lable at ScienceDirect

Regulatory Toxicology and Pharmacology 78 (2016) 24e36

Contents lists avai

Regulatory Toxicology and Pharmacology

journal homepage: www.elsevier .com/locate/yrtph

Best-practices approach to determination of blood alcoholconcentration (BAC) at specific time points: Combination of ante-mortem alcohol pharmacokinetic modeling and post-mortem alcoholgeneration and transport considerations

Dallas M. Cowan a, *, Joshua R. Maskrey b, Ernest S. Fung a, Tyler A. Woods a,Lisa M. Stabryla b, Paul K. Scott b, Brent L. Finley c

a Cardno ChemRisk, LLC, Aliso Viejo, CA, United Statesb Cardno ChemRisk, LLC, Pittsburgh, PA, United Statesc Cardno ChemRisk, LLC, Brooklyn, NY, United States

a r t i c l e i n f o

Article history:Received 5 October 2015Received in revised form24 March 2016Accepted 29 March 2016Available online 1 April 2016

KEYWORDS:AlcoholEthanolBlood alcohol concentration (BAC)Pharmacokinetic modelPost-mortem neoformationForensic toxicology

* Corresponding author. Cardno ChemRisk, LLC, 130CA, USA.

E-mail address: dallas.cowan@cardno.com (D.M. C

http://dx.doi.org/10.1016/j.yrtph.2016.03.0200273-2300/© 2016 Elsevier Inc. All rights reserved.

a b s t r a c t

Alcohol concentrations in biological matrices offer information regarding an individual's intoxicationlevel at a given time. In forensic cases, the alcohol concentration in the blood (BAC) at the time of death issometimes used interchangeably with the BAC measured post-mortem, without consideration foralcohol concentration changes in the body after death. However, post-mortem factors must be taken intoaccount for accurate forensic determination of BAC prior to death to avoid incorrect conclusions. Themain objective of this work was to describe best practices for relating ante-mortem and post-mortemalcohol concentrations, using a combination of modeling, empirical data and other qualitative consid-erations. The Widmark modeling approach is a best practices method for superimposing multiple alcoholdoses ingested at various times with alcohol elimination rate adjustments based on individual bodyfactors. We combined the selected ante-mortem model with a suggestion for an approach used toroughly estimate changes in BAC post-mortem, and then analyzed the available data on post-mortemalcohol production in human bodies and potential markers for alcohol production through decompo-sition and putrefaction. Hypothetical cases provide best practice approaches as an example for deter-mining alcohol concentration in biological matrices ante-mortem, as well as potential issues encounteredwith quantitative post-mortem approaches. This study provides information for standardizing BACdetermination in forensic toxicology, while minimizing real world case uncertainties.

© 2016 Elsevier Inc. All rights reserved.

1. Introduction

Alcohol (e.g., ethanol or ethyl alcohol), one of the mostcommonly consumed psychoactive drugs in theworld, is often usedto promote social interaction, is generally accepted and legal inmany countries. However, alcohol is a depressant that can impair aperson's ability to operate a motor vehicle; determining bloodalcohol concentration (BAC) is therefore one of the most prevalentforensic chemical analyses performed for criminal and medicalpurposes (Robinson and Harris, 2011). For example, a recent review

Vantis Suite 170, Aliso Viejo,

owan).

article evaluating 69 epidemiological studies found that 52% ofdriving-related fatalities and 35% of driving-related injuries wereassociated with positive alcohol tests (Schalast et al., 2011).

Although alcohol metabolism has been studied for over 100years, accurately predicting BAC following alcohol consumptionremains an active scientific research area (Nicloux, 1899; Hamill,1910). Precise estimation of the BAC at a given time point iscomplicated by individual variability in body and metabolismcharacteristics (e.g., age, body mass index, liver health, state ofnourishment, state of hydration and basal metabolic rate), vari-ability in mass or concentration of alcohol present in beverages(e.g., beer, wine, spirits), and the biological matrices sampled todetermine the BAC.

Determining BAC is particularly challenging when an impaireddriver is fatally injured in an accident. In such instances, the BAC

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e36 25

measured in a blood sample collected from the driver post-mortemis used to determine the level of the driver's impairment. However,various factors can affect post-mortem BAC measurements that donot typically affect ante-mortem measurements: alcohol meta-bolism phase, presence of a preservative in the collected sample,sample storage condition, variation in sampling media, putrefac-tion, and post-mortem alcohol neoformation. These factors areparticularly important in accident situations in which the body isnot recovered and promptly refrigerated. A direct post-mortem BACmeasurement may not accurately characterize a driver's impair-ment level at the time of death. In many instances, the BACmeasured after an accident is much higher than the level predictedby simple reconstruction of the driver's recent alcohol and foodconsumption (Wigmore, 2011).

The purpose of this paper is to present a best-practices ante-mortem alcohol modeling approach combined with a simple post-mortem alcohol concentration analysis to generate accurate BACpredictions before and after the time of death, thereby optimizingand standardizing forensic approaches in real world cases. Theobjectives of this study were to: 1) evaluate the relationships be-tween alcohol concentrations in various biological matrices; 2)generate an empirical modeling approach for correlating post-mortem alcohol concentrations with pharmacokinetic (PK)modeled ante-mortem concentrations up until the time of death; 3)describe factors associated with determining whether alcoholconcentrations measured post-mortem are due to ante-mortemingestion of alcohol or post-mortem synthesis of alcohol by mi-croorganisms; and 4) describe best practices for determining ante-and post-mortem alcohol concentrations with a focus on potentialsources of error.

2. Background

2.1. Human metabolism of alcohol

Alcohol (CH3CH2OH) is a small, polar molecule that accumulatesin water-rich areas of the body, and does not significantly diffuseinto fatty tissues. Following ingestion, alcohol is absorbed slowly inthe stomach and rapidly in the small intestines. The rate of alcoholabsorption is affected by the rate of gastric emptying, which in turnis influenced by various factors such as food ingestion (Holt, 1981;Holt et al., 1980; Sedman et al., 1976; Lin et al., 1976).

Various enzymes are responsible for alcohol metabolismincluding alcohol dehydrogenase (ADH) in the liver, and aldehydedehydrogenase (ALDH) and CYP2E1 in the brain and liver (Fig. 1)(Matsumoto and Fukui, 2002; Israel et al., 2013). Approximately90e98% of ingested alcohol is metabolized through the alcohol

Ethanol Acetaldehyde

Cytosol

Microsomes

Peroxisomes

Alcohol Dehydrogenase

Catalase

CYP2E1

Mitochondria

Circula on

NAD+ NADH

H2O2 H2O

NADPH + H+ + O2 NADP+ + H2O

Acetate

NAD+ NADH + H+

Aldehyde Dehydrogenase 2

Fig. 1. Metabolic pathway for elimination of alcohol in humans.

dehydrogenase þ aldehyde dehydrogenase pathway and otherphase II metabolic pathways, while the remaining 2e10% isexcreted un-modified in breath, sweat and urine (Jones, 2010). Incases of low exposure, alcohol is metabolized and eliminatedwithout significant physiological effects. The body's first-passmetabolism can prevent small doses of alcohol from reaching sys-temic circulation (Jones, 2010). However, once a threshold exposureis reached (which varies among individuals), the metabolic en-zymes are saturated, and excess alcohol begins to accumulate in thebloodstream. Alcohol in the blood will diffuse across the bloodbrain barrier, causing inebriation and impairment of physiologicalresponses. Alcohol's progressive physiological effects follow a dose-response relationship with respect to physiological effects indrinkers who do not suffer from alcoholism (Table 1) (Chong, 2014;Dubowski, 2006).

Alcohol concentration in the body changes as a function of time.BAC generally increases following an exponential curve to amaximum after initial alcohol ingestion as it is absorbed by thebody, then decreases linearly as it is eliminated until very low levels(<0.01e0.02%) of BAC, at which point the decrease becomesexponential (Jones, 2010). The increasing BAC phase is generallycalled the “absorption phase”, while the decreasing phase is calledthe “elimination phase”. The mass of alcohol ingested is importantin determining BAC, and the alcohol content varies widely by typeof drink. Additionally, the percentage of alcohol by volume (ABV)impacts the rate of absorption; drinks with 10e30% ABV areabsorbed the fastest; stronger or weaker drinks are absorbed moreslowly (Kelly and Mozayani, 2012). Also, during the absorptionphase, equilibrium is not reached, and the blood alcohol concen-tration may not fully reflect an individual's intoxication state(Wigmore, 2011). In the elimination phase, equilibrium is reached,and BAC is on the decline, thereby better reflecting the biologicalinfluence of alcohol (Wigmore, 2011).

2.2. Ante-mortem alcohol pharmacokinetic modeling approaches

Widmark presented an empirically-based formula in 1932 thatconsidered the exponential metabolic absorption rate constant, thezero-order elimination rate for alcohol, and the Widmark Factor(WF), an empirical rate constant accounting for the body's watercontent and volume alcohol distribution into body compartmentsas described below (inverse first-order dependence) (Posey andMozayani, 2007).

BAC ¼Aingested

�1� e�kt

�rW

� ðbtÞ

where,

BAC¼ Blood alcohol concentration (g/L)t ¼ Time since ingestion of alcohol (h)Aingested ¼ Mass of alcohol contained in the drink (g)r ¼ Widmark Factor (unitless)W¼ Body weight (kg)k ¼ Absorption rate constant (h�1)b ¼ Elimination rate ((g/L)/h)

The Widmark Equation remains the “gold standard” approachfor retrospectively estimating BAC (Posey and Mozayani, 2007;Widmark, 1932). Further developments in BAC estimation inrecent years included Derr's 1993 development of compartmentalphysiologically-based pharmacokinetic (PBPK) models for fourdifferent ethnicities, and Umulis et al., 2005 addition of reversibleenzyme kinetics (Derr, 1993; Umulis et al., 2005). PBPK models can

Table 1Relationship between blood alcohol concentration and reported physiological and behavioral effectsa.

