journal article
LitStream Collection
High Correlation between Ethanol Concentrations in Postmortem Femoral Blood and in Alternative Biological Specimens, but Large Uncertainty When the Linear Regression Model Was Used for Prediction in Individual Cases
Thelander, G; Kugelberg, F C; Jones, A W
doi: 10.1093/jat/bkaa018pmid: 32104897
Abstract In connection with medicolegal autopsies peripheral blood (e.g. from a femoral vein) is the specimen of choice for toxicological analysis, although alternative specimens are also sometimes submitted, such as bile, cerebrospinal fluid (CSF), vitreous humor (VH), bladder urine, pleural effusions and/or lung fluid. Ethanol concentrations were determined in duplicate in femoral blood and in various alternative biological specimens by headspace gas chromatography. The analysis was carried out on two different fused silica capillary columns furnishing different retention times for ethanol and both n-propanol and t-butanol were used as internal standards. The results were evaluated by linear regression using blood alcohol concentration (BAC) as dependent or outcome variable and the concentrations in an alternative specimen as independent or predictor variable. The Pearson correlation coefficients were all statistically highly significant (P < 0.001); r = 0.94 (bile), r = 0.98 (CSF), r = 0.97 (VH), r = 0.92 (urine), r = 0.94 (lung fluid) and r = 0.96 (pleural cavity effusions). When the regression model was used to predict femoral BAC from the mean concentration in an alternative specimen the mean and 95% prediction intervals were 1.12 ± 0.824 g/L (bile), 1.41 ± 0.546 g/L (CSF), 1.15 ± 0.42 g/L (VH), 1.29 ± 0.780 g/L (urine), 1.25 ± 0.772 g/L (lung fluid) and 0.68 ± 0.564 g/L (pleural cavity effusions). This large uncertainty for a single new observation needs to be considered when alcohol-related deaths are evaluated and interpreted. However, the analysis of alternative specimens is recommended in medical examiner cases to provide supporting evidence with regard to the origin of ethanol, whether this reflects antemortem (AM) ingestion or postmortem (PM) synthesis. Introduction Because of the existence of statutory blood alcohol concentration (BAC) limits for driving a motor vehicle, which range from 0.20 to 0.80 g/L (0.02–0.08 g%) in various countries, great care is needed when forensic toxicological results are interpreted (1). When legislators introduced these threshold alcohol limits, they were intended to apply to blood samples taken from living subjects (apprehended drivers) under sterile conditions and not medical examiner cases (death investigations). The enforcement of concentration per se statutes creates a razor sharp dividing line between being charged with a drink-driving offense or not. In PM toxicology, BAC can be determined with a high level of accuracy, precision and specificity, but the results must be interpreted carefully, because of various PM artifacts, not the least of which is the possibility that ethanol, at least in part, was produced in blood and tissues after death (2, 3). When a cadaver starts to decompose and putrefaction processes are advanced, glucose and/or other substrates in the blood ferment to produce ethanol and carbon dioxide (4). Whenever this takes place, the autopsy BAC has dubious validity and is unreliable as a reflection of consumption of ethanol and/or drunkenness before death. Much will depend on the PM interval, degree of trauma to the body, environmental temperature, the stage of decomposition and spread of bacteria from the gut (5). Exposure to subzero temperatures prevents bacterial growth and the concomitant fermentation of glucose. Likewise, with a short PM interval of <24 h, there is insufficient time and opportunity for microbes to generate ethanol by fermentation above analytical cutoff concentrations of 0.1–0.2 g/L (6). Interpreting a PM BAC is simplified if there are alternative biological specimens available for analysis, such as vitreous humor (VH) and urine (7), because PM synthesis of ethanol is less likely to occur in these than in blood (8). During autolysis, bacteria leaks from the bowel into the abdominal cavity then spreading into the major blood vessels and eventually invading most parts of the body (9). By comparing and contrasting the concentrations of ethanol determined in blood with alternative specimens (e.g. VH and urine) provides a means of judging whether PM synthesis was a problem to consider (10). Urine secreted by the kidney is enclosed within the bladder and its glucose content is negligible in healthy individuals (11), compared with blood, which contains 90–110 mg/100 mL glucose (0.09–0.11 g/100 mL). However, urine produced by people with untreated diabetes might be loaded with sugar thus providing a good substrate for PM synthesis of ethanol (12, 13). Although VH contains glucose, the favorable anatomically location of the eye sockets offers protection against the spread of bacteria from the gut after death. Hence VH is a desirable body fluid for analysis of ethanol and other drugs in PM toxicology (14). When neither VH nor urine is available, forensic pathologists might resort to sampling alternative specimens, such as bile from the gall bladder and/or cerebrospinal fluid (CSF) drained form cerebral ventricles or plural cavity effusions for toxicological analysis (15). This article presents a statistical