To our knowledge this paper represents one of the largest real-world evidence analyses of MC use in the pediatric population. Our results indicate low treatment persistence among children, despite the use of relatively low THC doses across most indications in contrast to prior research. Previous reports in pediatric cohorts have generally shown higher retention rates, even with longer follow-up periods (Aran et al., 2019; Barchel et al., 2024). From all the cannabis products, only Epidoilox is FDA approved for Dravet and Lennox–Gastaut syndromes (FDA Epidiolex Prescribing information).

These low persistence rates may derive from specific reasons in each indication: In epilepsy, the persistence rate is the highest, reflecting MC efficacy in this indication, as shown previously in Epidiolex studies (Devinsky et al., 2017; Thiele et al., 2019). Nevertheless, there is a relatively steep dropout after the first month of treatment. Since both Epidiolex and MC are used as add-on therapies for refractory epilepsy, lack of efficacy after 1 month of treatment is unlikely to be the primary cause for the observed drop in treatment discontinuation.

In two Epidiolex clinical trials conducted in children, a few subjects suffered significant clinical deterioration, as reflected in caregivers’ global clinical impression of change. If similar outcomes occur in real-life usage, they would account for some dropouts within the first month of treatment (Treves et al., 2021).

In pediatric cancer cases, dropouts were observed during the first few months of treatment. According to IMCA guidelines, specific indications for MC treatment in the pediatric cancer population include chemotherapy-induced nausea and vomiting and pain. They can be considered for symptoms such as depressed mood and sleep disturbances (Israeli Medical Cannabis unit, n.d.). The treatment regimen typically extends over a 3–6 month period and is not confined to the chemotherapy cycles; it also includes other antiemetics, such as ondansetron, and corticosteroids. Therefore, the length of treatment sessions alone should not explain the immediate dropout cases observed. In addition, this is the only indication in which high concentrations of THC are utilized and, therefore, may indicate severe adverse events (Simonian et al., 2020; Chhabra et al., 2023).

While several studies have described the utilization of MC therapy within the pediatric oncological population, to date, no regimen protocol has been established (Ofir et al., 2019; Andradas et al., 2021; Chapman et al., 2021). A systematic review concluded that the information in online literature on MC for the pediatric cancer population is “satisfactory” and reflects a pro-cannabis opinion (Yeung et al., 2020). However, it seems that the treatment regimen and cannabis formulation are a missing theme in the available literature. This may account for the low persistence and high expectations that are not met in reality.

In children with Tourette’s and ASD, the evidence for MC efficacy is limited. The positive indications in Tourette syndrome are focused on improvement of the patients quality of life (Barchel et al., 2024), whereas in ASD, the clinical improvement is observed in the behavioral comorbidities of aggressiveness and attention deficit, with high intersubject variability (Aran et al., 2021; Hacohen et al., 2022). Previous reports within these indications suggest that most patients are not treated with conventional medicines.

Many caregivers may have high expectations for significant clinical improvement with MC, which often is not materialized. Moreover, some cases experience early although transient in many of the adverse events of restlessness and agitation, that contribute to treatment discontinuation, as in the open-label study subgroup. To address this, healthcare providers should define clearer treatment objectives and provide enhanced comprehensive guidance on MC treatment, to better align therapy expectations and improve patients’ and caregivers’ communications.

The mean THC:CBD ratio in ASD cases increased in those who adhered to the treatment regimen, particularly in the older children. We found, as could be expected, that most MC treatment sessions start with a low THC:CBD ratio, reflecting the common clinical practice (Abuhasira et al., 2019; MacCallum et al., 2021). The mean THC:CBD ratio increases over time, although the number of patients for whom their THC:CBD ratio treatment was increased was similar to those for whom their THC:CBD ratio treatment was reduced. Some physicians believe that certain patients could potentially benefit from higher doses of THC concentrations, which may explain gradual increase in THC levels of dispensed MC over time. These patients are probably also those who experienced limited response to the treatment and likely present with severe behavioral challenges. The literature about the effects of THC on the ASD population is conflicted. Although some anecdotal evidence suggests THC’s beneficial effect on behavioral disturbances in ASD, many studies linked schizophrenia, ASD, and cannabis use, exhibiting genetic and biological mechanisms in which THC may trigger psychosis-related pathways (Bortoletto and Colizzi, 2022). The findings presented here illustrate a more complex picture, indicating that THC is not necessarily avoided and possibly even utilized in some cases.