BAC (%) Physiological effect

0.01e0.05 - Increased heart and respiration rates- Decreased functions in brain center- Slightly impaired judgment- Decreased inhibition- Mild euphoria- For some, effects are not apparent or obvious by ordinary observation- Inconsistent performance on special tests

0.06e0.10 (Legal limit ¼ 0.08) - Euphoria- Sociability, increased self-confidence- Decreased attention and alertness- Slowed reactions, impaired coordination, and reduced muscle strength- Reduced ability to make rational decisions and exercise good judgment- Increased anxiety and depression- Decrease in patience

0.11e0.15 - Emotional instability, loss of judgment- Dramatic slowing of reactions- Impairment of balance and movement- Impairment of some visual functions- Slurred speech- Vomiting- Drowsiness

0.16e0.29 - Severe sensory impairment, including reduced awareness of external stimulation- Increased pain threshold- Severe motor impairment (staggering gait)- Double vision and vertigo- Exaggerated emotional states and mental confusion- Lethargy

0.30e0.39 - Non-responsive stupor- Inability to stand or walk- No control of bladder and bowels- Vomiting- Loss of consciousness- Anesthesia comparable to that for surgery- Death

0.40 or above - Unconsciousness, coma-like state- Cessation of breathing- Death, usually due to respiratory arrest

a Adapted from Chong 2014 and Dubowski 2006.

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e3626

accurately estimate ethanol concentration over time in the bloodcompartment; however, they require complex solutions to differ-ential equations for little gain in accuracy over simple empiricalmodels, such as the Widmark Model (Derr, 1993).

The original WF was determined by averaging measured BACresults from a large group of individuals using a standard collectionof variables (e.g., age, sex, height, andweight) (Widmark,1932). TheWF has been modified and improved over the past 80 years toinclude more descriptive variables, such as body mass index (BMI),blood water content, and total body water (Watson et al., 1981;Posey and Mozayani, 2007; Forrest, 1986;Ulrich et al., 1987; Seidlet al., 2000). Posey and Mozayani (2007) recently modified theWidmark Model by including an empirical first order rate constantfor alcohol absorption in the GI tract, an average WF over multiplecalculation approaches, and an empirical elimination rate. Thecombination of these approaches allows the model to betterdescribe a specific individual's BAC using data specific to the indi-vidual (Posey and Mozayani, 2007).

The Widmark Equation, like any modeling approach, is limitedby input parameter accuracy. The magnitude of the WF is related tothe volume distribution of water in the body, which is a function ofgender and body weight. Body weight and BAC are inversely pro-portional; an individual with a greater body weight will thereforehave a lower BAC at a given dose (Kwo et al., 1998; Jones, 2010).Blood alcohol elimination rates (normalized over body weight)

depend on metabolic rate and gender; male elimination rates tendto be slightly lower than female elimination rates (Dettling et al.,2009; Pavlic et al., 2007). Also, the elimination rate of alcoholwithin persons of the same gender can vary: empirically measuredvalues have ranged between 0.096 and 0.241 g/kg/h in males and0.015e0.260 g/kg/h in females (Dettling et al., 2009; Pavlic et al.,2007). Another study measured similar elimination rates andfound a range of 0.106e0.217 g/L/h in males and 0.103e0.254 g/L/hin females (Pavlic et al., 2007). The accuracy of the Widmark modelcan be affected by the variability of elimination rates within thepopulation.

Also, elimination rates are not truly linear (or, zero-order withrespect to concentration for all alcohol concentrations). Indeed,alcohol elimination rates followMichaelis-Menten enzyme kinetics(Wagner, 1973; Mullen, 1977). For all concentrations greater than0.015e0.020 g%, the linear elimination assumption has very lowerror (Wagner, 1973; Posey and Mozayani, 2007). Most forensiccases involve BAC levels much greater than 0.02 g%; therefore, thissimplification is appropriate. However, in low-dose ante-mortemmodeling, we recommend use of a case-by-case Michaelis-Mentenkinetics approach such as those presented by Wagner and Patel orMullen (Wagner and Patel, 1972; Mullen, 1977).

Clearance rates of alcohol are a function of blood flow andmetabolic efficiency. Alcohol clearance rates are higher for olderpopulations than younger populations because elderly persons

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e36 27

have less volume of distribution for alcohol (Fiorentino andMoskowitz, 2013). Thierauf et al. compared theoretical WidmarkBAC calculations to observed BAC for persons of multiple ages andfound that a 75 year-old male reached a BAC of over 150% theintended endpoint; this unexpectedly high BAC was likely due toless total body water related to age (Thierauf et al., 2013). A morerecent study was performed to assess the accuracy of the WidmarkModel for elderly persons. The study included 51 individuals aged60 or over who had abstained from alcohol for two days. Theirblood alcohol concentrations were generally higher than predicted,and corrections to the WF for those in the study were suggestedbased on differences in body water volume (Bielefeld et al., 2015).

The absorption rate constant is affected by the presence of foodin an individual's stomach. The absorption rate constant for anempty stomach is roughly 2.3 h�1 compared to roughly 6.5 h�1 for afull stomach (Posey and Mozayani, 2007). Alcohol absorption isaffected significantly by fasting since alcohol is trapped in ingestedfood; the average absorption availability of alcohol was 97% withfasting, 94%with a light snack, and 72% and 66%with a meal in menand women, respectively (Sadler and Fox, 2011). Also, the absorp-tion of alcohol as a percentage can vary: the range for the fastingvalue was 87e108%, the range for the light snack value was81e112%, and the range for the gender-specific meal values were61e93% and 54e78% in men and women, respectively (Sadler andFox, 2011). The accuracy of the Widmark model can be affectedby varying absorption efficiencies (Sadler and Fox, 2011).

Othermodifying factors such as race are not explicitly obvious inthe Widmark Model. Individuals from different racial backgrounds,may contain different levels of alcohol dehydrogenase, affectingtheir ability to eliminate alcohol (Ehlers et al., 2012). Individuals ofHispanic descent generally have a higher alcohol tolerancecompared to other populations because of a heightened alcoholmetabolic rate (Caetano and Clark, 2000; Schwartz et al., 1996).Individuals of Asians and Native Americans do not produce suffi-cient ADH, therefore, the duration necessary for them to metabo-lize alcohol is longer compared to those of Caucasian origin (Israelet al., 2013). Individuals from these backgrounds therefore oftenexperience tachycardia, headache, nausea and facial flushingfollowing alcohol consumption.

Individual alcohol tolerance affects total alcohol metabolism,but the metabolic alcohol oxidation rate is not significantly affected(Palmer and Jenkins, 1982). Increased blood acetaldehyde levels areobserved in alcoholics after alcohol ingestion, however, the meanrate of alcohol oxidation (the first step of alcohol metabolism) didnot differ between alcoholics and non-alcoholics (Palmer andJenkins, 1982). Though alcohol dehydrogenase accounts for thegreater part of alcohol oxidation, other enzymes such as CYP2E1and catalase may be induced at high alcohol concentrations or afterlong term alcohol intake; though CYP2E1 and catalase may onlyaccount for a relatively small part of the total alcohol metabolism, itlikely contributes to the general variability of ethanol metabolismdue to alcohol tolerance (Quertemont, 2004). However, the activityof acetaldehyde dehydrogenase (the second step) was significantlylower in alcoholics; thus the Widmark Equation may be used foralcoholics with low error (Palmer and Jenkins, 1982).

2.3. Post-mortem BAC determination

Retrospectively determining BAC at the time of death can bechallenging because many additional variables must be considered.One such variable is the initial level of alcohol-generatingmicrobialcontamination and potential for environmental contamination overtime prior to collection of the sample by a forensic examiner. Bloodand other biological matrices can be potentially contaminated withsome species of bacteria, fungi and other agents capable of

generating alcohol (mostly from glucose) via putrefaction. Thiscontamination and neoformation of alcohol can confound identi-fication of BAC at time of death when derived from post-mortemblood samples.

2.3.1. Biological matrices for post-mortem alcohol determinationTesting multiple matrices to determine alcohol concentration is

common practice in forensic analyses. The most common biologicalmatrices tested include: blood, urine, and vitreous humor. In gen-eral, BAC is higher than urine alcohol concentration (UAC) andvitreous humor alcohol concentration (VAC) during the absorptionphase, and the reverse is true during elimination (Kelly andMozayani, 2012). Additionally, arterial blood may exhibit up to40% higher alcohol concentrations compared to venous bloodduring the absorption phase (Kelly and Mozayani, 2012).

A number of biological matrices including blood, vitreous hu-mor, muscle, urine, and internal organs have been previouslyevaluated forensically to determine level of alcohol intoxicationand cause of death. Bodily fluids such as bile, vitreous humor, urine,and synovial fluid have been studied to determine their accuracy inestimating BAC. Winek et al. (1993) determined the BAC:synovialfluid alcohol concentration ratio to be 0.98; however, variabilityexisted in samples with higher alcohol concentration, leading to aratio range of 0.4e1.72 (Winek et al., 1993). Stone and Rooney(1984) studied the viability of using bile, urine, and vitreous hu-mor to accurately determine BAC. The authors found that VAC:BACratios were consistently 0.77 for BAC >0.10% and 0.63 for BAC<0.10%; however, BAC: bile alcohol concentration and UAC: BACratios had larger variation for BAC <0.10% (Stone and Rooney, 1984).Additionally, Kugelberg and Jones, 2007 found insignificant varia-tion in VAC between the left and right eyes of one individual(Kugelberg and Jones, 2007). Blood, bile, vitreous humor, and urineare the four most common biological matrices used to determineBAC (Stone and Rooney, 1984).

