Despite the established benefit of CAN in palliating symptoms, there remains a substantial concern about the potential negative effect on clinical outcomes and the possible antagonistic interaction with antineoplastic agents, particularly IO. Preclinically, Xiong et al. ³¹ showed using a mouse model that THC reduces the therapeutic effect of PD-1 blockade by suppressing T cell-mediated antitumor immunity. On the other hand, preclinical investigations conducted by Waissengrin et al. detected no detrimental effect of concomitant use of CAN and IO. ²⁶ In our study, we sought to examine the relationship between CAN use and clinical outcomes in patients receiving IO. We extended upon the previously reported studies to evaluate the effect of modest doses of prescribed CAN on cancer outcomes, which is more clinically relevant in the United States. We found a significant inverse relationship between CAN use and OS, PFS, and DCR among patients receiving IO. This impact persisted despite adjusting for multiple variables such as age, gender, race, IO agent used, tumor type, site of metastasis, and PD-L1 expression. These findings are consistent with some but not all of the reported studies exploring this matter. Bala-Sela et al. conducted a prospective observatory study on 102 patients with advanced cancers receiving IO, 34 of whom received monthly CAN doses of 20–40 g. They reported lower OS, PFS, and DCR in CAN users. ³² Another retrospective study by Taha et al. evaluated 140 patients with advanced cancers receiving nivolumab, 51 of whom had a monthly CAN dose of ⩾20 g, and observed a lower response rate (RR) in the CAN users without effect on OS or PFS. ²⁵ Another study by Waissengrin et al. examined 201 patients with metastatic NSCLC treated with pembrolizumab and consumed a median monthly CAN dose of 30 g. Authors of this study reported no significant effect of CAN on time-to-progression or OS. The median time-to-progression was 5.6 months in the CAN users and 6.1 months in the CAN non-users (p = 0.386). The median OS was numerically higher in the CAN nonusers (54.9 vs 23.6 months) but did not reach statistical significance (p = 0.08). The trend toward worse OS in the CAN users was attributed to a higher disease and symptom burden. ³³ However, the difference in outcome between the groups is larger in OS compared to time-to-progression, which may indicate a potential impact of CAN use on survival that may have been veiled in this patient population. Of note, all the above studies used exceedingly high CAN doses (⩾20 g monthly, i.e., >666 mg per day). Such high doses are rarely used therapeutically in the United States. In our study, most patients received lower CAN doses of 5–10 mg daily (⩽0.31 g monthly) and we still noticed a similar adverse impact on clinical outcomes, indicating that even lower doses of CAN induce significant immunosuppression sufficient to counteract the antineoplastic effect of IO. In addition, most of the patients in our study received a prescribed oral formulation, which mainly contains THC. In the other studies, most of the patients received CAN via inhalation but some received CAN as an oil extract, which is likely CBD based.

There are several mechanisms through which THC suppresses immune signaling. These have been broadly classified into immune cell apoptosis, inhibition of proliferation, suppression of cytokine and chemokine production, and induction of T regulatory cells. ²³ THC is known to suppress CD8+ and cytotoxic T lymphocytes and inhibits lymphocyte proliferation, maturation, and response to exogenous stimulation. ³⁴ This subsequently leads to a remarkable immunosuppression, which could interfere with the function of IO. Such immune suppression has been linked to inhibition of cannabinoid receptor 2 and involvement of the JAK/STAT pathway, which mediates inhibition of T-cell activation. ³¹ On the other hand, while CBD also suppresses T-lymphocytes and modulates the cannabinoid receptors, resulting in immunosuppression, there have been suggestions that CBD may have an immunostimulatory effect.^35,36 This biphasic response to CBD is believed to be dependent on the dose and concentration of CBD and cell culture conditions, including immune stimulant and response to the immune stimulant. CBD’s ability to increase intracellular calcium is likely responsible for the immunostimulatory effect of CBD, which includes neutrophil degranulation, mast cell activation, and chemotaxis.^{35,37bibr38-25151355241309095–39} Therefore, it is possible that the action exerted by THC is responsible for this antagonistic effect to IO seen in our patients. In the study conducted by Waissengrin et al. ³³ that showed no significant adverse effect of CAN on clinical outcomes, 49% of the patients in this study used oil extracts, possibly CBD based. On the other hand, lower percentages of patients in the studies reported by Bela-Sela et al. and Taha et al. consumed CAN in an oil form only (25% and 35%, respectively).^25,32 Similar to our study, the predominance of THC in these studies could possibly explain the noticeable adverse effect of CAN on clinical outcomes. Moreover, Taha et al. performed an identification analysis of CAN characteristics using a high-performance liquid chromatography-diode array detector to precisely categorize phytocannabinoids. Although there was no significant difference in RR to IO between THC and CBD and between high and low doses of each, there was a notable numeric trend toward better RR when THC is used in high compared to low doses (29.6% vs 10%) and when CBD is used in low compared to high doses (41.7% vs 16%). Overall, the best RR was seen in patients who received low doses of CBD. ²⁵ While these findings did not reach statistical significance, the study is likely underpowered to assess the difference between these groups due to the small sample size. Nonetheless, this numeric difference in RR to THC and CBD may partly explain the discrepancy between the reported studies.

