A model that falls near a measured autopsy ethanol result has not necessarily identified how much ethanol was produced after death. In a 122-case study, “success” meant agreement within a specified percentage of the measured total. Understanding that endpoint is essential before treating the reported percentages as diagnostic accuracy.
The denominator was a selected set of autopsy specimens
Boumba and colleagues retrospectively selected 122 blood chromatograms from an archive. Each case required blood ethanol above 0.10 g/L and at least one detectable higher alcohol. Samples without ethanol or without any measured higher alcohol did not enter this evaluation.
The putrefaction categories were 27 without visible putrefaction, ten with early putrefaction and 85 with advanced putrefaction. This distribution is materially different from an unselected series of all deaths. The publication did not supply a uniform blood collection vessel, case-specific sampling delay, preservative history or storage history for these archived specimens.
Twenty-one equations were tested
The investigators applied seven E. coli equations, one E. faecalis equation, four K. pneumoniae equations, two S. aureus equations and seven C. albicans equations. Their predictors were higher-alcohol concentrations measured by headspace GC–FID; ethanol was the quantity predicted.
Units were not interchangeable: the model table expresses ethanol in g/L and higher alcohols in mg/dL. Entering all values in the same units without the appropriate equation-specific conversion would change the calculation.
The success threshold concerned measured total ethanol
The study calculated percentage error as 100 × (calculated ethanol − measured ethanol) ÷ measured ethanol, and used a band of approximately ±40% to classify agreement as successful. This was a study-defined acceptance criterion, not a confidence interval proving the microbial fraction.
For illustration, if measured total ethanol were 0.50 g/L, the ±40% band would extend approximately from 0.30 to 0.70 g/L. That arithmetic illustrates the criterion; it is not an additional observed case. A prediction in that band would not independently demonstrate that the entire 0.50 g/L was microbial rather than a mixture of sources.
Family-level results were not the accuracy of one equation
The E. coli model family achieved agreement in 76 of 122 cases, reported as 62%. Corresponding family results were 40/122 for K. pneumoniae, 25/122 for E. faecalis, 18/122 for S. aureus and 10/122 for C. albicans. No tested model succeeded in 26/122 cases.
Families contained different numbers of equations, and their successes could overlap. The percentages must not be added as mutually exclusive categories. Nor should 62% be described as the sensitivity of one fixed equation or as proof that 62% of the decedents had no predeath drinking. Models also worked across different microbial species in laboratory cultures, so agreement did not identify which organism was present.
What the comparison could not independently establish
The measured autopsy total was available, but an independently known microbial contribution was not established for each case. Agreement with the total is therefore a different target from accuracy against the unknown quantity formed after death.
The authors proposed that agreement across many bacterial models could support a microbial-origin inference and called for further work on applicability. The study did not establish sensitivity and specificity against independently verified drinking and formation histories, or validate a universal subtraction to reconstruct antemortem blood ethanol.
How the evidence can be used
Document the specimen and analytical limits, identify which equation was selected and why, and distinguish total-concentration agreement from source attribution. The paper’s laboratory and archive results support further evaluation; they do not directly establish impairment or cause of death.
For comparison with a separate laboratory regression, see the experimental 1-propanol model. Its different design and performance statistic should not be combined with this study’s family-level success percentage.
Primary source
Boumba VA, Velivasi G, Kourkoumelis N. Ethanol and Higher Alcohols’ Production in Fungal and Bacterial Laboratory Cultures and Significance for Forensic Samples. Microorganisms. 2024;12:462. doi:10.3390/microorganisms12030462.
Leave a Reply