Cardiovascular risk associated with the use of cannabis and cannabinoid

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Previous studies reported on potential cannabis-related cardiovascular outcomes. However, knowledge gaps remained on the magnitude of the associated risk for the people who use cannabis, particularly in the actual context of profound changes in use prevalence and characteristics of users.

WHAT THIS STUDY ADDS

• This systematic review and meta-analysis of real-world data outlines positive associations between cannabis use and major adverse cardiovascular events, with measured risk ratios of 1.20 (95% CI 1.13 to 1.26) for stroke, 1.29 (1.05 to 1.59) for acute coronary syndrome and 2.10 (1.29 to 3.42) for cardiovascular mortality.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The findings outlined by this meta-analysis should enhance the general awareness of the potential of cannabis to cause cardiovascular harm. They call for the systematic investigation of cannabis use in all patients presenting with clinical pictures of serious cardiovascular disorders.

Study design

The study protocol has been registered in PROSPERO under registration number CRD42023401401 (online supplemental appendix e1). We conducted a systematic review and meta-analysis of the scientific evidence made available between 1 January 2016 and 31 January 2023 in accordance with the international methodological recommendations (Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement from the Enhancing the QUAlity and Transparency Of Health Research Network guidelines) (online supplemental appendix e2).

Eligibility criteria were defined as follows based on the PICO approach:16

Population—Subjects from the general population likely to be exposed to cannabis or cannabinoids without restrictions on socio-demographic characteristics.

Intervention—Exposure to cannabis or cannabinoids.

Comparison—Subjects non-exposed to cannabis or cannabinoids.

Outcome—Occurrence of MACE, defined by the three-point composite outcome: cardiovascular death, non-fatal MI or non-fatal stroke.17 The primary outcome was to evaluate the risk of MACE associated with cannabis and cannabinoids based on real-life pharmacoepidemiological data. Four different subgroups of MACE were considered: acute coronary syndrome (ACS), stroke, composite outcome of ACS or stroke and cardiovascular mortality.

Study selection, data extraction and quality assessment

A search was performed on 20 September 2023 within PubMed, Web of Science and Scopus, following the search strategy detailed in online supplemental appendix e3.

Studies were independently selected by two investigators (WS and EJ) after title and abstract screening based on the defined inclusion and exclusion criteria. Each reviewer was blinded to the decision of the other, and any disagreement was resolved by discussion or by a third researcher (ML-M) in case of persisting disagreement. A cardiologist (ME) reviewed all included studies and clinically assessed the measured outcomes.

Studies were included when they contained original data with available risk estimates (relative risk RR, OR or HR). Case reports, systematic reviews, meta-analyses, in-vitro studies, animal studies, commentaries and editorials were excluded. Studies conducted specifically on subpopulations of disease patients (such as HIV positive cohorts) were excluded. Only studies written in the English language were included.

Data extraction aimed at collecting details on study design, duration, baseline characteristics, exposure levels and administration route when available, inclusion/exclusion criteria, sample size and effect size (HR, OR or RR) and the corresponding 95% CI, and any potential confounding factors including concomitant use of psychoactive substances. Two investigators (WS and EJ) were involved in carrying out data extraction and cross-checking the data. The online software Rayyan was used to manage the study selection process at all steps.18

The quality of each study was assessed using the ROBINS-E (risk of bias in non-randomised studies—of exposures) assessment tool, which provides a comprehensive and structured approach to assessing the risk of bias of non-randomised studies of exposure.19 The latter risk is approached systematically in seven distinct domains, that is, (1) bias due to confounding (D1), (2) bias arising from measurement of the exposure (D2), (3) bias in selection of participants into the study (or into the analysis) (D3), (4) bias due to post-exposure interventions (D4), (5) bias due to missing data (D5), (6) bias arising from measurement of the outcome (D6) and (7) bias in selection of the reported result (D7). This assessment was carried out independently by two researchers (WS and EJ).

Statistical analysis

A descriptive analysis was performed to characterise the included studies. Participants’ characteristics comprised the mean age, the proportion of males and the concomitant use of psychoactive substances. When these measures were unavailable in the original article, they were recalculated. When age was provided as a range, a weighted mean was calculated using the median for each age class.

The studies providing results in several outcomes were considered for each of the concerned outcomes in the quantitative analysis. Due to the rarity of the studied outcomes, all risk estimate measures were treated as equivalent for the meta-analysis. Pooled risk estimates were calculated using a DerSimonian and Laird random effects model with inverse variance weighting.20 Adjusted estimate measures were incorporated into the statistical model. The variance was calculated using the corresponding CI for each estimate measure. A subgroup analysis was performed according to the different outcome categories. Besides, sensitivity analyses were performed to evaluate the robustness of findings by excluding studies with a high risk of bias. The presence of between-study heterogeneity was assessed using the Cochran Q test and I² test. Heterogeneity was considered not important for I² values between 0% and 40%, moderate between 30% and 60%, substantial between 50% and 90% and considerable between 75% and 100%.21 Publication bias was assessed both visually through the examination of a funnel plot and statistically by applying Egger’s statistical test.22 Statistical analyses were conducted in R statistical software, V.4.3.0 (R Project for Statistical Computing).

