In our study of cancer patients admitted with severe sepsis, the CUD(+) cohort was more likely to be younger, male, black, and Medicaid enrollees. They had lower rates of cardiovascular comorbidities but higher rates of chronic pulmonary disease, substance use, and depression. They had lower odds of all-cause mortality and respiratory failure, but higher median cost of hospital stay compared to the CUD(-) cohort. In line with our findings, several authors have also previously examined the higher rates of cannabis use and dependence among young adults and blacks[4,23-25]. Furthermore, we report higher rates of CUD in hospitals based in the western United States. This is reflective of data from the Substance Abuse and Mental Health Services Administration’s 2021 National Survey on Drug Use and Health, where the Western states of Washington, Oregon, Nevada, and California reported the highest rates of marijuana use[26].

Our study revealed higher rates of comorbid alcohol use (13.1% vs 2.9%) and mood disorders (16.2% vs 9.8%) amongst the CUD(+) arm. Though concurrent alcohol and CUD are understudied, some authors have noted a significant two-way association between major depressive disorder and concurrent alcohol and cannabis use[27]. One primary care-based electronic health record study reported significantly higher odds for other substance use and most mental health conditions, including social anxiety, bipolar disorder, and depression, among those with CUD or cannabis use[28]. We, too, find that patients with CUD were more likely to have higher rates of concomitant depression. Animal studies have explored the neuromodulatory effects of cannabis on gamma-aminobutyric acid and glutamate neurotransmission, though this is still of unclear clinical significance[29,30].

A nationally representative 2018 survey of medical oncologists reported that nearly half (46%) recommend medical marijuana to their patients[31]. From a patient perspective, active users most often cited cannabis use for physical and neuropsychiatric symptoms, including pain, poor sleep or appetite, nausea, low mood and stress[32]. Interestingly, one study of cancer patients undergoing treatment noted that those using cannabis also tended to report more severe symptoms, though whether this is linked to cannabis use is unclear due to the cross-sectional design[25]. A multivariable analysis of different cancers found a statistically significant association between patients with gastrointestinal cancers and cannabis use[33]. A minority of patients discuss cannabis use with their healthcare providers or have medical authorization for its usage, with rates of around 25% and 30% respectively[33,34].

A cross-sectional study of 905 participants in Australia found comparable rates of CUD (32%) among those who used cannabis medically vs those who reported illicit use, with withdrawal and tolerance symptoms being the most common manifestations[35]. Another study found that around 80% of medical users also report recreational use and are more prone to daily usage, highlighting the overlap between medical and recreational cannabis use[36]. The significance of these study findings are underscored by data that suggest important gaps in medical literacy among regular cannabis users on its health effects, with a tendency to underestimate risks and overestimate benefits[37]. In fact, evidence-based studies regarding the potential risks and benefits of medical cannabis have been heterogeneous and inconsistent[38-40]. In part, this is due to its federal status as a Class I narcotic, limiting research funding[41].

Cannabinoids (CBs), the most well-researched chemical compound of cannabis, exert their action in the human body via the endocannabinoid system and are classified pharmacologically into the intoxicating tetrahydrocannabinol, non-intoxicating cannabidiol (CBD) and several other minor, less well-studied CBs[42]. Experiments investigating the cardiovascular effects of CBD in various pathological states have exhibited a wide range of effects. For example, reductions in stress related hypertension, not causing hypotension in animal hypertension models as well as vascular and cardiac protection in models of diabetes, sepsis and MI[43]. There is also a fair amount of basic science data supporting the ability of CBs to attenuate inflammatory pathways and oxidative stress by downregulating pro-inflammatory cytokines in sepsis[44-47]. Though much of this research has focused on CBD and tetrahydrocannabinol, more recent studies have also demonstrated similar anti-inflammatory properties during the lipopolysaccharide-induced, macrophage-mediated, cytokine storm of sepsis among minor cannabinoid groups like tetrahydrocannabivarin, cannabichromene, and cannabinol[48]. Some authors have also cited its anticancer potential through a diverse range of mechanisms like inducing apoptosis, inhibiting cancer signaling, cell proliferation, and angiogenesis via CB receptors expressed on tumor cells[45,49,50]. In contrast, a mendelian randomization study reported an increased risk for squamous cell lung carcinoma in individuals with genetic predisposition towards cannabis use[51]. An important consideration here also is the pharmacological interaction between CBs and conventional cancer therapies. Some studies have noted its ability to act synergistically alongside certain chemotherapy agents like paclitaxel, for example, through its effect on transport channels, increasing intracellular drug concentrations[52,53]. A 2021, phase 1b randomized, placebo controlled trial of 21 patients with recurrent glioblastoma being treated with temozolomide demonstrated improved one year survival rates in the intervention arm, which was concurrently administered the cannabis formulation, nabiximols (83% vs 44%, P = 0.042), though this study was not adequately powered to compare outcomes[54]. However, others have highlighted the contradictory effect of CBs in reducing the efficacy of certain agents, like immunotherapy and metal based drugs, through the modulation of T cell function and production of detoxifying proteins, respectively[53,55,56].

