Does CYP2C9*3 Lower THC-COOH Without Lowering Plasma THC? Evidence From 66 Driving Cases

Answer: In a study of 66 men with cannabis-related driving-under-the-influence histories, investigators found substantially lower plasma THC-COOH among carriers of the CYP2C9*3 variant than among CYP2C9 wild-type participants. The respective medians were 18.9 and 39.3 ng/mL (p = .001). Yet parent THC and 11-hydroxy-THC plasma concentrations did not differ significantly between the tested CYP2C9 allele groups. A higher THC-to-THC-COOH ratio was also observed in both CYP2C9*2 and CYP2C9*3 carriers. The results identify an important potential confounder in metabolite interpretation, but they do not provide a validated equation for dating an individual’s cannabis use or deciding whether a driver was impaired.

These observations are from the original, open-access forensic study by Gasse, Vennemann, Köhler and Schürenkamp, “Toxicogenetic analysis of Δ9-THC-metabolizing enzymes” (2020). Its distinguishing feature is that it combines independently assessed CYP2C9/CYP2C19 DNA variants with actual GC–MS plasma measurements obtained in authentic driving-related investigations. It is neither a THC administration experiment nor a direct performance study.

What was the specific forensic question?

THC is transformed into several chemical products, including 11-hydroxy-THC (11-OH-THC) and the carboxy metabolite THC-COOH. Because the proportions of these compounds change across people and time, it can be tempting to treat the ratio of parent THC to a metabolite as a simple clock or as a marker of consumption pattern. The investigation asked whether differences in genes coding for metabolizing enzymes might alter those measurements before anyone attempts that interpretation.

The distinction matters in an actual toxicology review. A THC-COOH value is an analytical observation about a particular analyte in a specified specimen. A claim that the value proves frequent use, a short interval since inhalation, a certain quantity consumed or impairment is an additional inference. If enzyme genetics shifts the relationship between the parent drug and its metabolites, otherwise plausible interpretive shortcuts may need to be reconsidered. This paper tested one possible source of variation, not all the variables necessary to reconstruct a person’s exposure.

Who were the 66 participants, and what was the study design?

Researchers recruited 66 consenting men at the Institute of Legal Medicine in Münster who were undergoing driving-aptitude evaluation in connection with driving under the influence of cannabis. The investigators obtained blood for DNA analysis and paired the genetic results with plasma cannabinoid results from the participants’ underlying driving cases. The men were not randomly assigned to genetic variants or given a controlled experimental THC dose for this study.

All participants were male. Fifty were reported as originating from Germany; others came from a range of countries. The researchers also collected demographic and reported consumption-pattern information, but the quantities and times of cannabis exposure in the driving cases were not established by controlled administration. Neither the reported absence of certain enzyme-inducing medicines nor broad demographic comparisons eliminates unknown exposure-time variability, product variability or confounding.

Each person contributed to the 66-person biological sample. A plasma analyte observation and a DNA-variant call from the same person do not create two independent people. The paper reports several overlapping allele-carrier summaries as well as mutually exclusive complete-genotype groups; those different denominators should not be added together.

What specimens and analytes were actually measured?

The toxicological method used plasma from the driving-related cases, not a reported direct forensic whole-blood THC result. Investigators quantified parent Δ9-THC, 11-OH-THC and THC-COOH using solid-phase extraction followed by gas chromatography–mass spectrometry (GC–MS). The analytical input was 1 mL of plasma, with deuterated internal standards and derivatization before selective-ion-monitoring detection. Concentrations were reported in ng/mL.

DNA testing was a distinct molecular procedure. Blood deposited onto cotton swabs was used for DNA extraction, and a multiplex PCR/SNaPshot single-base-extension assay was developed to call specified variants of CYP2C9 and CYP2C19. A genotype is a measured DNA sequence classification; it is not a direct measurement of active hepatic enzyme turnover in the same participant at the roadside.

The original analytical range among all 66 cases was 0.7–26.8 ng/mL for parent THC, not detected to 8.3 ng/mL for 11-OH-THC and 8.6–149 ng/mL for THC-COOH. Those are study-wide descriptive plasma values, not group thresholds or impairment bands. No hair, oral fluid, breath, vitreous humor or urine cannabinoid series was used to generate the key allele comparisons.

How sensitive was the plasma GC–MS procedure?

The source specified a limit of detection (LOD) of 0.2 ng/mL for both parent THC and 11-OH-THC, and 1.1 ng/mL for THC-COOH. The corresponding limits of quantification (LOQs) were 0.5 ng/mL for THC and 11-OH-THC and 3.2 ng/mL for THC-COOH. The reported linear intervals were 0.5–10 ng/mL for THC/11-OH-THC and 5–100 ng/mL for THC-COOH; specimens exceeding the calibrated range were reanalyzed after dilution.

