Cannabis use is reported in about 2.5% of the world’s population, and about 45% of individuals in the US report lifetime cannabis use. The prevalence is gradually rising with the widespread legalization of its medical and recreational use and is associated with a simultaneous increase in cannabis use disorder (CUD) [1]. This growing prevalence underscores the importance of understanding the potential health effects of cannabis use, particularly in individuals with preexisting conditions. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) characterizes cannabis use disorder by a problematic cannabis use pattern leading to clinically significant impairment or distress. This requires meeting 2 out of 11 criteria within a 1-year period, including social impairment, risky use, impaired control over use, tolerance, and withdrawal [2, 3].

Chronic pancreatitis (CP) is progressive pancreatic inflammation leading to irreversible damage and persistent abdominal pain. CP is associated with an elevated risk of pancreatic cancer (PC), a malignancy with one of the highest mortality rates among cancers [4].

CP patients may use cannabis as an alternative or adjunct therapy for pain management. While clinical efficacy data in CP specifically remain limited, medical cannabis has been studied for chronic pain conditions more broadly [5–7]. Despite the known association between CP and PC, a paucity of data explores the relationship between CUD and PC in individuals with CP. This study aims to address this by conducting a retrospective cohort study.

We conducted a retrospective cohort study utilizing TriNetX, a research network that provides real-time access to electronic health records from participating healthcare organizations. Most of these organizations are in the United States. Since it only provides de-identified data, TriNetX is compliant with the Health Insurance Portability and Accountability Act (HIPAA) and has received a waiver of review and informed consent from the Western Institutional Review Board. We identified patients using International Classification of Diseases, Tenth Revision (ICD-10) diagnostic codes available within the TriNetX research network.

The CUD cohort had 10,864 patients and the control cohort had 42,160 patients. The CUD cohort was a younger population (46.4 ± 13.0 years vs. 53.3 ± 18.1 years, SMD: 0.045) and had a higher proportion of females (50.9% vs. 34.9%, SMD: 0.501) and African American individuals (30.6% vs. 12.1%, SMD: 0.538). On the other hand, the proportion of white (52.9% vs. 64.8%, SMD: 0.324) and Asian individuals (1.0% vs. 3.5%, SMD: 0.114) was significantly lower in the CUD cohort. Prevalence of comorbid conditions such as cholelithiasis (12.1% vs. 7.7%, SMD: 0.393), type 2 diabetes (30.7% vs. 17.8%, SMD: 0.619), hypertension (56.8% vs. 28.9%, SMD: 0.883), hyperlipidemia (26.7% vs. 15.2%, SMD: 0.478), alcohol-related disorders (53.7% vs. 8.0%, SMD: 1.153), and nicotine dependence (64.9% vs. 9.0%, SMD: 1.337) were also significantly higher in the CUD cohort. Similarly, chronic kidney disease (13.8% vs. 6.7%, SMD: 0.422) and overweight/obesity (18.3% vs. 6.3%, SMD: 0.399) were significantly more prevalent in the CUD cohort. (Table 1)

Following matching, both cohorts included 6,858 patients, with appropriate balancing of all covariates (SMD < 0.1). (Fig. 1) The mean follow-up was 736 ± 422 days for the CUD cohort and 896 ± 368 days for the control cohort. Given the non-normal distribution of follow-up duration, medians are also reported: 642 days (IQR 381–1,012) in the CUD cohort vs. 811 days (IQR 501–1,219) in the control cohort.

The CUD cohort showed significantly lower rates of PC detection (67 vs. 274; HR 0.263, 95% CI: 0.202–0.344, log-rank p-value: <0.001). (Table 2) Conversely, the risk of AP flares was higher in the CUD cohort vs. controls (2,487 vs. 2,446; HR 1.102, 95% CI: 1.043–1.166, log-rank p-value: 0.001). In sensitivity analysis adjusting for opioid use disorder, results remained consistent with the primary analysis (PC: 62 vs. 328 h 0.281, 95% CI: 0.213–0.372, p < 0.001), demonstrating robustness of our findings.

