This study used prospectively obtained pharmacokinetic and pharmacodynamic data from healthy volunteers who received an IV bolus of THC or placebo to delineate the effects of THC and its primary active metabolite, THC-OH, on HR in humans. Several non-linear mixed-effect (NLME) pharmacokinetic/pharmacodynamic (PK/PD) models for THC and THC-OH were established and tested. These data contribute to a better understanding of the cardiovascular effects of THC.

The PK/PD models indicate one could explain the variability in the THC-induced HR change with the parent THC (model 1B1) or the metabolite THC-OH (model 2B2) or a combination of the two (GEDM). THC-COOH was not related to the change in HR. The EC50 of the THC Emax model was 0.53 µM, 25-fold the EC50 (0.02 µM) for the THC-OH Emax model. The GEDM model indicates that THC and THC-OH act synergistically to increase heart rate. Sex did not contribute to the variability of HR. The CYP2C9 phenotype did not contribute to the HR effects observed as the polymorphism effect is seen in the PK of the terminal metabolite (THC-COOH) only.

The ideal way to define the contribution of THC and THC-OH to the cardiovascular effects of THC would be to investigate the effect of THC alone and the effect of THC-OH alone. Since to date the metabolism of THC to THC-OH cannot be blocked, this is not feasible, and any investigation of the pharmacodynamic effects of THC is confounded by the presence of THC-OH and THC-COOH. Previous investigations have neglected or minimized the effect of THC-OH by noting that the THC-OH concentration is only 1/10 of the concentration of the parent drug [15]. This logic ignores the possibility that the metabolite may be pharmacologically more potent than the parent drug and ignores synergism or antagonism. Our modeling results indicate THC-OH must be from 10- to 25-fold more potent than THC to achieve the change in heart rate seen in our volunteers; plots of HR vs. THC-OH show counterclockwise hysteresis supporting the PK/PD model result, and our GEDM results indicate synergy of THC and THC-OH [20, 23].

Other investigations used plant-derived cannabis rather than THC alone. Thus, the pharmacologic effect is a result of the entourage effect that includes cannabidiol (CBD), other cannabinoids and other substances that comprise the 500 compounds in Cannabis sativa, not to mention the product of combustion when the drug is smoked [25].

The presented data are derived from an IV injection of THC alone and thus are free from the confounding effect of the entourage of other substances. Also, although the oral and inhalational routes are more commonly used, the IV route was chosen to avoid issues of uncertain bioavailability of oral and inhalation routes. The employed method of PK/PD modeling provides an objective method to compare the ability of the various models to predict the observed HR changes.

At least three distinct pharmacologic mechanisms could potentially cause the HR effect of cannabis. The first is simple binding as an agonist at the beta-adrenergic receptor, the second is inhibition of the acetylcholine signal to the sinoatrial node decreasing vagal tone, and the third is binding to the CB1 receptor in the myocardium and increasing HR via the endocannabinoid system [8, 26, 27]. There is no evidence to support THC binding to the beta-adrenergic receptor, but beta blockers can mitigate the effect of cannabis-induced tachycardia [26]. Regarding the cholinergic nervous system, THC could inhibit acetylcholine binding at the sinoatrial node, causing increased HR. In fact, atropine has been shown to enhance the HR effect of THC [26]. Autonomic effects have recently been investigated by Nardone et al. in 22 habitual recreational cannabis users who found cannabis flower vaporization decreased cardiac baroreflex and heart rate variability [28].

Evidence for the cannabinoid mechanism suggests the receptor in question is CB1 [8]. Studies that report the binding of THC to CB1 note a Ki of 18 to 32 nM, or about 1/10 of the THCEC50 from model 1B1 of our data [29–31]. This Ki is a free concentration, and since THC is > 90% protein bound, the discrepancy is explained by the protein binding [2]. The free concentration EC50 for THC in our study is 53 nM. The ke0 of 6 l/h is equivalent to an effect site equilibration half-life of 0.12 h, or 7 min, which matches the onset of tachycardia in our volunteer data. There is evidence that CB1 exists in the myocardium, and it may be a receptor for anandamide or 2-AG regulating HR [32]. Several CB1 receptor antagonists can competitively reduce the tachycardic effect of THC [33]. CB1 receptor binding is complex; it is a presynaptic G-protein-coupled receptor (GPCR). Endogenous ligands anandamide and 2-arachidonylglycerol are either partial antagonists or inverse antagonists and use retrograde signaling [34]. The CB1 receptor is buried in the lipid bilayer, and access to the receptor binding site means the agonist must diffuse across the lipid membrane. This could be consistent with the production of the hysteresis loop noted in our study, which reconfirms findings from other reports [15]. Anandamide is also known to cause vasodilation and tachycardia may be a reflex mechanism to maintain cardiac output as systolic and diastolic blood pressure is not altered after THC [8]. Reflex tachycardia may also be an explanation for the HR attenuating effect of beta blockers seen in the Benowitz study [26].

