This study has investigated the association between cannabis use and comprehensively characterized fat phenotypes derived from MR imaging, including visceral and subcutaneous fat as well as ectopic fat in the liver, pancreas and skeletal muscle. Further, this study has presented data on the association between cannabis use and several indices of insulin sensitivity and secretion (both fasting and post-prandial). All MR-derived measurements completed by two independent observers had excellent inter-rater reliability (ICC > 0.9). Further, the used statistical models accounted for potential covariates in both analyses of cannabis use with fat phenotypes and cannabis use with insulin traits. We report two main findings from this study. First, regular cannabis users had significantly greater IHFD (but not SFV, VFV, IPFD or SMFD) compared with never cannabis users. This remained significant after adjusting for age, sex, ethnicity, HbA1c, tobacco smoking, alcohol consumption and physical activity. Moreover, the amount of variance explained by cannabis use was greatest for IHFD (R² = 9.2%). Second, regular cannabis users had significantly greater insulin secretion (as defined by the insulinogenic index 60’) when compared with never cannabis users, whereas insulin sensitivity was not significantly altered in cannabis users.

This study presents novel findings and there are many unknowns regarding the mechanisms that underlie them. We speculate that, at least in part, our results may be attributed to upregulated neurohumoral stimulation from the central ECS, which is supported by evidence of the effect of cannabis use on autonomic regulation of β-cell secretion [18, 19]. The ECS is a neuromodulatory system that acts on both the central and peripheral nervous systems, as well as non-neural peripheral sites [18]. CB₁Rs are present in all sites involved in the control of energy homeostasis. While CB₁Rs are predominantly found in the central nervous system, lower levels are detected in skeletal muscle, liver, pancreas and gastrointestinal tract [18]. CB₁Rs are among the most abundant G-protein coupled receptors expressed in the brain and are present in the forebrain, basal ganglia, cerebellum, hippocampus and cerebral cortex [20]. The ECS is the main central system involved in appetite-regulation and food intake. Pharmacologic administration of CB₁ agonists increases food consumption and, in the case of Δ-9-THC, has been observed even in satiated animals [4, 5, 20]. Moreover, systemic CB₁R blockade by SR141716A (rimonabant) reduces food intake in both fed and fasted animals [20]. Its modulation of food intake is believed to be two-fold. First, through mesolimbic pathways and dopaminergic release, it activates the reward mechanisms for food consumption (usually of highly palatable foods). Second, through hypothalamic regulation of the level and/or activity of anorectic and orexigenic mediators that induce appetite. CB₁R expression in the hypothalamus, a key area for the regulation of energy balance, is relatively low but highly efficient, and activation stimulates the release of key neurotransmitters (including but not limited to noradrenaline and serotonin). Both noradrenaline and serotonin are key neurohumoral regulators of insulin secretion, and they can exert both stimulatory and inhibitory effects on the pancreatic β cells [21]. Acute increases in sympathetic neurotransmission initially inhibit insulin secretion but over longer periods elevate its plasma levels. While it is assumed that noradrenaline mediates the inhibitory actions of sympathetic stimulation on glucose-stimulated insulin secretion, it is accepted that it also has stimulatory actions through its direct activation of β₂-adrenoreceptors and α₂-adrenoreceprtors on both the β and α cells [22]. Serotonin can also influence insulin output, either by stimulating cellular receptors to enhance release of noradrenaline from islet sympathetic terminals, or through direct intracellular regulation of insulin release [21]. It could be speculated then that stimulation of the central ECS enhances autonomic neurotransmission to the pancreatic islet cells and induces supraphysiologic insulin secretion. Insulin drives glucose uptake into cells where it is either stored or used directly as fuel. The predominant sites of glucose uptake are the liver, skeletal muscle and adipocytes. Ectopic fat deposition typically reflects the failure of intracellular lipid homeostasis to prevent lipotoxicity as a consequence of accumulation of saturated lipid-storing white adipocytes [23]. In a hyperinsulinemic state, glucose uptake may exceed the energy requirements of the cells and lead to excess glucose storage, driving fatty infiltration into organs such as the liver. The above arguments suggest that significantly increased IHFD in regular cannabis users observed in the present study might be due to chronic stimulation of CB₁Rs promoting excess insulin secretion. Cannabinoids are also known to activate PPARγ, the isoform of PPARs that mediates some of the metabolic effects seen with both endo- and phyto-cannabinoids. Palomba and colleagues [24] recently recognized a potential role for PPARγ in the modulation of leptin activity by hypothalamic CB₁Rs, which further supports the effect of cannabis through CB_1­R mediation.

We found cannabis use was significantly associated with the insulinogenic index 60’ (reflecting increased insulin secretion at 60 min following mixed meal) but no significant changes in insulin sensitivity. The lack of observed difference in insulin sensitivity is unsurprising, given the overall studied population did not have marked abnormalities of glucose metabolism. Moreover, the group of regular cannabis users had lower median HbA1c (35 mmol/mol) than both non-regular (37 mmol/mol) and never (39 mmol/mol) users (Table 1). Increased insulin secretion may reflect an augmented insulin response following cannabis use and suggests the potential role of CB₁Rs in the gastrointestinal tract. DiPatrizo and colleagues [25] found that a diet high in fat altered intestinal endocannabinoid levels, which upregulated CB₁Rs in vagal nerve fibres. Investigations of β-cell function historically have used intravenous glucose. As far as the oral route is concerned, the relationship becomes more complex and is influenced by the rate of glucose absorption, neural activation, and incretin action. Incretins, such as glucose-dependent insulinotropic polypeptide (GIP) and glucagon like peptide-1 (GLP-1), enhance insulin secretion in response to nutrients and are responsible for the augmented response to glucose when given orally rather than intravenously [26]. Several animal models have isolated CB₁Rs on enteroendocrine cells in rodents and have suggested these play a role in the regulation of incretin secretion [27]. Chia and colleagues [26] demonstrated a significant increase in fasting and 60 min post-oral glucose tolerance test GIP levels following administration of a CB₁R agonist, but no change in GLP-1 levels. They also found an increase in GIP levels in obese individuals, suggesting increased GIP is likely a consequence of increased endocannabinoid levels. Further, rodents fed dietary linoleic acid (which increased weight gain) were found to have elevated endocannabinoid levels in both the liver and the gut [28]. Linoleic acid is also an endogenous activator of PPARs, which have further been implicated in the regulation of metabolism and energy homeostasis. The above arguments may explain why we found increased insulin secretion, but no significant change in insulin sensitivity.

