EVIDENCE LIBRARY
Can a Frozen Plasma Alcohol Retest Be Trusted After Five Months? An 80-Driver Study
Does freezing preserve plasma ethanol for a DUI retest? An 80-driver study at -20 °C found decreases after 2–5 months, with critical forensic limits.
Yes. Freezing an ethanol-positive plasma specimen at −20 °C did not prevent lower results on later analysis in a study of 80 samples from intoxicated drivers. In four separate groups of 20 specimens stored for two, three, four, or five months, measured plasma ethanol was lower after storage. The median reported percentage deviations were −7.8%, −11.2%, −20.5%, and −25.2%, respectively. Four of 20 four-month samples and 11 of 20 five-month samples exceeded the study’s ±25% analytical tolerance comparison. These findings concern plasma assayed enzymatically, not a validated correction for stored forensic whole-blood BAC.
The underlying study is Kocak, Isiklar, Kocak, and Meral (2015), Comparison of blood ethanol stabilities in different storage periods, Biochemia Medica 25(1):57–63, doi:10.11613/BM.2015.006. Its complete original published methods, results, tables, and discussion support a narrow but useful forensic conclusion: specimen preservation, analytical method, matrix, and time in storage must be considered separately before treating a later result as a direct replication of an earlier one.
What question did the frozen-plasma study actually test?
The investigators asked whether ethanol concentrations in plasma aliquots changed when the material remained frozen at −20 °C for different periods. The legal motivation was straightforward: retained specimens may be requested for reanalysis after the initial test. A numerical disagreement between original and later assays can matter, but it is not automatically evidence of misconduct, fabrication, or an invalid initial measurement.
There were 80 samples from drivers with initial ethanol above the assay’s approximately 0.1 g/L lower detection limit. Each was assigned to one of four different groups of 20 by storage duration. These were not the same 20 individuals followed serially through all four intervals. Consequently, the contrast across months is a comparison of different specimen groups and cannot be treated as a precise longitudinal decay curve for one person.
How were the specimens collected, frozen, and tested?
Collection and matrix
Blood was collected into 4-mL gray-stopper tubes containing sodium fluoride and K3EDTA. Samples were gently mixed and centrifuged; the experiment then used separated plasma aliquots, transferred into airtight polystyrene tubes. One aliquot was analyzed promptly and the companion aliquot was held at −20 °C. After thawing, the frozen material was reanalyzed. Samples were rejected for specified preanalytical defects, including inadequate fill, clotting, or improper tubes.
Analytical method
The laboratory measured plasma ethanol on a Roche Cobas C 501 using an alcohol-dehydrogenase enzymatic method, with duplicate testing and internal quality controls. It was not a study of headspace gas chromatography applied to sealed forensic whole-blood tubes. Its reported intra-assay coefficients of variation were 8.27% at a lower control level and 2.36% at a higher control level. Those quality characteristics matter when deciding how much of a small change is analytically distinguishable.
In casework, confirm the specimen and methodology rather than treating every reported “blood alcohol” value as interchangeable. See specimen collection and matrix documentation and the separate discussion of headspace GC-FID blood alcohol testing.
What were the actual results after two through five months?
The table below transcribes the primary study’s group medians from Table 1 and its separately calculated median percentage deviations from Table 2. Concentrations are in mmol/L, the units used in the original source. The percentage column is not calculated by simply dividing the displayed before-and-after group medians: it is the paper’s separately reported deviation statistic.
| Time at −20 °C | Specimens | Initial plasma ethanol, median | After storage, median | Reported median deviation | Individual results outside ±25% TEa |
|---|---|---|---|---|---|
| 2 months | 20 | 31.4 mmol/L | 29.1 mmol/L | −7.8% | 0 of 20 |
| 3 months | 20 | 34.3 mmol/L | 30.5 mmol/L | −11.2% | 0 of 20 |
| 4 months | 20 | 24.3 mmol/L | 19.2 mmol/L | −20.5% | 4 of 20 |
| 5 months | 20 | 39.3 mmol/L | 25.8 mmol/L | −25.2% | 11 of 20 |
Source: Kocak et al. (2015), original Tables 1–2 and accompanying Results; mmol/L values are medians, and the ±25% criterion is the study’s chosen CLIA proficiency-testing allowable-total-error comparison, not a universal forensic admissibility standard.