Generally, whole blood is used in post-mortem analysis due tothe difficulty in separating serum or plasma fractions post-mortem,while serum or plasma are used in ante-mortem analysis. The ratioof alcohol concentrations in serum to those in whole blood gener-ally ranges between 1.12 and 1.24, and the ratio of alcohol con-centrations in plasma to whole blood can range between 1.1 and1.35 (O'Neal and Poklis, 1996; Kelly and Mozayani, 2012; Bielefeldet al., 2015).

2.3.2. Sources of post-mortem alcohol neoformationAlcohol neoformation may result from microbial growth in

improperly stored samples. Post-mortem neoformation of alcoholby microorganisms can complicate analytical results from biolog-ical matrices, although it still may be possible to accurately inter-pret forensic results using best-practices approaches, such as thosepresented in this paper. Although vitreous humor and urine are lesslikely substrates than blood for microbial growth, such growthcannot be ruled out (Kelly and Mozayani, 2012).

The vitreous alcohol concentration (VAC) to BAC ratio generallyprovides valuable information regarding the alcohol metabolicstate, especially in forensic-related cases. Table 2 summarizesliterature-reported VAC:BAC ratios, along with the anatomicalsource of blood inwhich the BAC was measured. AVAC:BAC ratio ofless than one implies that the individual was in the absorptionphase prior to equilibrium; a ratio greater than one implies that theelimination phase was reached prior to death (Boonyoung et al.,2008). Deviation from typical VAC:BAC ratios (generally consid-ered 0.5 to 1.5) suggests alcohol consumption or production bymicroorganisms (de Lima et al., 1999). However, authors of somestudies have concluded that the VAC:BAC ratio is unreliable fordetermining the source of alcohol (Jollymore et al., 1984). Indeed,

Table 2Reported ratios of blood and vitreous alcohol concentration.

Reference Blood source Samplesize

VAC:BAC meanratio

VAC:BAC ratioSD

VAC:BACratio range

Elapsed time between deathand sample collection

Winek and Esposito, 1981 NR 30 0.94 0.17 NR NRBudd, 1982 NR 15 1.3 0.6 NR NRCaplan and Levine, 1990 NR 205 1.19 NR 0.10e1.91 NRChao and Lo, 1993 Femoral 68 1.06 0.23 NR Within 28 hSylvester et al., 1998 Femoral 9 1.06 0.10 NR NRJones and Holmgren, 2001 Femoral Vein 672 1.19 0.29 NR NRHoney et al., 2005 Femoral 203 1.24

Heart 5 1.19 NR 1.01e2.20 NRPleural Cavity 1 1.18

Boonyoung et al., 2008 Femoral 25 0.963e1.206 NR NR Within 24 hHassan, 2011 Femoral 43 1.13 0.43 NR NRHoffman et al., 2011 NR TBD 1.10e1.20 NR NR NR

NR¼Not reported; SD¼ Standard deviation; BAC¼ Blood alcohol concentration; VAC¼Vitreous alcohol concentration.

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e3628

autopsy analyses have led to the conclusion that extrapolation ofBAC from VAC results is a rough estimate at best, and that suchresults should be considered cautiously (Neil et al., 1985; Jollymoreet al., 1984).

Alcohol is generated in vitro via the glycolytic pathway utilizedby many microorganisms during fermentation (Kugelberg andJones, 2007; Skopp, 2009). High yields of alcohol can be producedfrom carbohydrates, with glucose being the substrate of choicewhere 1 mol of glucose is metabolized to 2 mol of alcohol (Sutlovicet al., 2013). Glucose content is thus often a primary determinant ofthe amount of alcohol produced (Boumba et al., 2012). The absenceof excess glucose, however, does not exclude post-mortem alcoholproduction, since alcohol may be produced from other substrates,such as mannitol, sucrose, mannose, lactose, ribose, and variousamino acids (O'Neal and Poklis, 1996; Skopp, 2009; Canfield et al.,2007). Substrate composition differences result in the productionof different amounts of alcohol from the same microorganism(Boumba et al., 2012). Generally, at least 58 species of bacteria, 17species of yeast, and 24 species of molds are capable of producingalcohol from sugars (O'Neal and Poklis, 1996). Information on 18species relevant to post-mortem alcohol generation in humanbodies is presented in Table 3. Twelve of the 18 species occurnaturally in the intestines, skin, mouth, and sputum (Corry, 1978).Post-mortem alcohol generation has been observed in blood, urineor vitreous humor samples as fast as 7 h after death and two daysafter sample collection, with levels ranging from 0.16 to 10.63 mg/mL (Boumba et al., 2012; Sutlovic et al., 2013; Corry, 1978).Generally, alcohol generation occurred in samples stored at roomtemperature and lacking preservatives. Additionally, low levels ofvarious biological indicators of putrefaction, such as short-chainalcohols, including 1-propanol, 2-propanol, 1-butanol, 3-methyl-1-butanol, 2-methyl-1-propanol, acetone, diethyl ether, acetalde-hyde, and formaldehyde, were identified (O'Neal and Poklis, 1996;Chikasue et al., 1988; Skopp, 2009; Sutlovic et al., 2013; Corry,1978). Specifically, 0.003e0.105 mg/mL of 1-propanol and0.002e0.119 mg/mL of 1-butanol was detected from various bio-logical matrices where putrefaction was suspected (Yajima et al.,2006; Boumba et al., 2012).

Microorganism presence in samples with detectable alcoholconcentration but no evidence of alcohol consumption suggestspost-mortem neoformation as the potential alcohol source. O'Nealand Poklis (1996) reported that 12% of alcohol detections in au-topsies were attributed to post-mortem generation; however, thisfigure rises to 40e50% in accident-related deaths (O'Neal andPoklis, 1996). Additionally, BAC values of up to 0.19% weremeasured in sailors killed in an explosion who had not been

drinking alcohol (Mayes et al., 1992). Studies suggest that samplehandling, storage conditions, and time between death and analysiscontribute to the potential for post-mortem neoformation alcoholproduction (Kugelberg and Jones, 2007).

Storage condition of the body or biological matrix sample is akey factor affecting post-mortem alcohol generation. Higher tem-peratures produce higher alcohol yields; refrigeration of the bodytherefore helps to prevent alcohol synthesis (Kugelberg and Jones,2007; Nanikawa et al., 1982). It is standard forensic practice to addsodium fluoride (NaF), a preservative with proven anti-microbialactivity, to blood samples in order to prevent microbial growth anddegradation. However Sutlovic et al. (2013), reported that alcoholconcentration continued to increase in blood and urine samplesstored at 4 �C despite adding NaF (Sutlovic et al., 2013). Alcoholproduction occurs at a higher rate under anaerobic conditions, suchas submersion in water (Hadley and Smith, 2003). Multiple studieshave found that alcohol concentrations are higher and may changeover time in submerged bodies (Kugelberg and Jones, 2007;Gonzales et al., 1954; Waller, 1972; Tomita, 1975). Post-mortemalcohol generation in submerged subjects started between 12 and24 h after submersion depending on the temperature (Kugelbergand Jones, 2007; Hadley and Smith, 2003; Wintemute et al., 1990).

Various biological indicators of putrefaction have been identi-fied, such as short-chain alcohols including n-propanol, n-butanol,3-methyl-1-butanol, and isopropanol, along with other non-alcohol indicators, such as acetone, diethyl ether, acetaldehyde,and formaldehyde (O'Neal and Poklis, 1996; Chikasue et al., 1988;Skopp, 2009; Sutlovic et al., 2013; Corry, 1978). Forensic labora-tories often use 1-propanol as a standard, making it invalid as amarker for post-mortem alcohol production in the cases where it isused; 2-methyl-2-propanol, which is not a natural indicator ofputrefaction, is also used as a standard by some laboratories (O'Nealand Poklis, 1996). The concentration of these indicators can be usedto quantitatively determine the alcohol concentration generated bymicrobes with considerable reliability.

Loss of alcohol concentration from blood samples may alsooccur during storage. Alcohol loss from samples occurs undervarious conditions, such as evaporation due to improper samplestorage, enzyme-mediated oxidation, and microorganism action(Brown et al., 1973;Dick and Stone, 1987). Brown et al. (1973) foundthat a temperature increase led to loss of alcohol content viaoxidation of alcohol to aldehyde, and Jones (2007) reported loss ofalcohol concentration in samples stored at 4 �C (Brown et al., 1973;Jones, 2007). Smalldon and Brown (1973) attributed the loss ofalcohol concentration from non-enzymatic oxidation involvingoxyhemoglobin (Smalldon and Brown, 1973).

Table 3Reported Cases where Alcohol was Generated Post-Mortem.