The link between inflammation, sepsis, and its subsequent adverse consequences like shock, metabolic acidosis and end-organ dysfunction is through the systemic inflammatory response syndrome, during which there is a hypermetabolic accumulation of hydrogen peroxide[47,57]. Hence, reducing maladaptive inflammation could be an important component of future therapeutic options aiming to improve outcomes in sepsis[43,58,59]. For example, a mouse sepsis experiment demonstrated decreased systemic inflammation, as well as cardiac and renal protective effects in mice injected with CBD compared to controls[60]. There have been other observational data with concurrent findings of paradoxically improved outcomes among cannabis users. One NIS study between 2005 and 2014 found that among 6073862 COPD admissions, those with cannabis use (0.4%) had statistically significantly lower odds of in-hospital mortality and pneumonia compared to those without cannabis use (99.6%). The cannabis use cohort also had lower odds of sepsis and respiratory failure, but this did not reach statistical significance[61]. In another cross-sectional analysis of the NIS database between 2007-2011, multivariable logistic regression showed lower in-hospital mortality among cancer patients with active marijuana use vs non-users (OR = 0.44, 95%CI: 0.35-0.55)[62]. In a retrospective cohort study of 510007 vascular surgery patients, those with CUD had a lower incidence of sepsis in the perioperative period (OR = 0.64, 95%CI: 0.47-0.85), though this association was not statistically significant on sensitivity analysis[63]. In contrast, there have also been studies showing cannabis use to be associated with higher risk for some adverse outcomes. For example, a mendelian randomization study of patients with genetic liability for cannabis use identified a greater risk for small vessel strokes and atrial fibrillation on multivariate analysis[64]. Another population-level cohort study comparing individuals reporting cannabis use within one year to propensity-matched controls found that the cannabis use group had significantly higher rates of emergency room visits and hospital admissions, though there was no difference in all-cause mortality (OR = 0.99, 95%CI: 0.49-2.02)[65].

The major strength of our study is that it utilizes the largest inpatient dataset, which encompasses hospitalizations across 47 states participating in Healthcare Cost and Utilization Project, approximating around 97% of the United States population, or around 7 million unweighted and 35 million weighted admissions nationwide. This allows for an analysis of trends and outcomes across various sociodemographic factors and co-morbidities. As a result, clinicians can better understand nationwide patterns, associations and disease burden. However, our study also has some limitations. The major limitation is that, as a population level administrative dataset, the NIS collects data from admission related International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification codes and not individual patients. Additionally, through the retrospective cohort design, there is a possibility of unmeasured confounding factors, selection, and sampling bias. Hence, no causality can be inferred regarding outcomes of CUD in sepsis among cancer patients. Furthermore, data regarding prior vs current cancer, chemotherapy regimen, age related variations in cancer profile, laboratory and microbiological data and the source of sepsis could not be assessed, which may have influenced the findings in our study by introducing uncontrolled confounders. Despite these limitations, we provide contemporary results from the largest database for outcomes of severe sepsis in cancer patients with CUD.

Prospective cohort studies with standardized data collection on cancer staging, treatment history, and microbiological data can help better account for these variables and improve the robustness of the study findings. Additionally, further research is needed to continue to elucidate the underlying immunological mechanisms of how cannabis use may influence the prognosis of critically ill patients. Furthermore, our study emphasizes the need for heightened awareness of CUD among cancer patients and its potential effects on sepsis outcomes. Healthcare providers need education on this relationship to deliver informed care. Importantly, policies should promote access to treatment for CUD treatment and support integrated care models for this vulnerable population.