A Systematic Review and Meta-Analysis of the In Vivo Haemodynamic Effects of Δ9-Tetrahydrocannabinol

∆9-Tetrahydrocannabinol (THC) has complex effects on the cardiovascular system. We aimed to systematically review studies of THC and haemodynamic alterations. PubMed, Medline, and EMBASE were searched for relevant studies. Changes in blood pressure (BP), heart rate (HR), and blood flow (BF) were analysed using the Cochrane Review Manager Software. Thirty-one studies met the eligibility criteria. Fourteen publications assessed BP (number, n = 541), 22 HR (n = 567), and 3 BF (n = 45). Acute THC dosing reduced BP and HR in anaesthetised animals (BP, mean difference (MD) −19.7 mmHg, p < 0.00001; HR, MD −53.49 bpm, p < 0.00001), conscious animals (BP, MD −12.3 mmHg, p = 0.0007; HR, MD −30.05 bpm, p < 0.00001), and animal models of stress or hypertension (BP, MD −61.37 mmHg, p = 0.03) and increased cerebral BF in murine stroke models (MD 32.35%, p < 0.00001). Chronic dosing increased BF in large arteries in anaesthetised animals (MD 21.95 mL/min, p = 0.05) and reduced BP in models of stress or hypertension (MD −22.09 mmHg, p < 0.00001). In humans, acute administration increased HR (MD 8.16 bpm, p < 0.00001). THC acts differently according to species and experimental conditions, causing bradycardia, hypotension and increased BF in animals; and causing increased HR in humans. Data is limited, and further studies assessing THC-induced haemodynamic changes in humans should be considered.


Introduction
∆ 9 -Tetrahydrocannabinol (THC) is the most abundant and widely studied phytocannabinoid, first discovered in 1964 [1]. THC is a partial agonist of both cannabinoid receptors CB 1 and CB 2 and other targets including G protein-coupled receptors GPR55 and GPR18 [2][3][4]. THC possesses interesting therapeutic potential as an antiemetic, appetite stimulant, and analgesic, and for the treatment of glaucoma, epilepsy, Parkinson's disease, and multiple sclerosis [5][6][7]. THC has been shown to be effective against refractory nausea and vomiting in cancer patients undergoing chemotherapy [8]. However, its use as a therapeutic agent is limited by its recognised psychogenic side effects including hallucinations, euphoria, dizziness, mood changes, nausea, and fatigue [8][9][10].
THC has numerous cardiovascular effects in animals and humans. In vitro studies have shown that THC causes endothelium-independent vasorelaxation of rabbit superior mesenteric arteries [11] and vasorelaxation of the rat mesenteric artery through sensory nerves via a CB 1 and CB 2 receptor-independent mechanism [12]. Other studies have found THC to activate a G protein-coupled receptor, inhibit calcium channels, and activate potassium channels in the rat mesenteric vasculature [13] and to cause endothelium-dependent and time-dependent vasorelaxation in the rat aorta [14,15]. In contrast, other studies have shown that THC causes vasoconstriction in guinea pig pulmonary arteries [16], rat mesenteric arteries and aorta [14,17], and rabbit ear arteries [18].
In vivo studies have reported different haemodynamic responses post-THC. An acute administration of THC caused hypotension and bradycardia in anesthetised dogs (intravenously; i.v.), conscious bats (intraperitoneal; i.p.), and humans (oral) [19][20][21]. In contrast, tachycardia and hypertension were reported in rats after i.p. administration of THC [22,23]. More complex effects on BP were induced by THC in anaesthetised rats [24]. The available evidence to date suggests that THC alters the haemodynamics in animals and humans, albeit with conflicting results variable with species, route of administration, and experimental conditions. Therefore, the aim this study was to systematically review and meta-analyse the in vivo literature assessing the effects of THC on the cardiovascular system in all species under different conditions.

Results
From the initial 2743 search results, 1935 relevant publications were identified and evaluated from three databases (Medline, EMBASE, and PubMed). Of these, 30 articles met the inclusion criteria and 1 article was added manually (Figure 1). A summary of the data extracted from the included studies is shown in Table 1. In contrast, other studies have shown that THC causes vasoconstriction in guinea pig pulmonary arteries [16], rat mesenteric arteries and aorta [14,17], and rabbit ear arteries [18]. In vivo studies have reported different haemodynamic responses post-THC. An acute administration of THC caused hypotension and bradycardia in anesthetised dogs (intravenously; i.v.), conscious bats (intraperitoneal; i.p.), and humans (oral) [19][20][21]. In contrast, tachycardia and hypertension were reported in rats after i.p. administration of THC [22,23]. More complex effects on BP were induced by THC in anaesthetised rats [24]. The available evidence to date suggests that THC alters the haemodynamics in animals and humans, albeit with conflicting results variable with species, route of administration, and experimental conditions. Therefore, the aim this study was to systematically review and meta-analyse the in vivo literature assessing the effects of THC on the cardiovascular system in all species under different conditions.

