A global ban during the first half of the 20th century saw hemp (Cannabis sativa) production relegated to the sidelines of industry and agriculture, despite its potential as a cheap, ecological and versatile crop1. The recent revision of cannabis regulations has triggered a fast-paced growth in the industrial hemp sector which has brought to the market many novel hemp-derived products1,2. This proliferation raises consumer safety issues because hemp contains cannabinoids, a class of substances that interact with the animal and human endocannabinoid system. Some of them—especially Δ9-tetrahydrocannabinol (Δ9-THC)—exert a psychoactive effect and others are only pharmacologically active, such as cannabidiol (CBD)3. Cultivation of industrial hemp with a maximum Δ9-THC content of 0.2% is allowed in the European Union4. In 2021, an increase to 0.3% was initiated in the European Union5.

A transfer of cannabinoids into foods of animal origin is conceivable when by-products of hemp production and the whole plant are used as feedstuffs6. So far, despite a few case reports, experimental data regarding the transfer of Δ9-THC from feed into the milk of cows are scarce and analytical techniques often failed to differentiate between psychoactive Δ9-THC and its non-psychoactive precursor Δ9-tetrahydrocannabinolic acid (Δ9-THCA)6. Given that Δ9-THCA can be up to nine times more abundant than Δ9-THC in hemp7, this differentiation is crucial to understand transfer processes and subsequently perform risk analysis.

In this article, we focus on the effects of industrial hemp silage feeding in lactating dairy cows with the aim of quantifying the transfer of cannabinoids into milk and determining possible effects on animal health and risks for consumer health. We collected and analysed milk, blood plasma and faeces, measured physiological parameters and observed animal behaviour. We employ a liquid chromatography–tandem mass spectrometry-based analytical technique that ensures differentiation between Δ9-THC and Δ9-THCA in various matrices, and enables quantification of the cannabinoids Δ8-tetrahydrocannabinol (Δ8-THC), Δ9-tetrahydrocannabivarin (Δ9-THCV), CBD, cannabinol (CBN), cannabidivarin (CBDV) and the two Δ9-THC metabolites 11-hydroxy-Δ9-THC (11-OH-THC) and 11-nor-9-carboxy-Δ9-THC (THC-COOH). The resulting data were used to develop a predictive toxicokinetic model, which can be used to simulate other exposure scenarios and to assess the transfer of different cannabinoids into cows’ milk when using industrial hemp as a dietary supplement for dairy cows.

Results and discussion

We conducted a feeding experiment with lactating Holstein Friesian dairy cows, where corn silage in the diet was first partially replaced with hemp silage made from whole plant hemp (very low cannabinoid concentration, hemp silage A) during an adaptation period followed by feeding hemp silage made from leaves, flowers and seeds only (higher cannabinoid concentration, hemp silage E) at two different supplementation levels (group L, low hemp, 0.84 kg dry matter per cow per day; group H, high hemp, 1.68 kg dry matter per cow per day) during an exposure period, and a subsequent hemp-free depuration period (Fig.1a). Partial replacement of corn silage with hemp silage of different nutritional composition also partially changed the total nutritional composition of the diets; however, the major difference was in total cannabinoid concentration (Table1and Supplementary Table1).

Fig. 1: Study design and results of the feeding experiment.
figure 1

a, Study design and sampling strategy. Grey arrows represent sampling days, striped lines in arrows indicate no sampling for this parameter on that day.aAll animals per group (n = 4).bTwo animals per, Feed intake (b), milk yield (c), respiratory rate (d) and heart rate (e) during the feeding experiment. Violin plots represent the density distribution; black error bars are the standard deviations of the means (black dot); red dots represent the median; blue dots (jittered) are raw data. Means followed by a common letter are not significantly different by a Tukey test.Pvalues (withtvalues and 95% confidence intervals for standardized effect size measures) are from multiple comparisons according to Tukey post hoc tests and are listed in Supplementary Table2and Supplementary Fig.1. To show differences more clearly, standardized effect size measures of animal health parameters were plotted between time periods (pairwise comparisons) with 95% confidence intervals.f, Pictures of cows during exposure period: (1) prolapsed and reddened nictitating membrane; (2) increased salivation with ponding; (3) pronounced tongue play; (4) nasal secretion.

Table 1 Cannabinoid concentration (mg per kg dry matter) of the different hemp silages

Feeding industrial hemp can affect animal health

The study shows that feeding up to 0.92 kg per cow per day (DM, dry matter) of industrial hemp silage A with very low cannabinoid concentration to dairy cows during the adaptation period had no effect on physiological parameters and health (Fig.1b–e; see Supplementary Table2and Supplementary Fig.1for statistical data).