It is reasonable to speculate that avoiding exposure to higher concentrations of THC in younger children is essential, given that this period witnesses the most significant advances in brain structure and behavior (Johnson, 2001). In addition, younger children’s responses to THC may be less predictable and tolerable compared with older children, who may exhibit a similar central nervous system response to that of adolescents and adults (Donovan et al., 2016). However, our results suggested that for many pediatric cases treatment trajectories involved reductions in monthly THC:CBD ratio and amount, as opposed to “start low, go slow,” implying the occurrence of adverse events.

When comparing the two datasets at two different time points, we observed that the persistence was much higher after 6 months of treatment in the open-label study compared to real world data. However, 6 months later, when medical follow-up were comparable, the persistence declined to similarly reduced levels in both cohorts. This difference can possibly be attributed to the medical and logistical support and recruitment bias in the open-label study and highlights the contribution of medical care and follow-up to therapy. Nonetheless, the Hawthorne effect may positively affect persistence in the open-label study group during follow-up (Fernald et al., 2012).

Assuming that the subgroup analysis represents of the broader ASD population, adverse events and ineffectiveness are the most likely causes for treatment discontinuation. Dose reductions manifested in the decreased amount of dispensed MC and decreased levels of THC:CBD ratio may suggest adverse events, which was confirmed at some level in the open-label dataset, as dose reductions were associated with psychiatric and neurological adverse events.

Based on a conservative evaluation, at least 20%–30% of ASD patients suffered from adverse events that changed their therapy projections (discontinuation or therapy adjustments). In addition, lack of treatment accessibility due to lack of supply, difficulties in license renewal, or the high cost of treatment were also reasons for discontinuation. Another possible reason for dropouts is the expectations for a profound improvement in the child’s status due to the therapy (Aran et al., 2019; Hacohen et al., 2022), and corresponding disappointments when these expectations are not met.

Although the evidence generally indicates that MC treatment has limited effectiveness in ASD, many caregivers testify that their lives have completely changed positively since they started treatment (Aran et al., 2019; Hacohen et al., 2022). This discrepancy highlights the need to better understand the precise benefits of MC among the pediatric population, including patient and family-relevant variables.

This study has several limitations. Some key data points were missing, such as clear outcomes indicating the effectiveness and safety of cannabis treatment. This limitation could explain the persistence rates and usage patterns of cannabis we observed in our findings. We tried to mitigate this issue by incorporating the data retrieved from the open-label dataset; however, it was quite limited.

Additionally, it is important to note that, as this is an observational study, mainly relying on dispensing data and not on actual administration. Cannabis oil in Israel is dispensed in standardized 10 mL bottles fitted with dropper inserts. The recommended route of administration is sublingual, with patients instructed to retain the oil under the tongue–a standard procedure also mandated by the open-label study protocol. However, due to practical challenges in maintaining sublingual administration, particularly among younger children and patients with ASD, oral ingestion mixed with fatty food at room temperature is commonly employed as an alternative delivery method. The specific method of administration utilized by individual patients was not systematically documented in the present datasets, which represents a limitation of this analysis, as variations in administration route may influence bioavailability and treatment adherence. Our findings can only provide insights regarding the real-world patterns of use of MC in the pediatric population, rather than claiming any suggestions on the efficacy or safety of MC therapy in children. Moreover, due to the observational design, other biases and confounders may have influenced our results.

Furthermore, since THC and CBD are the most extensively studied and documented cannabinoid agents in medical records, this study did not explore other active cannabinoid agents, which could potentially offer more options for personalized medical cannabis therapy. However, improved characterization, medical follow-up, and providing adequate information about outcomes to parents may enhance persistence without necessitating further rigorous research. Another limitation is that we did not have access to the concomitant medications the children had, so we could not assess possible interactions of these medications with the medical cannabis. These potential interactions may have contributed to the discontinuation of treatment. We also lacked long-term follow-up data, which could have revealed different patterns of treatment persistence.