Report: Efficacy and safety of medical cannabinoids in children: a systematic review and meta-analysis


Despite the increased use of medical cannabinoids, the efficacy and safety of the treatment among children remain uncertain. The objective was to study the efficacy and safety of medical cannabinoids in children. The search included studies through 11-May-2020. Selection criteria included studies evaluating efficacy and safety outcomes of medical cannabinoids (tetrahydrocannabinol, cannabidiol and other cannabis derivatives) versus control in children, independently assessed by two reviewers. Eight studies were included, all of which are randomized controlled trials. Cannabidiol is associated with 50% reduction in seizures rate (Relative Risk (RR) = 1.69, 95% CI [1.20–2.36]) and caregiver global impression of change (Median Estimated difference = (− 1), 95%CI [− 1.39–(− 0.60)]) in Dravet syndrome, compared to placebo. While cannabidiol was associated with a reduction in reported seizure events (RR = 0.59, 95% CI [0.36–0.97]), no association was found in products contained also tetrahydrocannabinol (RR = 1.35, 95% CI [0.46–4.03]). Higher dose of cannabidiol was associated with decreased appetite (RR = 2.40, 95% CI [1.39–4.15]). A qualitative assessment suggests that medical cannabinoids might be associated with adverse mental events. In conclusion, cannabidiol is associated with clinical improvement in Dravet syndrome. However, cannabidiol is also associated with decreased appetite. Adverse mental events were reported as well, however, more research should be performed to assess well this outcome.


The popularity of medical cannabinoids (MCs) treatment is growing, as more countries enable MCs treatment for various indications in adults1. This trend is also observed in the pediatric population, where MCs are authorized in children for refractory epilepsy, especially in Dravet syndrome and Lennox-Gastaut syndrome2. Dravet syndrome is a rare presentation of intractable epileptic encephalopathies associated with pleomorphic seizure activity, usually before the one year of age, while Lennox-Gastaut syndrome is characterized by multiple drug-resistant seizure with a unique EEG pattern. The infants suffer from cognitive decline, motor, and behavioral abnormalities3,4. Other presentations of epilepsy and additional indications, such as autism, and inflammatory bowel disease (IBD) have been reported recently5,6,7 The non-psychoactive substance cannabidiol (CBD) is used for these indications, whereas chemotherapy-induced nausea and vomiting (CINV) are treated with the psychoactive D-9-Tetrahydrocannabinol (THC) and its analogues8,9,10.

However, despite the increased use of MCs in children, there is limited data regarding their safety and efficacy in this population for those different indications11. Pharmacotherapy in children has other safety-benefit profiles for most drugs than those seen in adults due to differences in physical characteristics, developmental aspects, and different prognosis and disease manifestation12,13. In adults, cannabinoids-related adverse events were previously associated with physiological reactions, including hyperthermia, rhabdomyolysis, increased appetite, and hypoglycemia, mostly in exposure to THC14,15,16. Furthermore, exposure to cannabis for recreational use has been previously associated with cognitive and behavioral outcomes, namely induction of psychosis and schizophrenia, cognitive compromise in adolescents, and depression17,18,19,20. Reported symptoms of unintentional cannabis ingestion in children include tachycardia and mydriasis but are characterized mainly by neurologic abnormalities such as lethargy, ataxia, and prolonged coma21,22,23,24. Evidence suggests that in-utero exposure to cannabis may pose neurodevelopmental outcomes, such as autism and a possible association between marijuana detected in breast milk and decreased motor development in infants25,26.

A systematic review published in 2017 aimed to identify available data about the therapeutic effect of cannabinoids treatment in children and adolescents. The authors hereby found that evidence was limited and mainly was considered low quality, and concluded that additional research is needed to evaluate the cannabis risk–benefit balance27. Since 2017, several clinical trials and observational studies have examined the efficacy of MCs in the pediatric population for a few indications, most of them focused on CBD efficacy in Dravet syndrome28,29,30.