Reference Microorganism presentA BiologicalmatrixB

Samplesize

Storage durationafter death toanalysis (days)

Bloodglucoselevel (mg/mL)

Highestethanolconcentration(mg/mL)

Indicators of putrefaction

Othervolatiles andgasespresent

Max. concentrationdetected (mg/mL)

No Evidence of Alcohol Consumption

Hoisethet al.,2008

Escherichia coliNO, Clostridium bacteroidesNO,Prevotella species Escherichia coliNO, ClostridiumbacteroidesNO, Prevotella species

BloodFL,

RFBlood FL, RF1 11 days (10 days

after autopsy)9.96 in VH 3.50 NR NR

Appenzelleret al.,2008

Streptococcus familyNO, Lactococcus garvieae,Streptococcus familyNO, Lactococcus garvieae

BloodRTBlood

RT3 7 days 4.00 1.21 NR NR

AntonidesandMarinetti,2011

Staphylococcal aureusNO, CandidaalbicansNOStaphylococcal aureusNO, candida albicansNO

BloodRFBlood

RF1 30 þ days after

death2.17 in VH NR Acetone 0.540

Carbondioxide

NR

Boumbaet al.,2012

Clostridia perfrigensNOClostridia perfrigensNO BloodRTBlood

RTNR 29 days 2.00 0.16 1-propanol 0.007

1-butanol 0.0192-methyl-1-propanol

0.002

3-methyl-1-butanol

0.001

Boumbaet al.,2012

Clostridia sporogenesNOClostridia sporogenesNO BloodRTBlood

RTNR 5 days 2.00 0.89 1-propanol 0.105

1-butanol 0.1192-methyl-1-propanol

0.061

3-methyl-1-butanol

0.006

Boumbaet al.,2012

Escherichia coliNO BloodRT NR 22 days 2.00 0.56 1-propanol 0.0101-butanol 0.0022-methyl-1-propanol

0.001

3-methyl-1-butanol

0.001

Sutlovicet al.,2013

Citrobacter freundii, Enterococcus faecalis, SerratiamarcescensNO, Candida glabrata

BloodRF 1 17 days (14 daysafter autopsy)

5.85 2.34 Acetone 0.510

AntonidesandMarinetti,2011

Staphylococcal aureusNO, candida albicansNO UrineRF 1 30 þ days afterdeath

2.17 in VH 0.28 Acetone 8.020

Carbondioxide

NR

Sutlovicet al.,2013

Candida glabrata, Enterococcus faecalis, EscherichiacoliNO, Morganella morganii, KlebsiellapneumoniaeNO

UrineRF 1 17 days (14 daysafter autopsy)

0.07 10.63 Acetone 0.630

Evidence of Alcohol Consumption

Yajima et al.,2006

Candida albicansNO, Candida parapsilosisNO BloodRF 1 23 days (22 daysafter autopsy)

NR 4.90 1-propanol 0.002e0.003

Yajima et al.,2006

Corynebacterium sp.NO, Escherichia coliNO, CandidatropicalisNO

BloodNR 1 20 days NR 9.60 1-propanol 0.004e0.03

Yajima et al.,2006

Candida albicansNO BloodRT 4 2 days 7.00 2.10 1-propanol 0.030

A: NO - Naturally occurring in intestines, skin, mouth and sputum according to Corry (1978).B: FL - 0.21% w/v Fluoride ions added; NR - Not reported; RF - Stored at 2e10 �C; RT - Stored at room temperature; VH - Vitreous humor.

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e36 29

Trauma is often observed in forensic cases; introduction of mi-croorganisms from the environment or from within the gut intovarious biological matrices has been observed following trauma(references). Putrefaction and post-mortem alcohol productionmay occur more rapidly with trauma (Kugelberg and Jones, 2007;Canfield et al., 2007). Direct trauma to the gut is likely to result instomach rupture, leading to the spread of gastric contents,

including ingested alcohol and gut flora (Kugelberg and Jones,2007;Plueckhahn and Ballard, 1968; Winek et al., 1995). BACs of>5 mg/dL were observed in blood samples with gastrointestinalfluid contamination (Winek et al., 1995). Additionally, agonalevents resulting from trauma may lead to pulmonary aspiration ofstomach contents into the lungs, leading to direct alcohol diffusioninto the bloodstream (Kugelberg and Jones, 2007). Vitreous humor

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e3630

is often cited as the matrix least prone to microbial contaminationbecause of its biological remoteness and seclusion from the rest ofthe body, and is therefore also often cited as the most suitablematrix for determining BAC (de Lima et al., 1999; O'Neal and Poklis,1996; Jenkins et al., 1995; Hoffman et al., 2011). Furthermore, vit-reous humor is often preserved, even in the case of severe trauma(Coe and Sherman, 1970). Trauma to the eye itself, however, couldopen a pathway for microbial contamination of the vitreous humor.

3. Methods

3.1. Literature review

To more accurately estimate BAC at various time points ante-mortem and post-mortem (e.g., 2 h after two drinks; 12 h afteran alcohol related death, etc.), we evaluated the existing litera-ture and available methodologies regarding pharmacokineticmodeling of BAC and the empirical estimation of post-mortemBAC. A comprehensive literature search was performed usingPubmed, Elsevier, and Google Scholar to identify peer-reviewedstudies that evaluated relationships between alcohol concentra-tions in various biological matrices, applications of the WidmarkModel for modeling of alcohol concentrations in humans, andevidence of putrefaction and post-mortem alcohol generation byvarious microorganism species. The literature search helped usidentify a best-practices Widmark Model that was accurate in anumber of alcohol consumption scenarios. This best-practicesWidmark Model was then used to calculate BAC values in hy-pothetical scenarios that highlight its utility.

Particular emphasis was placed on studies containing infor-mation regarding the dependence of alcohol concentration inbodily matrices on gender, body weight, body mass index (BMI),height, age, ethnicity, elimination rate, absorption rate andtolerance for alcohol. The reliability of methods used to deter-mine alcohol concentration in post-mortem samples was alsoevaluated.

3.2. Specific parameter selections for case studies and modelingapproach

We next devised three hypothetical examples to demonstratethe Widmark Model's important features. In our first example, we

Table 4Explanation of the Widmark model and contributing factors.

Parameter Symbol Value range Unit

Ethanol absorption rate constant k 2.1e6.5 h-1

Ethanol elimination rate constant b 0.13e0.25 (g/L)/

Body weight W weight kgHeight H height mAge G age yearsWidmark factor - Watson et al., (1981) estimate rWatson See below e

Widmark factor - Forrest (1986) estimate rForrest See below e

Widmark factor - Seidl et al. (2000) estimate rSeidl See below e

Widmark factor - Ulrich et al. (1987) estimate rUlrich See below e

rWatsonðmalesÞ ¼ 0:39834þ 12:725HW � 0:11275G

W þ 2:8993W

rForrestðmalesÞ ¼ 1:0178� 0:012127WH2

rSeidlðmalesÞ ¼ 0:31608� 0:004821W þ 0:4632HrUlrichðmalesÞ ¼ 0:715� 0:00462W þ 0:22HrWatsonðfemalesÞ ¼ 0:29218þ 12:666H

W � 2:4846W

rForrestðfemalesÞ ¼ 0:8736� 0:0124WH2

rSeidlðfemalesÞ ¼ 0:31223� 0:006446W þ 0:4466HrUlrichðfemalesÞ NA

evaluated the differences in pharmacokinetics between the twogenders after ingesting three alcoholic drinks. In our secondexample, we performed several adjustments to the traditionalWidmark Equation to account for multiple drinks consumed overtime. In our third example, we combined ante-mortem modelingwith post-mortem considerations to estimate the BAC of an indi-vidual at the time of death, and the amount of alcohol generatedpost-mortem. Current literature is lacking in examples where ante-mortem modeling is combined with post-mortem considerationsregarding alcohol generation. As previously described, post-mortemgeneration is a concern in many forensic matters; we thus present athird example to address this concern and literature gap.

All ante-mortem Widmark parameters and equations weresolved as a function of time and displayed graphically usingMicrosoft Excel (Ver. 2013, Microsoft Corporation, Redmond,WA). Data were extracted from graphs presented in Yajima et al.,2006 using an open source digitization program. Microsoft Excelwas used to determine functions of best fit for post-mortemalcohol generation rates by particular species or groups of spe-cies using the empirical data presented in each of the followingstudies, where applicable: Yajima et al., 2006, Boumba et al.,2012,Antonides and Marinetti, 2011, Sutlovic et al., 2013,Hoiseth et al., 2008 and Appenzeller et al., 2008.

3.2.1. Example 1: application of the existing Widmark Modelingapproaches – gender and body weight differences

To illustrate the differences in BAC predicted by the WidmarkEquation between genders, a hypothetical case study was createdwith heights and weights representative of the general population,where a 1.8 m, 70 kg, 30 year old male and a 1.7 m, 55 kg, 30 yearold female both drink three shots (133.2 mL) of 80-proof liquor. TheWidmark Model was applied to estimating the BAC of these twoindividuals over a 180 min time period. The WF was calculatedusing the average of the approaches presented by Watson et al.(1981),Forrest (1986), Seidl et al. (2000) and Ulrich et al. (1987)as presented in Table 4 (Posey and Mozayani, 2007) was pre-sented in Table 4. The absorption rate constant was set equal to5 h�1 for both individuals to simulate a semi-full stomach (Poseyand Mozayani, 2007). Elimination rates were assumed to be0.162 g/L/h for the male and 0.179 g/L/h for the female (Pavlic et al.,2007).

Value factors

Depends mainly on BMI and personal metabolic factors (Posey andMozayani, 2007)

hr Depends on BMI, age, water content of body, fullness of stomach (Pavlicet al., 2007) Generally linear at BAC > 0.015e0.02%Dependent on the person's body weightDependent on the person's heightDependent on the person's ageSee belowSee belowSee belowSee below

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e36 31

3.2.2. Example 2: modified Widmark Model e superimposition ofmultiple BAC curves

A second hypothetical case study was created to illustrate themore precise BAC estimation over time by superimposition ofmultiple Widmark curves. In this case study, a 1.8 m, 70 kg, 30year old male consumes three shots (44.4 mL each) of 80-proofliquor over a 45 min period. The absorption rate constant ratewas set at 5 h�1 (Posey and Mozayani, 2007). The eliminationrate was assumed to be 0.162 g/L/h (Pavlic et al., 2007). Themale's BAC was modeled for 90 min. One model was runassuming that all three shots were consumed immediately (usingthe traditional Widmark Model), and the other model was runassuming that one shot was consumed every 15 min (using themodified Widmark Model).