analysis of the relationship between the concentration of ethanol determined in PM femoral venous blood and in various alternative specimens in connection with routine medical examiner cases in Sweden. The results are discussed in relation to the correlation between ethanol concentrations in blood and in the alternative specimens and the limits of uncertainty if and when femoral BAC is predicted from the concentration in one of the alternative specimens. Methods Forensic autopsies in Sweden All sudden, unnatural or suspicious deaths in Sweden are investigated by the police authorities and a forensic autopsy, including a comprehensive toxicological analysis, is ordered. Forensic toxicology in Sweden (population 10 million) is centralized to one laboratory, the National Laboratory of Forensic Genetics and Forensic Toxicology, located in the city of Linköping, Sweden. Medicolegal autopsies are carried out at the six university teaching hospitals in Umeå, Uppsala, Stockholm, Göteborg, Linköping and Lund. For the past 30 years, the annual number of forensic autopsies has remained fairly constant at just over 5000 cases/year and the vast majority of these includes the determination of ethanol and other drugs. Biological specimens The biological specimens submitted for toxicological analysis in medical examiner cases depends on the circumstances surrounding the death and the condition of the body. As a general rule, femoral blood, VH and bladder urine are the preferred specimens and these are taken by the pathologists or their assistants for toxicological analysis. However, other biological specimens are also sometimes received for analysis, such as bile, CSF, cardiac blood, pleural effusions or liquid from the lungs. Femoral blood samples were taken in a standardized way by making an incision in the right or left thigh, exposing the vein and then drawing blood before evisceration of the body. Care is taken not to obtain an admixture of venous and central blood by appropriate clamping. The blood aliquot is transferred to a container with enough potassium fluoride (KF) to give a final concentration of 1–2% w/v. When urine was present in the bladder, a suitable aliquot (>10 mL) was transferred to a clean and dry tube containing KF as preservative. VH was taken from one or both eyes and if necessary pooled to increase the volume of specimen, when sent for analysis of ethanol. Bile was obtained from the gall bladder with a needle and CSF was aspirated from the cerebral ventricle. Pleural cavity effusions were collected from the right pleural cavity and liquid from the right lung was collected by incising and squeezing the tissue. All biological specimens were shipped for toxicological analysis to the forensic toxicology laboratory in Linköping by rapid delivery and in refrigerated containers. Determination of ethanol The concentrations of ethanol in blood and other biological specimens were determined by headspace gas chromatography (HS-GC) as described in more detail elsewhere (16). For routine analysis, two HS-GC instruments from the same manufacturer (Perkin-Elmer, Waltham, WA, USA) were used; AutoSystem XL with TurboMatrix 110. The two instruments were fitted with fused silica capillary columns Rtx BAC-1 and Rtx BAC-2 purchased from Restek Corporation (Bellefonte, PA, USA). Two laboratory technicians were responsible for the toxicological analysis and they worked independently with different HS-GC instruments and diluter-dispenser equipment. The containers with blood and other specimens were rotated several times to ensure adequate mixing prior to removal of an aliquot (100 μL), which was dilution with 1 mL internal standard for GC analysis. The dilution of the blood and other specimens was done with equipment purchased form Eppendorf (diluter-dispenser N 5213). One of the GC methods made use of n-propanol as internal standard and the other used t-butanol. The diluted specimens were ejected into 22 mL glass vials, which were immediately made airtight with a butyl rubber stopper and crimped on aluminum cap. All specimens were analyzed in duplicate and the mean concentration of ethanol was used in the statistical analysis. The HS-GC responses were linear from 0.10 to 4.0 g/L and concentrations <0.10 g/L were reported as negative (not detected), which was therefore used as the limit of quantitation (LOQ) of the analytical method. Calibration of the HS-GC was carried out weekly using absolute ethanol (99.5%) supplied by Kemetyl (Haninge, Sweden) after making serial dilutions with distilled water to cover the concentration range expected in PM specimens. Furthermore, aqueous ethanol standards purchased from Cerilliant (Round Rock, Texas) were used as controls and positioned before and after a series of biological specimens within each analytical run. The standard deviation (SD) of this HS-GC method increased with concentration of ethanol in the samples, and with a mean concentration of 1.5 g/L the coefficient of variation was ~1–2% indicating a high analytical precision (17). Statistical analysis The statistical analysis of the results was done using MedCalc Statistical Software (version 19.1.3) purchased from MedCalc (Ostend, Belgium). The relationship between concentrations of ethanol in femoral blood and in each of the alternative PM specimens was evaluated by least squares linear regression and correlation