Skip to content
GLP Loss
← Research
Safety

GLP-1s and Seizure Medications: Nobody Has Drawn a Level

Phenytoin can be tipped into intoxication by a 10% change in dose, and carbamazepine is titrated to a 4 to 12 mcg/mL window. No GLP-1 label names either one, and no study has ever measured an antiseizure drug concentration in a person taking one of these injections. What has been counted is seizures — and they went down.

Owen Castellanos9 min read
Seizure drugs move weight in both directionsAnd nobody has measured one of their blood levels on a GLP-1Weight goes upValproate, long-term adult use:BMI higher by 2.73 kg/m²Weight gain is a labeled reactionWeight goes downTopiramate: weight loss andanorexia are among its mostcommon labeled reactionsInteraction studies pairing a GLP-1 with any of them: noneSix GLP-1 labels name no antiseizure drug, and no trial drew a levelWhat has been counted is seizures, and they went down.8,688 matched pairs: recurrence hazard ratio 0.82 against other diabetes drugs.Phenytoin is the caution: a 10% dose rise can produce intoxication.

People with epilepsy are prescribed weight-loss drugs at a rate their neurologists did not choose. Several of the medicines that control seizures also add weight, year after year, and the alternative to the one that adds weight is sometimes no alternative at all. So the pairing happens constantly. What has been published about it is a strange shape: a good deal about whether seizures get worse, and nothing whatsoever about whether the drugs holding them off still reach the bloodstream at the same concentration.

Two of these drugs have windows narrower than warfarin’s

Phenytoin is the clearest case, and its label explains the arithmetic in one sentence. Because the enzyme that clears it saturates, small increments matter disproportionately: “The steady-state level may be disproportionately increased, with resultant intoxication, from an increase in dosage of 10% or more.”[1] The effective serum concentration it is aimed at is 10 to 20 mcg/mL, and the label carries a separate warning about levels sustained above that range producing confusional states or, rarely, irreversible cerebellar dysfunction.[1] Carbamazepine is dosed to a usual adult therapeutic level of 4 to 12 mcg/mL, and its label states that monitoring of blood levels has increased the efficacy and safety of anticonvulsants.[2]

That is precisely the category both GLP-1 labels flag. The tirzepatide label asks prescribers to monitor patients on oral medications dependent on threshold concentrations for efficacy and those with a narrow therapeutic index, giving warfarin as its example.[3] The semaglutide label asks for increased clinical or laboratory monitoring for medications with a narrow therapeutic index or that require clinical monitoring.[4] Neither names an antiseizure drug. The worked example they do name is covered in the warfarin article, and the general mechanism behind both cautions is set out in the oral medications article.

What the mechanism predicts, and what it does not

The prediction is narrower than the caution sounds. Across this drug class the reliable effect of slowed gastric emptying is on the rate at which a swallowed drug arrives, not on how much of it eventually does. A 2026 scoping review of GLP-1 use in people with epilepsy states it in those terms: these drugs delay gastric emptying and may alter the timing of antiseizure medication absorption, but overall drug exposure appears largely unchanged in most cases.[5]

Timing is not nothing for this class of drug, though. A person who takes carbamazepine twice daily is living on a trough concentration, and a peak that arrives three hours late compresses the interval before the next dose. Whether that matters has not been measured in anybody. It is worth noticing that even the modeling work skipped these drugs: the 2025 physiologically based analysis that simulated what a motility delay would do to untested medicines covered statins, metformin, metoprolol, digoxin, an anticoagulant and a contraceptive, and included no antiseizure medication at all.[6]

The census, stated so it can be checked

Two searches define the gap. The first is over the labels: the full prescribing information for both tirzepatide products, both semaglutide injections, the semaglutide tablet and liraglutide contains no mention of any antiseizure medication, nor of the words antiepileptic, anticonvulsant or seizure in an interaction context. Six documents, zero entries.

The second is over the literature. A PubMed query crossing the five marketed GLP-1 molecules with phenytoin, carbamazepine, valproate, lamotrigine or levetiracetam and the term pharmacokinetics returns one record, and that record is a plasma protein-binding assay for liraglutide with no antiseizure drug in it. No trial, case series or cohort has reported an antiseizure medication concentration measured before and after a GLP-1 started. Whatever else is known here, that specific number does not exist.

