The Cannabinoid System as a Potential Novel Target for Alcohol‐Associated Liver Disease: A Propensity‐Matched Cohort Study
Department of Medicine Virginia Commonwealth University Richmond Virginia USA
Division of Gastroenterology, Hepatology, and Nutrition, Department of Internal Medicine Virginia Commonwealth University Richmond Virsginia USA
Division of Gastroenterology, Department of Medicine, Schulich School of Medicine Western University & London Health Sciences Centre London Ontario Canada
Department of Medicine St. Mary's Hospital Waterbury Connecticut USA
Departamento de Gastroenterología, Escuela de Medicina Pontificia Universidad Católica de Chile Santiago Chile
MASLD Research Center, Division of Gastroenterology and Hepatology University of California San Diego San Diego California USA
* Correspondence:Juan Pablo Arab (juanpablo.arab@vcuhealth.org)
ABSTRACT
Background
Alcohol‐associated liver disease (ALD) is a leading cause of liver‐related morbidity and mortality, yet effective therapeutic options remain limited. Preclinical data suggest that modulation of the hepatic endocannabinoid system, particularly via cannabidiol (CBD), may reduce alcohol‐induced liver injury. Due to CBD's limited clinical use, we sought to evaluate the association between cannabis use and ALD risk among patients with alcohol use disorder (AUD).
Methods
Using the TriNetX US Collaborative Network, we identified adult patients with AUD between 2010 and 2022. Three cohorts were constructed: cannabis use disorder (CUD), cannabis users without cannabis abuse or dependence (CU) and non‐cannabis users (non‐CU). Outcomes included ALD, hepatic decompensation and composite all‐cause mortality over 3 years. Incidence and hazard ratios were calculated using Kaplan–Meier analysis and Cox regression.
Results
After matching, 33 114 patients were included in each of the CUD and non‐CU groups. Compared to non‐CU, CUD was associated with a lower risk of ALD (HR 0.60, 95% CI 0.53–0.67; p < 0.001), hepatic decompensation (HR 0.83, 95% CI 0.73–0.95; p =0.005) and all‐cause mortality (HR 0.86, 95% CI 0.80–0.94; p < 0.001) among individuals with AUD. Although CU was associated with lower risks of ALD, its risks of hepatic decompensation and all‐cause mortality were similar to those of the non‐CU cohort with AUD.
Conclusion
In this propensity‐matched cohort study of patients with AUD, cannabis use was associated with a reduced risk of ALD, with the greatest risk reduction seen in patients with CUD compared to CU and non‐CU. Our findings suggest that modulation of cannabinoid receptors may offer a new target for the development of pharmacological therapies for ALD.
Article notes
B. Fakhoury , V. Jahagirdar , K. Rama , et al., “The Cannabinoid System as a Potential Novel Target for Alcohol‐Associated Liver Disease: A Propensity‐Matched Cohort Study,” Liver International 45, no. 11 (2025): e70401, 10.1111/liv.70401.
Footnote Group
Boxed Text
- Alcohol‐associated liver disease (ALD) is a serious medical condition that has limited therapies available.
- Although cannabidiol has shown promise in experimental studies for reducing alcohol‐related liver injury, its clinical use remains limited.
- To address this gap, we evaluated whether cannabis use is associated with a reduced risk of ALD among individuals with alcohol use disorder.
- Cannabis use was linked to lower risks of ALD, liver‐related complications and death compared to non‐cannabis users.
- These findings suggest the cannabinoid system may represent a promising therapeutic target for ALD.
Untitled section
- 2‐AG
- 2‐arachidonoylglycerol
- AC
- alcohol‐associated cirrhosis
- AEA
- anandamide
- AH
- alcohol‐associated hepatitis
- ALD
- alcohol‐associated liver disease
- AS
- alcohol‐associated steatosis
- AUD
- alcohol use disorder
- CB1
- cannabinoid receptor type 1
- CB2
- cannabinoid receptor type 2
- CBD
- cannabidiol
- CU
- cannabis users
- CUD
- cannabis use disorder
- DILI
- drug‐induced liver injury
- ECS
- endocannabinoid system
- FLI
- fatty liver index
- HR
- hazard ratio
- IBD
- inflammatory bowel disease
- ICD
- International Classification of Diseases
- MASLD
- metabolic dysfunction‐associated steatotic liver disease
- MELD
- Model for End‐stage Liver Disease
- NAFLD
- non‐alcoholic fatty liver disease
- non‐CU
- non‐cannabis users
- OR
- odds ratio
- PSM
- propensity score matching
- RCT
- randomised controlled trial
- SMD
- standardised mean difference
- STROBE
- Strengthening the Reporting of Observational Studies in Epidemiology
- THC
- delta‐9‐tetrahydrocannabinol
- TNF‐α
- tumour necrosis factor alpha
1Introduction
Alcohol‐associated liver disease (ALD) is a leading cause of liver‐related morbidity and mortality worldwide, with a global prevalence of approximately 3.5%. ALD comprises a spectrum of progressive liver injury, including steatosis, steatohepatitis and cirrhosis, and is the most common aetiology of liver cirrhosis and a primary indication for liver transplantation [1, 2]. Despite its growing burden, effective therapeutic options remain limited, underscoring the need to investigate novel pathogenic pathways and therapeutic targets to mitigate liver injury.
The endocannabinoid system (ECS) consists of cannabinoid receptors type 1 (CB1) and type 2 (CB2), endogenous ligands (i.e., endocannabinoids) such as anandamide (AEA) and 2‐arachidonoylglycerol (2‐AG), and its degradative enzymes [3]. CB1 receptors are predominantly expressed in central and peripheral neurons, while CB2 receptors are mainly found in peripheral lymphoid tissues and myeloid cells [3, 4, 5]. Both AEA and 2‐AG are key endocannabinoids that modulate CB1 and CB2, influencing sensory and autonomic signalling, energy homeostasis and immune responses [4, 6, 7]. Notably, both receptors are also expressed in hepatocytes, hepatic stellate cells and Kupffer cells, with upregulated expression observed in chronic liver disease [8, 9, 10]. Preclinical studies demonstrate a divergent role for these receptors in the liver, where CB1 activation promotes steatosis and fibrogenesis through stimulation of lipogenesis and hepatic stellate cell activation, whereas CB2 signalling exerts anti‐inflammatory and antifibrotic effects by modulating immune responses and suppressing stellate cell activation. These effects have been demonstrated in models of ALD, hepatitis and hepatic ischemia–reperfusion injury [9, 11, 12, 13, 14, 15].