The relationship between CUD and cancer risk remains complex and not fully established. With the growing prevalence of cannabis use, understanding its health implications is increasingly critical. Cannabis use has been associated with both an increased risk of some cancers, such as particular testicular cancers, and a decreased risk of others, such as bladder cancer [8, 9]. This variation in outcomes is often influenced by the type of cannabinoid being used and the mode of consumption. For instance, smoking cannabis is believed to increase the risk of cancers in the throat and respiratory tract due to the presence of carcinogens in cannabis smoke, which resemble those found in cigarette smoke [8, 10].

The relationship between cannabis use and cancer risk remains incompletely understood. Preclinical studies demonstrate that certain cannabinoids can inhibit tumor growth, induce apoptosis, and reduce angiogenesis in pancreatic cancer cell lines and animal models. However, several important caveats limit the interpretation of these findings. Most anti-cancer studies have been conducted in vitro or in animal xenograft models rather than clinical trials [11–14]. Cannabinoid receptors (CB1 and CB2) are expressed ubiquitously throughout the body and bind endocannabinoids, phytocannabinoids (such as THC and CBD), and synthetic cannabinoids—each with potentially different biological effects [15]. Studies demonstrating synergistic effects with chemotherapy have predominantly used synthetic cannabinoids rather than the cannabis products consumed by patients with CUD [16]. The cannabis used by patients in our cohort likely differs substantially from purified compounds studied in laboratory settings in terms of cannabinoid composition, THC: CBD ratios, route of administration, and dose.

We also found a modestly increased risk of AP flares in CP in the CUD. Whereas growing evidence suggests that cannabis use may contribute to idiopathic acute pancreatitis (AP), our study is unique in demonstrating an increased risk of AP flares, specifically within the CP population [17]. This is a pertinent novelty, as CP patients are inherently at higher risk for recurrent pancreatic inflammation, with a pressing need for the identification of modifiable risk factors in this vulnerable group.

Despite these mechanistic uncertainties, our study provides novel real-world epidemiological evidence from a large clinical population, addressing a critical knowledge gap. While preclinical data cannot be directly extrapolated to clinical outcomes, our findings generate important hypotheses that warrant prospective investigation through well-designed clinical studies and mechanistic research to clarify the relationship between cannabis use and pancreatic cancer risk in patients with chronic pancreatitis.

Our study’s strengths include the large sample size, propensity score matching to minimize confounding bias, and implementation of sensitivity analysis. The differential follow-up time between cohorts warrants consideration, with the shorter follow-up time in the CUD cohort being a potential factor for lower PC case detection. However, Cox proportional hazards regression inherently accounts for differential follow-up by estimating the instantaneous hazard at any given time point among those still at risk, rather than relying on cumulative event counts [18]. However, the study is limited by its observational nature, which means that only associations, not causal relationships, can be inferred. TriNetX relies on coded diagnostic data from electronic health records, which may be subject to coding inaccuracies and incomplete capture of clinical information. Moreover, the lack of detailed information on the frequency, duration, or type of cannabis use restricts the ability to determine dose-response relationships. Cannabinoid pharmacokinetics are influenced by numerous variables, and product labeling is often inaccurate, making the quantification of exposure particularly challenging [19–21]. Future studies incorporating more granular data on healthcare utilization patterns, socioeconomic variables, and provider-level factors would be valuable in clarifying the influence of potential unmeasured confounders on our findings.

In conclusion, these exploratory findings suggest an association between CUD and reduced PC detection during available follow-up in patients with CP. CUD was also associated with a modestly increased risk of AP flares, which aligns with emerging evidence linking cannabis use to acute pancreatitis episodes. Given the limitations of our retrospective observational study, these associations require validation in prospective studies with longer, standardized follow-up and comprehensive ascertainment of cancer outcomes before clinical implications can be determined. Future research should also investigate potential mechanisms and determine whether the observed differences reflect true biological effects or detection bias.

The authors would like to acknowledge West Virginia Clinical and Translational Science Institute for providing access to the TriNetX research network.

M.H.A.M., S.M., and M.M.A. conceptualized and designed the study. M.H.A.M. and S.M. performed data extraction and analysis. M.H.A.M., M.M.A., R.M.G., S.K., M.W., and I.Q. contributed to data interpretation and drafting of the manuscript. K.H. and A.A. supervised the project and provided critical revision of the manuscript for important intellectual content. All authors contributed to the manuscript’s writing and approved the final version for submission.

Data was accessed using TrinetX Database. More information is available at https://trinetx.com/.