What has been measured is the seizure itself

The outcome literature arrived quickly and it runs one way. The most directly relevant study is a retrospective cohort of adults who already had epilepsy and type 2 diabetes, drawn from a multinational health record network and matched on 82 covariates into 8,688 pairs. Against people starting other glucose-lowering drugs, GLP-1 initiation was associated with a lower risk of seizure recurrence (HR 0.82; 95% CI, 0.78 to 0.86), hospitalization (HR 0.35), all-cause mortality (HR 0.40) and status epilepticus (HR 0.75).[7]

A second cohort asked the safety question more narrowly and reported the answer honestly. Among 610 propensity-matched adults with epilepsy, obesity and type 2 diabetes, starting a GLP-1 was not associated with failure to control seizures (HR 0.92; 95% CI, 0.68 to 1.23), and the authors state that with the events observed the study could only have detected a hazard ratio of 1.53 or greater.[8] That sentence is the useful one: this is a study that could rule out a large harm, not a small one.

Behind those sit three larger datasets about people who did not have epilepsy yet. A cohort of 452,766 matched adults with type 2 diabetes found incident epilepsy lower on a GLP-1 than on a DPP-4 inhibitor (HR 0.84; 95% CI, 0.78 to 0.90), with semaglutide showing the strongest association at 0.68.[9] A target trial emulation reported adult-onset seizure risk at HR 0.44 against other glucose-lowering drugs and 0.48 against SGLT2 inhibitors, with mediation through HbA1c and body mass index close to zero.[10] And a meta-analysis of 27 randomized cardiovascular and renal outcome trials covering almost 200,000 patients found a 24% lower risk of late-onset seizures and epilepsy combined (RR 0.76; 95% CI, 0.62 to 0.95), an effect present only in the GLP-1 arms.[11]

The 2026 scoping review summarizes the whole set as relative risk reductions ranging from roughly 10% to 57% across studies, with no pharmacovigilance indication of seizure worsening.[5] Every one of those studies reads diagnosis codes. None of them drew a drug level. A reader whose question is “will my medication still be at the concentration my neurologist titrated it to” has not been answered by any of them, and the reassuring direction of the outcome data should not be allowed to stand in for an answer.

It is also worth noticing what the outcome studies could not separate. People who lose a fifth of their body weight sleep differently, and obstructive sleep apnea is itself a recognized aggravator of seizures, so a lower recurrence rate has several plausible routes that have nothing to do with any drug reaching a brain. The target trial emulation did test one version of that question and found mediation through body mass index at essentially zero, which argues against the simplest weight explanation without ruling out the others.[10]

The weight effect runs both ways, and it is sizable

The reason this pairing exists at all is on the other side of the prescription. A 2025 systematic review and meta-analysis of 28 studies covering 2,231 patients and 1,582 healthy controls found body mass index rising on valproate in adults during long-term therapy by a mean difference of 2.73 kg/m² (95% CI, 1.77 to 3.69), and by 0.58 in pediatric populations.[12] The valproate label lists weight gain and increased appetite among its adverse reactions.[13] Carbamazepine, oxcarbazepine and phenytoin turned up in the same analysis for cholesterol rather than weight, and levetiracetam came out neutral on lipids.[12]

Topiramate goes the other way, and its label says so without hedging: weight loss and anorexia are among the most common adverse reactions in both the adult and pediatric epilepsy trials, appearing in the highest-incidence list at the 400 mg daily dose.[14] That makes topiramate the one antiseizure drug likely to be doing a version of the same job as the injection, which is a conversation for a prescriber rather than a reason for anyone to assume the two add up neatly. Topiramate also reaches many people through migraine prevention rather than epilepsy, which is a separate reason a reader may be taking it without thinking of it as a seizure drug.

One study looked directly at whether the weight-gaining medicines blunt the injection’s effect. They did not: adjusted weight loss was 5.73% overall at twelve months, and 4.15% even among the participants still taking a weight-gaining antiseizure medication.[8] A smaller number, but not a canceled one.

What a reader on both prescriptions is actually left with

The honest summary has three parts and they do not resolve into a single instruction. The seizure outcome data are reassuring and observational. The absorption question has never been studied in a single person. And the drugs with the least room for error — phenytoin most of all, where a tenth of a dose can tip a level into intoxication — are the ones where a timing shift would show up first if it showed up anywhere.