Exogenous plant‐based cannabinoids, derived from cannabis, include more than 100 different cannabinoids of which delta‐9‐tetrahydrocannabinol (THC) and cannabidiol (CBD) are the most studied [16]. THC is a partial agonist at both CB1 and CB2, mimicking AEA and contributing to the psychotropic effects seen in cannabis use. In contrast, CBD modulates CB1 signalling and indirectly enhances CB2‐mediated anti‐inflammatory responses [14]. Both plant‐derived and synthetic cannabinoids have been investigated for therapeutic purposes, with several FDA‐approved indications, including treatment of seizure disorders, chemotherapy‐induced nausea and vomiting, and appetite stimulation in chronic illness [17, 18].
In gastrointestinal disease, RCTs evaluating cannabinoids in inflammatory bowel disease (IBD) have shown some clinical and endoscopic benefits, although findings remain inconclusive [19, 20, 21, 22]. In the context of liver disease, preclinical studies suggest that cannabinoids, particularly CBD, may attenuate alcohol‐related liver injury through several mechanisms, including enhanced mitochondrial activity to promote lipolysis, suppression of nuclear factor kappa B (NF‐κB) activation and NOD‐, LRR‐ and pyrin domain–containing protein 3 (NLRP3) inflammasome signalling to limit steatohepatitis, reduction of oxidative stress with stimulation of autophagy to facilitate lipid clearance, and modulation of inflammatory and fibrogenic pathways via CB2 and other receptor‐mediated mechanisms such as peroxisome proliferator‐activated receptor gamma (PPARγ) [23, 24, 25, 26, 27, 28]. However, clinical evidence remains limited to a small number of retrospective studies assessing self‐reported cannabis use in patients with metabolic dysfunction‐associated steatotic liver disease (MASLD) [29, 30], chronic viral hepatitis [31, 32, 33], and only one study examining its association with ALD [34]. Given the limited clinical application of CBD, assessing its impact on ALD remains challenging in the absence of RCTs. Therefore, we sought to evaluate the association between cannabis use (which contains CBD in varying concentrations) and the development of ALD using administrative data from a large, multi‐institutional United States (US)‐based cohort of patients with alcohol use disorder (AUD). Given the heterogeneity of cannabis exposure, we stratified patients into cannabis use disorder (CUD), cannabis users without CUD (CU) and non‐users (non‐CU). This allowed us to evaluate potential dose–response patterns, as CUD may serve as a clinical proxy for heavier or more sustained exposure compared to CU.
2Methods
2.1Study Design and Data Source
This retrospective cohort study used data from TriNetX (Cambridge, MA, USA), a federated health research platform that provides real‐time access to deidentified electronic health records (EHR) from included health care organisations (HCO). The TriNetX US Collaborative Network, which aggregates patient‐level data across 72 US HCOs, was used for this analysis. The dataset includes patient demographics; diagnoses (International Classification of Diseases, Tenth Revision, Clinical Modification [ICD‐10‐CM]); procedures (ICD‐10 Procedure Coding System and Current Procedural Terminology codes); medications (Veterans Affairs Drug Classification System and RxNorm); and laboratory values (Logical Observation Identifiers Names and Codes). Data are verified and quality‐checked prior to extraction from source EHRs and integration into the TriNetX platform. Complete data from all participating sites in the US Collaborative Network were extracted on 19 August 2025.
Cohort selection, including inclusion and exclusion criteria, is shown in Figure S1. Variables used to define the cohort, covariates included in propensity score matching (PSM), and outcome definitions along with their corresponding code sets are provided in Tables S1–S4. The study was deemed exempt from review by the Virginia Commonwealth University Institutional Review Board, as only de‐identified data were used. This study was conducted per Good Clinical Practice guidelines, the Declaration of Helsinki and applicable local laws. The study also followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines (Table S5).
2.2Inclusion and Exclusion Criteria
We included adult patients (aged ≥ 18 years) who were diagnosed with AUD between 1 January 2010 and 1 January 2022. AUD was identified using International Classification of Diseases (ICD) codes for alcohol abuse (ICD‐10: F10.1) and alcohol dependence (ICD‐10: F10.2), consistent with prior literature [35]. Patients were then stratified into three groups. The Cannabis Use Disorder (CUD) group included individuals with ICD codes for cannabis abuse (ICD‐10: F12.1) and cannabis dependence (ICD‐10: F12.2), as described in prior studies [36]. The Cannabis Users (CU) group included patients with a diagnosis of cannabis use, unspecified (ICD‐10: F12.9), but without any diagnosis of cannabis abuse or dependence. The Non‐Cannabis Users (non‐CU) group included individuals without any cannabis‐related diagnosis. To exclude patients with preexisting liver disease, we applied a comprehensive list of ICD‐10 codes, including: alcoholic liver disease (K70), toxic liver disease (K71), hepatic failure (K72), chronic hepatitis (K73), fibrosis and cirrhosis of the liver (K74), other inflammatory liver diseases (K75), other specified liver diseases (K76), liver disorders secondary to systemic disease (K77), viral hepatitis (B15–B19), ascites (R18.0), spontaneous bacterial peritonitis (K65.2) and oesophageal varices (I85).