A recent review of the literature of PK/PD modeling of cannabis commented on the complexity of the endocannabinoid system and claimed that because of the lack of specificity of ligand to receptor response and alterations that occur with disease, cannabinoid dose-response relationships may be relatively unpredictable [35].

Here, we summarize the results of related studies for comparison with the presented results: Strougo et al. modeled a concentration-effect relationship for THC using results from a two-way crossover study with 12 healthy male volunteers who were administered placebo or rising doses of THC compounded into an alcohol-based solution that was vaporized at 90-min intervals [15]. Results demonstrated a time delay (hysteresis) in observed effects compared with plasma concentrations of THC, but the author did not state the direction of the hysteresis. The baseline Emax model did not include THC-OH. Baseline effect E0 was 65.3 bpm, EC50 was 30.7 ng/ml (0.098 µM), and Emax was 41.5 bpm. Ke0 was 5.41 l/h, which is a t1/2 to the effect compartment of 7.7 min, similar to all THC models we report. The study concludes that the effect on HR was due to THC only and was mediated by CB1 in human atrial tissue. Although not reported in numeric form, using the graphical data we can infer the THC Cmax in the Strougo study was 100 ng/ml or 0.03 µM; THC-OH Cmax was 0.02 µM, almost 100-fold lower than that achieved in our study where the THC Cmax was 2.6 µM and the THC-OH Cmax was 0.1 µM. The Strougo study may have not reached a THC concentration sufficient to reach the maximum effect, which makes the Emax and EC50 estimates suspect [36]. A comparison of the EC50 from our study 0.57 µM to Strougo 0.098, and Emax 68 bpm vs. 41.5 further supports this hypothesis. No model shrinkage data were reported.

Guan et al. retrospectively reviewed the results of four CB1 antagonists in THC challenge tests in healthy volunteers [36]. THC was not from plant-based cannabis but was compounded into an alcohol-based solution that was vaporized. Population-based NLME Emax models for THC with an effect compartment and a term for a competitive antagonist were fit. Results for THC alone were EC50 0.23 µM (18% relative standard error) with Emax of 64 beats per minute; however, the maximum reported heart rate was 120. They observed dose-dependent reduction in THC-induced HR of 20% at 5 mg to 80% at 1000 mg for the CB1 antagonists drinabant, surinabant, rimonabant and TM38837. THC Cmax was 0.29 µM in the surinabant experiment [33]. The experiments with the four different CB1 antagonists reported by Guan et al. [33] may not have reached a THC concentration sufficient to reach the maximum effect, which makes the Emax and EC50 estimates suspect [36]. Also, THC-OH was not measured. As these drugs are selective CB1 antagonists, the experiments do lend substantial evidence to support the CB1 receptor pathway as the cause of THC-induced tachycardia.

Klumpers et al. [37] studied a tablet formulation of oral THC Namisol®. Healthy volunteers received 6.5 mg or 8.0 mg THC orally, and THC and THC-OH concentrations and HR were measured. They used population-based NLME modeling for compartmental PK and ANOVA for PD. THC Cmax was only 0.03 µM after the 8 mg tablet. Although no PD model was fit, they mention the HR effects matched the THC-OH concentration curve better than the THC curve.

Joerger et al. [17] fit an Emax model for HR to data from nine patients with amyotrophic lateral sclerosis who received single oral doses of THC 5 and 10 mg observed for 8 h. Model parameters were THC-OH EC50 = 3.2e − 4 µM, Emax = 93 bpm. The PD estimates of this model are again suspect because of failure to reach Emax because of low cannabinoid concentrations following oral exposure [36]. To our knowledge, this is the only report of a model for HR and THC-OH, and it found a substantially lower EC50 for THC-OH than we report (0.02 µM). Oral dosing with subsequent low plasma THC concentrations prevented Emax from being correctly assessed; they report an HR Emax of 93 bpm.

In summary, many of the PK/PD modeling attempts to date have suffered from one or more of the following issues: (1) plant-based cannabis confounding by the entourage effect [25]; (2) oral dosing route with low bioavailability or low doses preventing achievement of Emax [36]; (3) plasma concentrations were measured with methods with low precision and accuracy [38, 39], assuming THC and THC-OH have similar pharmacologic targets and or neglected THC-OH completely (see above); (4) no pharmacodynamic modeling or lack of model diagnostic information (see above).

In the presented study, delta-9-THC was administered intravenously, mitigating issues 1 and 2; we used a modern LC/MS assay and pharmacodynamic modeling of the parent and metabolite separately and in a combined model with reported model diagnostic information.