A microbial ethanol model that approximated measured blood ethanol in 35 of 93 selected autopsy cases did not establish that 38% of deaths involved postmortem ethanol production. The study measured agreement between a calculation and the observed ethanol concentration. It did not independently determine the microbial fraction in each specimen.
How were the models developed?
Velivasi and colleagues cultured three reference bacterial strains separately: Klebsiella pneumoniae, Staphylococcus aureus and Enterococcus faecalis. Brain Heart Infusion broth was inoculated to approximately 106 colony-forming units/mL. The main cultures were incubated at 25°C for 23 days, either anaerobically throughout or aerobically for five days followed by anaerobic conditions.
Stepwise linear regression related measured ethanol to the concentrations of other alcohols. The investigators measured 1-propanol, isobutanol, 1-butanol and two methyl-butanol isomers; the last two were combined for model reporting. These were culture-derived equations, rather than equations derived from people with known drinking doses and known amounts of postmortem production.
What counted as a successful application?
The authors set an error criterion of less than ±40% when comparing calculated ethanol with measured ethanol. Thus, their term “successful” allowed a substantial difference. It did not mean exact agreement, correct identification of the organism or validated reconstruction of blood alcohol during life.
For the autopsy application, the investigators retrospectively selected 93 blood chromatograms with ethanol above 0.10 g/L and detected higher alcohols and 1-butanol. This selected set was enriched for findings relevant to the proposed models. It was not a random sample of all deaths or all ethanol-positive autopsies. The paper did not provide one uniform blood collection site, death-to-collection interval or storage history for those 93 specimens.
The denominator and a numerical inconsistency
Table 3 reports that at least one K. pneumoniae model met the error criterion in 35 of 93 cases, approximately 38%. The corresponding totals were 18 of 93 for S. aureus and 25 of 93 for E. faecalis. These sets should not be added as if they were mutually exclusive diagnoses of three infections.
The higher-ethanol subgroup contained 54 cases. Table 3 gives 25 successful K. pneumoniae applications in that subgroup, about 46%; the lower subgroup had 10 of 39, about 26%. The discussion instead prints “35/54” beside 46%. That fraction is inconsistent with both the percentage and table. Using the table’s 25/54 preserves the distinction between the subgroup result and the overall 35/93.
Agreement with total ethanol does not isolate its source
In an authentic case, measured ethanol could contain antemortem ethanol, microbial ethanol or both. A model’s proximity to the total does not independently establish that the total was all microbial. Conversely, failure to match does not establish that no microbial production occurred.
The laboratory findings themselves showed dependence on organism, temperature and medium. Denatured blood did not reproduce every higher-alcohol pattern observed in broth. The authors also acknowledged variable microbial succession in real bodies and called for further work on applicability. The study supports investigation of these models, with limits on causal interpretation.
Analytical identification and model validation are separate
Alcohols were measured by headspace GC with flame-ionization detection. Table 1 lists ethanol detection and quantitation limits of 0.1 and 0.5 mg/dL, equivalent to 0.001 and 0.005 g/L, with calibration up to 4.0 g/L. The prose gives a different ethanol quantitation-limit figure, so that discrepancy should be resolved before reproducing the assay. None of those analytical limits establishes biological specificity for microbial ethanol.
A case review should state the selection criteria, model used, error tolerance and independent evidence of origin. The model-match percentage is not a prevalence estimate, an impairment measure or a cause-of-death determination. For another transfer problem, see ethanol models across broth, blood and plasma.
Primary source
Velivasi G, Kourkoumelis N, Sakkas H, Boumba VA. Modeling microbial ethanol production by S. aureus, K. pneumoniae, and E. faecalis under aerobic/anaerobic conditions — applicability to laboratory cultures and real postmortem cases. International Journal of Legal Medicine. 2021;135:2555–2565. doi:10.1007/s00414-021-02638-4.
Leave a Reply