Previous meta-analyses reported that CBD in epileptic adults and children in one cohort decreased the risk for seizures compared to placebo. However, these meta-analyses also reported the elevated risk for treatment withdrawals and adverse events such as somnolence, decreased appetite, diarrhea, and increased serum aminotransferases31,32.

This study aimed to assess the safety and efficacy of MCs in children using systematic review, meta-analysis, and advanced meta-analysis methods.


Search strategy

Systematic review and meta-analysis were performed to examine the efficacy and safety of MCs treatment among children. The systematic review was conducted according to the framework guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analysis (supplementary Table S1)33. The search included published and unpublished studies through 11-May-2020. The systematic review included MEDLINE, EMBASE, and ‘’ databases according to pre-selected keywords.

The protocol included a search strategy for keywords such as “medical cannabis”, “cannabis”, “medical marijuana”, “THC”, “CBD”, “dronabinol”, “nabiximols”, “adolescents”, “child”, “efficacy”, “safety”, and “adverse reactions”. Randomized control trials (RCTs) and observational studies were included in the search with no language nor date restrictions. The study protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) and was updated as the screening progressed (CRD-42019132383) (supplementary Method). Considering the study design, no ethical review board assessment was required.

Study selection

Two independent investigators (Noa Mor and Nir Treves) screened publications with Rayyan QCRI, a web and mobile app for systematic reviews34. The publications were screened based on their relevance to the selection criteria. The selection criteria included studies that evaluated the efficacy and safety outcomes of MCs treatment compared to placebo or other pharmacotherapy for any medical purpose among the pediatric population (≤ 18 years old). The screening was focused on studies that examined the efficacy and safety outcomes of active ingredients in cannabis, such as THC, CBD, and additional cannabinoids such as Nabilone. Due to scarce data and since the reported outcomes were expected to vary, no specific efficacy or safety outcomes were defined before data collection to facilitate a comprehensive evaluation of the available studies in this area. All potentially eligible studies were considered regardless of study design.

At initial screening, the studies were assessed independently for potential inclusion by title by the two investigators. A reason for exclusion for every individual study was documented. Disagreements were resolved by discussion. Following an initial screening, the included studies were reviewed once for evaluation based on the abstract, with disagreements resolved by discussion.

Following the abstract screening, the full text of eligible publications was examined, and a final decision for inclusion was made. At this point, studies with no comparison group or when exposure was not MCs were excluded from the meta-analysis. For eligible publications with missing data, the corresponding author was contacted to access additional data. If these attempts failed, the information as available was utilized to include the data when applicable. In addition, citations in the selected articles were reviewed independently by the two investigators (Noa Mor and Nir Treves) for identifying additional eligible articles.

Data extraction

Comprehensive data extraction from each article was conducted by the two investigators independently according to an extraction form created and agreed upon in advance (variables specified in Supplementary Method). Study quality was assessed for the risk of bias using the tool Risk of Bias in Non-Randomized Studies of Interventions for observational studies (Robin-I)35 and Revised Cochrane risk-of-bias tool for randomized trials (RoB-2)36. This assessment was also conducted separately, and differences were resolved by discussion.

Since only a limited number of studies, each assessing several outcomes was expected, the precise outcomes were defined during data extraction process, for avoiding any important trends and implications. Adverse events with common pharmacological mechanisms or standard system organ class were grouped in the same analysis. The following outcomes were examined:

