Article Text

Download PDFPDF

Research paper
Cannabis use is both independently associated with and mediates worse psychosocial health in patients with epilepsy
  1. Sandra Wahby1,
  2. Vikram Karnik1,
  3. Anita Brobbey2,
  4. Samuel Wiebe1,
  5. Tolulope Sajobi1,2,
  6. Colin Bruce Josephson1,2
  1. 1 Clinical Neurosciences, University of Calgary Cumming School of Medicine, Calgary, Alberta, Canada
  2. 2 Community Health Sciences, University of Calgary Cumming School of Medicine, Calgary, Alberta, Canada
  1. Correspondence to Dr Colin Bruce Josephson, Clinical Neurosciences, University of Calgary Cumming School of Medicine, Calgary, AB T2N 2T9, Canada; cbjoseph{at}


Objective To examine whether cannabis use is associated with or mediates psychosocial health in people with epilepsy.

Methods Consecutive adult epilepsy patients visiting the Calgary Comprehensive Epilepsy Programme clinic were administered validated patient-reported outcome measures (PROMs) including the Neurological Disorders Depression Inventory for Epilepsy (NDDI-E), Quality of Life in Epilepsy (QOLIE-10-P), EuroQOL five dimensions five level scale (EQ-5D-5L), Global Assessment of Severity of Epilepsy Scale, Global Assessment of Disability Associated with Seizures Scale and the Treatment Satisfaction Questionnaire for Medication scale. We used multiple regression analyses to investigate associations between cannabis use and PROMs. Mediation analyses were performed to determine the degree to which cannabis modulated the associations between current or past psychiatric disorders, monthly seizure frequency, and 1-year seizure freedom on psychosocial health.

Results Of 337 consecutive patients, 71 (21%) reported cannabis use. Cannabis use was independently associated with depression (NDDI-E score≥14; OR 3.90; 95% CI 2.01 to 7.59; p<0.001), lower quality of life (β=−16.73, 95% CI − 26.26 to − 7.20; p=0.001), worse epilepsy-related disability (OR 2.23, 95% CI 1.19 to 4.17; p=0.01) and lower satisfaction with antiepileptic medication (OR 0.41, 95% CI 0.23 to 0.72; p=0.002). Cannabis use mediates 7%–12% of the effect of a psychiatric history on depression, worse quality of life and worse health valuation.

Conclusions There is a strong and independent association between cannabis use and poor psychosocial health, and it partially mediates the deleterious effect of a psychiatric history on these same outcomes. Inclusion of PROMs in future cannabis trials is warranted.

  • epilepsy
  • cannabis
  • epidemiology
  • health outcomes
  • quality of life
View Full Text

Statistics from


Epilepsy is a common neurological condition that affects individuals of all ages.1 The primary aim of epilepsy therapy is complete cessation of seizures, an elusive goal for up to 30%–35% of patients.2 The pursuit of novel antiepileptic drugs (AEDs) has recently focused on the potential benefit of cannabis-based therapies.3 An open-label interventional trial has provided preliminary evidence of possible efficacy and tolerability of cannabidiol in children with epilepsy,4 and two randomised trials have demonstrated efficacy in Dravet syndrome5 and in Lennox-Gastaut syndrome.6

However, growing evidence suggests that psychosocial factors also exert an important influence on patient-reported quality of life that is independent of seizure frequency.7 Likewise, seizure frequency alone does not account for perception of overall health in a person with epilepsy; for many patients, medication adverse effects and psychosocial health, including depression, are also major contributors.8 Therefore, seizure freedom is a necessary, but insufficient, metric for epilepsy outcomes.

Patient-reported outcome measures (PROMs) are increasingly recognised as a critical means of assessing patients’ subjective experience of disease. Their routine use is associated with greater patient satisfaction and has helped improve outcomes beyond the domain for that which each scale reports.9 Hence, routine collection of patient-reported quality of life, mental health and health valuation is desirable and feasible.10 Despite this, a remarkable paucity of data exists examining the effect of cannabis use on PROMs including self-perceived quality of life, health valuation, depression and treatment satisfaction. With an increasing prevalence of cannabis use in both the epilepsy and general population, we sought to determine whether its use is associated with improved patient-reported indices of psychosocial well-being.