3.2.3. Example 3: combination of post-mortem alcohol generationconsiderations with the modified Widmark Model

To illustrate alcohol generation in the body after death, bothante-mortem and post-mortem BAC was estimated for a hypo-thetical scenario in which a 1.8 m, 70 kg, 30 year old male wasassumed to have consumed three shots (44.4 mL each) of 80proof liquor 15 min apart with all parameters the same as pre-sented in Example 2 before dying in an accident that occurred75 min after the first point of alcohol ingestion. For this example,microorganism-based neoformation of ethanol was assumed tobegin immediately at death. For simplicity, the broad range ofmicroorganisms present in the human gut was assumed to bewell represented by the 18 species presented in Table 3. Theblood-specific results for each of the six individual studies pre-sented in Table 3 were analyzed for the following: initial glucoseconcentration, temperature of study, real body or in vitro envi-ronment, and ethanol concentration in the medium as a functionof time. Following this analysis, studies identified as containingsufficient information to estimate an ethanol generation or lossrate were used to estimate the change in ethanol concentrationpost-mortem at 1 day after death, 7 days after death, and 14hours after death. It has been demonstrated in the literature thata logistic equation can be used to describe fermentation andethanol production processes by microbes (Wang and Liu,2014;Olaoye and Kolawole, 2013). A least-squares regressioncalculation was used to determine the logistical equation of best-fit for each data set of the form demonstrated below usingMicrosoft Excel (Ver. 2013, Microsoft Corporation, Redmond,WA).

BAChmgmL

i¼ A� A

1þ�time½day�

C

�B

where A, B, and C are constants. Results obtained by Hoiseth et al.(2008) were not modeled using the logistic model due to insuffi-cient sample size (n ¼ 3, resulting in 0 degrees of freedom). Thepost-mortem approximations for each data set were then added tothe existing BAC starting at the time of death to determine theethanol concentration over time. All regression values wereconfirmed, and equation coefficients were tested for statisticalsignificance using SYSTAT 11 (SYSTAT Software Inc, San Jose, CA).The data and parameters were input into a nonlinear regressionmodel, and the 95% Wald confidence intervals were evaluated todetermine if the model parameters A and B were significantlydifferent than 0, and if the model parameter C was significantlydifferent than 1.

4. Results

4.1. Identification, evaluation and modification of the standardWidmark Model

The Widmark Model appears to be the gold standard forquantitatively determining BAC at a given time point in a livinghuman. More complicated models are available in the literature(e.g., compartmental PBPK models). As previously stated, however,the pharmacokinetics of alcohol in humans is fairly simple, so thetraditional Widmark empirical model is sufficiently accurate toestimate BAC under most conditions. It has also been extensivelystudied and applied, and is simple to implement, making it a bestpractices approach for estimating BAC. Specifically, we used theadjusted empirical pharmacokinetic model presented by Posey andMozayani (2007), aWidmark-style approach adapted to account formultiple drinks over time.

4.2. The Widmark Model accurately predicts BAC

The standard and modified Widmark Model was used withinput parameters specific to the individual's body characteristics forExamples 1 and 2 described in the methods section. The results ofthe calculated BACs are described in detail below.

4.2.1. Gender and weight significantly affect BACThe difference between gender and body weight is demon-

strated in Fig. 2, where the BAC of a male and female individual isestimated after ingesting three shots (133.2 mL) of 80-proof liquor.The model predicts that the female reaches a maximum BAC of0.10%, while the male only reaches a maximum BAC of 0.064%,which is only 64% of the female maximum BAC. Additionally, themale reaches the maximum BAC at 36 min, while the female rea-ches the maximum at 40 min. The time to 0.02%, which is both thethreshold of non-linear elimination and a concentration at whichphysiological effects would be minimal (Table 1), was also deter-mined from the model. The time for BAC to return to 0.02% is212 min in the male, compared to 321 min in the female. Together,these findings indicate the gender-specific differences in elimina-tion rates and WFs affected by body mass and volume ofdistribution.

4.2.2. The modified Widmark Model accurately estimates BACWe compare two modeling approaches for an individual who

consumed three shots (44.4 mL each) of 80-proof liquor over a45 min period (Fig. 3). Fig. 3 contains the comparison of theinstantaneous dose and time-modified dose Widmark approaches.The instantaneous drinking approach resulted in a maximum BACof 0.064% at 36 min after the first drink, while the time-modifieddrinking approach resulted in a maximum of 0.043% within44 min after the first drink. In this case study, the traditionalWidmark Model predicted a maximum BAC that was 150% largerthan the maximum BAC predicted by the modifiedWidmarkModelthat accounts for drinks consumed at various times. The timerequired to reach the maximum BAC is 22% shorter in the originalapproach. The time-modified drinking approach results in a fasterdecline of BAC in the 60e90 min time range compared to the im-mediate consumption approach, because the non-linear term of theWidmark equation has larger magnitude over any specific timeperiod if the initial dose was larger. Specifically, in the instanta-neous approach, the BAC returned to 0.02% in 212min, compared to85 min with the time-modified drinking approach. After 85 min,Michaelis-Menten elimination kinetics should be considered. Thesewere not reflected in Fig. 3.

0.000%

0.020%

0.040%

0.060%

0.080%

0.100%

0.120%

0 20 40 60 80 100 120 140 160 180

Tota

l BAC

% (g

%)

Time Since 1st Drink (min)

Male

*Male is 1.8 m, 70 kg, 30 years old

**Female is 1.7 m, 55 kg, 30 years old Female

Fig. 2. Example comparison of BAC% Responses to Ingestion of Three Shots (44.4 mL) of 80-Proof Liquor in a Male and a Female.

0.000%

0.010%

0.020%

0.030%

0.040%

0.050%

0.060%

0.070%

0 10 20 30 40 50 60 70 80 90 100

Tota

l BAC

% (g

%)

Time Since 1st Drink (min)

1 Shot Consumed Every 15 Min

All 3 Shots Consumed Immediately*Male is 1.8 m, 70 kg, 30 years old

Fig. 3. Example comparison of BAC% Responses to Ingestion of Three Shots (44.4 mL) of 80-Proof Liquor at different time points.

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e3632

4.3. Alcohol produced by microorganisms in biological matrices

Alcohol present in biological matrices obtained during forensicinvestigations may signify alcohol consumption or post-mortemproduction by microorganisms. Information on 18 species rele-vant to post-mortem alcohol generation in human bodies werecollected (Table 3). Samples were collected between one to threedays after death, and analyzed between two to more than 30 daysafter death.

4.3.1. Combining the Widmark Model with post-mortem microbialgeneration considerations can roughly indicate whether microbialgeneration occurred

The three-parameter logistic model approximations for the datacollected in each study are described below. Additionally, the 95%Wald confidence intervals are presented in parentheses beside eachparameter and the overall coefficient of determination (R2) ispresented.

� Based on Boumba et al.’s results for C. sporogenes, C. perfrigens,and E. coli in vitro, the rate of alcohol generation after death wasmodeled with logistic functions of time as described above:� C. sporogenes (n ¼ 31): A ¼ 0.77 (0.74e0.79), B ¼ 7.0 (1.5e11),C ¼ 1.3 (1.0e1.6), R2 ¼ 0.92.

� E. coli (n ¼ 31): A ¼ 0.5 (0.48e0.52), B ¼ 1.5 (1.0e2.0), C ¼ 1.3(1.0e1.5), R2 ¼ 0.94.

� C. perfrigens (n ¼ 31): A ¼ 0.16 (0.14e0.17), B ¼ 1.8 (1.2e2.4),C ¼ 5.1 (4.2e6.0), R2 ¼ 0.94.

� Notably, E. faecalis was also tested, with no generation ofethanol.

� Based on Yajima et al.’s results (n ¼ 5) for C. albicans in vitro, theconcentration increased for two days and the rate of alcoholgeneration after death was modeled as a logistic function oftime as described above with parameters A ¼ 1.6 (1.2e2.0),B¼ 8.3 (�9.8e26), C¼ 1.1 (0.87e1.2), R2 ¼ 0.99. For this dataset,B and C are not statistically significant. Yajima et al. also notedthat the concentration began to decrease after two days anddescribed results for the concentration (n ¼ 4) over time; the

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e36 33

rate of loss of ethanol was unable to be fit to a logistic decaymodel. This model only had one degree of freedom.

� Hoiseth et al.’s measured results were not modeled using thethree parameter logistic model due to insufficient sample size.

� Based on Sutlovic et al.’s measured results (n ¼ 6) for repeatedanalysis of a blood sample from a deceased 75-year old femalewhich contained C. freundii, E. faecalis, S. marcesens andC. glabrata, the rate of alcohol generation after death wasmodeled as a logistic function of time as described above, withparameters: A ¼ 2.5 (2.3e2.8), B ¼ 6.0 (3.9e8.0), C ¼ 11 (10e12)(R2 ¼ 1.0).

� Antonides et al. measured no ethanol generation over 5 days in asample which contained S. aureus and C. albicans.

� The data presented by Appenzeller et al., 2008 showed that theblood concentration of ethanol in a sample of heart blood from a14-month old child who died suddenly was initially at 2.00 mg/mL decreased to 1.21 mg/mL after 7 days and 0.48 mg/mL ofin vitro fermentation.

Biological matrices are occasionally obtained a considerableperiod of time after an individual's death, making it difficult toaccurately determine the alcohol concentration attributed to post-mortem generation and the actual BAC at the time of death. Here,we estimated the BAC at various times post-mortem using best-fitequations based on the six studies presented in Table 3. The BAC attime of death is predicted by the Widmark Model to be 0.027%(75 min after consuming consumption the first drink, see Fig. 3),which is less than the US legal driving limit of 0.08%. The results ofapplication of the various lines of best fit cause a number ofdifferent results. For example, application of the C. albicans gen-eration/consumption curve demonstrates that the BAC can be ex-pected to reach amaximum at 2 days after death of 0.19% (0.027% atdeath þ 0.16% of generation), well above the US legal driving limit,but that it can be expected to decrease after two days (Yajima et al.,2006). Interpretation with respect to Boumba's results for the fourorganism studies over 30 days indicates a concentration thatlogistically increases for two to three days and then stabilizes.Individually, application of the three logistic equations of best fit fora time period of 5 days gives 0.10% (C. sporogenesgeneration þ ante-mortem), 0.071% (E. coli generation þ ante-mortem), or 0.035% (C. perfrigens generation þ ante-mortem),while selection of E. faecalics would imply a concentration of0.027%, or no change from the concentration at time of death(Boumba et al., 2012). Any other time period could be used in thesuggested logistic growth models. These data did not match theexperimentally determined values well; the Hoiseth et al., 2008measurements continued to increase over time, the Sutlovic et al.,2013 values demonstrated no decrease even after 10 days, unlikethe Yajima et al., 2006 curve with a decreasing portion, and theother cited values showed either no increase or a decrease inethanol concentration over time (Hoiseth et al., 2008; Sutlovicet al., 2013; Antonides and Marinetti, 2011; Appenzeller et al.,2008). Regardless, at 5 days, the fitted logistic equation for theSutlovic et al., 2013 data suggest an expected BAC of 0.12%.