Results
From the initial 2743 search results, 1935 relevant publications were identified and evaluated from three databases (Medline, EMBASE, and PubMed). Of these, 30 articles met the inclusion criteria and 1 article was added manually (Figure 1). A summary of the data extracted from the included studies is shown in Table 1.       Figure 2A,B). A cross-species analysis revealed that THC responses in the three species were significantly different in both BP (p < 0.00001) and HR (p = 0.01) (Figure 2A,B), and acute THC significantly reduced BP in rats and cats, but not in anesthetised dogs (p = 0.18, Figure 2A).  Figure 2A,B). A cross-species analysis revealed that THC responses in the three species were significantly different in both BP (p < 0.00001) and HR (p = 0.01) (Figure 2A,B), and acute THC significantly reduced BP in rats and cats, but not in anesthetised dogs (p = 0.18, Figure 2A).

Conscious Animal Models of Stress or Hypertension
Two publications [43,44] assessed the effect of THC administration on BP in hypertensive rats (n = 22), and one [42] in a rat model of stress (n = 30). Acute and chronic (4-10 days) THC dosing

Dose-Response to THC
Doses ranging from 0.0003 to 770 mg were used in different species. The animal analyses showed a trend in the reduction of BP with higher THC doses (p = 0.07), with no change in HR. In humans, THC caused dose-dependent tachycardia (p = 0.01) (Figure 7).  (HR, (B)), or heart rate (in humans only) (p = 0.01) (HR, (C)) is plotted against the log dose (mg) for each study. Error bars represent 95% confidence intervals (CI). Near-significant and significant dose-dependent effects on the blood pressure in animals (p = 0.07) and on the HR in humans (p = 0.01).

Dose-Response to THC
Doses ranging from 0.0003 to 770 mg were used in different species. The animal analyses showed a trend in the reduction of BP with higher THC doses (p = 0.07), with no change in HR. In humans, THC caused dose-dependent tachycardia (p = 0.01) (Figure 7).  (HR, (B)), or heart rate (in humans only) (p = 0.01) (HR, (C)) is plotted against the log dose (mg) for each study. Error bars represent 95% confidence intervals (CI). Near-significant and significant dose-dependent effects on the blood pressure in animals (p = 0.07) and on the HR in humans (p = 0.01).

Dose-Response to THC
Doses ranging from 0.0003 to 770 mg were used in different species. The animal analyses showed a trend in the reduction of BP with higher THC doses (p = 0.07), with no change in HR. In humans, THC caused dose-dependent tachycardia (p = 0.01) (Figure 7).

Dose-Response to THC
Doses ranging from 0.0003 to 770 mg were used in different species. The animal analyses showed a trend in the reduction of BP with higher THC doses (p = 0.07), with no change in HR. In humans, THC caused dose-dependent tachycardia (p = 0.01) (Figure 7).  (HR, (B)), or heart rate (in humans only) (p = 0.01) (HR, (C)) is plotted against the log dose (mg) for each study. Error bars represent 95% confidence intervals (CI). Near-significant and significant dose-dependent effects on the blood pressure in animals (p = 0.07) and on the HR in humans (p = 0.01).    ) is plotted against the log dose (mg) for each study. Error bars represent 95% confidence intervals (CI). Near-significant and significant dose-dependent effects on the blood pressure in animals (p = 0.07) and on the HR in humans (p = 0.01).

Quality
Among the 31 included publications, 6 publications used randomisation in their design and reported blinding assessment of outcome and measurements. Twenty publications assessed more than one outcome, 19 conducted dose-response relationships, 26 assessed a time window for intervention, 11 measured outcomes >24 h post-drug, and no publications provided incomplete data. There was no significant relationship between the quality score and any outcome (Spearman's rho coefficient of BP 0.22, p = 0.09; HR 0.27, p = 0.07 and BF 0.58, p = 0.3).

Publication Bias
Egger's test showed that bias was present in all studies except in studies in anaesthetised animals, conscious animals (p = 0.001), animal models of stress or hypertension (C) (p = 0.001), and humans (D) (p < 0.0001) (Appendix A, Figure A1).