analysis. Femoral BAC served as the dependent or outcome variable (y-variate) and each of the alternative specimens served as the independent or predictor variable (x-variate). This particular choice of x- and y-variates was made because the aim of the study was to estimate femoral BAC from the concentration determined in one of the alternative specimens (18). The linear regression model resulted in an equation of the form y = a + bx, where y = femoral BAC and x = concentration of ethanol in one of the alternative specimens. The coefficients in the equation a = y-intercept and b = regression coefficient and a Pearson correlation coefficient was also calculated for this linear relationship. The SD of residuals in the regression analysis allowed calculating the 95% confidence limits on the mean BAC as well as the 95% prediction interval for a single new observation. Results are also presented of mean and median concentration ratios of ethanol in femoral blood and one of the alternative biological specimens. However, these ratios were only relevant when the concentrations of ethanol in both types of biological specimens exceeded the analytical cutoff of 0.10 g/L (LOQ). Results Scatter plots of the relationship between femoral BAC and the concentrations in each of the alternative biological specimens are shown in Figure 1. The number of data points (N), the Pearson correlation coefficient (r) and the least squares linear regression equation (y = a + bx) are shown in the top-left corners of each plot. Figure 1 Open in new tabDownload slide Scatter plots, correlation coefficients and linear regression analysis of the relationships between femoral BAC (y-variate) and the concentrations determined in alternative biological specimens (x-variates). Figure 1 Open in new tabDownload slide Scatter plots, correlation coefficients and linear regression analysis of the relationships between femoral BAC (y-variate) and the concentrations determined in alternative biological specimens (x-variates). Figure 1 shows that femoral BAC and the concentrations in each of the alternative specimens are highly correlated and the Pearson correlation coefficients were all statistically highly significant (P < 0.001) and ranged from r = 0.92 (urine) to r = 0.98 (CSF). The corresponding coefficients of determination (R2) ranged from 85% (urine) to 96% (CSF), and this statistic represents the percentage of variance in BAC explained by the linear regression on the concentration in an alternative specimen. Table I Regression statistics for the linear relationship between concentrations of ethanol in femoral blood (y-variate) and in one of the alternative biological specimens (x-variate) Alternative specimen . N . Regression coefficienta . y-Intercept, g/L . r (R2)b . Residual SD g/Lc . Standard error SD/|$\sqrt{N}$| g/L . Pleural effusions 37 0.849 0.058 0.96 (92%) ±0.282 ±0.0464 Lung fluid 61 1.029 0.139 0.94 (88%) ±0.386 ±0.0494 CSF 48 0.858 0.013 0.98 (96%) ±0.273 ±0.0394 Bile 80 0.902 −0.013 0.94 (88%) ±0.412 ±0.0461 Urine 727 0.783 −0.022 0.92 (85%) ±0.390 ±0.0145 VH 134 0.796 0.079 0.97 (94%) ±0.216 ±0.0187 Alternative specimen . N . Regression coefficienta . y-Intercept, g/L . r (R2)b . Residual SD g/Lc . Standard error SD/|$\sqrt{N}$| g/L . Pleural effusions 37 0.849 0.058 0.96 (92%) ±0.282 ±0.0464 Lung fluid 61 1.029 0.139 0.94 (88%) ±0.386 ±0.0494 CSF 48 0.858 0.013 0.98 (96%) ±0.273 ±0.0394 Bile 80 0.902 −0.013 0.94 (88%) ±0.412 ±0.0461 Urine 727 0.783 −0.022 0.92 (85%) ±0.390 ±0.0145 VH 134 0.796 0.079 0.97 (94%) ±0.216 ±0.0187 aSlope of the regression line. br = Pearson’s correlation coefficient; R2 = coefficient of determination in percent. cSD = residual standard deviation. Open in new tab Table I Regression statistics for the linear relationship between concentrations of ethanol in femoral blood (y-variate) and in one of the alternative biological specimens (x-variate) Alternative specimen . N . Regression coefficienta . y-Intercept, g/L . r (R2)b . Residual SD g/Lc . Standard error SD/|$\sqrt{N}$| g/L . Pleural effusions 37 0.849 0.058 0.96 (92%) ±0.282 ±0.0464 Lung fluid 61 1.029 0.139 0.94 (88%) ±0.386 ±0.0494 CSF 48 0.858 0.013 0.98 (96%) ±0.273 ±0.0394 Bile 80 0.902 −0.013 0.94 (88%) ±0.412 ±0.0461 Urine 727 0.783 −0.022 0.92 (85%) ±0.390 ±0.0145 VH 134 0.796 0.079 0.97 (94%) ±0.216 ±0.0187 Alternative specimen . N . Regression coefficienta . y-Intercept, g/L . r (R2)b . Residual SD g/Lc . Standard error SD/|$\sqrt{N}$| g/L . Pleural effusions 37 0.849 0.058 0.96 (92%) ±0.282 ±0.0464 Lung fluid 61 1.029 0.139 0.94 (88%) ±0.386 ±0.0494 CSF 48 0.858 0.013 0.98 (96%) ±0.273 ±0.0394 Bile 80 0.902 −0.013 0.94 (88%) ±0.412 ±0.0461 Urine 727 0.783 −0.022 0.92 (85%) ±0.390 ±0.0145 VH 134 0.796 0.079 0.97 (94%) ±0.216 ±0.0187 aSlope of the regression line. br = Pearson’s correlation coefficient; R2 = coefficient of determination in percent. cSD = residual standard deviation. Open in new tab The least-square regression lines pass through the mean of the x- and y-variates, corresponding to femoral BAC and mean concentration in an alternative specimen. Table I shows the residual standard deviations (SDs) depicting the scatter of data points around the regression lines, which is an important statistic when the y-variable is predicted from certain values of the x-variable. The 95% confidence limits on the mean femoral BAC and the corresponding 