These limits describe the authors’ particular laboratory method and specimen. They should not be substituted for the cutoff in another jurisdiction’s whole-blood law or the reporting threshold of an unrelated laboratory. Nor does “not detected” for one metabolite demonstrate absence of that metabolite in every conceivable assay. The original paper reports satisfactory overall validation but a larger negative bias for one 11-OH-THC control, which should be considered when evaluating fine numerical differences.

The authors additionally investigated possible conversion of conjugated THC-COOH into free THC-COOH during sample preparation. That is a method-specific analytical artifact question, not a conclusion that a participant had recently used cannabis. The distinction between assay preparation and a person’s actual biology helps keep the reported parent/metabolite measurements in context.

Which CYP2C9 genotypes were observed?

Complete CYP2C9 genotypes were reported as 43 *1/*1, 8 *1/*2, 1 *2/*2, 10 *1/*3 and 4 *2/*3, totaling exactly 66 distinct men. No participant had the *3/*3 genotype. The small *2/*2 subgroup is literally one person; it cannot support a useful estimate of typical variation for that genotype.

The paper also analyzes allele-carrier groups: 43 CYP2C9 wild-type people, 13 carriers of *2 and 14 carriers of *3. Those latter groups overlap because four *2/*3 individuals carry both variant alleles. Thus 43 + 13 + 14 exceeds 66 without implying that additional participants existed. This is a particularly important denominator issue when an expert summarizes the study in a report or testimony.

All comparisons in the next table are cross-sectional group medians. They are not before-and-after changes within individual people, they are not adjusted drug-clearance rates, and they are not percentages of laboratory sensitivity or specificity.

Evidence table: CYP2C9 variant groups and the three measured cannabinoids

Source: original Table 2, Gasse et al. (2020). Variant-versus-wild-type p values were calculated by Mann–Whitney U tests. The IQR is the reported interquartile spread, not a confidence interval. Variant-carrier columns overlap for four *2/*3 participants.

Did CYP2C9*3 carriers really have lower THC-COOH?

Yes, as a group-level association in this sample. The observed plasma THC-COOH median was 18.9 ng/mL for the 14 *3 carriers versus 39.3 ng/mL for the 43 wild-type participants; the authors reported p = .001. That is an association between a measured metabolite and carrier category in this set of consented driving-related cases. It is not evidence that every *3 carrier had a lower result than every wild-type participant, and it does not identify the cause of any particular person’s low THC-COOH result.

The *2 carrier median, 22.9 ng/mL, was also numerically below 39.3 ng/mL. Its reported p = .068 did not reach the study’s .05 significance criterion. A directionally lower point estimate should not be recast as a statistically confirmed difference merely because it resembles the *3 result. Equally, a nonsignificant result is not proof that *2 has no biological effect in every setting.

Although genotype was associated with THC-COOH, the plasma parent THC and 11-OH-THC comparisons were nonsignificant across the allele groupings. The investigators noted unknown amount and time of use as plausible reasons the circulating parent-drug measurement did not show an analogous association. The paper did not isolate those explanations in a new randomized experiment.

What happened to the parent-THC/metabolite ratio?

The reported THC ÷ THC-COOH ratio used two measured concentrations from the same plasma matrix. It was calculated after measurement, not independently detected as a chemical. The median rose from 0.12 in wild-type participants to 0.20 for each variant-carrier group. Both tested contrasts were statistically significant: p = .011 for CYP2C9*2 and p = .032 for CYP2C9*3.

A higher quotient can arise because the denominator, THC-COOH, is lower; it is not necessary for the parent THC numerator to be independently elevated. That is exactly why the combination of these findings matters. Treating this quotient as a self-contained clock would ignore the possibility that the metabolite concentration is influenced by metabolic genetics as well as exposure timing, frequency, distribution and analytical method.

The ratio is dimensionless here because both source values were reported in ng/mL for the same plasma specimen. The result must not be silently translated into a serum, whole-blood or oral-fluid ratio, and it does not specify a legal driving limit. A population comparison of medians also cannot be rearranged into a validated equation for an individual driver.

Did the five CYP2C9 genotype groups show an ordered trend?

The authors additionally ordered the mutually exclusive genotypes as *1/*1 (n=43), *1/*2 (n=8), *2/*2 (n=1), *1/*3 (n=10) and *2/*3 (n=4), then used a Jonckheere–Terpstra trend test. On that prespecified sequence, the THC-COOH group pattern had p=.001 and the THC/THC-COOH ratio pattern had p=.007. Parent THC and 11-OH-THC again had nonsignificant ordered comparisons.