Study population

All patients receiving care in the Calgary Comprehensive Epilepsy Programme (CEP) clinic are enrolled in prospective data collection. The CEP is a tertiary care centre serving a region of 1.3 million people that prospectively and systematically collects data on adults (≥18 years of age) at their first and subsequent visits to the outpatient clinics. Every patient completes standardised forms pertaining to demographics, socioeconomic factors, basic health information and validated patient-reported health questionnaires. In addition, staff epileptologists complete standardised forms pertaining to seizure and epilepsy characteristics, use of AEDs, medical and psychiatric history and results of diagnostic investigations.11 Prospective data collection began in 2007. Data used in this study were obtained consecutively from all patients during their first visit to the CEP between June 2016 and October 2017. The starting period of June 2016 corresponds both with the introduction of the question on cannabis use and routine collection of PROMs in all patients. During the span of data collection, cannabis was legal in Canada for medicinal purposes only. Patients were excluded if the epileptologist deemed them to not have epileptic seizures based on clinical, neurophysiological and neuroimaging data.

Exposure and outcome measures

Cannabis use was the exposure of interest. This was assessed through self-report in a standard questionnaire that asks patients ‘Do you currently use marijuana? Yes or No’. Patients that reported cannabis use were then asked to select the mode of administration (smoking, ingestible oil or other). For the purposes of our evaluation, outcome measures were validated, self-administered questionnaires: (1) the Neurological Disorders Depression Inventory for Epilepsy (NDDI-E)12 using a binary cut-off of ≥14 to identify depression;13 (2) the Quality of Life in Epilepsy (QOLIE-10-P) scale14 in which higher scores indicate a greater self-perceived quality of life; (3) the EuroQOL five dimensions five level scale (EQ-5D-5L)15 where higher scores indicate greater health state valuation; (4) Global Assessment of Severity of Epilepsy (GASE)16 where higher scores indicate greater self-perceived epilepsy severity; (5) Global Assessment of Disability Associated with Seizures (GAD)17 where higher scores indicate greater self-perceived epilepsy disability and 6) a single-item addressing AED treatment satisfaction derived from the Treatment Satisfaction Questionnaire for Medication (TSQM) where higher scores indicate greater treatment satisfaction.18

Clinical and socioeconomic variables

Two senior authors (SW and CBJ) identified clinical and socioeconomic variables of interest a priori of the study. These included demographic characteristics (age, sex, marital status, education level and current employment), medical history (epilepsy type [generalised versus focal], monthly seizure frequency, seizure freedom over the last year, AED use, the epilepsy-specific comorbidity risk adjustment index19 and history of a psychiatric disorder as diagnosed by a clinician) and social habits (concurrent recreational drug use [aside from cannabis] and alcohol use). Seizure frequency consisted of the sum of all seizures reported by a patient per month irrespective of seizure type. AED use was measured as the mean defined daily dose (DDD) of all currently used AEDs.

Statistical analysis

We used parametric and non-parametric statistics to summarise continuous variables and frequency distributions for categorical variables. Categorical variables (marital status and education level) were converted to dichotomous variables (‘married/common-law’ vs ‘other’ and ‘college/university’ vs ‘high-school or below’). We performed univariable comparisons between cannabis users and non-users using χ² and Fisher’s exact test for categorical variables and the Mann-Whitney U test for continuous variables of all prospectively identified indices of interest.

We then used stepwise regression with backward elimination to select the most important predictors for each of the PROMs. Each model initially contained all relevant variables including cannabis use. We also added an interaction term between cannabis use and a history of a psychiatric disorder since the two may exert a synergistic effect.20 In order to develop parsimonious models, we used an alpha for entry of p≤0.10 and an alpha for removal of p≥0.15. Binary logistic regression analysis was used to model the association between depression (NDDI-E scores≥14) and the predictors identified from stepwise regression analysis. The predictive accuracy of the model was assessed using the C-statistic. Ordinal logistic regression analyses were used to model the association between the predictors derived from each stepwise regression analyses and the GASE, GAD and AED treatment satisfaction scales. The proportional odds assumption was assessed using the Brant test.21 Multiple linear regression analyses were used to model the association between QOLIE-10-P, and EQ-5D utility scores and the predictors derived from each corresponding stepwise regression analyses.