95% prediction intervals for a single new observation are shown in Table II. Confidence limits define the range of possible values for the sample mean and are therefore dependent on sample size. The values in Table II apply to the mean ethanol concentration in femoral blood and one of the alternative specimens, but in principle any value of the x-variate could be entered into the regression equation, although the formula for calculating uncertainty is than more complicated. At ethanol concentrations above and below the mean concentration, the confidence limits and prediction intervals are slightly wider. Table II Calculated mean femoral BAC at the mean concentration of ethanol in one of the alternative biological specimens along with 95% confidence limits on the mean and 95% prediction intervals for a single new observation Alternative specimen . N . Mean ethanol concentration, g/L . Mean femoral BAC g/La . 95% confidence limits on mean BACb . 95% prediction interval on new observationb . Pleural effusions 37 0.73 0.68 0.59–0.77 0.11–1.24 Lung fluid 61 1.08 1.25 1.15–1.35 0.48–2.02 CSF 48 1.63 1.41 1.33–1.49 0.86–1.96 Bile 80 1.26 1.12 1.03–1.21 0.30–1.94 Urine 727 1.67 1.29 1.25–1.31 0.51–2.07 VH 134 1.35 1.15 1.11–1.19 0.72–1.58 Alternative specimen . N . Mean ethanol concentration, g/L . Mean femoral BAC g/La . 95% confidence limits on mean BACb . 95% prediction interval on new observationb . Pleural effusions 37 0.73 0.68 0.59–0.77 0.11–1.24 Lung fluid 61 1.08 1.25 1.15–1.35 0.48–2.02 CSF 48 1.63 1.41 1.33–1.49 0.86–1.96 Bile 80 1.26 1.12 1.03–1.21 0.30–1.94 Urine 727 1.67 1.29 1.25–1.31 0.51–2.07 VH 134 1.35 1.15 1.11–1.19 0.72–1.58 aEstimated from regression equation data shown on Figure 1 and in Table I. bThe 95% confidence limits assumed a Student t-value of 2.0 for P = 0.95 for all alternative specimens (the actual t-value ranged from 1.96 to 2.0) Open in new tab Table II Calculated mean femoral BAC at the mean concentration of ethanol in one of the alternative biological specimens along with 95% confidence limits on the mean and 95% prediction intervals for a single new observation Alternative specimen . N . Mean ethanol concentration, g/L . Mean femoral BAC g/La . 95% confidence limits on mean BACb . 95% prediction interval on new observationb . Pleural effusions 37 0.73 0.68 0.59–0.77 0.11–1.24 Lung fluid 61 1.08 1.25 1.15–1.35 0.48–2.02 CSF 48 1.63 1.41 1.33–1.49 0.86–1.96 Bile 80 1.26 1.12 1.03–1.21 0.30–1.94 Urine 727 1.67 1.29 1.25–1.31 0.51–2.07 VH 134 1.35 1.15 1.11–1.19 0.72–1.58 Alternative specimen . N . Mean ethanol concentration, g/L . Mean femoral BAC g/La . 95% confidence limits on mean BACb . 95% prediction interval on new observationb . Pleural effusions 37 0.73 0.68 0.59–0.77 0.11–1.24 Lung fluid 61 1.08 1.25 1.15–1.35 0.48–2.02 CSF 48 1.63 1.41 1.33–1.49 0.86–1.96 Bile 80 1.26 1.12 1.03–1.21 0.30–1.94 Urine 727 1.67 1.29 1.25–1.31 0.51–2.07 VH 134 1.35 1.15 1.11–1.19 0.72–1.58 aEstimated from regression equation data shown on Figure 1 and in Table I. bThe 95% confidence limits assumed a Student t-value of 2.0 for P = 0.95 for all alternative specimens (the actual t-value ranged from 1.96 to 2.0) Open in new tab Taking urine as an example of an alternative biological specimen, the regression equation (Figure 1) was; Femoral BAC = −0.0223 + 0.783 UAC. At a mean urinary concentration of 1.67 g/L, the expected or predicted mean femoral BAC is 1.29 g/L. The SD of residuals was 0.39 g/L (Table I), so the 95% confidence limits on the mean BAC are given by 95% confidence limits = 1.29 ± 2 × (SD/|$\sqrt{N}$|), where N = 727. 95% confidence limits = 1.29 ± 2 × 0.01446 = 1.29 ± 0.029 g/L. 95% confidence limits range from 1.26 to 1.32 g/L. For a single new PM case when the concentration of ethanol in an alternative specimen is used to predict the femoral BAC, more relevant information is provided by the 95% prediction intervals; 95% prediction interval = 1.29 ± 2 × SD of residuals. 95% prediction interval = 1.29 ± 2 × 0.39 g/L = 1.29 ± 0.78 g/L. 95% prediction interval ranges from 0.51 to 2.07 g/L. In both calculations the mean femoral BAC is the same (1.29 g/L), but the limits of uncertainty are considerably wider for a single new observation ranging from 0.51 g/L to 2.07 g/L in 95 of 100 cases (Table II). This large uncertainty supports results from an earlier study when a linear regression model was used to predict BAC from UAC in PM specimens (19). Table III presents mean and median ethanol concentration ratios in alterative specimens compared with the concentration in femoral blood. The number of cases (N) is lower than that used in the linear regression analysis (Table I), because in some cases ethanol was <0.10 g/L in one or both specimens. The ratios indicate that concentrations of ethanol are ~20–30% higher in CSF, urine and VH compared with femoral blood, whereas the concentrations in bile, pleural cavity effusions and lung fluid were lower or the same as the femoral BAC. Table III Mean, median, the lowest and the highest concentration ratios of ethanol in alternative biological specimens compared with the concentration in femoral venous blood Alternative specimen . Na . Mean ethanol concentration ratio ± SD . Median ethanol concentration ratio . Highest and lowest ethanol concentration ratiosb . Pleural effusions 20 1.20 ± 1.44 1.04 0.08–7.1 Lung fluid 46 0.97 ± 0.49 0.85 0.23–2.6 