Did CBD-Rich Cannabis Reduce Plasma Cytokines? What a 23-Man Observational Pilot Found

Short answer: In a small study of legal-market cannabis flower, men assigned to a CBD-containing strain had a lower baseline-adjusted composite of three circulating inflammatory cytokines than men assigned to a THC-dominant strain (F = 6.77; p < .02). Yet the strains differed in more than CBD content, consumption was self-directed at home, and the investigators measured no clinical inflammatory disease outcome. Parent THC in separated plasma was numerically lower in the CBD-containing group, but its between-group test was not statistically significant (p = .45). Neither finding establishes that CBD alone caused the cytokine difference or that an individual user was unimpaired.

The evidence comes from Bidwell and colleagues’ original 2018 paper, “A Novel Observational Method for Assessing Acute Responses to Cannabis: Preliminary Validation Using Legal Market Strains” (DOI: 10.1089/can.2017.0038). The authors described a methodological pilot, not a validated clinical anti-inflammatory treatment or a forensic impairment test.

What exact question did the original experiment address?

Two commercially available cannabis flowers were compared after regular users had obtained and consumed the assigned strain in their ordinary environments. One was comparatively THC-dominant; the other contained both THC and substantial CBD. The immediate scientific questions were whether this practical design could produce analyzable cannabinoid measurements, whether subjective responses differed between strain assignments, and whether an exploratory panel of inflammatory cytokines varied between groups.

The question is narrower than the popular proposition that “CBD cancels THC.” CBD content, THC content, participants’ chosen amount, inhalation practices, time to sample collection, and background cannabis-use history were not isolated in a way that could assign any observed difference specifically to one molecule. The authors treated their findings as preliminary reasons for more controlled study. A forensic interpretation must retain that distinction rather than convert a pilot association into a universal pharmacological rule.

How many people were actually studied?

Twenty-four men, ages 21–34, were recruited. They reported using cannabis at least four times monthly over the preceding six months. One man assigned to the CBD-containing strain had no detectable CBD and a cannabinoid pattern inconsistent with his assignment; the investigators excluded his results. Consequently, the principal strain comparison involved 23 independent men: 11 assigned to THC-dominant flower and 12 assigned to THC-plus-CBD flower. That denominator, not 24, belongs beside the outcome estimates.

Each participant attended a baseline appointment and an experimental appointment after three days with his assigned flower. These two observations are not 46 independently recruited subjects. Repeated plasma biomarkers, questionnaire ratings, and memory responses from one participant are also not additional biological replicates. All participants were male, so the observed group differences cannot be assumed to apply in the same way to women, adolescents, older patients, or people with inflammatory illnesses.

Were participants randomly assigned and blinded to their cannabis?

Participants were randomly assigned to the two flowers, and the research assistants who met them were blinded to strain composition. The volunteers themselves could access information about the product and were asked to buy their assigned flower from a local dispensary. The protocol did not blind the participants to perceived differences in product composition, deliver an investigator-controlled THC or CBD dose, or directly observe each inhalation.

The team instructed the men to use the assigned product as they normally would for approximately three days. Before the final laboratory visit, participants used it once more at home and were transported to the facility; a blood specimen was obtained after arrival, described as within approximately 15 minutes of use. The tasks and self-reports then followed. This is a randomized strain assignment inside a naturalistic observational dosing design, not a conventional fixed-dose, placebo-controlled trial.

What were the two actual cannabis products?

The THC-dominant product, described as Blueberry Diesel, contained approximately 17% THC with no CBD. The CBD-containing product, Cannatonic, contained approximately 8% THC and 16% CBD. Percentages described the marketed plant material, not drug concentrations circulating in a participant’s body. They also did not define a delivered systemic dose.

For a given mass of plant, the two products differed in their nominal THC quantity before losses from combustion or vaporization and before user-dependent inhalation. A person could take more material or change puffing behavior. Therefore, the experiment compared two compound-rich products used according to participant choice; it did not add CBD to an otherwise identical fixed-THC exposure under controlled laboratory conditions. That distinction matters both for pharmacological attribution and for how an expert describes a “CBD protective effect.”

Was THC measured in whole blood, plasma, or something else?