2.3Covariates and Definitions
Table S2 outlines the covariates and corresponding codes used in the PSM. Given the shared risk factors among steatotic liver disease subtypes, we selected cardiometabolic risk factors including overweight and obesity, type 2 diabetes mellitus, hypertensive disease and dyslipidemia. To minimise the impact of missing data or misclassification, each covariate was captured using both ICD‐10 diagnostic codes and corresponding laboratory indicators, including body mass index, haemoglobin A1c, systolic blood pressure, triglyceride levels and high‐density lipoprotein levels. Baseline liver function was balanced using liver enzymes including aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, Model for End‐stage Liver Disease (MELD)‐related laboratory values; serum creatinine, total bilirubin and international normalised ratio along with additional indicators of hepatic synthetic function and portal hypertension such as platelet count and serum albumin. Given the challenges in quantifying alcohol intake and drinking patterns using structured data, we included proxy measures for risk‐taking behaviour, including psychiatric comorbidities (major depressive disorder, anxiety, bipolar disorder and schizophrenia) and substance use disorders (nicotine, cocaine, opioids and other stimulants). Lastly, we adjusted for potential initiation of pharmacologic treatment for alcohol use disorder before the index AUD diagnosis, including naltrexone, acamprosate, disulfiram, gabapentin, topiramate and baclofen.
2.4Outcomes and Definitions
Outcomes and corresponding ICD‐10 codes are detailed in Table S3. The primary outcome was the development of alcohol‐associated steatosis (AS) defined by K70.0 (alcoholic fatty liver) and K76.0 (Fatty [change of] liver, not elsewhere classified), alcohol‐associated hepatitis (AH) defined by K70.1 and K70.4 (alcoholic hepatitis or alcoholic hepatic failure), and alcohol‐associated cirrhosis (AC) defined by K70.30, K70.31, K74.60 and K74.69 (alcoholic cirrhosis with or without ascites, unspecified cirrhosis of liver or other cirrhosis of liver). The composite diagnosis of alcohol‐associated liver disease (ALD) included the following codes: K70.1 (alcoholic hepatitis), K70.2 (alcoholic fibrosis and sclerosis), K70.3 (alcoholic cirrhosis of the liver), K70.4 (alcoholic hepatic failure), K74.60 (unspecified cirrhosis of the liver) and K74.69 (other cirrhosis of the liver). Alcohol associated steatosis was excluded from the composite ALD definition to mitigate misclassification as previously described in the literature [37].
Secondary outcomes included hepatic decompensation, defined by the presence of ascites, spontaneous bacterial peritonitis, variceal bleeding or hepatic encephalopathy, using a combination of ICD‐10 codes previously validated in the literature [38]. Finally, composite all‐cause mortality was also assessed.
The index date was defined as the first recorded diagnosis of AUD, and patients were followed for up to 3 years from the index date. We conducted a multilevel comparison of outcomes, first comparing patients with CUD to those without CU, then comparing CU to non‐CU and finally comparing CUD to CU.
2.5Sensitivity Analysis
To assess the robustness of our findings, we conducted several complementary sensitivity analyses, each involving a single modification of the analytic approach, and the outcomes were compared between the CUD and non‐CU groups.
- Metabolic restriction: To address the residual metabolic imbalance after PSM among the CUD versus Non‐CU groups, we restricted the cohort to patients with BMI < 30 kg/m2 and HbA1c < 6%.
- Gamma‐glutamyl transferase (GGT)‐adjusted analysis: Given the lack of direct alcohol intake assessment, we used GGT as a surrogate marker of alcohol consumption. We restricted the analysis to patients with a recorded GGT value prior to the index date of AUD diagnosis. The GGT values were subsequently incorporated as a covariate in PSM.
- Additional sensitivity analyses: (a) subgroup analysis among patients aged ≥ 50 years to account for the predominance of younger adults in the original population; (b) applied narrower ICD‐10 definitions for each outcome to increase specificity and limit misclassification (Table S3); (c) excluded patients with comorbid substance use disorders (nicotine, opioids, cocaine, stimulants and sedatives) to reduce the potential confounders (Table S5); (d) extended follow‐up to 10 years to assess long‐term outcomes; (e) performed positive and negative control outcome analyses; schizophrenia (ICD‐10 code F20) was used as a positive control outcome, given its established association with cannabis use, whereas stroke (ICD‐10 codes I60–I69) was selected as a negative control outcome, given the absence of a known association.
2.6Statistical Analysis
Categorical variables were summarised as frequencies and percentages and compared using Pearson's Chi‐squared test, while continuous variables were presented as means with standard deviations and compared using independent‐sample t tests. PSM was performed using logistic regression based on the previously described covariates. One‐to‐one greedy nearest‐neighbour matching was conducted using a calliper width of 0.1. Standardised mean differences (SMDs) < 0.1 were considered indicative of adequate covariate balance.
Time‐to‐event outcomes were evaluated using Kaplan–Meier estimates and Cox proportional hazards models to calculate hazard ratios (HRs) with 95% CIs. Proportional hazards assumptions were tested using Schoenfeld residuals. The log‐rank test was used to compare survival curves. Incidence rates per 1000 person‐years were calculated by dividing the number of new events by the estimated total person‐years at risk. The total person‐years at risk were estimated by multiplying the total number of patients in each group by the mean follow‐up time provided by the TriNetX platform for that group. All tests were two‐sided, with statistical significance set at p < 0.05. Analyses were conducted on 10 May 2025, using the TriNetX platform and Stata 18 (StataCorp LLC College Station, TX).
3Results
3.1Cannabis Use Disorder Versus Non‐Cannabis Users
Table 1 outlines the baseline characteristics after PSM in AUD patients with CUD, CU and non‐CU cohorts. Prior to matching, patients with CUD were younger (mean age 35.8 vs. 45.3 years), more likely to be Black (31.0% vs. 16.0%), and had higher rates of psychiatric comorbidities and other substance use disorders compared to non‐CU. After PSM, baseline characteristics were well balanced, with SMDs less than 0.1 across all key variables except for BMI (SMD = 0.15).