CSF 40 1.20 ± 0.21 1.19 0.79–1.9 Bile 57 1.22 ± 0.78 1.00 0.26–5.0 Urine 727 1.52 ± 1.06 1.32 0.24–13.0 VH 134 1.18 ± 0.35 1.19 0.21–2.2 Alternative specimen . Na . Mean ethanol concentration ratio ± SD . Median ethanol concentration ratio . Highest and lowest ethanol concentration ratiosb . Pleural effusions 20 1.20 ± 1.44 1.04 0.08–7.1 Lung fluid 46 0.97 ± 0.49 0.85 0.23–2.6 CSF 40 1.20 ± 0.21 1.19 0.79–1.9 Bile 57 1.22 ± 0.78 1.00 0.26–5.0 Urine 727 1.52 ± 1.06 1.32 0.24–13.0 VH 134 1.18 ± 0.35 1.19 0.21–2.2 aN is smaller than in Tables I and II because only cases with ethanol >0.10 g/L in both blood and alternative specimens were included. bThese extreme values should not be used for prediction, owing to the inclusion of outlying values. Open in new tab Table III Mean, median, the lowest and the highest concentration ratios of ethanol in alternative biological specimens compared with the concentration in femoral venous blood Alternative specimen . Na . Mean ethanol concentration ratio ± SD . Median ethanol concentration ratio . Highest and lowest ethanol concentration ratiosb . Pleural effusions 20 1.20 ± 1.44 1.04 0.08–7.1 Lung fluid 46 0.97 ± 0.49 0.85 0.23–2.6 CSF 40 1.20 ± 0.21 1.19 0.79–1.9 Bile 57 1.22 ± 0.78 1.00 0.26–5.0 Urine 727 1.52 ± 1.06 1.32 0.24–13.0 VH 134 1.18 ± 0.35 1.19 0.21–2.2 Alternative specimen . Na . Mean ethanol concentration ratio ± SD . Median ethanol concentration ratio . Highest and lowest ethanol concentration ratiosb . Pleural effusions 20 1.20 ± 1.44 1.04 0.08–7.1 Lung fluid 46 0.97 ± 0.49 0.85 0.23–2.6 CSF 40 1.20 ± 0.21 1.19 0.79–1.9 Bile 57 1.22 ± 0.78 1.00 0.26–5.0 Urine 727 1.52 ± 1.06 1.32 0.24–13.0 VH 134 1.18 ± 0.35 1.19 0.21–2.2 aN is smaller than in Tables I and II because only cases with ethanol >0.10 g/L in both blood and alternative specimens were included. bThese extreme values should not be used for prediction, owing to the inclusion of outlying values. Open in new tab The median CSF/blood ratio was 1.19, median VH/blood ratio was 1.19 and the urine/blood ratio was 1.32 (Table III). The corresponding median blood-to-alternative specimen concentration ratios for bile was 1.0, pleural cavity effusions 1.04 and lung fluid 0.85. These ratios can be used to provide rough estimates of the femoral BAC by analyzing an alternative specimen, although the regression analysis is preferred because estimates of uncertainty are also obtained. Discussion After drinking alcoholic beverages, the ethanol they contain is absorbed into the bloodstream and distributed throughout the total body water compartment and there is no evidence of any binding to plasma proteins or other biomolecules (20). Accordingly, when absorption and distribution of ethanol in all body fluids and tissues are complete, the concentrations present should depend on their relative water contents (21). Accordingly, CSF, VH and urine, which contain >95% water, are expected to contain higher concentrations of ethanol than femoral blood, which is ~80–85% w/w water (22). The scatter plots in Figure 1 verify positive correlations between femoral BAC and the concentrations in each of the alternative biological specimens, being highest for CSF (r = 0.98) and lowest for urine (r = 0.92). The square of the correlation coefficient represents the proportion of variation in the dependent variable (femoral BAC) that is explained by its regression on the independent variable, here the concentration in one of the alternative specimens (see Table I). These percentages ranged from 85% for urine to 96% for CSF. The main use of the regression model is to predict the dependent y-variable (femoral BAC) associated with the concentration of ethanol determined in one of the alternative specimens, the independent or x-variable (23). In some death investigations, there might only be a urine sample available for toxicological analysis and the question arises—what does this say about the deceased person’s BAC? For example, if the UAC was 1.67 g/L, the regression equation (BAC = −0.0223 + 0.783 UAC) predicts a femoral BAC of 1.29 g/L, but in any individual case this might range from 0.51 to 2.07 g/L (95% prediction interval in Table II). Various strategies are available to evaluate the source of ethanol in PM blood, that is, whether this reflects AM ingestion or PM formation. One method is to compare and contrast concentrations of ethanol in blood with those in alternative specimens (3, 7). Finding an elevated BAC but a low or undetected concentration in CSF, VH and/or urine speaks toward PM synthesis of ethanol in the vascular system (3). However, if the concentration of ethanol in VH, CSF or urine is ~20% higher than femoral BAC, this supports the contention that the deceased had consumed alcohol during life. However, care is necessary when ethanol concentrations in urine are interpreted, because the deceased might have suffered from diabetes and excreted sugar through the kidneys giving a good substrate for the production of ethanol after death (24). When a dead body starts to decompose, blood sugar and other endogenous substrates (e.g. amino acids) are converted into ethanol by fermentation. Thus, a normal blood glucose of 1.0 g/L can theoretically yield 0.50 g/L ethanol along with CO2 being liberated in the process. Furthermore, depending on amount of fermentable substrate and circumstances surrounding the death, such as the