Discussion
The aim of this study was to determine the effect of THC on haemodynamics in vivo in animals and cannabis-naïve humans. Our analysis has shown that an acute dosing of THC reduced BP and HR, and increased BF in animals of different models. Chronic dosing of THC tended to increase BF in anaesthetised animals and reduced BP in animal models of stress or hypertension. The data concerning the effects of THC in humans was limited to HR only, revealing a dose-dependent increase, suggesting further work is required to determine the full haemodynamic effects of acute and chronic THC administration in humans, especially given the different effects of THC on HR observed across species.
Our meta-analysis showed that acute THC dosing in anaesthetised animals reduced BP and HR, while a subgroup analysis revealed that there was no effect on BP or HR of anaesthetised dogs. However, Cavero et al. (1972Cavero et al. ( , 1974 reported that intravenous administration of THC induced hypotension and bradycardia in dogs anaesthetised with pentobarbital caused by a reduction in the cardiac output and venous return mediated by the autonomic system [19,[25][26][27]. Similarly, Schmeling reported that the reduction in sympathetic activity induced by THC in cats may cause hypotension and bradycardia [34]. It is suggested that the vagus nerve and the sympathetic outflow play a role in these effects induced by THC [36] and can be inhibited by the administration of a CB 1 antagonist [50]. The administration of THC for seven days subcutaneously reduced the increase in HR induced by pentobarbital anaesthetic agent in dogs, suggesting that THC antagonises the pentobarbital effect on the parasympathetic system (inhibiting the vagal tone) [30]. In rats anesthetised with pentobarbital, hypotension was reported after the administration of THC [35]; on the contrary, hypertension was reported in rats anesthetised with urethane post-THC [36], suggesting that THC may act differently with different anaesthetic agents. These studies suggest that the effects of THC in anaesthetised animals (hypotension and bradycardia) are induced through a central mechanism via the activation of CB 1 receptors.
In conscious animals under normal conditions, THC caused a variety of effects: hypotension was observed in bats, an effect which may be related to a change in venous activity [20], whereas another study in rats reported that THC induced tachycardia and hypertension, which are centrally mediated by increasing the level of adrenaline in the circulation [22]. However, studies in rat models of stress and hypertension, showed that THC lowered BP effectively [42][43][44]. The mechanism of the antihypertensive effect of THC in these models still needs to be studied.
Our meta-analysis in cannabis-naïve humans highlighted the limited number of studies investigating the effect of THC in humans (6 publications, n = 123 participants) with insufficient data to meta-analyse BP or regional BF. Studies in cannabis-naïve volunteers showed that the administration of THC orally or by inhalation caused tachycardia [46][47][48][49]51]. Tachycardia is also reported in humans after smoking cannabis [52][53][54] which may indicate that tachycardia induced post-cannabis smoking is caused by THC. The increase in HR caused by THC can be inhibited by CB 1 antagonism [55], suggesting that CB 1 activation may play a role in the haemodynamic effect of THC in humans. A greater number of studies investigating the haemodynamic effect of THC and its mechanisms under normal and pathological conditions in humans are required.
Several studies have reported that phytocannabinoids such as cannabidiol (CBD) may alter the effect of THC. For example, Borgen and Davis suggested that CBD may act as a potential antagonist of the THC effect on HR in rabbits and rats [38] and protects against some of the negative effects of THC in humans with potentially opposite effects on regional brain functions [56,57]. The combination of CBD and THC such as in Sativex ® , a licenced agent for the symptomatic treatment of spasticity in multiple sclerosis, has shown that CBD inhibits the tachycardia effect induced by THC in humans [58].
Dose-response analyses showed a relationship between THC dose and effect size on BP, but not HR, in different animal models, and on human HR. Dose-dependent effects on BP were also observed post-THC in anaesthetized rats [24,36], cats [28], and dogs [26]. A dose of 100 and 200 mg caused a dose-dependent reduction on the BP of conscious bats, but not on HR [20]. HR dose-dependent reduction was reported in anaesthetized dogs [26,27] and conscious monkeys [39]. In human studies, doses between 2.5 and 25 mg were used. A dose-dependent increase in HR was observed in humans after oral THC administration of 5, 10, and 20 mg [21,49]. Over-intoxication has been reported after 20 mg of oral administration of THC in 5 of 21 healthy volunteers [48].
There are a number of limitations to consider in this analysis. First, the principal intention of 10 of the included studies was not to assess the cardiovascular effects of THC administration; therefore, the data extracted through secondary haemodynamic outcomes in this meta-analysis is for hypothesis-generating purposes. Second, the results should be interpreted with caution because of the heterogeneity between studies in terms of THC dose, time, and route of administration; the responses to THC will clearly be dependent upon peak plasma concentration, which are not easily comparable across studies. Indeed, a significant statistical heterogeneity was observed in the majority of the meta-analyses. Third, only 6 out of 31 articles used randomisation and described a masked assessment of outcomes, factors that can influence the reported outcomes. However, we found no significant correlation between study quality and effect size in this review.
In conclusion, this study has summarised the in vivo cardiovascular effects of THC administration. Our analysis demonstrates that THC acts differently according to species, causing tachycardia in humans, and bradycardia, hypotension, and an increase in regional BF in animals under different conditions. THC may be a potential future treatment for cardiovascular disorders, though its use as a single agent will be limited by CB 1 mediated psychogenic side effects, events that could be counterbalanced with other agents such as CBD. Data from human studies using THC alone is limited to heart rate only, thereby further good quality, randomised, blinded studies investigating the haemodynamic effects of THC in humans should be considered.
Prespecified inclusion and exclusion criteria were used to prevent bias; the studies had to be in vivo, assess haemodynamics (BP, HR or BF), be original articles, be controlled studies, and use cannabis-naïve participants. Therefore, the exclusion criteria were: in vitro studies, mixtures of ∆ 9 -THC with other cannabis extracts, studies investigating the interaction of THC with other drugs or cannabinoids, studies not assessing haemodynamics (BP, HR, or BF), review articles, editorials, and uncontrolled studies.

