting extrapolation below 0.06‰. The numerical model table labels the interval through 0.15‰; a case at the boundary requires explicit attention to that textual/table difference rather than a silently expanded rule.
| Approach discussed by authors | Time treatment | Interpretive status |
|---|---|---|
| Unadjusted conservative model | Use entire event-to-collection interval at 0.1‰/h | May overestimate minimum earlier BAC if terminal decline is slower. |
| “Realistic” model | About 14, 10, or 6 minutes deducted for measured bands 0.06–0.09, 0.10–0.12, or 0.13–0.15‰ | Exploratory interval-specific correction; not authors’ final maximum-protection rule. |
| “Extreme” model | Deduct approximately 43 minutes once, then use 0.1‰/h | Authors’ proposed more protective minimum-BAC method within studied conditions. |
| Measured starting BAC below 0.06‰ | Do not retrograde-extrapolate under their proposal | Model-support boundary, not a statement that ethanol was absent. |
What did the authors’ worked example actually show?
The original article considers a measured BAC of 0.06‰ five hours after an incident. An unadjusted 0.1‰/h calculation yields 0.56‰. Deducting about 14 minutes under the realistic approach gives approximately 0.54‰; the one-time 43-minute extreme deduction gives approximately 0.49‰.
These figures are outputs of three assumed models applied to an illustrative input. None is a directly assayed specimen from the incident; none demonstrates that a particular defendant actually had that concentration at the time. The calculation presumes that a suitable post-absorptive decline applies across the relevant interval. If the event preceded completion of absorption, the mathematical backward extension can be misleading regardless of the numerical refinement.
What is the evidentiary meaning of the simulations?
For the realistic model, the investigators used Monte Carlo simulations with 1,000 draws per starting-concentration category to examine variation in modeled time adjustments. Simulation output quantifies how a chosen statistical model behaves given its assumptions and source data. It does not independently measure hidden past BACs in real-world defendants, nor does a narrow simulated confidence interval establish a dependable individual prediction interval.
The authors described the 43-minute extreme correction as avoiding overestimation in the available data set. The word “guarantee” in the original discussion must be understood within that modeled empirical domain; it is not proof against unseen distributions of elimination rates, analytical error, timing errors, recent re-drinking, or a different population. The paper explicitly says this extreme model can yield unrealistically low back-calculated values in many cases.
Does a low measured BAC prove the person was unimpaired?
No. Nor does a larger modeled earlier concentration independently prove that the person was functionally impaired at the incident. This study measured or reanalyzed ethanol concentration-time behavior. It was not designed to validate a psychomotor test, driving observation, standardized field sobriety examination, toxicological causal mechanism for a collision, or individual criminal responsibility.
Concentration, time, and impairment form different inferential steps. Breath instruments, venous whole-blood testing, and any clinical serum or plasma value must be identified separately. A numerical result can be analytically sound and yet fail to establish what was occurring hours earlier; a model of earlier exposure can likewise be reasonable without independently proving motor coordination, judgment, ability to drive, or the cause of a crash.
What case-specific facts must be checked before using this model?
At minimum, an expert should verify the event time, collection time, actual measured analyte and matrix, reporting units, number of analytical replicates, calibration and quality controls, storage/preservative and chain of custody, uncertainty at low concentration, drinking chronology, food and absorption, any post-incident alcohol intake, medical context, and whether serial measurements establish the declining phase. A disputed timeline can be more consequential than adjusting a fitted slope by a few minutes.
Review the original laboratory analytical records using the site’s headspace blood-alcohol testing guide, the specimen collection framework, and the pharmacokinetics and uncertainty overview. These are background resources, not substitutes for the method used in the particular case.
In particular, the paper’s German minimum-BAC practice aims to avoid overstatement to a defendant’s disadvantage. It is not a universally adopted elimination rate or a judicial instruction in another jurisdiction. Before assigning a legal conclusion, counsel and the expert must distinguish scientific feasibility, jurisdiction-specific standards, and the burden of proof.
What limitations are important in cross-examination?
External validation is missing. The researchers pooled earlier experiments, then derived and assessed their algorithms largely from those observations. The studies were not designed as a blinded independent forensic validation across laboratories, all specimen matrices, diverse medical conditions, and uncertain incident timelines. The lowest observed slope is selected from a finite set; future persons can fall outside it. Their finding that different beverage groups lacked a statistically detected difference is not proof that beverage, dose, food, or interindividual differences never matter.
Near-zero analytic effects remain influential. A value near the reporting limit is more sensitive, in relative terms, to instrument imprecision, calibration bias, reporting thresholds, and how a regression treats a result as zero or near zero. Those issues cannot be solved merely by choosing a very conservative elimination constant.
There are internal editorial discrepancies in the published article. One displayed confidence-interval equation appears to contain a misplaced decimal, while the adjoining prose and numerical table give the coherent bounds. The abstract specifies 0.06–0.14‰, whereas a model table includes 0.15‰. These details should be evaluated against the complete original tables before using the model as a courtroom calculation rule; neither discrepancy should be silently normalized into a new empirical finding.
Which forensic question can this study answer?
It offers a serious challenge to the categorical claim that no retrospective BAC modeling is scientifically conceivable from any result below 0.15‰. It does not establish that a 43-minute correction is correct for everyone, that all low-positive tests can be back-calculated, or that a modeled estimate proves impairment. The best supported conclusion is narrower: the terminal data support further case-qualified low-BAC modeling, and the authors have proposed a deliberately conservative adjustment requiring validation, sound analytical records, and careful limits on its application.
Frequently asked forensic questions
Is 0.15‰ the same as a U.S. DUI limit of 0.08%?
No. They are numerically different scales and may employ different specimen mass/volume conventions. A paper’s proposed low-concentration threshold should not be equated with a statutory percentage without a documented unit and matrix conversion.
Does the study establish one universal human ethanol elimination rate?
No. The 0.083‰/h terminal mean, 0.13‰/h broader-interval mean, and 0.156‰/h full-curve mean summarize different segments of variable experimental curves; none is a direct individual measurement for an unidentified driver.
Can the laboratory add 43 minutes to the blood-draw timestamp?
No. The proposed procedure shortens the modeled elapsed interval once; it does not alter collection records or observed measurements.
Would a 0.04‰ result qualify for the paper’s extreme model?
No. The authors recommend not applying their new retrograde method below a starting measured concentration of 0.06‰.
Was the 43-minute method independently validated in criminal cases?
No independent prospective criminal-case validation is reported. It was constructed from and evaluated against reanalyzed controlled-drinking observations and model-based comparisons.
Can this model prove impairment at the incident time?
No. It estimates an earlier concentration under assumptions and does not measure a driver’s performance, signs, tolerance, or crash causation.
Can an expert apply the equation without the original analytical record?
That would be unsound. The exact specimen, method, validated low-end analytical uncertainty, reportable result, drinking chronology, timing, and all relevant case facts must be established before any retrospective opinion.
Primary scientific source
Mihan ATR, Toennes S, Paulke A, Verhoff MA. On the Possibility of retrograde calculation of blood alcohol concentrations below 0.15‰. International Journal of Legal Medicine. 2026;140:2833–2844. Original primary research: doi:10.1007/s00414-026-03838-6. Complete original publisher article. This article is an original educational analysis of the complete study, not individualized toxicological or legal advice.