degree of trauma, open blood vessels and with long PM interval, much higher concentrations of ethanol might be produced, sometimes reaching 1.5 g/L or more (25, 26). During PM examination of nonputrefied bodies, if biological specimens are taken according to an established protocol, and then properly stored (refrigerated or frozen) with a fluoride chemical preservative added (1–2% NaF or KF), the concentration of ethanol determined by HS-GC-FID is not expected to exceed the BAC at the time of death. Accurate assignment of PM ethanol to AM ingestion cannot be overemphasized, owing to the existence of statutory BAC limits for driving or engagement in safety-sensitive work that stipulates abstinence from alcohol. During the investigation of fatal road-traffic crashes or workplace accidents when toxicology results indicate a BAC above some permissible limit might have serious consequences for the family and next of kin of the deceased (1). For example, a road-traffic fatality in Sweden, Norway or Poland showing that the driver had a BAC at autopsy above the statutory limit of 0.20 g/L means that vehicular insurance claims are null and void. Other approaches to investigate whether ethanol might have been generated in the body after death comes from inspecting the gas chromatograms for presence of other volatile substances, such as n-propanol and/or n-butanol, which if present speaks toward PM synthesis of ethanol as well (27). Finding small peaks on the chromatograms at retention times of these other volatile alcohols is therefore a warning flag that the concentration of ethanol in blood might, at least in part, be a PM artifact. The concentrations of the other volatile alcohols (e.g. n-propanol) are always considerably lower than the concentration of ethanol generated by microbial action (28). Identifying the nonoxidative metabolites of ethanol, such as ethyl glucuronide (EtG) or ethyl sulfate (EtS), in blood and/or urine gives evidence that ethanol had undergone hepatic metabolism and therefore consumption of alcoholic beverages during life (29). Finding an elevated autopsy BAC but negative EtG or EtS (below detection limits) leads to a conclusion that ethanol was probably produced in the body after death (30). Whether these nonoxidative metabolites are stable or become degraded as PM interval increases has not been thoroughly investigated. Analysis of the urinary metabolites of serotonin and calculating the 5-hydroxytryptophol/5-hydroxyindoleacetic acid (5-HTOL/5-HIAA) ratio is another well-proven method of resolving the source of ethanol in PM blood (31). Normally 5-HTOL is a minor metabolite (~1%) excreted in urine and 5-HIAA is the dominant (~99%) metabolite resulting in very low 5-HTOL/5-HIAA ratios (32). Acetaldehyde generated during the metabolism of ethanol competes with the intermediate aldehyde involved in the metabolism of serotonin, which causes a shift toward the reduced product (5-HTOL) rather than the expected oxidative 5-HIAA metabolite. The reduction pathway is also facilitated by an excess of the reduced coenzyme (NADH) generated during the metabolism of ethanol (33). Accordingly, when the 5-HTOL/5-HIAA in PM urine is >15, this is convincing evidence the person had consumed alcoholic beverages during life. This approach was utilized in aviation medicine when plane crashes were investigated and when there was massive trauma and burnt bodies recovered, because in such cases late recovery promotes PM synthesis of alcohol in the corpse (34). The analysis of drugs or ethanol in bile increases the window of detection by several hours or longer (35). Support for this comes from the present study, because ethanol was reported as negative (<0.1 g/L) in 18 PM blood specimens, whereas mean and median concentrations in bile were 0.30 g/L and 0.17 g/L, respectively. In five other instances bile was reported as <0.10 g/L, whereas femoral blood contained 0.10. 0.11, 0.19, 0.19 and 0.25 g/L. These latter cases might indicate PM synthesis in blood had occurred (36). The concentration of ethanol in CSF was <0.10 g/L in three cases when BAC was 0.11, 0.12 and 0.14 g/L, also perhaps indicative of ethanol being synthesized in blood after death. When only a blood specimen is available and the toxicological results indicate low concentrations of ethanol, such as <0.5 g/L, this should be considered with caution, especially if the condition of the body shows outward signs of decomposition (37). Increases in BAC can arise from mycotic activity and fermentation processes and the ubiquitous substrate being glucose. There is much to recommend use of a higher analytical BAC cutoff concentration, such as 0.20 g/L, before reporting positive toxicological results for ethanol (3). Low concentrations of ethanol (0.10–0.20 g/L) are often observed in PM blood when the concentrations in VH and/or urine are reported as negative (37). A recent study suggested that ethanol might be generated in PM blood in the sampling tubes shipped to a laboratory for analysis, but it was not made perfectly clear how much fluoride preservative, if any, was added to the specimen tubes (38). A fluoride preservative is an essential component and the concentration should be 1–2% to be effective as enzyme inhibitor when specimens are shipped to a laboratory for analysis (39). Whenever possible the shipment should also be refrigerated or frozen, because this