Literature search and inclusion process

The database query yielded 3012 records. After screening titles and abstracts, we identified 119 articles, 110 of which were deemed eligible for inclusion (figure 1). Ultimately, our systematic review included a total of 24 articles following the inclusion procedure.

Characteristics of included studies

Table 1 depicts the included studies. They were of cross-sectional (n=17, 70.8%), cohort (n=6, 25.0%) and case-control (n=1, 4.2%) study designs. All assessed the potential association between recreational use of cannabis and MACE. 14 studies were based on the exploration of three databases: the National Inpatient Sample (NIS) (n=8 studies), the Behavioural Risk Factor Surveillance System (BRFSS) (n=4) and the National Health and Nutrition Examination Survey (NHANES) (n=2), with potential overlaps. The 24 studies involved a total of 432 245 972 patients.

The mean age across studies ranged from 19 to 59 years, as reported in 16 studies (table 1). The corresponding weighted average age was 38.4. Mean age calculation was not possible, unavailable or limited to cases for the remaining eight studies. Similarly, the calculation of male proportion was not possible or unavailable in five studies. Cannabis users were predominantly males (54–100%) in studies for which this data was available (n=14) and tended to be younger than non-users.

Risk of bias within studies

The risk of bias assessment retrieved a high rating for most studies (n=20, 83.3%), while the remaining four (16.7%) raised some concerns (figure 2). The most frequent causes of overall risk of bias were uncontrolled confounding factors (ROBINS-E risk of bias domain 1) and misclassification of exposure (domain 2). The risk of bias due to postexposure interventions (ROBINS-E Domain 4) was irrelevant to the selected studies, leading to a systematically low risk of bias in this domain.

Risk of bias across studies

No asymmetry was evidenced after visual inspection of the funnel plot (figure 3) and the Egger’s test was not significant (p=0.0829) for funnel plot asymmetry, suggesting the absence of systematic bias towards the reporting or publication of studies with more favourable results.

Cannabis use and risk of cardiovascular diseases

Included studies evaluated stroke (n=14),23–36 ACS (n=7),26 37–42 cardiovascular mortality (n=3)36 43 44 and the composite endpoint of ACS and stroke (n=2).45 46 Two studies measured different outcomes, and their results were considered in each of the concerned outcomes.26 36

The measured ORs of ACS, stroke and cardiovascular mortality associated with cannabis use were 1.29 (95% CI 1.05 to 1.59), 1.20 (1.13 to 1.26) and 2.10 (1.29 to 3.42), respectively, and that of the composite ACS/stroke outcome, 1.04 (0.54 to 1.99) (figure 4). Heterogeneity between studies was substantial to considerable with I² values ranging from 79.2% to 89.3% across subgroups except cardiovascular mortality. The sensitivity analysis restricted to cohort studies yielded comparable results to the primary model (RR=1.32, 1.01 to 1.73) (figure 5). The limited number of studies did not allow us to perform other sensitivity analyses.

Main findings

24 studies evaluated the occurrence of MACE in the context of exposure to cannabis, including one to medical cannabis and none to other cannabinoids. The quantitative analysis suggests a positive association between cannabis use and MACE. Findings from the sensitivity analysis restricted to cohort studies were consistent with the primary analysis. These results cohere with other studies published outside of the time window of the present meta-analysis, including those from various cohorts in France or in the USA, respectively, showing an independent association between cannabis and in-hospital MACE,47 48 or between daily cannabis use and MI, stroke and the composite of coronary heart disease, MI and stroke.49 The only study on medical cannabis among those included in the meta-analysis also highlighted such a positive association.45

Cerebrovascular disorders

The studies centred on the assessment of stroke provide divergent results, whether suggesting or not a significant association between cannabis use and stroke. First, no association was found between cannabis use in young adulthood and early stroke (HR: 1.59, 0.59 to 4.28) in a study among a cohort of 50 000 men included during the compulsory military service in Sweden.35 In a case-control study among US adults younger than 50, the odds for stroke were found to be similar in subjects ever exposed to cannabis than in those never exposed.34 In both studies, estimation of exposure is likely biased since use of cannabis was measured at inclusion with no follow-up data in the first study; and the inclusion of single use over the lifetime in the second. No association was found in another cohort of US subjects included between the ages of 18 and 30 and followed up more than 25 years in the Coronary Artery Risk Development in Young Adults cohort (OR=0.57, 0.17 to 1.93).36 Age when cerebrovascular accidents occurred is not provided. Interestingly, no significant association was emphasised in the focused analysis on recent cannabis use, possibly due to a lack of power. Similar conclusions were provided from two additional studies among adults over 18 despite an overall adjustment on relevant covariables.31–33 In contrast, several studies within large cohorts found a higher risk of stroke in cannabis users, persistently significant after adjustment on relevant cardiovascular risk factors.23–25 28 Among those based on the exploration of the NIS, the largest database of US inpatients, the study by Vin-Raviv et al included all hospitalised patients (OR=1.60, 1.44 to 1.77), whereas in those by Kalla et al, Desai et al, Parekh et al and Rumalla et al, only patients under the age of 55 were included. In the latter study, ischaemic stroke was significantly associated with use of cannabis, with a marked increase in the 25–34 age range.28 The exploration of the PATH through life study cohort in Australia outlined that elevated stroke/transient ischaemic attack was specific to participants who used cannabis at least weekly (IRR=4.7, 2.1–10.7). Similarly, a higher proportion of stroke was emphasised only among subjects aged 18–74 who used cannabis frequently.