Every storage-duration group showed a statistically significant initial-versus-retest change in Table 1 (reported p < .001). That does not mean every individual sample lost 25%, or that all changes were unacceptable against the same error allowance. The four-month group is especially important: although its median deviation remained within 25%, four individual specimens exceeded that allowance.
Why are statistical significance and analytical significance different?
A paired statistical test asks whether a consistent shift in paired results is unlikely to have arisen under a statistical null model. It does not answer whether a difference is large enough to violate an analytical acceptance criterion or decide a forensic issue. The authors compared deviations against a ±25% total allowable error borrowed from Clinical Laboratory Improvement Amendments proficiency-testing limits. That is a defined comparator in this study—not a general rule for every forensic BAC method, jurisdiction, or laboratory.
For example, the two-month median deviation was −7.8%, yet the table reported p < .001 for the paired concentration change. Conversely, the four-month median was −20.5%, with 4/20 individual samples beyond the ±25% criterion. Replacing these different questions with “statistically significant means the evidence was ruined” would misread the experiment.
Does the paper prove plasma is safe to retest for four months?
No universal four-month guarantee follows. The paper’s broad conclusion suggests that under its specific procedures, plasma at −20 °C may be sufficiently stable for approximately three to four months. But the detailed Results are more qualified: the four-month cohort had four individual deviations beyond the study’s own ±25% comparison. Its abstract emphasizes the five-month group, while the full Results identify exceptions in the four-month group. The individual-result statement is critical in a defensible forensic reading.
Equally, the observation of a median −25.2% deviation after five months does not prove that any arbitrary specimen loses exactly one-quarter of its ethanol. Participants started at different concentrations; tubes, closures, headspace, freeze–thaw conditions, assay variation, and specimen history may differ from those in a particular criminal investigation. The investigators did not establish a reliable individual back-correction formula.
Why plasma cannot simply be converted into forensic whole-blood BAC
Plasma is the liquid fraction remaining after separation of cells. Forensic whole-blood BAC ordinarily reports ethanol in the full blood matrix, including cellular constituents. The two matrices have different water fractions and may yield different concentrations, even before accounting for assay methods or storage conditions. It would be incorrect to present the study’s plasma enzymatic percentages as experimentally verified losses from a preserved whole-blood headspace GC specimen.
A case-specific evaluation should first identify whether the tested material is whole blood, serum, or plasma; its units; any authorized matrix conversion; the relevant method validation; and the date and conditions of each analytical run. Readings in mmol/L are neither driving-impairment scores nor interchangeable with a jurisdiction’s whole-blood legal threshold without appropriate unit and matrix analysis.
Does sodium fluoride or freezing eliminate all preanalytical change?
Neither intervention makes time irrelevant. The authors collected blood with sodium fluoride and K3EDTA and kept separated plasma frozen, yet later measurements still declined. They did not demonstrate new ethanol formation during these experiments. The absence of an observed increase under their protected frozen protocol should not be generalized to decomposed, contaminated, unpreserved, or improperly stored specimens.
Possible changes in stored ethanol specimens have multiple mechanisms; this study did not isolate a case-specific contribution from volatility, tube permeability, handling, chemical change, or measurement differences. Its data therefore support scrutiny of the actual preservation process, not a confident assignment of a mechanism in an unrelated tube. For the role and limits of preservation, see sodium fluoride in forensic blood alcohol testing.
What evidence matters when a retained DUI sample is reanalyzed?
The practical forensic question is not just whether the retest differs. It is whether the analyst can account for the specimen, custody, storage, and analytical comparability sufficiently to interpret that difference. Relevant records ordinarily include original collection time and tube identification, preservatives and fill volume, any separation or aliquoting, seal and opening history, laboratory receipt, continuous freezer-temperature documentation, freeze–thaw episodes, initial and repeat instrument methods, calibration, controls, replicate agreement, detection and quantitation limits, and documented uncertainty.