Finally, we explored whether cannabis existed on a causal chain between specific exposure variables (monthly seizure frequency, 1 year seizure freedom and a current or past history of psychiatric disease) and non-ordinal outcomes (depression as defined by a NDDI-E score ≥14, quality of life [QOLIE-10] and health utility [EQ-5D]) using mediation analyses. Statistically significant mediation was determined using an alpha level of significance of 0.05. The degree (proportion of total effect) of the indirect effect mediated through cannabis is teased out from the total effect exerted by the exposure (eg, a psychiatric disease) on the outcome (eg, QOLIE-10) through regression modelling of the causal pathway.22

All statistical analyses were conducted using Stata V.13.023 and SPSS V.24.0.24 Statistical significance was set at p≤0.05 (2-sided).

All CEP data are collected, managed, stored and extracted using REDCap,25 an electronic data capture tool hosted by the Clinical Research Unit at the University of Calgary.


We identified 337 consecutive patients, of whom 71 (21%) reported current cannabis use, 231 (69%) reported not using cannabis and 35 (10%) chose not to answer the question (non-responders; figure 1; table 1). Of those using cannabis, 54 (76%) reported smoking it, 27 (38%) reported ingesting it as cannabis oil and 9 (13%) reported using other methods or formulations, including cannabidiol or tetrahydrocannabinol (THC) infused food products and vaporising. Compared with those that abstained, cannabis users were more likely to be male (61% vs 46%; p=0.03) and to use concurrent recreational drugs (20% vs 1%; p<0.001), alcohol (51% vs 32%; p<0.001) and cigarettes (37% vs 9%; p<0.001) and to have a current or past diagnosis of a psychiatric disorder (47% vs 27%; p=0.003). Additionally, cannabis users had significantly more seizures per month (median 0.42, IQR 0.08–4.17 vs 0.08, IQR 0.02–1.08; p=0.02), but seizure freedom over the last year was not statistically different between users and non-users (21% vs 30%; p=0.21). Cannabis users were less likely to be in a married or common-law relationship (27% vs 42%; p=0.03) or to have achieved a postsecondary education (37% vs 53%; p=0.02).

Figure 1

Flowchart of our prospective open cohort study of adults presenting to the Calgary Comprehensive Epilepsy Programme Clinic. PROMs, patient-reported outcome measures.

Table 1

Demographic and clinical characteristics of the cohort

Cannabis exposure and self-reported health outcomes

Cannabis was a significant predictor of four of the six self-reported health outcomes of interest (figure 2). After adjusting for other important variables, cannabis exposure was associated with higher odds of depression (OR 3.90, 95% CI 2.01 to 7.59; p<0.001; figure 3), lower quality of life (β=−16.73, 95% CI −26.26 to −7.20; p=0.001; table 2), higher perceived epilepsy disability (OR 2.23, 95% CI 1.19 to 4.17; p=0.01; figure 4) and lower satisfaction with AEDs (OR 0.41, 95% CI 0.23 to 0.72; p=0.002; figure 5). However, cannabis use was not significantly associated with self-perceived epilepsy severity (OR 1.27, 95% CI 0.71 to 2.26; p=0.42 (online supplementary table S1) or health state valuation (failed to meet inclusion criteria in stepwise regression analysis (online supplementary table S2).

Supplementary data

Figure 2

ORs for the association between cannabis use and self-reported health outcomes. Multiple linear regression results of cannabis associations with quality of life and health state valuation cannot be incorporated in the figure. ORs>1 represent worse outcomes, except for AED treatment satisfaction, where OR<1 represents worse outcome. The dashed line is at the null value (OR=1). AED, antiepileptic drug.

Figure 3

Factors independently associated with depression according to NDDI-E scores in multiple logistic regression analyses. The vertical line is at the null value (OR=1). *P≤0.05. NDDI-E, Neurological Disorders Depression Inventory for Epilepsy.

Figure 4

Factors independently associated with perceived epilepsy disability according to GAD scores in multiple logistic regression analyses. the vertical line is at the null value (OR=1). *P≤0.05. AED, antiepileptic drug; DDD, defined daily dose; GAD, Global assessment of disability associated with seizures.

Figure 5

Factors independently associated with AED treatment satisfaction according to TSQM scores in multiple logistic regression analyses. The vertical line is at the null value (OR=1). *P≤0.05. AED, antiepileptic drug; DDD, defined daily dose; TSQM, Treatment Satisfaction Questionnaire for Medication.

