Reviews/Ketogenic and Low Carb Diets/Reduces seizures in drug-resistant epilepsy
Kw04 · KETO-NEU-01 · v2
Borderline: within 4 points of the Established/Likely boundary, so a re-grade could land either side.
Several randomised trials that compared a ketogenic-style diet against carrying on without the diet found that children had far fewer seizures on the diet [8][1][5][11], and two non-randomised comparison studies pointed the same way [10][16]. The result being measured is seizures themselves, not a stand-in marker, which makes the finding more direct. The main weaknesses are that nobody in these trials could be kept unaware of who was dieting, and seizures were counted by families, so belief in the diet could nudge the counts. The adult evidence is thinner and less convincing: one adult trial missed its main target [2] and a small adult study saw no change in seizures [13]. One trial in infants found the diet no better than simply adding another medicine [6]. The panel judged the direction of benefit well supported but held back from the top rating because of these open-label and adult-data limits.
Randomised trials with a non-diet or usual-care control consistently favour the diet in children: [8] reported baseline seizures of 62.0% versus 136.9% in controls (75% decrease, p<0.0001) and 38% versus 6% achieving >50% reduction; [1] gave class I evidence with 50% versus 18.2% responders (P=0.024); [5] showed >50% reduction in 26.2% versus 2.5% (p<0.001); and [11] found 6/20 versus 0 responders (p=0.02). Controlled cohorts agree, with [10] reporting effective rates near 79% versus 59% (P<0.05). The outcome measured is seizure frequency directly rather than a surrogate. Against this, the adult trial [2] missed its primary endpoint (median difference −7.0, 95% CI −37.0 to 3.0; P=.21), [13] saw no seizure-frequency change, and infant trial [6] found the diet no better than a further medicine (IRR 1.33, 95% CI 0.84-2.11). The contested point: The division is between a consistent, often large benefit in children from controlled trials [8][1][5][11] and a weaker, inconsistent adult/infant picture where primary endpoints were non-significant [2], no change was seen [13], or the diet did not beat an added drug [6]. The whole base is also open-label with self-reported seizure counts, so the panel contested how much of the effect could reflect expectation rather than the diet itself.
All 17 papers behind this claim with their DOIs and what each does to it, the panel’s arbitration, the evidence gap map and the open research questions. Free, and sent to your inbox.
We’ll also send you other claims from Ketogenic and Low Carb Diets, one at a time, and tell you when a release or a pivotal trial changes a grade. Unsubscribe by replying.
One reviewer's only job is to attack the finding. The challenges below were weighed before the grade was set.
Every controlled trial in the corpus is open-label; seizure counts are caregiver/patient-reported, so expectation bias could inflate effects systematically across the whole evidence base.
Efficacy is overwhelmingly paediatric; the adult evidence is mixed-to-null — [2]'s primary endpoint was non-significant (P=.21) and [13] found no change in seizure frequency — yet the claim reads as all-ages DRE.
Where it remains uncertain. Every trial is open-label with caregiver-reported counts, and adult/infant evidence is mixed-to-null [2][13][6], so the effect's size and generalisability remain uncertain.
Aleatory
Seizure response varied considerably between individuals in [2], and responder rates rather than universal benefit are reported throughout [8][5][11], so outcome at the patient level is inherently variable.
Epistemic
Every controlled trial is open-label with caregiver/patient seizure diaries [8][6][11], leaving expectation bias on the outcome unmeasured across the whole base.
Inferential
[3][4][7][9] lack a non-diet comparator, so within-group reductions cannot be causally attributed to the diet versus regression-to-mean or co-interventions.
The load-bearing papers from a corpus of 17, each classified by what it does to the claim. Every DOI is taken from the corpus the claim was graded on.
Class I RCT vs care-as-usual confirming higher responder rate on KD.
“In an intention-to-treat analysis, 13 patients (50%) treated with the KD and four patients (18.2%) of the CAU group were responders.”
Panel auditInternal validity: high · scope of inference: Children and adolescents aged 1-18 with refractory epilepsy not eligible for surgery, over 4 months.Key limitations: small sample · two outliers removed from KD analysis · open-label (no blinding) · short 4-month horizon
Counter-evidence: primary seizure-frequency endpoint non-significant in adults, tempering the magnitude and generalisability of the claim.
“The median difference between the groups was -7.0 (95% confidence interval [CI] -37.0-3.0; P = .21).”
Panel auditInternal validity: medium · scope of inference: Adults >16 years with drug-resistant focal epilepsy over 12 weeks.Key limitations: primary endpoint not significant · high dropout · wide CIs · possible drop in antiepileptic drug serum concentrations
Extends the controlled effect to adolescents and adults with DRE.
“At 6 months, >50% seizure reduction was seen in 26.2% in the intervention group vs 2.5% in the control group (95% CI 13.5-33.9; p < 0.001).”
Panel auditInternal validity: medium · scope of inference: Adolescents and adults aged 10-55 with nonsurgical drug-resistant epilepsy over 6 months.Key limitations: open-label · authors classify as Class III evidence · single center
Both arms received a ketogenic-type diet with no non-diet control, so the within-arm reductions cannot be attributed to diet versus no diet.
Panel auditInternal validity: medium · scope of inference: Children aged 1-18 with intractable epilepsy comparing KD vs MAD over 3-6 months.Key limitations: no non-diet control · between-diet differences not significant overall · subgroup finding only in <2 year olds
4 of 17 papers shown. The full review lists all of them, with the panel’s arbitration →
Claims are graded at each quarterly release, and re-graded between releases only when a pivotal trial lands. Every change is a new version of the claim.
Converging controlled RCTs carry the direction of the claim robustly, but pervasive open-label bias and weaker adult data cap confidence below the Established band.
Release 2026-q4 re-graded every claim from its corpus under protocol v0.19. An earlier release's grade is shown for the record; it came from an earlier protocol and is not a movement.
Assumptions that carry weight but are not themselves established.
The panel records the gap that would most move the claim, and the study that would close it.
Blinded outcome ascertainment. All effects rest on unblinded, caregiver-reported counts [8][6][11], the Sceptic's highest-severity concern. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: An RCT using an objective or blinded-adjudicated seizure measure, or a credible sham/placebo diet comparator, to separate diet effect from expectation.
Open question: How much of the observed seizure reduction survives when outcome assessment is blinded? Why it is still open: Every controlled trial in the corpus is open-label with self-reported diaries [8][6][11], so expectation bias is unmeasured.
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 21 Oct 2026 · protocol v0.19