Cannabis is produced from Cannabis sativa, known as the marijuana plant, which contains more than 60 active chemical compounds called cannabinoids[3,13]. The primary effects of cannabinoids come from activation of cannabinoid receptors: The two types of G-protein coupled cannabinoid receptors, cannabinoid receptor type 1 (CB1) and cannabinoid receptor type 2 (CB2)[7,14]. CB1 is primarily expressed in the brain while CB2 in immune tissue, and under normal physiologic conditions both CB1 and CB2 are very weakly expressed by the liver[15]. However, an upregulation of these receptors are markedly induced in the human liver with cirrhosis[16,17]. The levels of CB1 are six times greater in patients with chronic hepatitis C compared to control while twice greater in patients with cirrhosis compared to those at a lower fibrosis state[18-20]. CB1 and CB2 exert opposite effects on the liver; CB1 has profibrogenic properties while activation of CB2 in mice has been associated with antifibrotic and anti-inflammatory properties[7].

The most active and best-studied cannabinoids are tetrahydrocannabinol (THC) and cannabidiol (CBD)[7,13]. Both THC and CBD act on CB-1 and CB2, but they have different affinities for the receptors. CB1 activation has been shown to have pro-inflammatory effects and to be involved in hepatic fibrosis and steatosis[3,21]. For instance, in CB1 null mice, CB1 blockers increase fatty acid oxidation and decrease hepatic inflammation and lipogenesis[22]. In contrast, CB2 exerts anti-inflammatory effects. CB2 activation has also been shown to suppress obesity and steatohepatitis and to protect the liver from ischemia-reperfusion injury[3]. THC preferentially activates the CB1 pathway, whereas CBD triggers the dominant activation of CB2[3]. The clinical impact of receptor expression and the ratio of activated CB1 and CB2 are unclear among patients with NAFLD. Given the previous evidence of opposite hepatic effects of CB1 and CB2, it is conceivable to suspect the overall cannabinoid effects from both receptors. Ascites is the only liver-related complication that was significantly different between cannabis and non-cannabis groups out of the complications tested in our study, (OR 1.25; 95%CI: 1.02-1.51). Therefore, a possible explanation for higher prevalence of ascites is the dominant hepatic effects of activated CB1 compared to CB2.

Because marijuana use remains illegal in many states in the United States, most marijuana consumed is often not produced under a controlled setting. Therefore, marijuana varies in its contents and has a risk of containing other chemical contaminants, and it also varies in the amounts of active cannabinoids[23]. Different modes of administration along with dynamic pharmacokinetic processes of cannabinoids lead to different bioavailabilities of active cannabinoids ingested. The most common administration route is smoking through cigarettes, pipes, or water pipes[24]. Smoking is the principal route of cannabis administration as it provides a rapid and efficient delivery to the brain, and the ability to titrate dose to the desired effect[25]. The different inhalation methods lead to inconsistent delivery of cannabinoids in the body, thus making control of the dose of active cannabinoids difficult[23]. Oral administration of cannabis involves a slower absorption, and therefore it has a slower, more-delayed peak of THC, one of the primary psychoactive components of cannabis[25]. Wall et al[26] reported oral THC bioavailability as 10%-20%. Therefore, many cannabis users prefer smoking due to a quick onset of effects, but smoking is generally not recommended for therapeutic use[25]. Unfortunately, the mode of cannabis use is not systematically recorded in the NIS, and our retrospective study would be unlikely to reliably capture mode data to compare different modes in the cannabis group. A large prospective study in which mode and dose of cannabis would be ascertained and monitored would be of interest to study the effects of dose and mode on the clinical outcomes.

The patients who used cannabis had higher prevalence of ascites compared to non-cannabis group (OR 1.25; 95%CI: 1.02-1.51). Possible explanations for this finding can be related to 1) hypoalbuminemia due to the toxicity of cannabis use or 2) lower body mass, which is a potential indicator for inadequate nutritional status. Hypoalbuminemia was described as a spectrum of toxic reactions from intravenous cannabis use along with fulminant gastroenteritis, toxic hepatitis, acute renal failure, electrolyte disturbances, leukocytosis, anemia, and relative thrombocytopenia from a study of 4 cases of intravenous cannabis use by Payne et al[27] in 1975. Serum albumin is the most abundant plasma protein and therefore binding to albumin is a key determinant of the drug efficacy, distribution and possible toxicity of cannabinoids[28]. The increase in plasma albumin may reduce the unbound fraction of cannabinoids, which further reduces the efficacy of the drug[28]. Along the same reasoning, it is conceivable to hypothesize more cannabinoid toxicity in patients with hypoalbuminemia due to more unbound cannabinoids. A study by Blüml et al[29] showed an inverse relationship between body mass index (BMI) and illicit drug use including cannabis use among young males, and therefore along with the toxicity of high-dose cannabis use, hypoalbuminemia from poor nutritional status may explain higher rate of ascites in cannabis users due to the low oncotic pressure from hypoalbuminemia. The forementioned explanation for hypoalbuminemia can partly explain the isolated finding of higher rates of ascites in cannabis users without higher rates of portal hypertension, varices and variceal bleeding and cirrhosis. Another explanation of such finding can be related to a lack of long-term follow-up. As the NIS is limited to hospitalized data, a spectrum of clinical presentations of decompensated cirrhosis such as varices and variceal bleeding and actual clinical diagnosis of cirrhosis outside of hospital may not be captured in our study design. Further studies with a long-term follow-up investigating cannabis use in patients with NAFLD are warranted to further evaluate the rates of clinical manifestations of decompensated cirrhosis.