| Variables a | CUD vs. non‐CU after PSM | CU vs. non‐CU after PSM | CUD vs. CU after PSM | ||||||
|---|---|---|---|---|---|---|---|---|---|
| CUD | Non‐CU | SMD | CU | Non‐CU | SMD | CUD | CU | SMD | |
| N = 33 114 | N = 33 114 | N = 11 339 | N = 11 339 | N = 11 292 | N = 11 292 | ||||
| Demographics | |||||||||
| Age at index, years | 36 ± 13.2 | 36.5 ± 12.7 | 0.04 | 38.5 ± 14.5 | 39.5 ± 14 | 0.07 | 38.4 ± 14.2 | 38.4 ± 14.5 | 0.002 |
| Sex, female, n (%) | 10 835 (32.7%) | 11 380 (34.4%) | 0.03 | 3926 (34.6%) | 4104 (36.2%) | 0.03 | 3965 (35.1%) | 3908 (34.6%) | 0.01 |
| Race, n (%) | |||||||||
| White | 18 517 (55.9%) | 18 586 (56.1%) | 0.004 | 6517 (57.5%) | 6641 (58.6%) | 0.02 | 6482 (57.4%) | 6472 (57.3%) | 0.002 |
| Black | 10 158 (30.7%) | 10 491 (31.7%) | 0.02 | 3344 (29.5%) | 3310 (29.2%) | 0.006 | 3312 (29.3%) | 3343 (29.6%) | < 0.001 |
| Hispanic | 3067 (9.3%) | 3004 (9.1%) | 0.006 | 900 (7.9%) | 822 (7.2%) | 0.03 | 886 (7.8%) | 901 (8.0%) | 0.005 |
| Asian | 393 (1.2%) | 363 (1.1%) | 0.009 | 107 (0.9%) | 128 (1.1%) | 0.02 | 123 (1.1%) | 107 (0.9%) | 0.01 |
| Other race | 1144 (3.5%) | 1072 (3.2%) | 0.01 | 387 (3.4%) | 321 (2.8%) | 0.03 | 383 (3.4%) | 387 (3.4%) | 0.002 |
| Cardiometabolic risk factors | |||||||||
| Overweight and obesity, n (%) | 3894 (11.8%) | 4195 (12.7%) | 0.03 | 1504 (13.3%) | 1563 (13.8%) | 0.02 | 1452 (12.9%) | 1501 (13.3%) | 0.01 |
| BMI, kg/m2 | 26.6 ± 6.68 | 27.7 ± 6.96 | 0.15 | 26.5 ± 6.89 | 27.5 ± 6.93 | 0.14 | 26.2 ± 6.66 | 26.5 ± 6.9 | 0.04 |
| BMI > 25 kg/m2, n (%) | 14 683 (44.3%) | 15 649 (47.3%) | 0.06 | 5005 (44.1%) | 5347 (47.2%) | 0.06 | 4935 (43.7%) | 4995 (44.2%) | 0.01 |
| Type 2 diabetes mellitus, n (%) | 2896 (8.7%) | 3047 (9.2%) | 0.02 | 1157 (10.2%) | 1245 (11.0%) | 0.03 | 1133 (10.0%) | 1148 (10.2%) | 0.004 |
| HbA1c, % | 6.02 ± 1.88 | 6.14 ± 2.02 | 0.06 | 6.06 ± 2.06 | 6.27 ± 2.13 | 0.09 | 6.07 ± 1.86 | 6.06 ± 2.06 | 0.004 |
| HbA1c ≥ 5.7%, n (%) | 3234 (9.8%) | 3440 (10.4%) | 0.02 | 1281 (11.3%) | 1377 (12.1%) | 0.03 | 1245 (11.0%) | 1276 (11.3%) | 0.009 |
| Hypertensive diseases, n (%) | 8167 (24.7%) | 8995 (27.2%) | 0.06 | 3426 (30.2%) | 3740 (33.0%) | 0.06 | 3365 (29.8%) | 3388 (30.0%) | 0.004 |
| Blood pressure, systolic, mmHg | 124 ± 18.2 | 126 ± 18.6 | 0.08 | 126 ± 19.2 | 127 ± 18.7 | 0.09 | 125 ± 18.9 | 125 ± 19.1 | 0.02 |
| Blood pressure, systolic ≥ 130 mmHg, n (%) | 20 986 (63.4%) | 22 500 (67.9%) | 0.09 | 8011 (70.7%) | 8398 (74.1%) | 0.08 | 7992 (70.8%) | 7964 (70.5%) | 0.005 |
| Dyslipidemia, n (%) | 3576 (10.8%) | 3954 (11.9%) | 0.04 | 1473 (13.0%) | 1677 (14.8%) | 0.05 | 1451 (12.9%) | 1453 (12.9%) | < 0.001 |
| Triglycerides, mg/dL | 131 ± 111 | 140 ± 149 | 0.06 | 128 ± 94.5 | 138 ± 136 | 0.08 | 126 ± 96.2 | 128 ± 94.6 | 0.02 |
| Triglycerides ≥ 150 mg/dL, n (%) | 3835 (11.6%) | 4098 (12.4%) | 0.03 | 1196 (10.5%) | 1327 (11.7%) | 0.04 | 1199 (10.6%) | 1194 (10.6%) | 0.001 |
| HDL, mg/dL | 48.8 ± 18.6 | 49.4 ± 20.3 | 0.03 | 50.3 ± 19.9 | 51.7 ± 21 | 0.06 | 51.4 ± 20.3 | 50.2 ± 19.9 | 0.06 |
| HDL ≤ 50 mg/dL, n (%) | 6871 (20.8%) | 7117 (21.5%) | 0.02 | 2104 (18.6%) | 2182 (19.2%) | 0.02 | 2063 (18.3%) | 2105 (18.6%) | 0.009 |
| Liver function and laboratory values | |||||||||
| AST, U/L | 31.6 ± 89.5 | 33.4 ± 73.8 | 0.02 | 32.2 ± 49.6 | 35 ± 61.7 | 0.05 | 32.9 ± 117 | 32.2 ± 49.7 | 0.008 |
| ALT, U/L | 31.1 ± 73.6 | 34 ± 92 | 0.04 | 31 ± 54.7 | 33.9 ± 57.6 | 0.05 | 31 ± 67.3 | 31 ± 54.8 | 0.001 |
| ALP, U/L | 80.3 ± 39.7 | 81.1 ± 41.5 | 0.02 | 81.4 ± 45.3 | 82.2 ± 41.3 | 0.02 | 80.5 ± 36 | 81.2 ± 42.6 | 0.02 |