Table 2

Multiple linear regression of variables associated with QOLIE-10-P total scores

Mediation analyses

A history of a psychiatric disease or disorder was significantly associated with depression (figure 3), quality of life (table 2) and health valuation (online supplementary table S2), but there was no evidence that cannabis acted as an effect measure modifier in these relationships (p>0.05). Hence, mediation analyses were performed. Cannabis use accounted for 12% (95% CI 1% to 23%) of the total effect of a history of a psychiatric disease or disorder on self-reported depression (p=0.03), 11% (95% CI 8% to 19%) of the total effect on quality of life (p<0.01) and 7% (95% CI 4% to 17%) of the total effect on health valuation (p=0.002; figure 6). In contrast, cannabis use did not significantly mediate the effects that either monthly seizure frequency or 1-year seizure freedom exerted on self-reported depression (p>0.05), quality of life (p>0.05) and health valuation (p>0.05).

Figure 6

Graphical representation of the mediation effect of cannabis use on patient-reported Depression (NDDI-E), quality of life (QOLIE-10) and health valuation (EQ-5D utility Score). The total effect of a past or current history of a psychiatric disorder on these outcomes is decomposed into a direct effect and an indirect effect mediated through cannabis use. NDDI-E, Neurological Disorders Depression Inventory for Epilepsy; QOLIE, Quality of Life in Epilepsy.


Cannabis derivatives, in particular cannabidiol, have shown promise in epilepsy, especially in paediatric populations.3–6 It is therefore not surprising that patients are increasingly likely to request prescriptions of cannabis for various conditions, including epilepsy.26 Our analyses of over 300 people with epilepsy at first presentation to a tertiary care clinic demonstrate a significant and independent association between cannabis use and worse patient-reported health measures including depression, quality of life, epilepsy disability and AED treatment satisfaction. This association remained strong and robust even after controlling for clinically relevant demographic, clinical, psychological and socioeconomic factors such as monthly seizure frequency, polytherapy and mean DDD of AEDs. Cannabis use also exerts a significant mediation effect on the relationship between a history of a psychiatric disorder (clinician diagnosed) and self-reported depression (12%; 95% CI 1% to 23%), quality of life (11%; 95% CI 8% to 19%) and health valuation (7%, 95% CI 4% to 17%) but does not mediate the relationship between more somatic characteristics, such as monthly seizure frequency or 1-year seizure freedom, and patient-reported psychosocial health.

Overall, the effect of cannabis use on psychosocial health, as assessed in a variety of medical conditions, remains inconclusive.27 For instance, using a similar cohort design from a tertiary care centre, cannabis use in 148 patients with head and neck cancer coincided with improved health valuation on pain/discomfort and anxiety/depression subscales of the EQ-5D.28 Likewise, use of inhaled cannabis appears associated with improved quality of life measurements in patients with inflammatory bowel disease.29 Other studies, though, have uncovered links between cannabis use and lower mental health and quality of life ratings in people with anxiety disorders30 and HIV.27 Similar to patients with anxiety, our results indicate that cannabis use is associated with lower measures of health-related quality of life. The potentially deleterious cognitive and affective effects of cannabis31 may preferentially exacerbate neurological disease, such as epilepsy and anxiety, thus explaining why it mediates the relationship between comorbid psychiatric disease and worse psychosocial health while appearing to have no influence on the somatic manifestations of epilepsy (seizure frequency and seizure freedom) and their relationship with the same outcome measures. Ultimately, the innate disposition to psychiatric disease seems to dictate the markedly divergent response to cannabis exposure on psychosocial health in patients with epilepsy compared with those suffering from conditions such as non-neurological cancer.

Interestingly, cannabis use does not appear to interact with a prior history of a psychiatric disorder when predicting worse psychosocial outcomes but rather exists on the causal pathway partially mediating the effect. This finding is pertinent for patients with epilepsy given the high comorbid risk of depression, anxiety and psychosis.32 Hence, our data suggest that patients with epilepsy may have a greater susceptibility for cannabis-related negative psychosocial outcomes. These data are critical to the ongoing evaluation of the use of cannabinoids in epilepsy since, to date, we have accrued information that is insufficient and of too low quality to make any definitive statements about its role on patient satisfaction and psychosocial health.33 Therefore, validated PROMs, such as those used in this study, should form a central component of future randomised controlled trials (RCTs) assessing the risks and benefits of cannabinoids in epilepsy. This is especially imperative given that approximately 60% of patients with epilepsy have expressed a willingness to participate in clinical trials of cannabinoids.34