The smallest categories are a major obstacle to an individualized interpretation. A genotype group represented by one person yields that person’s result, not a stable distribution; a group represented by four people can shift considerably with an additional case. Also, a significance test for a particular ordered sequence establishes a statistical trend under those chosen group definitions. It is not a measurement of the precise percentage reduction in that participant’s hepatic clearance.

The source’s genotype boxplots show spread and outliers within major categories. Those plots matter because the question in court is often about one person, while the study’s reported significance tests operate across distributions. The observed overlap prevents the median differences from being treated as a categorical diagnostic boundary.

What did the CYP2C19 analysis show?

To reduce the competing influence of CYP2C9, the researchers examined 43 people with CYP2C9 *1/*1 when comparing CYP2C19 variants. Within these same 43 people, 16 were assigned the CYP2C19 wild-type group, 13 carried *2 and 17 carried *17; three carried both *2 and *17 and appear in both variant-carrier columns. No CYP2C19*3 carrier was observed in this subset.

Across parent THC, 11-OH-THC, THC-COOH and the THC/THC-COOH ratio, the paper reported no significant CYP2C19 variant-versus-wild-type comparisons. That finding is useful because it prevents an interpretation that all tested CYP enzymes showed the same cannabinoid pattern. Nevertheless, the overlapping categories and small subgroup sizes mean a nonsignificant contrast cannot prove absolute equivalence of CYP2C19 function.

The tested DNA variants and the chemical outcomes also address different layers of biology. Even if an enzyme is known to contribute to a pathway in vitro, this specific observational plasma comparison cannot measure how much actual conversion occurred in each person’s liver or eliminate other enzymes and distribution factors.

Evidence table: CYP2C19 comparison restricted to CYP2C9 wild type

Source: original Table 4, restricted to CYP2C9 wild-type participants. Three people carry both CYP2C19*2 and *17. “Nonsignificant” accurately describes the Mann–Whitney tests, not proof that the biochemical effects are mathematically zero.

Why is plasma not interchangeable with forensic whole blood?

Plasma is the liquid fraction obtained from anticoagulated blood after separation of blood cells. Whole blood retains its cellular component. Differences in cannabinoid binding and distribution can make measured concentrations dependent on the chosen matrix. The numeric median of 18.9 ng/mL in this study’s plasma therefore is not automatically 18.9 ng/mL in a later defendant’s whole blood. Nor does a DNA genotype itself have ng/mL concentration units.

The source begins with blood collected in driving-related proceedings, but the key direct analyte table is expressly a plasma analysis. Any forensic comparison with a statutory whole-blood cutoff, an oral-fluid screening result or a specimen taken at another time first requires specimen-specific methodology and an appropriately supported comparison. See the published guide to biological matrices in forensic toxicology.

Similarly, the investigation did not assay actual 11-OH-THC glucuronide formation, hair deposition, exhaled THC, urine metabolite elimination or vitreous drug partitioning. Those are different biological questions requiring separate sampling and analytical procedures. A direct GC–MS plasma concentration is stronger evidence of the analyte in that specimen than an inferred history, but it remains limited to the measured analyte and collection context.

Can genotype correct a THC-COOH concentration or establish last use?

Not with a validated correction from this paper. The researchers propose that CYP2C9 variation could be considered when interpreting cannabinoid metabolism, particularly when an expert attempts to distinguish more frequent from occasional use based on THC-COOH. They expressly call for controlled inhalation studies with known administration timing and exposure patterns before calculating how much that genetic factor should change forensic assessment.

One possible error would be to treat a low THC-COOH value as compelling evidence of occasional use while assuming identical metabolism in everyone. Another error would be to invert the association and claim that anyone with a CYP2C9*3 variant must have used cannabis earlier, later or less frequently. The study supports neither categorical claim. Its participants were selected from driving-related cases, and their controlled absorption curves, doses and last-use intervals were not known.

The presence of *3 also did not yield a significantly higher measured parent THC concentration in this sample. Therefore a statement such as “this allele necessarily doubles active THC” is not supported by the observed results. The paper does not supply sensitivity, specificity or individual prediction intervals for such a claim. For broader context on population and individual variability, see pharmacokinetics and forensic interpretation.

What limits generalization of these results to a legal case?

Exposure was not experimentally standardized. People encountered in driving-aptitude proceedings may have used different cannabis products and quantities at different times. The main toxicokinetic question therefore cannot be resolved by treating all 66 participants as if they received the same inhaled THC dose at the same clock time.

All participants were male. The cohort does not directly evaluate the same source question among women or a representative range of clinical backgrounds. The study’s geographic origins do not establish a broadly representative reference distribution for a future defendant, and allele frequencies varied within its small selected population.