  • The outcomes of 50% reduction in seizure rate from baseline; Caregiver Global Impression of Change (CGIC), a subjective assessment of the caregiver’s view of the patient’s global functioning, measured in an ordinal scale37; and reported seizures events were measured as epilepsy efficacy outcomes in CBD treatment vs. placebo.
  • The mean difference of vomits and retching events in nabilone (synthetic THC) vs. control was evaluated as an efficacy outcome in CINV.
  • Number of patients who suffered from serious adverse events (SAEs), as described in the studies’ articles, (supplementary Method) or listed in the critical medical events list of the European Medical Agency38.
  • Decreased appetite was defined as a binary outcome.
  • Signs and symptoms of gastrointestinal hyperactivity were grouped, including events of diarrhea, upset stomach, nausea, retching and vomit. Due to the nature of reports in the included studies, more than one event can occur in one participant.
  • Adverse mental events. Since various adverse mental events can be interpreted differently by the patients, caregivers and the researchers, a united outcome of adverse mental events was measured. Due to the nature of reports in the included studies, more than one event can occur in one participant. This outcome included the following events: fatigue, somnolence, unresponsive to stimuli, vagueness, lightheadedness, excitability, irritability, hysteria crying, aggression, abnormal behavior, psychomotor hyperactivity, hallucinations, elevation of mood, mood changes, euphoria. The full list of adverse mental event is detailed in the supplementary Methods.
  • The outcome of infections included the infective events from pathogenic agents reported in the studies (supplementary Method).
  • Pyrexia, as reported in the included studies.

Statistical analysis

The possible impact of variations in study design and different formulations was addressed by measuring heterogeneity and utilizing random-effects models. I2 was calculated as the degree of heterogeneity observed in the analysis. The p-value < 0.1 was considered statistically significant. When heterogeneity was observed, a further investigation was performed in subgroup analyses, divided by products composite: CBD products, CBD and THC mixed products, and THC\THC-like products. As meta-regression was not applicable due to small number of studies and available dosages a network meta-analysis (NMA) was conducted to estimate the dosage impact on the measured results in a head-to-head method and heterogeneity.

Pooled data analysis was performed with R programming language with R-studio platform, using “meta”, “netmeta”, “metamedian”, “metafor”39,40,41,42. The pooled risk-ratios was calculated and 95% confidence of intervals (CI) to summarize the results for the dichotomous outcomes assessed for children treated with MCs versus children who were treated with placebo or control treatment. Sidik-Jonkman method was used, which provides conservative results and wider confidence intervals (Supplementary Method). Trial sequential analysis (TSA) was utilized to quantify the required sample size to determine the effect of outcomes while adjusting the threshold for statistical significance, with the Sidik-Jonkman model and O’brein Fleming boundaries function, using the trial sequential analysis program v.0.9-beta ( Two-sided trial sequential analysis was conducted to maintain a risk of 5% for type I error and a power of 80%, using the retrieved data from included studies and to estimate the required information size (IS). The information size is the total number of subjects and events that are necessary to detect or reject an assumed intervention effect in a meta-analysis44.

For pooling data of additional efficacy measured by an ordinal variable, the quantile estimation method was used. Analysis of results and mean difference calculation were extracted solely from studies with a crossover design and was performed by Comprehensive Meta-Analysis (CMA) software to consider the premises of this unique study design during pooled data calculations45. Outcomes evaluated by only one of the included studies are not shown in the meta-analysis. The outcome of adverse mental events was not analyzed as a pooled analysis since the number of events were higher than number of the participants in some of the comparison groups. Therefore, a summary table of adverse mental events is presented with Conditional maximum likelihood estimate of Rate Ratio and Fischer exact test46. Since less than ten studies were included, funnel plots were not used to assess publication bias47.


Out of 9133 results, ten articles met the inclusion/exclusion criteria. Three studies were excluded because of missing data, despite trying to contact the authors, and one additional study was found in the references of eligible articles (Supplementary Fig. S1). Eventually, eight remaining studies were RCTs conducted with MCs indicated for epilepsy (CBD extract), severe behavioral problems (CBD extract), CINV (nabilone) and spasticity (nabiximols, a 1:1 formulation of THC and CBD), and autism spectrum disorder (ASD) (whole-plant cannabis extract, purified CBD, and THC, both contained CBD and THC in ratio 20:1 respectively), with a total of 642 patients28,48,49,50,51,52,53,54. Table1 presents the included studies in the meta-analysis. According to quality assessment (Supplementary Fig. 2), five of the included studies posed a low risk of bias, one posed some concerns for bias, while the remaining two were judged to be at high risk of bias. Supplementary Table S2 presents a summary of findings of all meta-analyses conducted.



Primary Sponsor

Top Marijuana Blog