A later lower concentration might be compatible with storage-associated change. It does not by itself invalidate the initial result, identify the exact cause of the change, or determine the concentration at an earlier driving time. Similarly, one frozen-plasma study cannot substitute for validation of the forensic laboratory’s own preserved-whole-blood retesting protocol.
What can—and cannot—be inferred from the study?
| Evidence or conclusion | Study support | Important limit |
|---|---|---|
| Direct ethanol measurement | Paired plasma concentrations before freezing and after thawing, using one enzymatic assay | Not forensic whole-blood GC or direct breath ethanol |
| Storage association | Lower plasma ethanol after 2–5 months at −20 °C in different 20-specimen cohorts | Not one person’s serial degradation curve |
| Analytical tolerance | 0/20, 0/20, 4/20, and 11/20 beyond a chosen ±25% comparator | That comparator is not a universal criminal-law cutoff |
| Retest reconstruction | Evidence that a stored aliquot can change | No validated calculation restoring an individual original BAC |
| Alcohol exposure or impairment | Ethanol was analytically present in the included driver specimens | No measured driving performance, individual functional impairment, crash causation, or timing model |
The distinction between analytically measured material, a population/group description, a mathematical reconstruction, and a functional impairment opinion should remain explicit at each step of a forensic report.
How should the study be used in expert testimony?
The narrow, supportable use is to explain why a frozen plasma retest may not reproduce the immediate result and why the magnitude of disagreement depends on actual sample conditions and assay characteristics. An opposing expert should not simply take the four- or five-month median percentage and multiply an individual’s reported BAC by its reciprocal. That would convert group observations into an unvalidated individual correction, while also importing a different specimen matrix and possibly a different analytical technology.
Cross-examination can appropriately examine which claim the witness is making: (1) the laboratory directly measured a concentration in a named specimen, (2) storage may have changed that specimen, (3) a model is being used to estimate an unmeasured earlier concentration, or (4) a conclusion about personal driving ability is being offered. Each transition needs a separate evidentiary foundation. None is established solely by the publication’s four group medians.
Frequently asked forensic questions
Did all frozen plasma samples lose ethanol?
The authors reported lower group results at every storage interval. Their publication does not provide a basis to claim a particular percentage loss for every individual specimen outside these cohorts.
Were the same 20 drivers retested at two, three, four, and five months?
No. Four different groups of 20 specimens were used for the four storage intervals, partly to avoid repeated opening of sample tubes.
Does a p value below .001 show that a retest was analytically unacceptable?
No. A statistical change and exceeding a particular allowable-error limit are different findings. Four of 20 four-month specimens exceeded the chosen ±25% limit even though that group’s median deviation was −20.5%.
Does the study establish that plasma remains unchanged for three months?
No. A median decline of −11.2% was reported at three months; none of those 20 specimens exceeded the study’s ±25% comparator.
Can the five-month −25.2% finding correct an individual’s original BAC?
No. It is a group summary from selected plasma samples, not an individual correction equation validated against preserved forensic whole blood.
Would the conclusions be identical for headspace GC testing of whole blood?
That cannot be assumed. The investigators used separated plasma and an alcohol-dehydrogenase analyzer; whole-blood gas chromatography requires its own specimen- and method-specific stability evaluation.
Can a lower frozen-sample retest establish whether the driver was impaired?
No. The investigation addressed laboratory stability. It did not measure driving tasks, cognitive or motor performance, tolerance, time-of-driving BAC, or causal contributions to an event.
Primary scientific source
Kocak FE, Isiklar OO, Kocak H, Meral A. Comparison of blood ethanol stabilities in different storage periods. Biochemia Medica. 2015;25(1):57–63. doi:10.11613/BM.2015.006. The complete original publication, including Tables 1–2, is available from the journal’s original PDF.
Scope: Scientific and forensic education about plasma specimen stability; individual conclusions require the complete case record, laboratory validation, and applicable jurisdiction-specific evidence.