The patients with cannabis group had a higher rate of ascites compared to the non-cannabis group despite higher baseline rates of metabolic risks in non-cannabis group such as diabetes, hyperlipidemia, and obesity. This suggests that cannabis may not be a magic bullet for the management of NAFLD. In cannabis group, higher comorbidity index was an independent risk factor for ascites, and this is expected as older patients with more comorbidity are associated with worse prognosis in patients with chronic liver diseases. In non-cannabis group, age and higher comorbidity were again independent risk factors for ascites. However, strikingly obesity (aOR 0.61; 95%CI: 0.46-0.81) and hyperlipidemia (aOR 0.59; 95%CI: 0.44-0.88) were independent protective factors against ascites in patients with NAFLD who did not use cannabis. A possible explanation for this finding can be related to nutritional status in patients with ascites. Metabolic derangements such as obesity and insulin resistance predispose to NAFLD[30], and therefore they are risk factors for development of NAFLD. However, at terminal stage of liver disease with evidence of decompensation such as ascites, the increased BMI and appropriate albumin levels can be protective against developing ascites. Previous studies showed that muscle wasting was worse in obese patients[31] with cirrhosis, and these patients are at high risk for fat-and water-soluble vitamin depletion[32]. Therefore, our finding of the association between obesity and reduced prevalence of ascites is surprising. However, a study by Li et al[33] showed that patients with higher BMI had lower rates of liver-related mortality compared to lower BMI among patients with cirrhosis and hepatocellular carcinoma. We were not able to measure nutritional status in our study; therefore although obesity often co-exists with malnutrition, we can speculate that our study population with an obesity diagnosis may have better nutrition than those without obesity.

Patients with cirrhosis are at higher risk for poor tolerance of fasting, and therefore aggressive energy restriction is avoided in these patients[34]. Due to the risk of sarcopenia with weight loss, patients who are advised to lose weight should be monitored for changes in body muscle mass and muscle strength[34]. Therefore, increased BMI in the setting of end-stage liver disease may suggest better nutritional status compared to non-obese patients. In the meantime, sarcopenic obesity is associated with a higher rate of mortality in patients with cirrhosis, and therefore further distinction between sarco obesity and obesity is warranted. The impairment of liver lipogenesis is prominent in decompensated NAFLD, and subsequently low levels of cholesterol in advanced NAFLD is not surprising. Therefore, in patients with decompensated NAFLD as seen with ascites, hyperlipidemia and obesity can be associated with less prevalence of ascites due to a better nutritional status in this vulnerable population with advanced liver disease.

Among cannabis users, hyperlipidemia was an independent risk factor for ascites (aOR 1.53; 95%CI: 1.08-2.17). Previous few studies showed significantly diminished level of serum high-density lipoprotein, low-density lipoprotein, and total cholesterol in liver cirrhosis[35,36]. Similarly, some studies showed a decrease in triglyceride in cirrhosis[36]. Liver biosynthesis is reduced with the progression of cirrhosis[37]. Although there were mixed findings of the relationship between cholesterol values and cirrhosis, our findings are compatible with the abovementioned relationship.

This is the national retrospective study to evaluate the clinical effects of cannabis among hospitalized patients with NAFLD at the national level. Strengths of our manuscript are a large sample size as well as the use of case-control matching analysis, where the groups are matched on age, sex, race, comorbidity. To address potential confounding factors, we also examined diabetes, hyperlipidemia, obesity as co-variates in multi-variate analysis.

Despite these strengths, our study has some limitations that are mainly associated with the nature of large population database studies, in which patients are typically not routinely tested for cannabis use upon admission, and the diagnosis of cannabis use is often made from patient reports. Therefore, unless patients are forthcoming with their caregivers regarding cannabis use, cannabis use may be missed or under-coded. The NIS also relies on accurate billing by clinicians to accurately record diseases and complications, which may lead to an underestimate of diagnosis. In addition, monitoring long-term clinical outcomes, such as liver-related complications not recorded in NIS, remains challenging.

Another limitation of this study is the inability to characterize different concentrations or modes of cannabis administration. 94.1% of the cannabis users were coded under non-dependent use, and due to the limited sample size, we could not provide dose-dependent data. A possible explanation for the low baseline percentage of dependent cannabis use is a lack of available data, limited current routine clinician assessment, or truly low number of patients who abuse cannabis. We were unable to characterize the concentrations and effects of each cannabinoid, as well as the dose-dependent effects of these cannabinoids. Long-term randomized controlled studies with different levels of cannabinoid types and amounts are warranted to better understand each cannabinoid’s effects on the cannabinoid system in the body.