In contrast, feeding of cannabinoid-rich industrial hemp silage E during the exposure period had a significant effect for both group L and group H. A Kruskal–Wallis test showed that feed intake (χ2 = 28.9,P < 0.001, d.f. = 5) and consequently milk yield (χ2 = 44.5,P < 0.001, d.f. = 5) decreased significantly from the second day of exposure period in both experimental groups (Fig.1b,c). The average ingested doses of ∆9-THC and CBD were 1.6 ± 0.3 and 10.7 ± 1.9 mg per kg body weight for group L, and 3.1 ± 0.7 and 20.4 ± 4.4 mg per kg body weight for group H, whereas the dose of the other cannabinoids was ≤1.1 mg per kg body weight (Supplementary Table3). Similarly, the respiratory (χ2 = 50.2,P < 0.001, d.f. = 5) and heart rate (χ2 = 77.4,P < 0.001, d.f. = 5) decreased significantly within hours in both groups due to the feeding of hemp silage E (Fig.1d,e) and in individual animals fell below physiological reference values, allowing classification as either bradypnea or bradycardia. Concomitantly, changes in animal behaviour and appearance were evident (Fig.1f), such as pronounced tongue play, increased yawning, salivation, nasal secretion formation, prolapse and reddening of the nictitating membrane, and somnolent appearance. Some animals from group H displayed careful, occasionally unsteady gait, unusually long standing and abnormal posture. All changes observed disappeared within two days of discontinuing cannabinoid feeding. Milk constituents (fat, protein, lactose, dry matter, somatic cell count and urea), body temperature and body weight were unaffected in all periods (Supplementary Tables4and5).

For humans, the lowest-observed-adverse-effect-level for ∆9-THC is 0.036 mg per kg body weight6. On average, the cows ingested up to 86 times more ∆9-THC during the exposure period, probably explaining the health effects they exhibited. Another study in ruminants reported no health effects, although this involved feeding only 35 g of industrial hemp, a much lower exposure than the present study8. In animals, only a few studies have reported a reduction in feed consumption due to administration of ∆9-THC and CBD, whereas other studies showed no significant decrease or even an increase in feed consumption9,10,11,12,13. Other factors in the silages, such as fermentation acids or lignin (Supplementary Table6), may also have led to the drop in feed intake, but were not observed when feeding industrial hemp silage A during the adaptation period. In contrast, hemp silage E was markedly richer in fat, which could also have led to a reduction in feed intake14. Studies on the suitability of industrial hemp components, other than seeds and oil, as animal feed are rare15, complicating the interpretation of our results. The reduction in daily milk yield is presumably caused by the decrease in feed intake. A decrease in water consumption (not measured) due to cannabinoids in the feed may have promoted the decrease16,17,18. Bradypnea and bradycardia are rare symptoms in cows that only occur in the course of serious illnesses or can be pharmacologically induced. In animal studies, administration of ∆9-THC (but not CBD) influenced respiratory and heart rate and caused drowsiness, slow movements, ataxia and salivation12,13,19,20,21. A reddening of the conjunctiva after ∆9-THC intake is known in humans22,23.

Although it is clear that the observed effects of industrial hemp silage feeding on animal health were mainly caused by the cannabinoids, it cannot be clearly defined which cannabinoid was responsible. Due to its high concentration in the cannabinoid-rich silage, ∆9-THC is the most likely cause, but combination effects may also play a role16,17,24,25. Finally, it should be considered that hemp plants produce other phytochemicals (for example, terpenes, flavonoids) that may have an impact on the changes observed26,27. Similarly, an effect due to minor changes in nutrient concentration of the diets by replacement of corn silage with hemp silage (Supplementary Table1) could not be completely ruled out.

Our study shows that the feeding of industrial hemp silage to dairy cows, even in small amounts, is associated with health consequences. These seem to be dependent on the cannabinoid concentration of the silage, which is influenced by, among other factors, the variety, the parts of the plant from which the silage is derived and the time of harvest7,28. Due to the manifold parameters influencing the cannabinoid concentration of hemp and derived feed, the innocuousness of hemp cannot be reliably assessed without prior cannabinoid analysis.

Cannabinoid transfer from feed to cow’s milk

Feeding hemp silage resulted in measurable levels of ∆9-THC, ∆9-THCA, ∆9-THCV, CBD, CBN and CBDV in cow’s milk at the end of the adaptation period and during the exposure period. Concentrations of up to 316 µg ∆9-THC and 1,174 µg CBD were detected per kg milk (Fig.2). Maximum values for other cannabinoids in milk were 1.9 (∆9-THCA), 8.0 (∆9-THCV), 2.5 (CBN) and 10.1 µg per kg body weight (CBDV). On the last day of the depuration period, ∆9-THC (group L: 1.4 ± 0.4 µg per kg; group H: 5.0 ± 0.6 µg per kg) and CBD (group L: 7.0 ± 1.9 µg per kg; group H: 16.2 ± 2.6 µg per kg) were still detectable in milk. The other cannabinoids analysed were undetectable in milk during all periods (limit of detection (LOD): 0.1 µg per kg milk for ∆8-THC, 0.5 µg per kg milk for 11-OH-∆9-THC and THC-COOH). At the end of the exposure period, the levels in milk were 6–26 times higher for ∆9-THC, 3–5 times higher for ∆9-THCV and 11–32 times higher for CBD than the corresponding blood plasma levels, pointing to the accumulation potential of the substances in cow’s milk. In contrast, ∆9-THCA did not accumulate in milk and CBDV showed no clear trend (Supplementary Table7). Since CBN was undetectable in plasma, no ratio could be calculated. Urine was collected during the experiment, but due to analytical challenges, that is, inconsistent results after glucuronidase treatment, the data could not be evaluated.