Data Acquisition
Data on BP (mmHg), HR (beats per minute, bpm), and BF (% change from baseline or mL/min) were extracted from the included papers, and the changes in haemodynamics 2 h post-drug after acute THC dosing were used for the analyses. This time point was selected as the peak plasma time is between 30 min and 4 h after oral administration and it was the most common time point when haemodynamics were measured throughout the articles. If there were no measurements taken at this time point (2 h post-drug), the closest time point to 2 h was used for the analyses. In chronic studies, the measurements taken at the end of the studies were used for the analyses. If the exact number of animals used in each drug group was not available, the lowest number of animals within the range given was used for the experimental group (THC), and the highest number was used for the control group. If a crossover design was used in a study, the total number of humans was distributed equally to the two groups. Articles were excluded if data were not available. Grab application (version 1.5) was used to extract values from the figures given in published articles if no values were stated within the text. If the published articles used multiple groups (e.g., to assess dose-dependent effects) with one control group, then the number of humans or animals per control group was divided into the number of comparison groups. For the dose-response analysis, the total dose of the drug administrated up to the time when the haemodynamics was measured was used.

Quality
Eight-point criteria derived from Stroke Therapy Academic Industry Recommendations (STAIR) [59][60][61] and the Cochrane collaborations tool [62] were used to identify the risk of bias. Each of the following criteria was equal to 1 point: randomisation, blinding of outcome assessment, blinding of personnel and participant, assessment of more than one outcome, dose-response relationship, therapeutic time window, assessment of outcome >24 h, and incomplete outcome data.

Data Analysis
The studies were divided into acute and chronic groups. The data from human and animal studies were analysed separately. The animals were divided into two groups, anaesthetised and conscious, as the autonomic nervous system may respond differently in the two conditions [63], then grouped before the analysis in normal and abnormal (i.e., models of stress or hypertension) models and then subgrouped by species (mice, rats, dogs, etc.). For the THC dose-response analysis, the data were grouped according to the endpoint (BP, HR, or BF), and then subgrouped according to the dose. The data from each group were analysed as forest plots using the Cochrane Review Manager software (Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014), and as funnel plots using Stata (StataCorp. 2009. Stata Statistical Software: Release 11. College Station, TX, USA). Funnel plot asymmetry (publication bias) was assessed by Egger's test [64]. Stata was also used for meta-regression that described the relationship between THC dose and effect size. PRISM 7 (GraphPad, Software, La Jolla, CA, USA) was used to produce the figures of dose-response. Since heterogeneity was expected between the study protocols (different species, models, dose, and time) random-effect models were used. The results of continuous data are expressed as mean difference (MD) with 95% confidence intervals (CIs). The studies were weighted by sample size, and statistical significance was set at p <0.05.   Figure A1. Funnel plots for each outcome evaluating the publication bias. The standard error (SE) of the mean difference (MD) in haemodynamics (MD, y axis) for each study is plotted against its effect size (horizontal axis). There was significant bias in conscious animals (B) (p = 0.001), animal models of stress or hypertension (C) (p = 0.001), and humans (D) (p < 0.0001). No significant bias in anaesthetised animals (A).