also hinders or prevents formation of ethanol in vitro by microbial action. Without being sure that PM specimens contained 1–2 g% NaF or KF the toxicological results from ethanol analysis should be considered suspect. In conclusion, this study showed that femoral BAC and the concentrations of ethanol in alternative biological specimens were highly correlated. The mean femoral BAC could be predicted from the linear regression model with a good precision, but there was considerable uncertainty of the estimated value in any individual new case. The principal reason for analyzing ethanol in alternative biological specimens in PM toxicology is to give supporting evidence and strengthen the conclusion that the deceased had consumed an alcoholic beverage before death. Declaration There was no external funding applied for or received to undertake this study and none of the authors consider that they have any conflicts of interest in publishing this study and placing the information in the public domain. Acknowledgement We are grateful to Dr. Wolfgang Giebe and other pathologists working at the Department of Forensic Medicine in Stockholm, Sweden where the autopsies were performed and biological specimens obtained for toxicological analysis. References 1. Jones , A.W. , Morland , J.G., Liu , R.H. ( 2019 ) Driving under the influence of psychoactive substances - a historical review . Forensic Science Review , 31 , 103 – 140 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 2. Skopp , G. ( 2004 ) Preanalytic aspects in postmortem toxicology . Forensic Science International , 142 , 75 – 100 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Kugelberg , F.C. , Jones , A.W. ( 2007 ) Interpreting results of ethanol analysis in postmortem specimens: A review of the literature . Forensic Science International , 165 , 10 – 29 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Corry , J.E. ( 1978 ) A review. Possible sources of ethanol ante- and post-mortem: Its relationship to the biochemistry and microbiology of decomposition . Journal of Applied Bacteriology , 44 , 1 – 56 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Ziavrou , K. , Boumba , V.A., Vougiouklakis , T.G. ( 2005 ) Insights into the origin of postmortem ethanol . International Journal of Toxicology , 24 , 69 – 77 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Kalant , H. ( 1968 ) Interpretation of post-mortem ethanol concentrations . Aerospace Medicine , 39 , 633 – 637 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 7. O'Neal , C.L. , Poklis , A. ( 1996 ) Postmortem production of ethanol and factors that influence interpretation: A critical review . The American Journal of Forensic Medicine and Pathology , 17 , 8 – 20 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Pounder , D.J. , Jones , A.W. Postmortem alcohol - aspects of interpretation. In: Karch S.B. (ed). Drug Abuse Handbook , second edition. Taylor and Francis CRC press : Boca Raton, FL , 2007 ; pp. 376 – 401 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 9. Boumba , V.A. , Economou , V., Kourkoumelis , N., Gousia , P., Papadopoulou , C., Vougiouklakis , T. ( 2012 ) Microbial ethanol production: Experimental study and multivariate evaluation . Forensic Science International , 215 , 189 – 198 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Petkovic , S.M. , Simic , M.A., Vujic , D.N. ( 2005 ) Postmortem production of ethanol in different tissues under controlled experimental conditions . Journal of Forensic Sciences , 50 , 204 – 208 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Fine , J. ( 1965 ) Glucose content of normal urine . British Medical Journal , 5444 , 1209 – 1214 . Google Scholar Crossref Search ADS WorldCat 12. Coe , J.I. ( 1993 ) Postmortem chemistry update. Emphasis on forensic application . The American Journal of Forensic Medicine and Pathology , 14 , 91 – 117 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Hess , C. , Musshoff , F., Madea , B. ( 2011 ) Disorders of glucose metabolism-post mortem analyses in forensic cases: Part I . International Journal of Legal Medicine , 125 , 163 – 170 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Kraut , A. , Purchase , C.A. ( 1984 ) Vitreous humor/blood and urine/blood alcohol ratios as an aid to the forensic investigator . Canadian Society of Forensic Science Journal , 17 , 159 – 166 . Google Scholar Crossref Search ADS WorldCat 15. Backer , R.C. , Pisano , R.V., Sopher , I.M. ( 1980 ) The comparison of alcohol concentrations in postmortem fluids and tissues . Journal of Forensic Sciences , 25 , 327 – 331 . Google Scholar Crossref Search ADS PubMed WorldCat 16. Jones , A.W. , Schuberth , J. ( 1989 ) Computer-aided headspace gas chromatography applied to blood-alcohol analysis: Importance of online process control . Journal of Forensic Sciences , 34 , 1116 – 1127 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 17. Jones , A.W. , Ericsson , E. ( 2016 ) Decreases in blood ethanol concentration during storage at 4°C for 12 months were the same for specimens kept in glass and plastic tubes . Practical Laboratory Medicine , 4 , 76 – 81 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Linnet , K. ( 1993 ) Evaluation of regression procedures for methods comparison studies . Clinical Chemistry , 39 , 424 – 432 . Google Scholar Crossref Search ADS PubMed WorldCat 19. Kuroda , N. , Williams , K., Pounder , D.J. ( 1995 ) Estimating blood alcohol from urinary alcohol at autopsy . The American Journal of Forensic Medicine and Pathology , 16 , 219 – 222 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Jones , A.W. ( 2010 ) Evidence-based survey of the elimination rates of ethanol from blood with applications in forensic casework . Forensic Science International , 200 , 1 – 20 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Jones , A.W. , Hahn , R.G., Stalberg , H.P. ( 1990 ) Distribution of ethanol and water between plasma and whole blood; inter- and intra-individual variations after administration of ethanol by intravenous infusion . Scandinavian Journal of Clinical & Laboratory Investigation , 50 , 775 – 780 . Google Scholar Crossref Search ADS WorldCat 22. Lentner , C. (ed) In: Geigy scientific tables: units of measurement, body fluids, composition of the body, nutrition . Ciba-Geigy : Basel , 1981 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 23. Westgard , J.O. , Hunt , M.R. ( 1973 ) Use and interpretation of common statistical tests in method-comparison studies . Clinical Chemistry , 19 , 49 – 57 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Alexander , W. ( 1998 ) Postmortem urinary alcohol is unreliable in diabetes . British Medical Journal , 317 , 206 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Mayes , R. , Levine , B., Smith , M.L., Wagner , G.N., Froede , R. ( 1992 ) Toxicologic findings in the USS Iowa disaster . Journal of Forensic Sciences , 37 , 1352 – 1357 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Pounder , D. ( 1998 ) Dead sober or dead drunk? British Medical Journal , 316 , 87 . Google Scholar Crossref Search ADS PubMed WorldCat 27. Boumba , V.A. , Kourkoumelis , N., Ziavrou , K., Vougiouklakis , T. ( 2019 ) Estimating a reliable cutoff point of 1 propanol in postmortem blood as marker of microbial ethanol production . Chinese Journal of Forensic Medicine , 5 , 141 – 146 . Google Scholar Crossref Search ADS WorldCat 28. Liang , H. , Kuang , S., Guo , L., Yu , T., Rao , Y. ( 2016 ) Assessment of the role played by n-propanol found in postmortem blood in the discrimination between antemortem consumption and postmortem formation of ethanol using rats . Journal of Forensic Sciences , 61 , 122 – 126 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Hoiseth , G. , Karinen , R., Christophersen , A., Morland , J. ( 2010 ) Practical use of ethyl glucuronide and ethyl sulfate in postmortem cases as markers of antemortem alcohol ingestion . International Journal of Legal Medicine , 124 , 143 – 148 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Sundstrom , M. , Jones , A.W., Ojanpera , I. ( 2014 ) Utility of urinary ethyl glucuronide analysis in post-mortem toxicology when investigating alcohol-related deaths . Forensic Science International , 241 , 178 – 182 . Google Scholar Crossref Search ADS PubMed WorldCat 31. Helander , A. , Beck , O., Jones , A.W. ( 1995 ) Distinguishing ingested ethanol from microbial formation by analysis of urinary 5-hydroxytryptophol and 5-hydroxyindoleacetic acid . Journal of Forensic Sciences , 40 , 95 – 98 . Google Scholar Crossref Search ADS PubMed WorldCat 32. Voltaire , A. , Beck , O., Borg , S. ( 1992 ) Urinary 5-hydroxytryptophol: A possible marker of recent alcohol consumption . Alcoholism, Clinical and Experimental Research , 16 , 281 – 285 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Helander , A. , Beck , O., Jacobsson , G., Lowenmo , C., Wikstrom , T. ( 1993 ) Time course of ethanol-induced changes in serotonin metabolism . Life Sciences , 53 , 847 – 855 . Google Scholar Crossref Search ADS PubMed WorldCat 34. Johnson , R.D. , Lewis , R.J., Canfield , D.V., Blank , C.L. ( 2004 ) Accurate assignment of ethanol origin in postmortem urine: Liquid chromatographic-mass spectrometric determination of serotonin metabolites . Journal Chromatography B Analytical Technological Biomedical Life Sciences , 805 , 223 – 234 . Google Scholar Crossref Search ADS WorldCat 35. Ferner , R.E. , Aronson , J.K. ( 2018 ) The toxicological significance of post-mortem drug concentrations in bile . Clinical Toxicology , 56 , 7 – 14 . Google Scholar Crossref Search ADS PubMed WorldCat 36. Lewis , R.J. , Johnson , R.D., Angier , M.K., Vu , N.T. ( 2004 ) Ethanol formation in unadulterated postmortem tissues . Forensic Science International , 146 , 17 – 24 . Google Scholar Crossref Search ADS PubMed WorldCat 37. Ehrlich , E. , Kästner , J., Hegewald , C., Rießelmann , B. ( 2010 ) Alkoholbefunde bei fäulnisveränderten Leichen . Rechtsmedizin , 20 , 258 – 261 . Google Scholar Crossref Search ADS WorldCat 38. Quintas , M.J. , Costa , P., Melo , P., Castro , A., Franco , J.M., Teixeira , H.M. ( 2017 ) Postmortem in vitro ethanol production - it could be more common than we think! Forensic Science International , 274 , 113 – 116 . Google Scholar Crossref Search ADS PubMed WorldCat 39. Jones , A.W. , Hylen , L., Svensson , E., Helander , A. ( 1999 ) Storage of specimens at 4 degrees C or addition of sodium fluoride (1%) prevents formation of ethanol in urine inoculated with Candida albicans . Journal of Analytical Toxicology , 23 , 333 – 336 . Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2020. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) © The Author(s) 2020. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]