5. Discussion

The three examples presented in this analysis serve as a guide todemonstrate how BAC results should be interpreted with consid-eration for individual body parameters, superposition of multiplealcoholic beverages, and postmortem generation of alcohol ifapplicable. Selection of individual parameters and interpretation ofpost-mortem considerations depends on the details of individualforensic cases. In real-world scenarios, parameter selections and/orprinciples applied should vary based on the details of the case,

which is not always done. Example 3 demonstrates how bestpractices ante-mortem modeling can be combined with post-mortem considerations to determine whether alcohol found insamples was generated by microorganisms or was ingested prior todeath. The ante-mortem Widmark model was adapted from Poseyand Mozayani (2007) and utilized the time-modified drinkingapproach, allowing for more accurate modeling of BAC overextended periods of alcohol ingestion. Also, the adapted ante-mortem Widmark Model used an average WF from four empiri-cally determined relationships (Watson et al., 1981; Forrest, 1986;Ulrich et al., 1987; Seidl et al., 2000). This approach averages outany biases present in each group, such as a narrow age range or anarrow body weight range. For example, the Watson et al. (1981)WF approach for males included a slight dependence on age dueto the wide age range in their sample, while the other three ap-proaches did not.We evaluated the applicability of thesemodels forgender differences and time-stepped modeling differences usingthe first two hypothetical examples.

This analysis of the Widmark Model indicates that it is a rela-tively accurate approach for determining BAC as a function of timefor an individual until the time of death, provided that adjustmentsare made for body mass index, age, individual metabolic rate, ab-sorption of ethanol and elimination of ethanol above 0.02% BAC.The variation between BAC predictions made with general pa-rameters and time-adjusted parameters highlights the importanceof proper adjustment of input parameters on a case by case basis.Adjustments for these factors are well described, yet few forensicinvestigators have incorporated these parameters together. Addi-tionally, superimposition of multiple Widmark BAC curves is areasonable method for estimating BAC over time for a personconsuming multiple alcohol-containing beverages at differenttimes, and this approach results in a more precise estimate ofmaximum BAC and the time to maximum BAC than does assumingall drinks were consumed at once. A key difference between thetraditional and modified superimposition approach is an in-dividual's maximum BAC. The traditional Widmark Model maypredict a maximum BAC above the legal limit, whereas the modi-fied superimposition approach may predict a maximum BAC belowlegal limits.

It is common knowledge among forensic investigators thatalcohol present in post-mortem body matrices originates fromalcohol consumption prior to death, post-mortem alcohol produc-tion by microorganisms, or some combination of both (Posey andMozayani, 2007; Kelly and Mozayani, 2012). The phenomenawhere cadaver BAC was elevated in the apparent absence of ante-mortem alcohol ingestion has been noted (Mayes et al., 1992;Boumba et al., 2012; Antonides and Marinetti, 2011). Further-more, putrefaction or decomposition follows shortly after death,depending on conditions surrounding the body, and autolysis oc-curs during putrefaction. During autolysis, tissue decompositionoccurs and endogenous enzymes are released. Further decompo-sition by endogenous enzymes lead to putrefactive bacteria releasefrom the gut (Boumba et al., 2012). In total, we highlight 18 speciesof bacteria or fungi with demonstrated ability to generate alcohol inbody matrices. Indicators or biomarkers are produced in concertwith alcohol during alcohol neoformation. Key indicators of post-mortem neoformation of alcohol from the processes of putrefac-tion include: 1-propanol, n-butanol, and 2-propanol. Concentrationof these indicators can be used to quantitatively estimate theconcentration of alcohol produced, as demonstrated by Boumbaet al. for 1-propanol and n-butanol (Boumba et al., 2012). However,confounding is possible, as these substances may be ingested aswell. For example, some researchers have estimated the half-life of2-propanol in blood following isopropanol ingestion in attemptedsuicides, and others have described a case of ingestion of acetone

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e3634

(Ramu et al., 1978; Daniel et al., 1981). All cases described involvedalcohol dependent or alcoholic persons, which may indicate thatalcoholism is a risk factor for ingestion of other solvents.

No quantitative models for post-mortem neoformation ofethanol in human blood existed in the literature using input pa-rameters such as time since death, initial ethanol concentration,etc. However, empirical data existed, and models of best fit for in-dividual species did not match experimental data well. Addition-ally, experimental data did not match other sets of experimentaldata well, as evidenced by the widely different predictions for BAC(modeled to start at death at 0.027%) between the Hoiseth et al.,2008 data (which were not approximated by a model, butincreased rapidly) and the Sutlovic et al., 2013 fitted equation(0.028%). The use of the empirical equations based on the literaturewas meant to correct the concentration measured during the au-topsy for post-mortem neoformation or loss, but the results wereoften different on a case-by-case basis. This exercise demonstratedthe massive variety of potential results that can be estimated usingliterature: themeasured BACwithin 11 days of death can peak up to0.38% based on the values observed by Hoiseth et al. (2008), andmay decrease to 0 within the same time period based on Yajimaet al., 2006. Therefore, it is recommended that the forensic practi-tioner rely more heavily on ante-mortem modeling if the amountconsumed is determinable, and qualitatively measures the poten-tial for post-mortem generation of alcohol by sampling the blood orother medium for: 1) presence of microorganisms, 2) amount ofpreservative and 3) presence of other markers of putrefaction(which can be used semi-quantitatively to estimate ethanol pro-duction, per Boumba et al., 2012).

Similar pharmacokinetic modeling to that presented in thisstudy for ethanol can be performed for other alcohols and indus-trial solvents such as methanol, acetone, 2-propanol, or acetalde-hyde. Bouchard et al. (2001) developed and validated a biologicallybased dynamic model that described the time trajectories ofmethanol and its metabolites in whole blood and other biologicalmatrices (e.g.,urine; expired air) (Bouchard et al., 2001). The modelcan quantitatively relate the parent compound or metabolites inthe biological matrix to the absorbed dose and tissue burdens atany point in time for different exposure scenarios (Bouchard et al.,2001). In a separate study, the coherence between occupationalexposure limits and biological limit values was evaluated, andblood and urine concentrations of 2-propanol and acetone weredetermined after inhalation exposure by human pharmacokineticmodeling. The acetone and 2-propanol model could be used tocreate a time profile of the commonly used solvents in differentbiological matrices for various exposure scenarios (Huizer et al.,2014). Umulis et al., 2005 developed a compartmental PBPKmodel that predicts the time evolution for ethanol's majormetabolite, acetaldehyde, in the blood by having derived averageenzymatic rate laws for alcohol dehydrogenase and acetaldehydedehydrogenase. This approach was novel in that it combinedethanol and acetaldehyde PBPK modeling, which correlatedstrongly with the experimentally observed ethanol and acetalde-hyde concentration results for healthy individuals and those withreduced acetaldehyde dehydrogenase activity. This model alsoaccounted for the reverse reaction of acetaldehyde back intoethanol, keeping acetaldehyde levels 10-fold lower than if irre-versible (Umulis et al., 2005). All of the aforementioned solventscould serve as a starting point for additional areas of research incombining ante- and post-mortem approaches in order to deter-mine a time profile for solvent concentration before or at the timeof death.

This research did not focus on modeling of VAC. However, inforensic investigations, common practice is to obtain blood from afemoral source, determine its BAC, and compare it to VAC to

determine an individual's intoxication and alcohol metabolic state.Our literature review yielded 10 peer-reviewed papers that re-ported both BAC and VAC values, as well as VAC:BAC ratios(Table 2). Sample sizes ranged from one sample to 672 samples,with one paper not reporting the sample size. The mean VAC:BACratio ranged from 0.94 to 1.30, with standard deviations rangingfrom 0.10 to 0.60. Multiple studies have reported that VACs aretypically 10e20% higher than BACs during the elimination phase; aBAC larger than or equal to VACs is expected before equilibrium orsignifies post-mortem production of alcohol. Interestingly, theduration between death and sample collection was not reported inall but two studies. The time to collection is specifically importantbecause potential for microorganisms to generate alcohol in theblood exists post-mortem. Two studies indicated the time fromdeath to collection, which was only listed as within 24 and 28 h,respectively. According to this body of literature, all 1276 subjectsincluded in these ten papers had some exposure to alcohol throughingestion, and the mean VAC:BAC ratio ranged from 0.94 to 1.30.These results indicate that the VAC:BAC ratio can vary significantlybased on the specific ingestion aspects, particularly by metabolismphase (e.g., absorption or elimination). Therefore, caution must beexerted when comparing VAC to BAC results.

The models and equations used in this study are likely appli-cable to a wide range of scenarios associated with alcohol con-sumption and BAC determination. The modifications to thetraditional Widmark Model allow for a better reflection of BAC, anda better understanding of an individual's intoxication state bymoreaccurately estimating BAC at the time of death. Thus, the combi-nation of the modified Widmark Model and qualitative consider-ations for post-mortem alcohol generation allows for a morecomprehensive interpretation of BAC obtained in forensicinvestigations.