| Total bilirubin, mg/dL | 0.585 ± 0.471 | 0.572 ± 0.555 | 0.03 | 0.594 ± 0.48 | 0.57 ± 0.42 | 0.05 | 0.59 ± 0.519 | 0.59 ± 0.48 | 0.007 |
| sCr, mg/dL | 0.924 ± 1.1 | 0.904 ± 1.3 | 0.02 | 0.939 ± 0.721 | 0.897 ± 0.668 | 0.06 | 0.92 ± 0.647 | 0.939 ± 0.721 | 0.03 |
| INR | 1.1 ± 0.877 | 1.09 ± 0.357 | 0.02 | 1.1 ± 0.355 | 1.09 ± 0.336 | 0.04 | 1.09 ± 0.298 | 1.1 ± 0.356 | 0.04 |
| PLT, ×109/L | 257 ± 80.7 | 258 ± 80.4 | 0.02 | 261 ± 84.4 | 258 ± 85.1 | 0.04 | 256 ± 82.8 | 261 ± 84.3 | 0.07 |
| Albumin, g/dL | 4.12 ± 0.572 | 4.1 ± 0.569 | 0.04 | 4.11 ± 0.59 | 4.1 ± 0.568 | 0.03 | 4.11 ± 0.587 | 4.11 ± 0.589 | 0.007 |
| Psychiatric and substance use comorbidities | |||||||||
| Major depressive disorder, n (%) | 5247 (15.8%) | 5395 (16.3%) | 0.01 | 1449 (12.8%) | 1458 (12.9%) | 0.003 | 1459 (12.9%) | 1451 (12.9%) | 0.002 |
| Depressive episode, n (%) | 13 346 (40.3%) | 14 227 (43.0%) | 0.05 | 4179 (36.9%) | 4277 (37.7%) | 0.02 | 4170 (36.9%) | 4175 (37.0%) | < 0.001 |
| Other anxiety disorders, n (%) | 12 727 (38.4%) | 13 694 (41.4%) | 0.06 | 4539 (40.0%) | 4627 (40.8%) | 0.02 | 4535 (40.2%) | 4517 (40.0%) | 0.003 |
| Bipolar disorder, n (%) | 7152 (21.6%) | 7239 (21.9%) | 0.006 | 1929 (17.0%) | 1798 (15.9%) | 0.03 | 1949 (17.3%) | 1930 (17.1%) | 0.005 |
| Schizophrenia, n (%) | 3479 (10.5%) | 3167 (9.6%) | 0.03 | 911 (8.0%) | 770 (6.8%) | 0.05 | 930 (8.2%) | 914 (8.1%) | 0.005 |
| Nicotine dependence, n (%) | 19 677 (59.4%) | 20 655 (62.4%) | 0.06 | 7022 (62.0%) | 7351 (64.8%) | 0.06 | 7033 (62.3%) | 6984 (61.8%) | 0.009 |
| Cocaine related disorders, n (%) | 7413 (22.4%) | 6726 (20.3%) | 0.05 | 1408 (12.4%) | 1280 (11.3%) | 0.03 | 1472 (13.0%) | 1412 (12.5%) | 0.02 |
| Opioid related disorders, n (%) | 5460 (16.5%) | 5169 (15.6%) | 0.02 | 1203 (10.6%) | 1173 (10.3%) | 0.008 | 1245 (11.0%) | 1204 (10.7%) | 0.01 |
| Other stimulant related disorders, n (%) | 4410 (13.3%) | 3773 (11.4%) | 0.06 | 1233 (10.9%) | 1045 (9.2%) | 0.06 | 1277 (11.3%) | 1237 (11.0%) | 0.01 |
| Sedative, hypnotic or anxiolytic‐related disorders, n (%) | 2205 (6.7%) | 1674 (5.1%) | 0.07 | 322 (2.8%) | 280 (2.5%) | 0.02 | 304 (2.7%) | 323 (2.9%) | 0.01 |
| Medications for alcohol use disorder and adjunctive therapy | |||||||||
| Naltrexone, n (%) | 556 (1.7%) | 530 (1.6%) | 0.006 | 165 (1.455%) | 152 (1.341%) | 0.009 | 155 (1.4%) | 164 (1.5%) | 0.007 |
| Disulfiram, n (%) | 206 (0.62%) | 176 (0.53%) | 0.01 | 20 (0.176%) | 22 (0.194%) | 0.004 | 21 (0.2%) | 20 (0.2%) | 0.002 |
| Acamprosate, n (%) | 160 (0.5%) | 162 (0.5%) | < 0.001 | 19 (0.2%) | 26 (0.2%) | 0.01 | 13 (0.1%) | 19 (0.2%) | 0.01 |
| Gabapentin, n (%) | 4751 (14.3%) | 4947 (14.9%) | 0.02 | 1876 (16.5%) | 1926 (17.0%) | 0.01 | 1866 (16.5%) | 1863 (16.5%) | < 0.001 |
| Topiramate, n (%) | 1002 (3.0%) | 1032 (3.1%) | 0.005 | 320 (2.8%) | 313 (2.8%) | 0.004 | 302 (2.7%) | 320 (2.8%) | 0.01 |
| Baclofen, n (%) | 809 (2.4%) | 865 (2.6%) | 0.01 | 338 (3.0%) | 334 (2.9%) | 0.002 | 316 (2.8%) | 336 (3.0%) | 0.01 |
Comparing CUD to non‐CU, CUD was associated with a significantly lower risk of AS (HR 0.71, 95% CI 0.5–0.78, p < 0.001), AH (HR 0.52, 95% CI 0.44–0.61, p < 0.001), AC (HR 0.63, 95% CI 0.54–0.73, p < 0.001) and composite ALD (HR 0.60, 95% CI 0.53–0.67, p < 0.001). CUD was also associated with a reduced risk of hepatic decompensation (HR 0.83, 95% CI 0.73–0.95, p = 0.005) and all‐cause mortality (HR 0.86, 95% CI 0.80–0.94, p < 0.001) (Table 2, Figures 1, 2, 3 and Figures S2–S4).
