CBD Probably Reduces Seizures in Refractory Epilepsy, But Raises Risk of Serious Adverse Events

A systematic review and meta-analysis of seven randomized controlled trials finds moderate-certainty evidence that cannabidiol at standard doses meaningfully reduces seizure frequency in treatment-resistant epilepsy, while simultaneously increasing the risk of serious adverse events, and concludes that evidence for all other cannabinoids remains far too weak to support clinical conclusions.

Why This Matters

Treatment-resistant epilepsy affects roughly one-third of all epilepsy patients and carries substantially elevated mortality, cognitive decline, and diminished quality of life. Cannabidiol has moved rapidly from preclinical promise to regulatory approval in several jurisdictions, yet the cumulative trial evidence for both its benefits and harms has not been synthesized with rigorous certainty grading until recently. This review arrives at a moment when clinicians are fielding increasing patient demand for cannabinoid therapies while navigating incomplete safety data and a crowded landscape of poorly characterized cannabis derivatives.

Clinical Summary

Treatment-resistant epilepsy, defined as inadequate seizure control despite two or more appropriately chosen antiseizure medications, remains a major unmet clinical need. A 2025 systematic review and meta-analysis published by researchers following Cochrane Handbook and PRISMA 2020 standards pooled data from seven randomized controlled trials to evaluate the efficacy and safety of cannabidiol and other cannabinoid-based products in this population. CBD exerts anticonvulsant effects through multiple proposed mechanisms, including modulation of GPR55 signaling, TRPV1 channel desensitization, and enhancement of endocannabinoid tone, though the precise pathways responsible for clinical efficacy in humans remain incompletely delineated. The included trials enrolled patients with Dravet syndrome, Lennox-Gastaut syndrome, and other treatment-resistant epilepsy phenotypes across a range of ages.

At 20 mg/kg/day, CBD approximately doubled the proportion of patients achieving at least a 50% reduction in monthly seizure frequency compared to placebo (RR 1.92; 95% CI 1.49 to 2.46; four RCTs; n=575; moderate GRADE certainty). The 10 mg/kg/day dose showed a nearly identical effect size (RR 1.94; 95% CI 1.32 to 2.86; two RCTs; n=280; moderate certainty). However, serious adverse events were probably increased with CBD 20 mg/kg/day (RR 2.30; 95% CI 1.36 to 3.89; moderate certainty) and may be increased with 10 mg/kg/day (RR 1.62; 95% CI 0.92 to 2.84; low certainty, with the confidence interval crossing the null). For all other cannabinoid comparisons and secondary outcomes including quality of life and cognition, evidence certainty ranged from very low to low, precluding reliable conclusions. The authors emphasize that additional RCTs are needed, particularly for non-CBD cannabinoids and for patient-reported outcome measures.

Dr. Caplan’s Take

This review does what the field needed: it applies transparent GRADE methodology to the accumulated RCT data and confirms that CBD probably works for seizure reduction in treatment-resistant epilepsy, while making the safety costs equally visible. Patients and families seeking cannabinoid therapies often arrive with enormous hope, sometimes fueled by anecdotal reports that dramatically overstate what controlled data actually show. An honest conversation requires acknowledging that the benefit is real but so is the risk, and that the roughly twofold increase in serious adverse events at the higher dose is not a footnote.

In practice, when I evaluate a patient with refractory epilepsy who asks about CBD, I focus the discussion on pharmaceutical-grade cannabidiol products with established dosing, not artisanal or dispensary-sourced preparations. I counsel that the evidence supports a meaningful seizure benefit but that hepatotoxicity monitoring and drug interaction screening, especially with clobazam and valproate, are non-negotiable. For other cannabinoid products, I am direct: the evidence simply does not exist at a level that supports their use for seizure control.

Clinical Perspective

This systematic review sits near the top of the evidence hierarchy for evaluating CBD in treatment-resistant epilepsy, and its moderate-certainty findings for both benefit and harm should inform clinical conversations. The nearly identical point estimates at 10 and 20 mg/kg/day are notable: they suggest that the lower dose may achieve comparable efficacy with a potentially more favorable safety margin, though the confidence interval for serious adverse events at 10 mg/kg/day is wide and crosses the null. Clinicians should not interpret this as proof of equivalent safety; it reflects insufficient statistical power rather than demonstrated tolerability. For all non-CBD cannabinoids, the evidence remains effectively unestablished, and recommendations for their use in epilepsy cannot be supported by current data.