Fig. 2: Cannabinoid concentration in milk.
figure 2

Mean ± s.d. ∆9-THC and CBD levels. Days 1–6 and 8–13: no milk samples analysed. *Data interpolated. During the experimental periods, milk samples were taken on days 7, 14–24, 26 and 28 from both groups L and H (n = 4 per group) for cannabinoid analysis. To compile a sample for analysis, the evening and the following morning milk of each cow were mixed according to their proportion of daily production. The resulting composite samples were stored at −20 °C until HPLC–MS/MS analysis.

In contrast to an earlier study in Holstein calves in which only ∆9-THCA and traces of CBD could be detected in the plasma after hemp feeding8, we were able to detect ∆9-THC, ∆9-THCV, CBD and CBDV, presumably due to the higher dose. Studies on the transfer of cannabinoids from feed into milk are rare and, up to now, only ∆9-THC and THC-COOH have been found in ruminant milk after oral ∆9-THC-exposure29,30. In humans, the transfer of ∆9-THC and CBD into breastmilk after marijuana use has been demonstrated, whereas 11-OH-THC and THC-COOH have not always been found31,32,33. The wide range of reported milk/plasma ratios for ∆9-THC is consistent with an earlier study31.

Toxicokinetic model

We set up two-compartment models, as shown in Fig.3a, with either two (model A) or one (model B) excretion outputs from the central compartment. Model A was used for Δ9-THC and CBD, for which enough data were available, and the simplified model B for the other cannabinoids with fewer data. The model reproduces milk data with a Pearson coefficient of determinationr2of 0.92 (Supplementary Fig.2). The model can approximate Δ9-THC and CBD plasma and faeces levels with a Pearsonr2of 0.93 and 0.90, respectively, even though few data points were available. Figure3b,eshows the relative fate of Δ9-THC and CBD, respectively. Only a small fraction ends up in milk, with most of the substance (77% Δ9-THC and 64% CBD) being ‘eliminated’ in processes such as putative metabolism, biochemical transformations in the gastrointestinal tract and urinary excretion. This hints at a high bioavailability for the compounds, with 20–35% being excreted into faeces. Figure3c,d9-THC) and Fig.3e,f(CBD) show the milk predictive model, obtained using the average of the single model parameters estimates, plotted against the data of all animals in the experiment. The simulation was performed to simulate the same intake scenarios as in the experiment. The predictive models for Δ9-THC and CBD can predict the milk levels in each scenario (L/H) with an average underestimation between 12% and 26% for the exposure period. Nevertheless the model predictions are in line with the observed data range (Pearsonr2between 0.59 and 0.88), such that they appear sufficient for risk assessment. Model B was successfully used to predict cannabinoid milk levels and to provide an estimate for the transfer rates at steady state. The figures showing the results for Δ9-THCA, Δ9-THCV, CBN and CBDV are included in Supplementary Fig.3.

Fig. 3: Overview of the modelling results.
figure 3

a, Model structure schema. Model A (used for Δ9-THC and CBD) has three output routes: faeces, elimination and milk. Model B (used for the other cannabinoids) is a simplification of model A, with only two outputs: total elimination and milk.b, Modelled fractions of Δ9-THC output. Most Δ9-THC is eliminated, some is excreted with faeces with only a small amount found in milk.c,d, Predictive model of Δ9-THC excretion in milk for group L (c) and group H (d). The plots show the Δ9-THC amount (mg d−1) excreted in milk against the experimental data of the cows in the group, in a simulation of the experiment with low (c) and high (d) dose.e, Modelled fractions of CBD output. Compared to Δ9-THC, CBD is excreted in higher amounts in the faeces, hinting to lower bioavailability.f,g, Predictive model of CBD excretion in milk for group L (f) and group H (g). The plots show the CBD amount (mg d−1) excreted in milk against the experimental data of the cows in the group, in a simulation of the experiment with low (f) and high (g) dose.

Both the data and the models show that cannabinoids display biphasic elimination: there is a rapid decline in milk levels following discontinuation of cannabinoid feeding and further decline over a longer period.

Cannabinoids are metabolized in the liver of mammals, where they undergo biotransformation. Our model lumps these processes into ‘elimination’ of the central compartment (model A) and total elimination (model B), without details. Having information on these processes would allow for a more refined model and unify the models for individual cannabinoids into a single one including interconversion. Such processes could be quantified using both in vivo and in vitro techniques. Finally, it should be noted that there is still no information on the fate of cannabinoids in the gastrointestinal tract of the cow prior to absorption.

Read the full article at