Running CBD and the Ketogenic Diet Together: What a 58-Patient Johns Hopkins Cohort Found

A multicenter retrospective cohort published in Epilepsy Research describes 58 children and adults who received cannabidiol and ketogenic diet therapy concurrently. Seizure outcomes resembled what is reported for either therapy used on its own, and the authors position the combination as a reasonable additive option rather than a step change.

The signal the authors flag is one of sequencing: greater seizure reduction was seen when CBD was started before the ketogenic diet rather than after. This was an observed pattern within an uncontrolled cohort, not a randomized comparison.

Roughly one in three people with epilepsy continues to have seizures despite adequate trials of two or more appropriately chosen medications. For that group, cannabidiol and ketogenic diet therapy are two of the few remaining evidence-supported options, and many families end up on both. Whether the two interfere with each other has been an open practical question rather than a theoretical one.

This cohort answers a narrower question than the headline suggests. It does not tell us that the combination outperforms either treatment alone. It tells us what happened, biochemically and clinically, in 58 real patients who were given both, which is information clinicians and families have been working without.

The paper, Concurrent use of ketogenic diet therapy and cannabidiol in children and adults with epilepsy: A multicenter retrospective cohort study, appeared in Epilepsy Research on May 5, 2026. The authors are Lauren M. Hinkley, Mackenzie C. Cervenka, Stacey K. Bessone, Zahava Turner, and Eric H. Kossoff, working at the Johns Hopkins University School of Medicine and Johns Hopkins All Children’s Hospital. Kossoff and Turner have spent years running one of the larger ketogenic diet programs in the United States, which is worth knowing when reading their description of how these patients were managed.

The team identified 58 patients treated with cannabidiol and ketogenic diet therapy at the same time between 2018 and 2025. Median age at the start of combination therapy was 2 years, with a range from infancy to 60 years, so this is a heavily pediatric population. Average duration of combined treatment was 17.4 months. Forty-four patients (76 percent) received prescription CBD, meaning Epidiolex, and 14 (24 percent) received artisanal CBD products. Thirty-five (60 percent) were on the classic ketogenic diet rather than a modified variant.

One structural point matters for interpretation and is easy to lose in secondhand coverage. Every patient in this cohort received both treatments. There was no internal CBD-only group and no internal diet-only group. When the authors describe efficacy as similar to either treatment alone, they are comparing against what the published literature reports for monotherapy, not against a control arm they recruited.

At one month of combination therapy, 40 percent of patients had greater than 50 percent seizure reduction. At three months, 28 percent did. Those numbers sit in the same territory as published responder rates for cannabidiol alone and for ketogenic diet therapy alone, which is the basis for the authors’ conclusion that the combination performs comparably rather than synergistically.

The drop from 40 percent to 28 percent between one and three months deserves comment rather than alarm. Attrition of responder rates over time is a familiar pattern in epilepsy cohorts and can reflect tolerance, diet adherence, changes in concurrent medications, or the natural fluctuation of seizure frequency that makes short observation windows unreliable. A retrospective chart review cannot separate those explanations.

The sequencing observation is the part of this paper most likely to influence practice. Thirty-three patients (57 percent) started CBD first and then added the ketogenic diet. Those patients showed greater seizure reduction than patients who went the other way. The authors present this as a reason to consider the combination, particularly when the diet is layered onto established CBD, and they are careful not to call it a causal effect.

For families already on the ketogenic diet, the first worry about adding CBD is usually whether the carbohydrate content of an oil-based product or the drug itself will break ketosis. This cohort offers reassurance on that point. Eighty-seven percent of patients reported moderate to large urinary ketosis while on both treatments. Among the subgroup who added CBD to an existing ketogenic diet, 8 of 13 had no change in measured ketosis at all.

That is a small subgroup and urinary ketones are a crude measure compared with serum beta-hydroxybutyrate. Still, it is a direct look at what happens to ketosis when CBD is introduced, and it argues against the assumption that the two are biochemically incompatible.

A second practical finding concerns liver enzymes. Among patients who had liver function tests both before and during combination therapy, five (15 percent of those tested) showed increased values, described by the authors as mostly minor. Transaminase elevation is a known effect of cannabidiol, particularly alongside valproate, and this cohort does not suggest the ketogenic diet amplifies it dramatically. It does support checking enzymes rather than assuming they are fine.

Cannabidiol and the ketogenic diet reduce seizures through routes that have little overlap. Cannabidiol is a poor direct agonist at CB1 and CB2 and appears to act instead through targets including GPR55 antagonism, modulation of TRPV1 channels, and inhibition of adenosine reuptake. It also raises clobazam levels through CYP interactions, which accounts for part of its observed benefit in some trials.

The ketogenic diet works metabolically. Shifting the brain’s fuel supply toward ketone bodies alters mitochondrial function, reduces glycolytic flux, and appears to shift the balance between glutamatergic and GABAergic signaling. Neither mechanism is a version of the other, which is the pharmacologic rationale for using them together.