Our study benefits from rigorous, prospective data collection and the use of validated PROMs in a consecutive cohort of patients. This standardised approach to collecting data, combined with the large referral base of patients with first seizures, drug responsive and drug-resistant epilepsy in our tertiary care centre, partially mitigates the risk of selection bias. Another strength of the study is that information bias is limited through common approaches to data acquisition. All patients complete an independent questionnaire and then undergo thorough assessments by clinicians who use patient history and medical records to establish concurrent diagnoses and relevant demographic features and confirm patient reports. Additionally, it is reassuring that the associations between cannabis use and adverse psychosocial outcomes remain consistent and robust across analyses. Likewise, additional clinical and socioeconomic variables influenced patient-reported outcomes in ways that were consistent with clinical intuition, thus strengthening confidence in our results.

Although our study reveals a strong and consistent association between cannabis use and worse patient-reported health outcomes, we cannot establish a cause-effect relationship due to the cross-sectional nature of the study. Longitudinal research is necessary to examine causality as well as to investigate whether patient-reported health outcomes vary significantly according to consistency and intensity of cannabis use. Though we were able to record mode of cannabis use, we were not able to explore dose-response relationships or the impact of cannabis formulation and duration of use on psychosocial health. If the acute psychotropic effect of THC were present at the time of assessment, it might exert anti-anxiety and antidepressive properties that would be expected to improve perceptions of health and, if anything, could dilute the strength of associations that we found since the groups would be more homogenous.35 Another potential limitation is that we are unable to determine reasons for cannabis use. If patients were self-medicating, then residual confounding could result in inflated estimates since these patients were exposed to a heavier seizure burden. Despite this, it is reassuring to note that, in our cohort, cannabis users and non-users did not differ in the proportions achieving seizure freedom over the last year. Furthermore, since likely many patients were using cannabis recreationally (though a subset may have been using it for medicinal purposes), the Tetrahydrocannabinol (THC) to Cannabidiol (CBD) ratio is almost certainly high. Therefore, though we cannot make direct extrapolations to purified CBD, it is reassuring to note that the psychotropic effects of THC would be expected to have antidepressive and antianxiety properties which would result in conservative estimates of effect in this study. We also rely on self-reporting of cannabis use, though a single screening question used to identify recreational drug use, including cannabis, is highly sensitive (100%) and specific (73%) in primary care settings.36 Additionally, the accuracy of self-reported cannabis use is high, even in sensitive populations.37 Underreporting, if present, is expected to produce conservative estimates of a potentially larger effect size, since there would be a spuriously high degree of homogeneity among groups.

Ultimately, we have demonstrated a strong and independent association between cannabis use and poor patient-reported outcomes at baseline in patients referred for epilepsy care at large referral centres. Furthermore, cannabis partially mediates the effect that psychiatric disorders, but not monthly seizure frequency or seizure freedom, exert on validated patient-reported psychosocial outcomes. Although causation cannot be established, the association is consistent enough that cannabis use should be screened at all epilepsy clinic visits. Future studies will indicate whether this simply is a surrogate marker or whether direct intervention on cannabis use will affect quality of life estimates. Special attention should also be paid to psychiatric and social management of these patients admitting to cannabis use in future clinical trials. Ultimately, the association is compelling enough to advocate for uniform inclusion of patient-reported outcomes in ongoing and future RCTs evaluating the anticonvulsant effects of cannabis products such as cannabidiol.


View Abstract


  • Collaborators Calgary Comprehensive Epilepsy Programme collaborators: Yahya Agha-Khani; Paolo Federico; Alexandra Hanson; Nathalie Jette; William Murphy; Neelan Pillay; Shaily Singh.

  • Contributors CBJ and SWi conceived and designed the study. CBJ, SWi and the Calgary Comprehensive Epilepsy Programme collaborators all contributed to the acquisition of data. CBJ, SWi, SWa, VK, AB and TS all contributed to analysis and interpretation of the data. SWa, VK and CBJ wrote the initial draft of the manuscript and designed the figures. All authors revised it critically for important intellectual content. All authors gave final approval for publication of this version of the manuscript. All authors agree to be accountable for all aspects of the work and ensure that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval The study was approved by the University of Calgary’s Conjoint Health Research Ethics Board and Alberta Health Services (REB17-0369).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.