Cardiac disorders

Seven studies investigated the potential implication of cannabis in the occurrence of ACS, including five focused on acute MI, which demonstrated an independent association with the use of cannabis after adjustment for tobacco smoking and abuse of cocaine and amphetamine.37–41 In the study by Desai et al exploring the NIS database by millions of participants, the measured association was barely significant (OR=1.03, 1.02 to 1.05, p<0.001), raising the question of the clinical significance of statistically significant results.41 Three of the other four studies were also conducted on large electronic health databases,38–40 including that by Patel et al, which specifically explored this association in a younger population aged 15–22 (OR=1.36, 1.16 to 1.59). Similar results were found in a study examining the BRFSS: higher odds of acute MI were observed in patients who used cannabis more than once a week (OR=2.31, 1.18 to 4.50) but not in less frequent users (OR=1.48, 0.52 to 4.21).38 Surprisingly, the third study, which explored the French administrative hospital discharge database, concluded that among illicit drugs, cannabis was a predictor for MI, unlike cocaine and opioids.40 The authors hypothesise that their results lacked power due to the lower prevalence of cocaine and opioids than cannabis use in France. The non-significant association between the use of cocaine and MI may also illustrate the limitation of hospital databases to accurately measure exposure to illicit drugs. The fifth study highlighted a positive association within a small cohort of 85 men younger than 40.37 Considering the low number of included patients, caution is required to interpret the high OR value (OR=13.9, 3.4 to 57.1). Finally, one study specifically investigated the association between cannabis use and ACS in nearly 15 000 patients aged 18–54 and found no significant association in the overall sample but a higher risk in the subgroup of patients aged 18–36 (OR=5.24, IC 95% 1.85 to 16.94).42

Cardiovascular mortality

Cannabis use significantly increased all-cause mortality and cardiovascular mortality in a cohort of patients diagnosed with MI before the age of 50, after adjustment for age, cardiovascular risk factors including tobacco smoking and other health conditions.44 These findings are consistent with results from studies included in our previous review in which cannabis was statistically associated with increased middle-term but not long-term mortality in subjects with a history of acute MI.50 51 These were further supported by a more recent study exploring data from the NHANES which revealed a significant association between cannabis use and death from cardiovascular causes (HR=2.29, 1.10 to 4.78).43 No significant association was found for all-cause mortality (HR=1.14, 0.81 to 1.59).

Interestingly, an analysis conducted in 2024 from the UK Biobank population emphasised a sex difference regarding cardiovascular mortality related to heavy cannabis use, with a significantly higher risk for women unlike men.52

Strengths and limitations

The main strength of our study lies in its methodology, which aligns with international recommendations. To our knowledge, this is the first meta-analysis examining the potential association of cannabis use and MACE, performed from observational data and applying the highest-quality methodological standards. This approach better reflects the real-world scenario of cannabis use and the corresponding associated risks. Our study has several limitations. First, cannabis exposure was poorly reported in the included studies, which prevented our meta-analysis from assessing it. Second, a significant portion of included studies was at moderate to high risk of bias, primarily due to a lack of information regarding missing data. Concerns were also raised about the risk of misclassification of exposure, particularly in studies from medical databases, which have a low sensitivity for non-medical drug use. Studies that relied on patient surveys faced substantial bias regarding exposure and outcome misclassification when patients assessed these data themselves. Furthermore, most included studies (n=19) were cross-sectional, a design providing a poor level of evidence unable to establish the causal link between outcome and exposure. Third, several of the included studies used the same data source, sometimes overlapping in the period, with the risk of including the same patients. Fourth, our data collection was limited to between 1 January 2016 and 31 January 2023. Therefore, our results provide a fully comprehensive report of the recent situation towards the cardiovascular health of cannabis users. It is worth noting that our research team previously conducted a review that encompassed data until 2016.9 Considering the current situation and recent trends in cannabis use, the need to specifically address these recent developments was critical.

Further research is warranted to address the methodological limitations of pharmacoepidemiological studies on cannabis-related adverse events. In particular, observational studies with an accurate measure of cannabis exposure are lacking.

Patient consent for publication

Not applicable.

Ethics approval

Not applicable.