6. Best practices for determination of BAC at various timepoints

Based on our analysis, the best methods for determining BAC ata given time point and a given set of parameters are describedbelow:

� Matrices Collection:� Sample immediately to avoid any microbial contamination� Collect liquids in glass tubes� Addition of 1e2% of NaF� Store at 0e4 �C

� Modeling Parameters:� Use the modified Widmark Model proposed by Posey andMozayani (2007), which allows for superimposition of mul-tiple drinks consumed at various times.

� Use aWF approach specific to gender, weight, height, and age.Calculate the WF (r) from multiple empirical approaches anduse the average value.

� Use gender-specific and weight-specific empirical data forelimination rate (b), which is generally available in peer-reviewed literature.

� For cases involving low-BAC modeling (<0.02%), use aMichaelis-Menten kinetics approach to estimation of theelimination rate (b).

� Combine the ante-mortem approach with post-mortemethanol generation considerations. If available, use dataregarding other markers of putrefaction in combination withthe methodology presented by Boumba et al. (2012) to gaugeethanol production. Also, check the sample for microbialcontamination, and concentration of preservative.

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e36 35

� If multiple measurements are taken from various media(whole blood, serum, plasma, vitreous humor, urine) correctthe concentrations for variability between media using pub-lished literature values.

Conflict of interest

The authors report no conflicts of interest. Funding for thismanuscript was provided entirely by Cardno ChemRisk, LLC, aconsulting firm that provides scientific advice to the government,corporations, law firms, and various scientific/professional organi-zations. This paper was prepared and written exclusively by theauthors without review or input by any outside sources. Two of theauthors (DMC, BLF) have served as an expert witness regardingalcohol toxicology and PBPK modeling of alcohol.

Acknowledgments

The authors wish to thank Nekisa Heghitat for referencingassistance.

Transparency document

Transparency document related to this article can be foundonline at http://dx.doi.org/10.1016/j.yrtph.2016.03.020.

References

Antonides, H., Marinetti, L., 2011. Ethanol production in a postmortem urine sample.J. Anal. Toxicol. 35 (7), 516e518.

Appenzeller, B.M., Schuman, M., Wennig, R., 2008. Was a child poisoned by ethanol?discrimination between ante-mortem consumption and post-mortem forma-tion. Int. J. Leg. Med. 122 (5), 429e434.

Bielefeld, L., Auwarter, V., Pollak, S., Thierauf-Emberger, A., 2015. Differences be-tween the measured blood ethanol concentration and the estimated concen-tration by Widmark's equation in elderly persons. Forensic Sci. Int. 247, 23e27.

Boonyoung, S., Narongchai, P., Junkuy, A., 2008. The relationship of alcohol con-centration in epidural or acute sub-dural hematoma compared with vitreoushumor and femoral blood. J. Med. Assoc. Thai 91 (5), 754e758.

Bouchard, M., Brunet, R.C., Droz, P.O., Carrier, G., 2001. A biologically based dynamicmodel for predicting the disposition of methanol and its metabolites in animalsand humans. Toxicol. Sci. 64 (2), 169e184.

Boumba, V.A., Kourkoumelis, N., Ziavrou, K.S., Fragkouli, K., Vougiouklakis, T., 2012.Patterns of the most abundant volatiles detected in post-mortem blood. Rom. J.Leg. Med. 20, 147e154.

Brown, G.A., Neylan, D., Reynolds, W.J., Smalldon, K.W., 1973. The stability of ethanolin stored blood. I. Important variables and interpretation of results. Anal. Chim.Acta 66 (2), 271e283.

Budd, R.D., 1982. Ethanol levels in postmortem body fluids. J. Chromatogr. 252,315e318.

Caetano, R., Clark, C.L., 2000. Hispanics, blacks and white driving under the influ-ence of alcohol: results from the 1995 National Alcohol Survey. Accid. Anal.Prev. 32 (1), 57e64.

Canfield, D., Brink, J., Johnson, R., Lewis, R., Dubowski, K., 2007. Clarification ofethanol-positive case using urine serotonin metabolite ratio. J. Anal. Toxicol. 31(9), 592e595.

Caplan, Y.H., Levine, B., 1990. Vitreous humor in the evaluation of postmortemblood ethanol concentrations. J. Anal. Toxicol. 14 (5), 305e307.

Chikasue, F., Yashiki, M., Miyazaki, T., Okamoto, I., Kojima, T., 1988. Abnormally highalcohol concentration in the heart blood. Forensic Sci. Int. 39 (2), 189e195.

Chong, L.Y., 2014. A Three-pronged Approach to Drink-driving Studies. Departmentof Civil Engineering, University of Hong Kong, Hong Kong.

Chao, T.C., Lo, D.S., 1993. Relationship between postmortem blood and vitreoushumor ethanol levels. Am. J. Forensic Med. Pathol. 14 (4), 303e308.

Coe, J.I., Sherman, R.E., 1970. Comparative study of postmortem vitreous humor andblood alcohol. J. Forensic Sci. 15 (2), 185e190.

Corry, J.E., 1978. Possible sources of ethanol ante-and post-mortem: its relationshipto the biochemistry and microbiology of decomposition. J. Appl. Bacteriol. 44(1), 1e56.

Daniel, D.R., McAnalley, B.H., Garriott, J.C., 1981. Isopropyl alcohol metabolism afteracute intoxication in humans. J. Anal. Toxicol. 5 (3), 110e112.

de Lima, I. V., Midio, A.F., 1999. Origin of blood ethanol in decomposed bodies.Forensic Sci. Int. 106 (3), 157e162.

Derr, R.F., 1993. Simulation studies on ethanol metabolism in different humanpopulations with a physiological pharmacokinetic model. J. Pharm. Sci. 82 (7),

677e682.Dettling, A., Witte, S., Skopp, G., Graw, M., Haffner, H.T., 2009. A regression model

applied to gender-specific ethanol elimination rates from blood and breathmeasurements in non-alcoholics. Int. J. Leg. Med. 123 (5), 381e385.

Dick, G.L., Stone, H.M., 1987. Alcohol loss arising from microbial contamination ofdrivers' blood specimens. Forensic Sci. Int. 34 (1e2), 17e27.

Dubowski, K.M., 2006. Stages of Acute Alcoholic Influence/Intoxication. The Uni-versity of Oklahoma Department of Medicine, Oklahoma City, Oklahoma.

Ehlers, C.L., Liang, T., Gizer, I.R., 2012. ADH and ALDH polymorphisms and alcoholdependence in Mexican and native Americans. Am. J. Drug Alcohol Abus. 38 (5),389e394.

Fiorentino, D.D., Moskowitz, H., 2013. Breath alcohol elimination rate as a functionof age, gender, and drinking practice. Forensic Sci. Int. 233 (1e3), 278e282.

Forrest, A., 1986. The estimation of Widmark's factor. J. Forensic Sci. Soc. 26 (4),249e252.

Gonzales, T., Vance, M., Helpern, M., Umberger, C.J., 1954. Legal Medicine: Pathologyand Toxicology, second ed. Appleton-Century-Crofts: Inc, New York.

Hadley, J.A., Smith, G.S., 2003. Evidence for an early onset of endogenous alcoholproduction in bodies recovered from the water: implications for studyingalcohol and drowning. Accid. Anal. Prev. 35 (5), 763e769.

Hamill, P., 1910. Cardiac metabolism of alcohol. J. Physiol. 39 (6), 476e484.Hassan, H.M.A., 2011. An Evaluation of the Stability and Prevalence of Alcohol and

Related Biomarkers in Biological Matrices with Applications to the Interpreta-tion of Medico-Legal Cases (Doctoral dissertation, University of Glasgow).

Hoffman, D.B., Zedeck, B.E., Zedeck, M.S., 2011. Fundamental issues of postmortemtoxicology. Forensic Chem. Handb. 457.

Hoiseth, G., Kristoffersen, L., Larssen, B., Arnestad, M., Hermansen, N.O., Morland, J.,2008. In vitro formation of ethanol in autopsy samples containing fluoride ions.Int. J. Leg. Med. 122 (1), 63e66.

Holt, S., 1981. Observations on the relation between alcohol absorption and the rateof gastric emptying. Can. Med. Assoc. J. 124 (3), 267e277, 297.

Holt, S., Stewart, M.J., Adam, R.D., Heading, R.C., 1980. Alcohol absorption, gastricemptying and a breathalyser. Br. J. Clin. Pharmacol. 9 (2), 205e208.

Honey, D., Caylor, C., Luthi, R., Kerrigan, S., 2005. Comparative alcohol concentra-tions in blood and vitreous fluid with illustrative case studies. J. Anal. Toxicol.29 (5), 365e369.

Huizer, D., Huijbregts, M.A., van Rooij, J.G., Ragas, A.M., 2014. Testing the coherencebetween occupational exposure limits for inhalation and their biological limitvalues with a generalized PBPK-model: the case of 2-propanol and acetone.Regul. Toxicol. Pharmacol. 69 (3), 408e415.

Israel, Y., Rivera-Meza, M., Karahanian, E., Quintanilla, M.E., Tampier, L., Morales, P.,Herrera-Marschitz, M., 2013. Gene specific modifications unravel ethanol andacetaldehyde actions. Front. Behav. Neurosci. 7, 80.

Jenkins, A.J., Levine, B.S., Smialek, J.E., 1995. Distribution of ethanol in postmortemliver. J. Forensic Sci. 40 (4), 611e613.

Jollymore, B., Fraser, A., Moss, M., Perry, R., 1984. Comparative study of ethyl alcoholin blood and vitreous humor. Can. Soc. Forensic Sci. J. 17 (2), 50e54.

Jones, A.W., 2007. Are changes in blood-ethanol concentration during storageanalytically significant? Importance of method imprecision. Clin. Chem. Lab.Med. 45 (10), 1299e1304.