| Outcome | Events | Incidence rate (1000PY) | HR (95% CI) | p | Proportionality | ||
|---|---|---|---|---|---|---|---|
| CUD | Non‐CU | CUD | Non‐CU | ||||
| N = 33 114 | N = 33 114 | ||||||
| AS | 720 | 983 | 8.87 | 12.11 | 0.71 (0.5–0.78) | < 0.001 | 0.10 |
| AH | 223 | 422 | 2.75 | 5.20 | 0.52 (0.44–0.61) | < 0.001 | 0.55 |
| AC | 263 | 410 | 3.24 | 5.05 | 0.63 (0.54–0.73) | < 0.001 | 0.49 |
| ALD | 431 | 703 | 5.31 | 8.66 | 0.60 (0.53–0.67) | < 0.001 | 0.72 |
| Hepatic decompensation | 419 | 494 | 5.16 | 6.08 | 0.83 (0.73–0.95) | 0.005 | 0.55 |
| Composite all‐cause mortality | 1065 | 1207 | 13.13 | 14.87 | 0.86 (0.80–0.94) | < 0.001 | 0.98 |
3.2Cannabis Users Versus Non‐Cannabis Users
Before matching, CU patients were younger (mean age 38.5 vs. 45.3 years), had higher rates of nicotine and other substance use, and were more likely to have psychiatric comorbidities. Post‐matching, covariates were well balanced with SMDs < 0.1 across all variables except for BMI (SMD = 0.14).
In the comparison between CU and non‐CU, CU was associated with a significantly lower risk of AS (HR 0.86, 95% CI 0.75–0.97, p = 0.04), AH (HR 0.63, 95% CI 0.49–0.80, p < 0.001) and composite ALD (HR 0.81, 95% CI 0.68–0.97, p = 0.02), but not AC (HR 0.96, 95% CI 0.77–1.2, p = 0.73), hepatic decompensation (HR 0.89, 95% CI 0.74–1.08, p = 0.24) or mortality (HR 0.94, 95% CI 0.83–1.07, p = 0.38) (Table 3).
| Outcome | Events | Incidence rate (1000PY) | HR (95% CI) | p | Proportionality | ||
|---|---|---|---|---|---|---|---|
| CU | Non‐CU | CU | Non‐CU | ||||
| N = 11 339 | N = 11 339 | ||||||
| AS | 333 | 412 | 13.26 | 15.17 | 0.86 (0.75–0.97) | 0.04 | 0.33 |
| AH | 101 | 170 | 4.02 | 6.26 | 0.63 (0.49–0.80) | < 0.001 | 0.12 |
| AC | 150 | 166 | 5.97 | 6.11 | 0.96 (0.77–1.2) | 0.73 | 0.03 |
| ALD | 219 | 285 | 8.72 | 10.50 | 0.81 (0.68–0.97) | 0.02 | 0.005 |
| Hepatic decompensation | 195 | 232 | 7.76 | 8.54 | 0.89 (0.74–1.08) | 0.24 | 0.08 |
| Composite all‐cause mortality | 435 | 494 | 17.32 | 18.18 | 0.94 (0.83–1.07) | 0.38 | 0.79 |
3.3Cannabis Use Disorder Versus Cannabis Users
Before matching, patients with CUD were slightly younger and had higher rates of stimulant, opioid and inhalant‐related disorders compared to CU. After matching, the groups were well balanced, with SMDs < 0.05 across all covariates.
Comparing CUD to CU, CUD was associated with a significantly lower risk of AH (HR 0.68, 95% CI 0.49–0.92, p = 0.01), AC (HR 0.77, 95% CI 0.60–0.97, p = 0.03) and composite ALD (HR 0.78, 95% CI 0.64–0.95, p = 0.01), with no significant difference in hepatic decompensation or all‐cause mortality (Table 4).
| Outcome | Events | Incidence rate (1000PY) | HR (95% CI) | p | Proportionality | ||
|---|---|---|---|---|---|---|---|
| CUD | CU | CUD | CU | ||||
| N = 11 292 | N = 11 292 | ||||||
| AS | 275 | 334 | 10.02 | 13.29 | 0.75 (0.64–0.88) | < 0.001 | 0.96 |
| AH | 76 | 101 | 2.83 | 4.02 | 0.68 (0.49–0.92) | 0.01 | 0.22 |
| AC | 123 | 148 | 4.48 | 5.89 | 0.77 (0.60–0.97) | 0.03 | 0.14 |
| ALD | 183 | 217 | 6.67 | 8.64 | 0.78 (0.64–0.95) | 0.01 | 0.09 |
| Hepatic decompensation | 198 | 193 | 7.22 | 7.68 | 0.95 (0.78–1.16) | 0.63 | 0.22 |
| Composite all‐cause mortality | 440 | 429 | 16.03 | 17.06 | 0.95 (0.83–1.08) | 0.43 | 0.24 |
3.4Sensitivity Analysis
Several sensitivity analyses, outlined in Table 5 and Tables S6–S11, confirmed the robustness of these findings.
- Metabolic restriction: After excluding patients with HbA1c > 6% or BMI > 30 kg/m2, 4662 patients were matched per group with excellent covariate balance (all SMDs < 0.1). In this restricted cohort, CUD was associated with significantly lower risks of AS (HR 0.63, 95% CI 0.51–0.78, p < 0.001), AH (HR 0.53, 95% CI 0.38–0.73, p < 0.001), AC (HR 0.56, 95% CI 0.40–0.78, p < 0.001) and composite ALD (HR 0.55, 95% CI 0.42–0.71, p < 0.001), with no differences in hepatic decompensation or all‐cause mortality (Table 5).
- GGT‐adjusted analysis: Among patients with available baseline GGT values, 902 matched per group with adequate balance among all covariates, including GGT (mean 76.1 ± 111 in CUD vs. 80.6 ± 113 in non‐CU; SMD = 0.08). CUD was associated with lower risks of AS (HR 0.53, 95% CI 0.38–0.73, p < 0.001), AH (HR 0.60, 95% CI 0.38–0.95, p = 0.03) and composite ALD (HR 0.69, 95% CI 0.49–0.96, p = 0.03), with no significant differences in AC, hepatic decompensation or all‐cause mortality (Table S6).