The laboratory drew venous blood into an EDTA tube, then centrifuged it at 400 g for 15 minutes to separate plasma. The separated plasma was frozen and analyzed by high-performance liquid chromatography/mass spectrometry using an API5500 multiple-reaction-monitoring method. Reported chemical concentrations in the relevant source table were plasma ng/mL values, not directly assayed forensic venous whole-blood THC. The laboratory measured parent Δ9-THC, CBD, and distinct cannabinoid metabolites, including 11-hydroxy-THC and THC-COOH.

Whole blood contains cellular material and has a different matrix from separated plasma. Substituting a plasma parent-THC value as if it were a same-number whole-blood result would add an unsupported matrix assumption. Similarly, the fact that the sample originated from a venous blood draw does not change what material the assay actually quantified. The published biological matrices guide explains why a defensible report names the processed specimen rather than merely the collection procedure.

What did the original plasma cannabinoid table show?

The following group values are from the experimental-session plasma rows of the original study’s Table 1. Each figure is the source-reported mean with its standard deviation. The p-values compare the two groups; they do not estimate equivalence, prove equal THC delivery, or establish the absence of individual variation.

The study verified substantially different plasma CBD group patterns, consistent with the assigned products for the retained participants. It did not demonstrate statistically different group means for parent THC, THC-COOH, 11-hydroxy-THC, or the authors’ combined arithmetic summary. But a nonsignificant difference is not proof that those concentrations are biologically interchangeable, that individual exposures were matched, or that a specified small effect has been excluded. The large standard deviations underline the dispersion in this small sample.

Why is an average of THC and metabolites not a forensic THC dose?

For exploratory analysis, the investigators calculated the arithmetic mean of measured plasma parent THC, THC-COOH, and 11-hydroxy-THC. They used it as a concentration-related covariate because parent THC changes rapidly after inhalation while metabolites offer additional exposure information. That analytical choice is understandable within a feasibility study, but the three chemical species are not toxicologically interchangeable.

Parent THC is psychoactive; 11-hydroxy-THC is also pharmacologically active; THC-COOH is principally an exposure-associated carboxylated metabolite rather than a direct measure of current intoxication. Simply averaging their numeric ng/mL concentrations is not a measured dose, a validated “total active THC” quantity, a blood-alcohol equivalent, or a legally established impairment threshold. A reader cannot use 22.49 ng/mL from that arithmetic average to declare that a driver had 22.49 ng/mL of parent THC.

The primary chemical measurements remain useful when each analyte and its matrix are stated separately. For interpretive background on timing, distribution, and elimination without an invented one-person clock, see the published specimen collection guide.

Did people using the CBD-containing flower consume more cannabis?

Within the THC-plus-CBD assignment, the reported amount on the experimental day increased from approximately 0.259 grams for a usual use occasion to 0.380 grams (p < .05; reported effect size d = 0.7). The THC-dominant group was not reported to increase its amount relative to its typical pattern. This pattern is compatible with the hypothesis that participants altered use in response to product strength, but self-titration was not directly demonstrated as a controlled physiological mechanism.

The grams were participant-reported flower quantity, not weighed and independently observed consumption at the final session. The actual inhaled dose depends on product content, use method, combustion or aerosolization, puffing and retention. The reported increase therefore cannot be transformed into a measured milligram amount of THC reaching the bloodstream. Nor does a similar group mean on one blood-draw occasion prove that both groups had identical exposure over the prior three days.

Were the men less subjectively intoxicated after CBD-containing flower?

The investigators adjusted their regression analysis for the mean-centered plasma THC-plus-metabolite summary. At its average level, men assigned to THC-plus-CBD flower reported less desire to use additional cannabis (p < .05), feeling less mentally stoned (p < .005), and feeling less physically stoned (p < .02) than men assigned to THC-dominant flower. These were outcomes from participant self-report scales, not clinician-adjudicated impairment diagnoses.

Within the THC-dominant group, the investigators observed positive correlations between their averaged cannabinoid concentration measure and desire (r = .60), mental intoxication (r = .79), and physical intoxication (r = .61), all with reported p-values below .05, and mental intoxication below .005. Within the THC-plus-CBD group, the corresponding correlations were negative but nonsignificant (r = −.30, −.25, and −.29). Such small, group-specific correlations can be unstable. A nonsignificant within-group slope should not be described as proven zero effect, and different within-group p-values do not by themselves quantify a formal interaction.