IO agents are thought to modulate cancer-induced immunosuppression, which occurs by interrupting cell evasion of immune surveillance. The interaction between the antigen-presenting cells and T-lymphocytes results in T-cell activation, a mechanism that is co-stimulated with the expression of multiple molecules on the T-cell surface, ultimately inducing tumor rejection.^25,31,40 On the other hand, CAN induces immunosuppression by impairing the function of the tumor-specific T cell, NK cells, dendritic cells, and macrophages, enhancing T-cell apoptosis and reducing cytokine production in a dose-dependent pattern.^{31,41bibr42-25151355241309095bibr43-25151355241309095–44} These contradictory effects on the immune system could explain the negative clinical outcomes seen when these agents are used concurrently.

Racial disparity in oncology care has been present for decades and has been primarily attributed to access to care. In our study, we performed an exploratory analysis and noted a significant differential effect on CAN among different races. Black patients had no significant detriment in PFS and OS when used CAN. On the other hand, White patients had significantly worse PFS and OS when using CAN compared to White patients who did not use CAN. White patients using CAN had the worst PFS and OS among all racial subgroups (Figure S1). Although this could be attributed to the small number of Black patients in this study, genetic variability in the metabolism of CAN could be responsible for this difference. A study reported by Thethi et al. noted that African American patients have a higher prevalence of endocannabinoid receptor type 1 gene 3813A/G and fatty acid amide hydrolase 385 polymorphisms, and a lower prevalence of endocannabinoid receptor type 1 gene 4895A/G polymorphisms. The difference in genetic compositions of patients of various races could affect the metabolism of CAN, and therefore, affects how CAN interacts with IO among various racial groups. ⁴⁵

Tobacco use is strongly associated with worse lung cancer outcomes and survival. ⁴⁶ The interaction between tobacco and CAN is complex. In our study, CAN users were more likely to be tobacco users as well. This finding concurs with some of the previously published studies on this topic.^25,27 In fact, one of the previously published studies on this topic was criticized for not accounting for this important potentially confounding factor. ²⁴ In our study, tobacco use did not significantly influence the effect of CAN use on PFS and OS.

This study has several limitations. This study is a retrospective with a small sample size, which limits the power of the study. The patients included in this study, the IO agent used, and the CAN agent used are heterogeneous, which could limit the generalizability of the results and may give opposing results at times. The exact dose used and the compliance with the use of CAN could not be assured and was mainly based on prescription logs and providers’ documentation. In addition, some patients may not report or underreport recreational use of CAN (smoked, edible, and CBD formulations), which could influence the results of the study. Underreporting of CAN use could have occurred due to concerns about legal litigation and public stigma as most of the patients were treated before the legalization of CAN in Michigan in November 2018. Public stigma appears to be particularly heightened in cancer patients, resulting in altered behavior and underreporting of CAN use. ⁴⁷ While the Prescription Drug Reporting Program tracks the use of several controlled drugs, to date prescription CAN is not tracked through this system; thus, this tool is not helpful in monitoring prescribed CAN use. It is also possible that CAN users have higher symptoms and disease burden that could have confounded the outcomes. We attempted to account for several but not all disease and host-related confounding factors, such as age, diagnosis, metastatic site, and stage. However, the impact of unmeasured factors could potentially confound the results of this study. Finally, the study used RECIST criteria to assess response, which was the institutional method used at that time. iRECIST could result in a better evaluation of response in patients undergoing treatment with IO.

In summary, our study showed a negative impact of CAN use on OS, PFS, and DCR in cancer patients undergoing treatment with IO, which could be explained by the antagonistic interaction between CAN and IO. Our exploratory data suggest that the negative effect of CAN is primarily restricted to White patients, and sparing Black patients, which could possibly be explained by the genetic composition of different racial groups and its effect on the metabolism of CAN. A randomized clinical trial is needed to rigorously assess the interaction between CAN and IO and further explore the potential racial disparity in response to CAN and its interaction with IO.