- Additional sensitivity analyses:
- Restricting the analysis to patients aged ≥ 50 years, CUD remained significantly associated with lower risk of ALD (HR 0.56, 95% CI 0.48–0.67, p < 0.001), hepatic decompensation (HR 0.72, 95% CI 0.59–0.88, p < 0.001) and all‐cause mortality (HR 0.84, 95% CI 0.75–0.94, p = 0.02) (Table S7).
- Using more specific ICD‐10 definitions, CUD remained significantly associated with a lower risk of the composite ALD (HR 0.61, 95% CI 0.53–0.70, p < 0.001), hepatic decompensation (HR 0.81, 95% CI 0.71–0.93, p < 0.001) and all‐cause mortality (HR 0.84, 95% CI 0.77–0.91, p < 0.001) (Table S8).
- After excluding patients with other substance use disorders, CUD was still associated with substantially lower risks of composite ALD (HR 0.48, 95% CI 0.37–0.61, p < 0.001), hepatic decompensation (HR 0.61, 95% CI 0.46–0.78, p < 0.001); though no significant difference was observed in all‐cause mortality (HR 0.95, 95% CI 0.78–1.16, p = 0.62) (Table S9).
- Extending the follow‐up to 10 years demonstrated sustained associations, with HRs of 0.71 for composite ALD (95% CI 0.65–0.78, p < 0.001), 0.83 for hepatic decompensation (95% CI 0.75–0.91, p < 0.001) and 0.92 for all‐cause mortality (95% CI 0.87–0.98, p = 0.005) (Table S10).
- Positive and negative control outcome analyses demonstrated a higher risk of schizophrenia among the CUD group (HR 1.23, 95% CI 1.17–1.30, p < 0.001), while no association was observed with cerebrovascular disease (HR 0.98, 95% CI 0.90–1.14, p = 0.36) (Table S11).
| Outcome | Events | Incidence rate (1000PY) | HR (95% CI) | p | Proportionality | ||
|---|---|---|---|---|---|---|---|
| CUD | Non‐CUD | CUD | Non‐CUD | ||||
| N = 4662 | N = 4662 | ||||||
| AS | 140 | 218 | 11.10 | 17.35 | 0.63 (0.51–0.78) | < 0.001 | 0.64 |
| AH | 54 | 101 | 4.28 | 8.05 | 0.53 (0.38–0.73) | < 0.001 | 0.10 |
| AC | 55 | 97 | 4.36 | 7.73 | 0.56 (0.40–0.78) | < 0.001 | 0.66 |
| ALD | 93 | 167 | 7.37 | 13.30 | 0.55 (0.42–0.71) | < 0.001 | 0.20 |
| Hepatic decompensation | 86 | 111 | 6.82 | 8.84 | 0.76 (0.58–1.01) | 0.06 | 0.58 |
| Composite all‐cause mortality | 165 | 192 | 13.07 | 15.28 | 0.85 (0.69–1.05) | 0.12 | 0.66 |
4Discussion
In this large, multi‐institutional cohort of individuals with AUD, cannabis use was associated with a 40% hazard reduction in the composite ALD, including alcohol‐associated steatosis, hepatitis, fibrosis and cirrhosis, as well as a 17% reduction in hepatic decompensation, and a 14% reduction in all‐cause mortality. The risk reduction was observed across the ALD stages with a gradient of effect between CU and CUD. This pattern may suggest a dose–response relationship, though its interpretation remains uncertain, as it relies on diagnostic codes with no direct measures of cannabis consumption. Furthermore, while our findings suggest a potential protective association between cannabis use and ALD, this must be interpreted with caution, given the well‐established health risks of cannabis, including psychiatric disorders such as schizophrenia and cognitive impairment [39]. Although our study evaluated cannabis use, which contains more than 100 distinct cannabinoids, preclinical evidence increasingly supports the hepatoprotective potential of CBD, a non‐psychoactive component, particularly in the context of ALD.
4.1The Interaction Between Cannabis and Alcohol Consumption
To explore potential mechanisms underlying the observed reduction in ALD, it is important to consider patterns of alcohol consumption among individuals with AUD who use cannabis. With the expanding legalisation of cannabis, its use is no longer limited to individuals engaging in high‐risk behaviours, potentially shifting longstanding assumptions regarding its interaction with alcohol consumption. From a behavioural perspective, the relationship between cannabis and alcohol use in this population is complex and incompletely understood. Several studies have explored this interaction with mixed findings. For instance, a secondary analysis of a multi‐center RCT (the COMBINE study) examined cannabis use in 1383 participants receiving outpatient AUD treatment, comparing cannabis users to non‐users. At 16 weeks, any cannabis use reduced percent days abstinent by 4.35%, with no effects on drinks per drinking day or heavy drinking days. No significant differences were observed at 1‐year follow‐up, and no consistent dose–response relationship was identified across cannabis use quartiles [40]. Furthermore, a prospective nationwide study of US adults evaluated the risk of persistent AUD among cannabis users compared to nonusers after 3 years of follow‐up. Among those with baseline AUD, cannabis use was associated with a 32% increased risk of persistent AUD (aOR = 1.32, 95% CI: 1.17–1.50) [41]. In contrast, a secondary analysis of a RCT in Colorado examined alcohol consumption patterns among 96 heavy‐drinking cannabis users and found that alcohol consumption was 29% lower and the likelihood of binge drinking was reduced by 50% on cannabis‐use days compared to non‐use days [42].