Mechanistic independence is an argument for plausibility, not for benefit. Two treatments that act differently can still fail to add anything when combined, and this cohort, with its comparable rather than superior responder rates, is consistent with that possibility. The honest reading is that combining them does not appear to cancel either one out.

If a patient is stable on cannabidiol and seizures remain inadequately controlled, this paper supports a conversation about adding ketogenic diet therapy rather than treating the two as alternatives. It does not establish that the combination will work, and it does not identify who is likely to respond.

If a patient is already on the diet and CBD is being considered, the ketosis data here are the most directly useful part of the paper. Most patients in this cohort maintained ketosis. That is worth knowing before a family declines CBD out of concern about disrupting a diet that took months to establish.

Ketogenic diet therapy requires supervision by a team that includes a dietitian, and cannabidiol requires attention to drug interactions and liver enzymes. Running both at once compounds the monitoring burden. Nothing in this study changes that, and the median age of 2 years in this cohort is a reminder that most of this experience comes from specialized pediatric epilepsy centers.

This is descriptive evidence from a specialized center, and it should be read as a careful case series rather than a comparative study. Retrospective chart review with no randomization, no control arm, and no blinding cannot establish that either treatment caused the observed seizure changes, let alone that the combination did.

Within that ceiling, the work is solid and useful. The cohort is real, the follow-up averaged well over a year, the biochemical monitoring is documented, and the authors describe their limitations rather than writing around them. For questions about feasibility and tolerability, this is reasonable evidence. For questions about efficacy, it is a starting point for a trial, not a substitute for one.

The efficacy comparison is indirect. The conclusion that combination therapy performs similarly to either treatment alone rests on comparing this cohort’s responder rates against published monotherapy figures from other populations. Those populations differ in age, syndrome, seizure type, and concurrent medications. Cross-study comparison of responder rates is one of the weaker forms of inference in epilepsy research.

The sequencing finding, which is the most quotable part of the paper, is also the most fragile. Patients whose clinicians started CBD first may differ systematically from those who started the diet first, in seizure severity, in syndrome, in age, or in family resources. With 58 patients split across sequencing groups, the subgroups are small and confounding by indication is likely. The ketosis subgroup of 13 patients is smaller still.

Responder rate at three months was measured in a cohort with an average treatment duration of 17.4 months, so the reported outcomes capture an early window rather than the full course. Urinary ketone measurement is also a coarse proxy for metabolic state.

This study does not show that adding the ketogenic diet to cannabidiol works better than continuing cannabidiol alone. It contains no comparison group that would allow that claim.

It does not identify which patients, syndromes, or seizure types respond to the combination, nor does it establish a CBD dose, a diet ratio, or a titration schedule. It does not demonstrate that starting CBD first causes better outcomes, only that better outcomes were observed in the group that happened to be sequenced that way.

It is not a safety study. Five patients with liver enzyme increases in a cohort of 58, with no systematic adverse event capture, cannot characterize the risk profile of concurrent use.

Cannabidiol entered mainstream epilepsy care through three randomized programs in Dravet syndrome, Lennox-Gastaut syndrome, and tuberous sclerosis complex, which produced FDA approval of a purified formulation. Ketogenic diet therapy arrived by a different route, with decades of center-based experience and a smaller randomized literature. The two have coexisted in the same clinics for years without a published account of what happens when they are combined.

That gap is characteristic of drug-resistant epilepsy generally. Most patients in this population are on several therapies simultaneously, and almost none of those combinations have been tested as combinations. Descriptive cohorts like this one are how the field usually learns that a pairing is feasible, well before anyone funds a trial to learn whether it helps.

The result families will care about most is the one about ketosis, not the one about seizures. I have watched parents who spent months getting a child stable on a ketogenic diet refuse to consider cannabidiol because they were told, or assumed, that an oil-based product would undo the work. This cohort suggests that concern is mostly unfounded. Eighty-seven percent stayed in moderate to large ketosis, and most of the small group who added CBD to an existing diet saw no change at all.

The sequencing finding is interesting and I would not build a treatment plan on it yet. In a retrospective cohort, the order in which treatments were started is a decision a clinician made for reasons that are not in the chart, and those reasons tend to track with how sick the patient was. If the effect is real, a prospective study will find it. If it is confounding, a prospective study will find that too.

What I take from this paper clinically is permission to stop treating these two as competing options. They are not. They work through different biology, they appear to coexist without wrecking each other, and for a family that has run out of medications, having two additive tools instead of a choice between them is worth something.

Cannabidiol and ketogenic diet therapy can be run concurrently without apparent loss of ketosis in most patients, and liver enzyme changes in this cohort were uncommon and mostly minor. Seizure outcomes looked like what either treatment produces on its own. Treat this as a feasibility and monitoring paper, and monitor liver function when both are in use.

The finding worth carrying into a clinic visit is that concurrent use appears biochemically tolerable and that ketosis generally holds. The finding not to carry forward is any claim that the combination is more effective than either therapy alone, or that starting cannabidiol first causes better seizure control. Neither claim is supported by a cohort of this design.