Did CBD-rich flower significantly protect delayed verbal memory?

No statistically significant strain-by-concentration interaction was found for delayed recall errors. The participants completed the International Shopping List Task, with delayed recall approximately 30 minutes after an orally presented 12-item list and a different list at the two laboratory visits. Within the THC-dominant group, the study’s cannabinoid summary correlated with delayed recall errors (r = .60; p < .05); within the THC-plus-CBD group, that correlation was essentially absent and nonsignificant (r = −.07; p > .05).

The authors speculated that the memory analysis might have been underpowered. That possibility does not turn the reported nonsignificant interaction into positive evidence of CBD neuroprotection. Comparing one significant correlation with one nonsignificant correlation is not a substitute for demonstrating a difference between the correlations. Even a genuine group memory benefit, if confirmed, would not directly establish driving fitness or the accuracy of one person’s eyewitness recollection.

What is the evidence for the lower cytokine composite?

Plasma was separately tested for tumor necrosis factor alpha (TNF-α), interleukin-6 (IL-6), and interleukin-1β (IL-1β) using a multiplex immunoassay. Because the three marker measurements were highly correlated, the investigators analyzed their average as a composite. A baseline-adjusted group comparison found a lower composite following the CBD-containing flower assignment: F = 6.77; p < .02.

This was an exploratory endpoint. The reported test concerns the three-marker average, not a separate statistically confirmed reduction in every named cytokine. It does not measure a clinically validated inflammation diagnosis, immune competence, wound healing, chronic disease activity, or a drug’s therapeutic efficacy. The paper’s printed table and graph also use inconsistent absolute concentration-unit labels, so the responsible interpretation here reports the original adjusted comparison without asserting an unverified absolute cytokine concentration.

Which findings were significant, and which were not?

Reporting the null results matters because a promotional summary could otherwise present every finding as evidence that CBD offsets THC. The design also produced multiple exploratory outcomes; significance levels in one small study should be considered in the context of statistical multiplicity and replication. The primary paper did not supply a validated diagnostic model assigning case-specific probability of impairment or inflammation to one plasma result.

Does the trial prove that CBD itself suppressed immune signaling?

No. CBD-containing flower was associated with a lower baseline-adjusted plasma cytokine composite, but the groups differed in flower THC percentage, flower CBD percentage, reported amount consumed, and likely inhaled dose and timing. The experimental-day dose was not controlled or observed. The analysis did not randomize purified CBD against otherwise identical exposures, nor measure a chronic clinical inflammatory outcome. A causal claim specifically attributing the composite difference to CBD requires additional evidence.

One could instead identify the observed exposure as a product assignment under naturalistic use. That is a defensible statement: men assigned to one commercial product had a different average laboratory marker after a short observation period. It is not the same proposition as “CBD treats inflammation” or “a certain blood CBD concentration prevents cytokine release in this person.” Study design determines which inference can be supported.

How do timing, baseline measures, and self-report limit forensic interpretation?

The participants were instructed to abstain from cannabis for 24 hours before the baseline visit, but that instruction does not establish cannabinoid clearance. The original Table 1 reported nonzero baseline mean plasma parent THC in both groups, consistent with residual presence in some users. At the experimental visit, participants had just used cannabis in an uncontrolled location; the venous sample was collected after travel, and the self-reported and memory measures followed. Parent THC changes rapidly, so individual intervals between final puff, draw, and tests are consequential.

The authors did not obtain a same-day, immediately pre-use plasma sample or pre-use cognitive task. Their design was useful for feasibility but could not definitively isolate an acute change from fluctuations associated with expectancy, prior use, practice, sleep, or cannabis product characteristics. It also excluded people with various medical and psychiatric conditions, limiting extrapolation to clinical populations. A case expert should not conceal those design facts merely because an exploratory group p-value was below .05.

Can these data be used to declare an individual driver unimpaired?