Another possible explanation for our findings is that peripheral CB₁Rs may be preferentially distributed in particular sites and result in augmented physiological effects from stimulation of the ECS. Peripheral CB₁Rs play an important role in glucose homeostasis by modulating lipogenesis in the liver and adipose tissue. CB₁R activation enhances de novo lipogenesis and triglyceride accumulation in hepatocytes and adipocytes, while minimizing lipolysis and fatty acid oxidation [28]. Liu and colleagues [29] demonstrated a reversal of diet-induced IHFD through peripheral CB₁R blockade in rodents. They also found increased fatty acid oxidation with CB₁R blockade in mice both with and without diet-induced obesity. This suggests a possible propensity for cannabis-related effects on hepatocytes over other peripheral cell types, due to more concentrated CB₁Rs than in other peripheral sites. Alternatively, the differing effects of Δ-9-THC on fat deposition may be due to variable expression of CB₁ and CB₂ receptors in the peripheral ECS. While CBR expression has been shown in the islets of Langerhans, evidence is conflicting on the receptor subtypes in each of the different islet cells [19]. Previous in vivo experiments on rodents demonstrated metabolic derangement with reduced glucose uptake following administration of CB₁ agonists, but the opposite with CB₂ agonists [30]. Given the lack of research conducted on CB₂Rs, both in their relative distribution and functions, we can only speculate their function in both metabolic dysfunction and ectopic fat deposition. Greater understanding of the interplay between the two types of cannabinoid receptors is needed to understand how cannabis differentially affects various organs and systems in human studies. Given that the relative expressions of CB₁ to CB₂ receptors in the peripheral tissues are currently unknown, variability in the level of expression in the liver compared with the pancreas or skeletal muscle may explain the increased IHFD compared with other fat phenotypes.

The present study has several limitations. First, the questionnaire asked about cannabis use did not specify the type of cannabis product or the mode of administration. Second, the relative amounts of Δ-9-THC versus non-Δ-9-THC cannabinoids in the used cannabis preparations are unknown. However, our question referred to “marijuana”, which is most frequently referred in the literature as recreational cannabis (with much greater Δ-9-THC than non-Δ-9-THC) [4]. Therefore, we assumed the effects seen were more the consequence of Δ-9-THC and not the CBD components (which are frequently used for therapeutic purposes). Third, there is a risk of under-reporting when asking about substance use. This would have resulted in our study yielding more conservative findings. Fourth, this was a post-hoc analysis of data on participants recruited into a study on sequelae of pancreatitis. Although all the 120 individuals had a history of mild pancreatitis, no individual had signs of acute inflammation on their MR scans and all had complete clinical resolution of their pancreatitis prior to MR acquisition [31, 32]. This also provided a relatively homogenous group of individuals who had a detailed clinical phenotype [33, 34]. They were also potentially a population with greater pain concerns due to their history of pancreatitis and subsequent impaired quality of life [35]. Hence, they could be considered an at-risk group for cannabis use or dependence, or potentially could be considered a group that may benefit from the therapeutic use of cannabis if delivered in controlled conditions [36]. Fifth, some individuals with diabetes received insulin, which may have influenced abdominal body fat composition. However, the effect is unlikely to be material as none of the participants in the regular cannabis use group and only three participants in the other two groups received insulin. Sixth, the number of regular cannabis users was rather limited. This study is considered pilot and hypothesis-generating. Power calculations of future studies in the field will benefit from the data presented here. Seventh, the analyses did not adjust for BMI or waist circumference. Whilst it is acknowledged that these parameters might have influenced the studied associations, it is also acknowledged that these are less accurate than MR-derived visceral and subcutaneous fat compartments (used as dependent variable in the present study) for metabolic health. Adjusting the studied associations for BMI or waist circumference would have introduced multicollinearity into the regression analyses and therefore undermined the statistical assumptions [13]. Last, given the cross-sectional nature of the study, inference of causality between cannabis use and fat phenotypes or insulin traits cannot be made. However, to the best of our knowledge, no longitudinal study has been conducted in humans on cannabis use. Future studies are now warranted to track temporal changes in fat phenotypes with regular cannabis use, both recreationally and therapeutically.

In conclusion, our study presents first data on the differential association between cannabis use and MR-derived fat phenotypes, including SFV, VFV, IPFD, IHFD and SMFD. Our results show significantly greater IHFD (but not SFV, VFV, IPFD or SMFD) in regular cannabis users compared with never cannabis users, both in the unadjusted and adjusted analyses. Cannabis use contributes most to variance in IHFD, and least to IPFD. Our results show significantly altered insulin secretion (but not insulin sensitivity) in regular cannabis users compared with never cannabis users. These findings provide novel insights into the relationship between cannabis use, different abdominal fat phenotypes, and insulin traits, the knowledge on which will likely increase in the future.