4.3Clinical Evidence for Cannabinoids in Liver Disease
Clinical evidence on the role of cannabinoids in liver disease remains extremely limited. A multicenter pilot RCT conducted in the UK evaluated the effects of CBD at 100 mg twice daily, both alone and in combination with other cannabinoids, among patients with type 2 diabetes. A secondary endpoint was liver triglyceride content, assessed by magnetic resonance spectroscopy over a 13‐week follow‐up period, with no significant changes observed [50]. Several factors may explain these findings. First, the study population did not have preexisting liver disease and thus may have lacked the upregulation of cannabinoid receptors observed in chronic liver injury. Second, the CBD dose (200 mg/day) was relatively lower than doses used in other clinical trials, including those for seizure disorders and IBD. Third, the study was limited by a small sample size, relatively short follow‐up duration and absence of histological assessment [51]. A Spanish prospective longitudinal cohort study evaluated the association between cannabis use and liver steatosis among 159 patients with the first episode of non‐affective psychosis. After a 3‐year follow‐up, cannabis users had significantly lower fatty liver index (FLI) scores and were less likely to meet criteria for liver steatosis, though no significant differences were observed in fibrosis scores based on FIB‐4 scores and NAFLD fibrosis scores [30]. A population‐based cross‐sectional study using the HCUP‐NIS database evaluated the association between cannabis use and MASLD among nearly 6 million hospitalised adults. Both non‐dependent and dependent cannabis use were significantly associated with reduced prevalence of MASLD (aOR 0.84, 0.45, respectively) [29]. The same group conducted a similar study design evaluating the association between cannabis use and ALD among 319 514 patients with a diagnosis of AUD. After adjusting for demographics and comorbidities, the study found that cannabis use was significantly associated with a reduced prevalence of all stages of ALD, including AS, AH, AC and HCC (aOR 0.55, 0.57, 0.45 and 0.62, respectively) in a dose–response manner [34].
In the context of chronic viral hepatitis, observational studies in chronic hepatitis C virus (HCV) have yielded conflicting results. In a retrospective cohort of 204 patients with paired liver biopsies, daily cannabis use (N = 28) was independently associated with increased fibrosis progression (aOR 6.78; 95% CI 1.89–24.3; p = 0.003) [52]. However, another study using a similar design reported no significant association with fibrosis severity [53]. Conversely, in cohorts of patients with HIV/HCV co‐infection or HIV/cured HCV, regular or daily cannabis use was associated with a significantly lower risk of hepatic steatosis, as estimated by the FLI (aOR 0.45; 95% CI 0.22–0.94; p = 0.03) [32], a finding replicated in an independent study [33]. Taken together, these mixed findings suggest that the impact of cannabinoids may be disease specific, particularly in the setting of active viral hepatitis. It is biologically plausible that the suppressed immune response imposed by cannabinoids impairs host responses to infections. This is supported by preclinical studies demonstrating cannabinoid‐mediated immune suppression in infections such as Legionella pneumophila , Herpes simplex virus‐2 and Listeria monocytogenes [54, 55, 56].
4.4Risks and Considerations
While the use of CBD in liver disease appears promising, it is not without potential risks. Liver enzyme elevation has been reported in patients receiving CBD in randomised controlled trials for epilepsy. A recent systematic review and meta‐analysis evaluated 28 clinical trials and found a pooled incidence of liver enzyme elevation of 7.4%, with DILI occurring in 3.0% of cases. The risk of hepatotoxicity was significantly higher in patients receiving high‐dose CBD (≥ 1000 mg/day) and those on concomitant valproate therapy, while no cases of DILI were reported at doses below 300 mg/day, and most liver enzyme abnormalities resolved with dose reduction or discontinuation [57]. However, quantifying CBD exposure in real‐world cannabis users remains challenging due to variability in product composition and usage patterns.
4.5Strengths and Limitations
This study leveraged a large, multi‐institutional US cohort to evaluate the association between cannabinoid use and the development of ALD, an area previously underexplored in clinical research despite the growing evidence demonstrated in pre‐clinical studies. The observed protective association was consistent across the ALD spectrum, even among patients with lower cardiometabolic risk profiles. Furthermore, the inclusion of both positive and negative control outcomes strengthens the internal validity of the findings.
Several limitations should be acknowledged. First, it is a retrospective observational analysis relying on administrative billing codes, which may introduce outcome misclassification. To mitigate this, we used previously validated code sets and conducted sensitivity analyses using more specific diagnostic definitions. Second, although preclinical studies support CBD as a protective agent in chronic liver disease, its limited clinical use precluded direct assessment in this study. Instead, cannabis use, which involves a mixture of cannabinoids in varying concentrations, was used as a proxy. This approach may be suboptimal and could underestimate or overestimate CBD's potential effect due to interacting mechanisms from other bioactive compounds found in cannabis. Third, there was residual imbalance in cardiometabolic risk factors between the CUD and non‐CUD cohorts (BMI SMD = 0.15). Sensitivity analyses yielded consistent findings, supporting the robustness of our results. Furthermore, comparisons between the CUD and CU cohorts demonstrated an excellent covariate balance, with a persistent reduction in ALD risk. Fourth, behavioural factors such as alcohol and cannabis consumption patterns, frequency and quantity were not directly captured and may represent unmeasured confounders. We attempted to address this by matching on proxy indicators of risk‐taking behaviour, including substance use and psychiatric comorbidities. Notably, even when quantitative alcohol intake data are available, underreporting, impaired recall and variability in consumption over time remain major limitations inherent to all studies of ALD. The use of objective biomarkers such as phosphatidylethanol may be the most reliable approach to mitigate this bias; however, such data are rarely available in administrative databases [58, 59]. Lastly, a potential source of bias is that patients with a diagnosis of CUD may have more frequent healthcare encounters, leading to earlier recognition and management of other comorbid conditions such as AUD and ALD.
5Conclusion
In this large, multi‐institutional cohort, cannabis use was associated with a reduced risk of ALD, consistent with prior observations by Adejumo et al. Based on preclinical evidence, the observed protective effect is possibly attributable to CBD. Given the expanding body of experimental data supporting the hepatoprotective properties of CBD and its favourable safety profile, further studies evaluating its impact in ALD, using appropriate dosing and treatment duration, are both justified and warranted.
Ethics Statement
This study was conducted using de‐identified data from the TriNetX US Collaborative Network and was deemed exempt from Institutional Review Board (IRB) review by the Virginia Commonwealth University.
Consent
The authors have nothing to report.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Data Availability Statement
The data supporting the findings of this study were obtained from the TriNetX US Collaborative Network. The de‐identified, aggregated data and analytical methods are available upon request from the corresponding author.