Rare genotypes were absent or extremely sparse. There were no CYP2C9*3/*3 participants, one CYP2C9*2/*2 participant and four *2/*3 participants. The source discusses mechanistic expectations and previous research, but stronger predictions about *3/*3 cannot be inferred as an observed result in these 66 people.

Carrier comparisons are overlapping. Four *2/*3 people contributed to both variant-carrier summaries. CYP2C19 carrier categories also overlap in three cases. Any statistical presentation that treats these combined totals as unrelated people artificially inflates the apparent sample size.

Outcome and statistical scope were limited. The paper used several analyte, genotype and ratio comparisons, mostly with nonparametric group tests. It was not a prospective validation of a classification algorithm, nor a trial of psychomotor performance or a calibration of individual time-since-use predictions. Lack of a significant finding in a subgroup should not be mischaracterized as laboratory proof of no possible effect.

The assay’s conditions matter. Parent and metabolite concentrations were generated by a plasma GC–MS method with defined calibrators, LOQs and preparation procedures. A sample collected in another matrix or measured with a different protocol may not be numerically interchangeable. Method validation is not a substitute for a forensic case’s own quality-control record, which is discussed in the published guide to confirmatory drug testing.

What should a careful forensic report actually say?

A restrained description would say that the original n=66 all-male DUID-related cohort showed lower plasma THC-COOH among CYP2C9*3 carriers and higher plasma THC/THC-COOH ratios among CYP2C9*2 and CYP2C9*3 carrier groups, while tested parent THC and 11-OH-THC group differences were nonsignificant. CYP2C19 comparisons within the restricted CYP2C9 wild-type subset were also nonsignificant. The specific observations come from group data, not individually calibrated genotype-dependent drug profiles.

Before using those observations in litigation, the reviewer would identify which specimen actually yielded the case concentration; how parent drug and metabolites were confirmed; whether a genotype was actually measured; whether the person carries one or more tested variants; and how dose, route, prior use and sampling time were established. If information is absent, the uncertainty should be stated rather than silently resolved using a group median.

Finally, a valid drug measurement and a valid DNA test still do not substitute for evidence about a person’s driving behavior, capacity or the cause of a collision. The study contains no observed roadside impairment classification, no crash-reconstruction test and no validated legal cutoff for CYP2C9 or CYP2C19. Its useful contribution is to identify a possible source of variability in metabolite interpretation, not to award a retrospective diagnosis to a person whose case details were never studied.

Frequently asked questions

Does CYP2C9*3 always lower THC-COOH in an individual?

No. The median plasma THC-COOH was lower in the 14 *3 carriers than in the 43 wild-type men, with p=.001, but individual distributions overlapped. The study did not establish a universal value, an individual clearance multiplier, or a genotype-only interpretation rule.

Was parent THC significantly higher in the CYP2C9 variant groups?

No. Neither parent THC nor 11-OH-THC differed significantly in the reported CYP2C9 carrier comparisons. The significant ratio results were accompanied by lower THC-COOH measurements rather than a demonstrated parent-THC elevation.

Does p=.068 establish a significant CYP2C9*2 effect on THC-COOH?

No. The *2 carrier THC-COOH median was numerically lower, but the paper’s p=.068 exceeded its .05 statistical criterion. It is accurate to describe a nonsignificant direction of difference, not a confirmed positive contrast.

Why do the CYP2C9 groups total 70 when there were only 66 men?

Four men had CYP2C9*2/*3, so they were counted in both the *2-carrier and *3-carrier summaries. The five mutually exclusive complete genotypes sum to 66; the overlapping carrier-group counts must not be interpreted as 70 separate people.

Did the research measure THC in the drivers’ whole blood?

The central quantitative comparisons were expressly made on plasma by GC–MS, alongside separate DNA genotyping from blood material. A source describing a DUID “blood sample” does not justify relabeling its plasma concentrations as contemporaneous statutory whole-blood results.

Did CYP2C19*2 or *17 show the same association?

No statistically significant differences in parent THC, 11-OH-THC, THC-COOH or the ratio were observed in the CYP2C19 comparisons restricted to 43 CYP2C9 wild-type participants. Small and overlapping groups limit stronger exclusionary conclusions.

Can this study prove a driver’s last cannabis use, impairment or crash causation?

No. The study did not administer a uniform THC dose, establish all last-use times, follow standardized concentration-time curves or validate an impairment classifier. Those conclusions require independent case-specific evidence and cannot be recovered from a carrier-group plasma median alone.

Primary original research

Gasse A, Vennemann M, Köhler H, Schürenkamp J. Toxicogenetic analysis of Δ9-THC-metabolizing enzymes. International Journal of Legal Medicine. 2020;134:2095–2103. Original complete open-access study: DOI 10.1007/s00414-020-02380-3.