Pharmacological vigilance is essential when prescribing CBD alongside existing antiseizure regimens. CBD is a potent inhibitor of CYP3A4 and CYP2C19, and its co-administration with clobazam leads to elevated N-desmethylclobazam levels, which may account for a substantial proportion of both therapeutic and adverse effects observed in trials. Liver function monitoring is mandatory, particularly in patients receiving concurrent valproate. One concrete, actionable recommendation: when initiating pharmaceutical-grade CBD in a patient with treatment-resistant epilepsy, begin at 5 mg/kg/day and titrate to 10 mg/kg/day before considering 20 mg/kg/day, with serial hepatic panels and clobazam level monitoring at each dose adjustment.

Study at a Glance

Study Type

Systematic review and meta-analysis of randomized controlled trials

Population

Patients of any age with treatment-resistant epilepsy, including Dravet syndrome and Lennox-Gastaut syndrome

Intervention

Cannabidiol at 10 mg/kg/day and 20 mg/kg/day; other cannabis derivatives and synthetic analogs

Comparator

Placebo

Primary Outcomes

Proportion achieving at least 50% reduction in monthly seizure frequency; proportion experiencing at least one serious adverse event

Sample Size

Seven RCTs (specific pooled n varied by comparison: up to 583 for CBD 20 mg/kg/day safety analyses)

Journal

Not fully specified in extracted text

Year

2025

Registration

PROSPERO CRD42024508610

Funding Source

Not specified in extracted text

What Kind of Evidence Is This

This is a pre-registered systematic review and meta-analysis of randomized controlled trials, conducted according to Cochrane Handbook and PRISMA 2020 standards, with evidence certainty graded using the GRADE framework. It occupies the highest tier of the evidence hierarchy for questions of therapeutic efficacy and safety. The single most important inference constraint is the small number of eligible RCTs (seven total), which limits pooled estimate precision, heterogeneity assessment, and the feasibility of formal publication bias evaluation via funnel plots.

How This Fits With the Broader Literature

These findings are broadly consistent with the pivotal trials that supported regulatory approval of pharmaceutical-grade CBD (Epidiolex) for Dravet syndrome and Lennox-Gastaut syndrome, including the landmark RCTs by Devinsky et al. (2017) and Thiele et al. (2018). What this review adds is a formal GRADE assessment that makes the certainty of both the benefit and the harm signal explicit, rather than leaving clinicians to weigh individual trial findings in isolation. The near-identical efficacy estimates at 10 and 20 mg/kg/day echo earlier dose-ranging observations but now carry pooled statistical support. Importantly, the review also clarifies what remains unknown: no other cannabinoid preparation has generated sufficient controlled trial data to permit even low-certainty conclusions about seizure efficacy, a finding that challenges the broader narrative of “medical cannabis” as a unified therapeutic category for epilepsy.

Common Misreadings

The most likely overinterpretation is to read this review as a straightforward endorsement of CBD for epilepsy without weighing the concurrent safety signal. The roughly twofold increase in serious adverse events at 20 mg/kg/day carries the same moderate GRADE certainty as the efficacy estimate, meaning the evidence for harm is approximately as robust as the evidence for benefit. A second common misreading is to generalize these findings to all cannabis products. The review explicitly found that evidence for non-CBD cannabinoids is very low to low certainty, which means dispensary products, full-spectrum preparations, and synthetic analogs cannot be assumed to share CBD’s benefit-risk profile.

Bottom Line

Pharmaceutical-grade cannabidiol at 10 and 20 mg/kg/day probably produces clinically meaningful seizure reduction in treatment-resistant epilepsy, but this benefit comes with a probable increase in serious adverse events that demands careful monitoring. For all other cannabinoid products, controlled trial evidence is insufficient to support clinical use. This review does not change current practice so much as it sharpen the terms on which CBD should be prescribed: with clear-eyed acknowledgment of both benefit and risk, and with rigorous pharmacovigilance.

References

1. Devinsky O, Cross JH, Laux L, et al. Trial of cannabidiol for drug-resistant seizures in the Dravet syndrome. N Engl J Med. 2017;376(21):2011-2020. doi:10.1056/NEJMoa1611618
2. Thiele EA, Marsh ED, French JA, et al. Cannabidiol in patients with seizures associated with Lennox-Gastaut syndrome (GWPCARE4): a randomised, double-blind, placebo-controlled phase 3 trial. Lancet. 2018;391(10125):1085-1096. doi:10.1016/S0140-6736(18)30136-3
3. Cochrane Handbook for Systematic Reviews of Interventions, version 6.4. Cochrane Collaboration; 2023. Available at: https://training.cochrane.org/handbook
4. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. doi:10.1136/bmj.n71
5. PROSPERO registration: CRD42024508610. Available at: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=508610