No individual driving outcome was observed. The study collected self-ratings and one verbal delayed-recall task, not actual highway driving, lane-position variability, emergency-response testing, collision records, or a validated roadside decision rule. The fact that a participant reports feeling less “stoned” does not establish that divided attention, visual scanning, or vehicle control is normal. Conversely, detectable parent THC in plasma does not alone demonstrate that a particular person failed a driving standard at a specified time.

Forensic interpretation requires the actual specimen fraction and collection timeline, chemical identities, analytical calibration and uncertainty, relevant functional observations, exposure history, and legally applicable question. Presumptive screening and chemical confirmation must likewise remain distinct, as explained in the published confirmatory drug testing guide. The Bidwell study validated neither a THC/CBD concentration threshold for impairment nor a mathematical conversion from a cytokine composite to driving risk.

What can opposing experts legitimately ask about this research?

Relevant methodological questions include whether the source measured plasma or whole blood, how the 24 recruited people became 23 analyzed people, whether reported flower mass is equivalent to delivered THC dose, which statistical result pertains only to a three-marker cytokine average, whether the verbal-memory interaction actually reached significance, and how variable post-use sampling time affected measured parent THC. It is also appropriate to ask if a claim extrapolates beyond all-male regular users or substitutes subjective ratings for a driving endpoint.

The sound distinction is between what was observed and what the study did not test. Statistical significance in one exploratory laboratory outcome does not warrant an individual diagnosis; lack of significance for another does not prove that an effect is absent. Either party should apply that principle symmetrically, whether arguing that CBD protected a driver, that any measurable THC necessarily established impairment, or that one verbal-memory score settles credibility.

Frequently asked forensic questions

1. Did the researchers study 24 or 23 cannabis users?

They recruited 24 men but excluded one with a plasma cannabinoid profile inconsistent with his CBD-containing assignment. The main analyzed groups were 11 THC-dominant users and 12 THC-plus-CBD users, for 23 people.

2. Were the THC values measured in forensic whole blood?

No. Venous blood was drawn, then processed to obtain plasma for HPLC/mass-spectrometric measurement. A plasma concentration is not numerically interchangeable with a legally defined whole-blood THC measurement.

3. Did nonsignificant plasma THC differences prove the two groups had identical THC exposure?

No. Parent THC group means were 16.35 and 11.97 ng/mL with p = .45. That result fails to establish a statistically significant mean difference in this small sample; it does not demonstrate equivalence of exposure or matched inhaled doses.

4. Was every inflammatory cytokine independently shown to decrease?

No. The reported significant exploratory comparison concerns the baseline-adjusted average of TNF-α, IL-6, and IL-1β (F = 6.77; p < .02). It is not three independently confirmed cytokine effects or proof of treatment of an inflammatory illness.

5. Did the CBD-containing flower significantly prevent delayed memory errors?

No significant strain-by-concentration interaction was demonstrated for delayed recall errors. Differing within-group correlations do not replace the nonsignificant formal memory comparison.

6. Can the authors’ THC-plus-metabolite average establish recent use or impairment?

No. The computed arithmetic average joins parent THC with chemically distinct metabolites measured in plasma. It is neither a dose calculation nor a validated use-time, impairment, or whole-blood statutory threshold.

7. Does feeling less mentally stoned establish safe driving?

No. That was a subjective rating in an uncontrolled home-use observational design. No actual driving, road safety, collision, or individualized functional impairment classification was validated by this study.

Bottom line for expert testimony

The original 23-person study reports a lower exploratory plasma cytokine composite and lower subjective intoxication ratings among men using the CBD-containing flower, alongside nonsignificant between-group parent-THC and metabolite means and no significant delayed-recall interaction. Those findings are worth studying; they do not show that CBD itself independently prevented inflammation, memory impairment, or dangerous driving. They do not establish a patient’s inflammation diagnosis, a user’s precise THC dose, or a driver’s collision causation.

Primary research: Bidwell LC, Mueller R, YorkWilliams SL, Hagerty S, Bryan AD, Hutchison KE. A Novel Observational Method for Assessing Acute Responses to Cannabis: Preliminary Validation Using Legal Market Strains. Cannabis and Cannabinoid Research. 2018;3(1):35–44. doi:10.1089/can.2017.0038.