A study containing 166 autopsies did not provide 166 paired observations for every ethanol comparison. Issa and Khattab’s 2024 report used 125 cases for comparisons involving vitreous and 127 for the blood–urine correlation. Keeping those denominators attached to the findings is essential when evaluating specimen coverage and claims about alternative matrices.
Which comparisons were actually available?
The overall series contained 649 alcohol-positive cases: 166 deceased and 483 living people. For postmortem blood versus vitreous, Table 4 reports 125 cases, Pearson r = 0.831 and p = 0.001. For postmortem urine versus vitreous, it reports 125 cases, r = 0.973 and p = 0.0001. The blood–urine comparison used 127 cases, r = 0.819 and p = 0.001.
The living blood–urine comparison used 483 cases, with r = 0.489 and p = 0.01. These are four distinct comparisons. Neither 649 nor 166 should be substituted as the denominator for all the reported correlations.
The paper does not fully explain the missing-specimen pattern or demonstrate that people with available paired samples were representative of those without them. Consequently, the results should not be presented as complete three-matrix measurements from every autopsy or as proof that missing specimens were random.
Who and what did the autopsy findings cover?
The records spanned 2017–2020. The study included positive alcohol cases and generally collected autopsy specimens within 24 hours of death. It excluded putrefied bodies, collection beyond 24 hours, environmental contamination, harsh environmental conditions and open wounds that could hasten bacterial invasion.
Postmortem blood was femoral venous blood. Bladder urine was extracted suprapubically, and approximately 2 mL clear vitreous was sufficient for testing under the described procedure. Blood was collected with 1% sodium fluoride, vitreous into dry tubes and urine into sterile containers. The stated storage conditions were 2–8°C, with prompt vitreous analysis or refrigeration if delayed until the next day.
These conditions constrain generalization. A correlation in relatively early, selected autopsies does not validate substitution in advanced decomposition or establish performance in specimens contaminated by trauma.
A strong correlation is not an individual prediction interval
The particularly high urine–vitreous correlation indicates that concentrations tended to vary together in the available paired cases. It does not mean the measurements were equal. Pearson r does not, by itself, specify a slope, intercept, systematic difference or uncertainty for predicting one new person’s blood concentration.
Statistical significance also does not establish which matrix best reflects the concentration during life. No independent antemortem reference concentration was supplied for each of these autopsy comparisons. The correlation cannot identify what proportion of a positive finding arose from drinking or from a postmortem process.
The living and deceased blood–urine correlations differed numerically. That does not make the stronger postmortem correlation a validated conversion rule. Different selection criteria, histories and collection circumstances remain relevant.
Analytical scope and reporting limits
Ethanol was measured by headspace GC–FID, with a reported calibration span of 10–400 mg/dL and a reporting cutoff of 10 mg/dL. Values above 400 mg/dL were diluted and reanalyzed, with the dilution factor applied. Other drug screening and confirmatory procedures described in the paper should not be confused with this ethanol assay.
The methods call the calibration six-point but elsewhere list eight concentrations. They also define detection and quantitation limits without giving clear numerical values for both in each matrix. Those limits cannot simply be inferred from the reporting cutoff. The main case counts total 649, whereas the discussion once states 549; the specimen-specific counts in Table 4 are the relevant reported counts for its correlations.
How should a reviewer use the study?
Keep the total case count, available pair count, exact matrices, collection interval and statistical measure together. Ask whether a proposed clinical or forensic inference was actually tested, and whether missingness could affect the comparison. Detection, correlation, quantitative prediction and attribution of impairment are separate claims.
For another alternative-matrix question, see urine-to-blood ethanol interpretation in 378 autopsies. For individual uncertainty, see vitreous ethanol prediction intervals.
Primary source
Issa SY, Khattab A. Ethanol concentrations in various biological specimens: Living and postmortem forensic toxicology analysis and comprehensive literature review. Journal of Forensic and Legal Medicine. 2024;106:102737. doi:10.1016/j.jflm.2024.102737.
Leave a Reply