Jones, A.W., 2010. Evidence-based survey of the elimination rates of ethanol fromblood with applications in forensic casework. Forensic Sci. Int. 200 (1e3), 1e20.

Jones, A.W., Holmgren, P., 2001. Uncertainty in estimating blood ethanol concen-trations by analysis of vitreous humour. J. Clin. Pathol. 54 (9), 699e702.

Kelly, A.T., Mozayani, A., 2012. An overview of alcohol testing and interpretation inthe 21st century. J. Pharm. Pract. 25 (1), 30e36.

Kugelberg, F.C., Jones, A.W., 2007. Interpreting results of ethanol analysis in post-mortem specimens: a review of the literature. Forensic Sci. Int. 165 (1), 10e29.

Kugelberg, F.C., Jones, A.W., 2007. Interpreting results of ethanol analysis in post-mortem specimens: a review of the literature. Forensic Sci. Int. 165 (1), 10e29.

Kwo, P.Y., Ramchandani, V.A., O'Connor, S., Amann, D., Carr, L.G., Sandrasegaran, K.,Kopecky, K.K., Li, T.K., 1998. Gender differences in alcohol metabolism: rela-tionship to liver volume and effect of adjusting for body mass. Gastroenterology115 (6), 1552e1557.

Lin, Y., Weidler, D.J., Garg, D.C., Wagner, J.G., 1976. Effects of solid food on bloodlevels of alcohol in man. Res. Commun. Chem. Pathol. Pharmacol. 13 (4),713e722.

Matsumoto, H., Fukui, Y., 2002. Pharmacokinetics of ethanol: a review of themethodology. Addict. Biol. 7 (1), 5e14.

Mayes, R., Levine, B., Smith, M.L., Wagner, G.N., Froede, R., 1992. Toxicologic findingsin the USS Iowa disaster. J. Forensic Sci. 37 (5), 1352e1357.

Mullen, P.W., 1977. The metabolism and pharmacokinetics of alcohol in man.J. Forensic Sci. Soc. 17 (1), 49e55.

Nanikawa, R., Ameno, K., Hashimoto, Y., Hamada, K., 1982. Medicolegal studies onalcohol detected in dead bodiesealcohol levels in skeletal muscle. Forensic Sci.Int. 20 (2), 133e140.

Neil, P., Mills, A., Prabhakaran, V., 1985. Evaluation of vitreous humor and urinealcohol levels as indices of blood alcohol levels in 75 autopsy cases. Can. Soc.Forensic Sci. J. 18 (2), 97e104.

Nicloux, M., 1899. Sur le passage de l'alcool ingere de la mere au foetus, en par-ticulier chez la femme. Comp. Rend. Seance Soc. Biol. Fill 51, 980e982.

O'Neal, C.L., Poklis, A., 1996. Postmortem production of ethanol and factors thatinfluence interpretation: a critical review. Am. J. Forensic Med. Pathol. 17 (1),8e20.

Olaoye, O., Kolawole, O., 2013. Modeling of the kinetics of ethanol formation from

D.M. Cowan et al. / Regulatory Toxicology and Pharmacology 78 (2016) 24e3636

glucose biomass in batch culture with a non structured model. Int. J. Eng. Res.Appl. 3 (4), 562e565.

Palmer, K.R., Jenkins, W.J., 1982. Impaired acetaldehyde oxidation in alcoholics. Gut23 (9), 729e733.

Pavlic, M., Grubwieser, P., Libiseller, K., Rabl, W., 2007. Elimination rates of breathalcohol. Forensic Sci. Int. 171 (1), 16e21.

Plueckhahn, V.D., Ballard, B., 1968. Factors influencing the significance of alcoholconcentrations in autopsy blood samples. Med. J. Aust. 1 (22), 939e943.

Posey, D., Mozayani, A., 2007. The estimation of blood alcohol concentration.Forensic Sci. Med. Pathol. 3 (1), 33e39.

Quertemont, E., 2004. Genetic polymorphism in ethanol metabolism: acetaldehydecontribution to alcohol abuse and alcoholism. Mol. Psychiatr. 9 (6), 570e581.

Ramu, A., Rosenbaum, J., Blaschke, T.F., 1978. Disposition of acetone following acuteacetone intoxication. West J. Med. 129 (5), 429e432.

Robinson, S., Harris, H., 2011. In: Dunstan, S. (Ed.), Smoking and Drinking AmongAdults, 2009: a Report on the 2009 General Lifestyle Survey. UK StatisticsAuthority.

Sadler, D.W., Fox, J., 2011. Intra-individual and inter-individual variation in breathalcohol pharmacokinetics: the effect of food on absorption. Sci. Justice 51 (1),3e9.

Schalast, N., K€osters, C., Demmerling, R., Mushoff, S., 2011. Drei prognostisch undtherapeutisch relevante Gruppen alkoholabh€angiger Patienten imMaßregelvollzug gem€aß x 64 StGB. Psychiatr. Prax. 38 (01), 31e37.

Schwartz, J.G., Salman, U.A., McMahan, C.A., Phillips, W.T., 1996. Gastric emptying ofbeer in Mexican-Americans compared with non-hispanic whites. Metabolism45 (9), 1174e1178.

Sedman, A.J., Wilkinson, P.K., Sakmar, E., Weidler, D.J., Wagner, J.G., 1976. Food ef-fects on absorption and metabolism of alcohol. J. Stud. Alcohol 37 (9),1197e1214.

Seidl, S., Jensen, U., Alt, A., 2000. The calculation of blood ethanol concentrations inmales and females. Int. J. Leg. Med. 114 (1e2), 71e77.

Skopp, G.A., 2009. Postmortem Toxicology: Artifacts. Wiley Encyclopedia ofForensic Science.

Smalldon, K.W., Brown, G.A., 1973. The stability of ethanol in stored blood. II. Themechanism of ethanol oxidation. Anal. Chim. Acta 66 (2), 285e290.

Stone, B.E., Rooney, P.A., 1984. A study using body fluids to determine blood alcohol.J. Anal. Toxicol. 8 (2), 95e96.

Sutlovic, D., Nestic, M., Kovacic, Z., Gusic, S., Mlinarek, T., Salamunic, I., Sardelic, S.,2013. Microbial ethanol production in postmortem urine sample. Med. Sci. Law53 (4), 243e246.

Sylvester, P.A., Wong, N.A., Warren, B.F., Ranson, D.L., 1998. Unacceptably high sitevariability in postmortem blood alcohol analysis. J. Clin. Pathol. 51 (3), 250e252.

Thierauf, A., Kempf, J., Eschbach, J., Auwarter, V., Weinmann, W., Gnann, H., 2013.A case of a distinct difference between the measured blood ethanol concen-tration and the concentration estimated by Widmark's equation. Med. Sci. Law53 (2), 96e99.

Tomita, K., 1975. On putrefactions and floatations of dead bodies under water.Hiroshima J. Med. Sci. 24 (2e3), 117e152.

Ulrich, L., Cramer, Y., Zink, P., 1987. Die Beriicksichtigung individueller Parameter beider Errechnung des Blutalkoholgehaltes aus der Trinkmenge. Blutalkohol 24,192e198.

Umulis, D.M., Gurmen, N.M., Singh, P., Fogler, H.S., 2005. A physiologically basedmodel for ethanol and acetaldehyde metabolism in human beings. Alcohol 35(1), 3e12.

Wagner, J.G., 1973. A modern view of pharmacokinetics. J. Pharmacokinet. Bio-pharm. 1 (5), 363e401.

Wagner, J.G., Patel, J.A., 1972. Variations in absorption and elimination rates of ethylalcohol in a single subject. Res. Commun. Chem. Pathol. Pharmacol. 4 (1),61e76.

Waller, J.A., 1972. Nonhighway injury fatalities. I. The roles of alcohol and problemdrinking, drugs and medical impairment. J. Chronic Dis. 25 (1), 33e45.

Wang, Y., Liu, S., 2014. Kinetic modeling of ethanol batch fermentation by Escher-ichia Coli FBWHR using hot-water sugar Maple wood extract hydrolyzate assubstrate. Energies 7 (12), 8411.

Watson, P.E., Watson, I.D., Batt, R.D., 1981. Prediction of blood alcohol concentra-tions in human subjects. Updating the Widmark Equation. J. Stud. Alcohol 42(7), 547e556.

Widmark, E.M.P., 1932. Die theoretischen Grundlagen und die praktische Ver-wendbarkeit der gerichtlich-medizinischen Alkoholbestimmung. Urban &Schwarzenberg.

Wigmore, J.G., 2011. Wigmore on Alcohol: Courtroom Alcohol Toxicology for theMedicolegal Professional. Irwin Law.

Winek, C.L., Bauer, J., Wahba, W.W., Collom, W.D., 1993. Blood versus synovial fluidethanol concentrations in humans. J. Anal. Toxicol. 17 (4), 233e235.

Winek Jr., C.L., Winek, C.L., Wahba, W.W., 1995. The role of trauma in postmortemblood alcohol determination. Forensic Sci. Int. 71 (1), 1e8.

Wintemute, G.J., Teret, S.P., Kraus, J.F., Wright, M., 1990. Alcohol and drowning: ananalysis of contributing factors and a discussion of criteria for case selection.Accid. Anal. Prev. 22 (3), 291e296.

Winek, C.L., Esposito, F.M., 1981. Comparative study of ethanol levels in blood versusbone marrow, vitreous humor, bile and urine. Forensic Sci. Int. 17 (1), 27e36.

Yajima, D., Motani, H., Kamei, K., Sato, Y., Hayakawa, M., Iwase, H., 2006. Ethanolproduction by Candida albicans in postmortem human blood samples: effects ofblood glucose level and dilution. Forensic Sci. Int. 164 (2e3), 116e121.