Cannabis Use Is Associated With a Small Increase in the Risk of Postoperative Nausea and Vomiting

A Retrospective Machine-learning Causal Analysis

Wendy Suhre; Vikas O'Reilly-Shah; Wil Van Cleve


BMC Anesthesiol. 2020;20(115) 

In This Article


This two-center retrospective cohort study identified an association between chronic cannabis use and an increased risk of postoperative nausea and vomiting. Using modern statistical techniques for estimating causal effects, we observed a mean increase in relative risk of PONV associated with daily cannabis use of 1.20, with a 95% Bayesian credible interval of 1–1.45. This estimate is supported by the fact that we calculated a nearly identical estimate in a second hospital with a different patient population and different providers. Many providers might assume that chronic cannabis use exerts some form of lasting antiemetic effect; however, our analysis indicates the potential for an increased risk of postoperative nausea among such patients. Although a dose-response effect may exist with respect to the dose of daily exposure to cannabis, this could not be tested retrospectively with the data available to us.

In our analysis, we observed that the association between cannabis use and PONV appeared to decrease with increasing Apfel score (i.e. Apfel score exerted a moderating effect in the model). One interpretation of this observation is that daily cannabis is a relatively "weak" risk factor for PONV when compared to the classical risks measured in the simplified Apfel score. While the credible intervals for relative risk begin to include 1 (no increase in risk), one advantage to Bayesian modeling is that a credible interval that includes 1 does not indicate a non-significant effect, but rather an increased probability of a non-positive (or even negative) association. We would argue that classical models for PONV may under-measure the complexity of interaction between risk factors, and advocate for statistical approaches like BART that permit a data-driven, non-parametric approach to data analysis that can reveal additional complexity.

Cannabinoids exert a well documented antiemetic effect, though it remains less clear whether they are similarly effective at preventing nausea.[20,21] The antiemetic effects of cannabinoids are thought to be mediated by activation of CB-1 receptors in the area postrema of the nucleus tractus solitarus and the "vomiting center" of the medulla.[22] The observation that cannabis acted as an antiemetic and the intractability of nausea in patients receiving chemotherapeutic agents for cancer treatment stimulated the development of synthetic cannabinoids to specifically treat CINV. Two such synthetic cannabinoids, dronabinol and nabilone, have been shown to be effective treatments for CINV.[23,24] In a more recent study, dronabinol was found to be as effective as ondansetron in reducing the incidence of nausea and vomiting in patients on highly emetogenic chemotherapy.[25] In that study, patients receiving dronabinol also reported decreased severity of their nausea and retching. Nabiximols (trade name Sativex), a whole plant extract of cannabis, available as an oro-mucusol spray in Canada and Europe, has been demonstrated to be superior to placebo in decreasing CINV.[5,26]

Though cannabis and synthetic cannabinoids are used to treat CINV, their use to treat PONV has not been established. In a recent randomized controlled trial, Kleine-Brueggeney and colleagues compared intravenous THC prior to emergence from general anesthesia to placebo, but the study was discontinued due to unacceptable side effects, including sedation and psychotropic phenomena.[9] In another trial comparing nabilone to placebo in patients at high risk of PONV receiving a standardized regiment of other antiemetics, the authors concluded that nabilone did not decrease the incidence of PONV.[8] Notably, nabilone also failed to improve pain scores, opioid consumption, or side effects. The use of cannabinoids for intractable PONV has also been described in a case report in which a young woman who underwent laparoscopic gastric bypass surgery experienced intractable postoperative nausea lasting weeks following surgery.[10] After multiple admissions and treatments, the patient was finally given dronabinol and experienced a significant improvement in her nausea within 1–2 days.

Paradoxically, cannabinoids can also elicit nausea, as seen in Cannabinoid Hyperemesis Syndrome (CHS), a condition associated with heavy daily use of cannabis. CHS has been theorized to be caused by either a buildup of toxic chemicals found in cannabis or the downregulation of CB1 receptors in chronic cannabis use.[21] Genetic differences in the P450 enzyme family responsible for cannabinoid metabolism may also play a role.[26] Withdrawal symptoms after cessation of chronic cannabis use include nausea, irritability, anxiety, sleep disturbances, restlessness, depressed mood, and physical discomforts such as abdominal pain, and typically begin within 24–48 h, with onset depending upon the type of cannabinoid used, the route of ingestion, and the frequency and amount of consumption. Withdrawal timing and severity may also have a genetic component. Kebir et al. recently described the presence of a polymorphism in a cannabinoid transporter which can significantly alter THC levels in the blood and body stores resulting in more severe withdrawal symptoms.[27] Interestingly, Schilienz et al. found cannabinoid withdrawal symptoms to be more severe in females, specifically nausea, though nausea was less common than other physical symptoms of withdrawal.[28]

Several possibilities could explain why patients chronically using cannabis in an outpatient setting demonstrated a higher risk of developing postoperative nausea and vomiting in our study. The simplest hypothesis is that patients were demonstrating symptoms of cannabis withdrawal. While cannabis withdrawal symptoms generally take several days to appear, the exposure to emetogenic stimuli (e.g. anesthetic and analgesic drugs, peritoneal stretch) combined with reduction or abstention from cannabis use in the perioperative period might unmask withdrawal symptoms earlier than they might be expected. Another possibility is that patients using cannabis choose to do so in part because of the drug's antinausea properties. In this conception of risk, cannabis itself is not emetogenic, but rather a marker for a patient at elevated risk of PONV who is chronically self-medicating.

Our study's observations are strengthened by our use of a modern statistical technique for obtaining estimates for causal inference that avoids some of the classical problems associated with matching and propensity score estimates. Further, we performed both internal and external validation analyses, a process which we believe strengthens our results. As is true of any non-randomized study of an intervention, we are limited by potential associations between our predictor (cannabis use) and outcome (PONV) that are not appropriately managed by our statistical methods. We find it unlikely that a randomized study to answer this question will ever be conducted, and therefore hope that other groups with comparable datasets will explore this question and provide additional independent analyses that would provide further confirmation or spur debate as to the reliability of our findings.