The instrument that would settle it already exists and is already routine: a serum level, which people on phenytoin and carbamazepine have drawn periodically anyway. The gap is not laboratory capacity. It is that a level is ordered by a neurologist who may not know a weekly injection was bought from a telehealth seller, and the weeks when a timing shift would be largest are the weeks of dose escalation, when nothing prompts anybody to check. Compounded semaglutide and tirzepatide are also not approved by the FDA and are not reviewed by the agency for safety, effectiveness or quality before a pharmacy dispenses them, so the general caution about narrow-therapeutic-index drugs quoted above does not travel with the vial. Nothing here is medical advice or a reason to change any dose; how claims on this site are established is set out in the methodology.

Frequently asked

Do GLP-1 labels say anything about seizure medications?
No. The prescribing information for both tirzepatide products, both semaglutide injections, the semaglutide tablet and liraglutide names no antiseizure drug at all. What they carry is a general instruction to monitor medications with a narrow therapeutic index or that depend on threshold concentrations, with warfarin given as the worked example. Phenytoin and carbamazepine fit that description and are simply not mentioned.
Has anyone measured whether a seizure drug level changes on a GLP-1?
No. A PubMed search crossing the five marketed GLP-1 molecules with phenytoin, carbamazepine, valproate, lamotrigine or levetiracetam and pharmacokinetics returns one record, and it is a protein-binding assay containing no antiseizure drug. Even the physiologically based modeling study that simulated untested medicines covered statins, metformin, metoprolol, digoxin, an anticoagulant and a contraceptive, and included no antiseizure medication.
Do GLP-1s make seizures more likely?
The published data point the other way. A cohort of 8,688 matched pairs of adults who already had epilepsy found seizure recurrence lower after starting a GLP-1 than after starting another glucose-lowering drug, at a hazard ratio of 0.82. A separate safety-focused cohort found no association with loss of seizure control, and three larger datasets found incident seizures and epilepsy lower. All of it is observational, and none of it involved drawing a drug level.
Why is phenytoin singled out?
Because its clearance saturates, so the relationship between dose and blood level stops being proportional at the top of the range. Its label states that the steady-state level may be disproportionately increased, with resultant intoxication, from an increase in dosage of 10% or more. That is a narrower margin than almost any other drug discussed on this site, and it is why a timing change would show up there first if it showed up anywhere.
Will a seizure medication that causes weight gain cancel out the injection?
Not in the one study that checked. Among 610 matched adults with epilepsy, obesity and type 2 diabetes, adjusted weight loss at twelve months was 5.73% overall and 4.15% among those still taking a weight-gaining antiseizure medication. Less, but far from nothing. For scale, the same meta-analysis that documents the problem found body mass index higher by 2.73 kg/m² in adults on long-term valproate.
What about topiramate, which already causes weight loss?
Its label lists weight loss and anorexia among the most common adverse reactions in the epilepsy trials at the higher dose studied, so it is the one antiseizure drug already doing a version of what the injection does. Many people take it for migraine prevention rather than epilepsy and may not think of it as a seizure medication at all. Whether the two effects should be combined is a question for the prescriber who titrated the first one.

Sources

  1. [1] Viatris Specialty LLC (2026). DILANTIN (extended phenytoin sodium) capsules — Dosage and Administration 2.3, Warnings and Precautions 5.15, and Clinical Pharmacology 12.3 DailyMed, U.S. National Library of Medicine. Source
  2. [2] Novartis Pharmaceuticals Corporation (2026). TEGRETOL (carbamazepine) tablets and suspension — Clinical Pharmacology, and Precautions: Laboratory Tests DailyMed, U.S. National Library of Medicine. Source
  3. [3] Eli Lilly and Company (2026). ZEPBOUND (tirzepatide) injection — Drug Interactions 7.2: Oral Medications DailyMed, U.S. National Library of Medicine. Source
  4. [4] Novo Nordisk Pharmaceutical Industries LP (2026). WEGOVY (semaglutide) injection — Drug Interactions 7.2: Oral Medications DailyMed, U.S. National Library of Medicine. Source
  5. [5] Pinto LF, Guerreiro CAM (2026). GLP-1 receptor agonists in people with epilepsy: A scoping review of clinical data, safety, and therapeutic implications. Seizure. PMID 42202441
  6. [6] Hooper L, Liu S, Pai MP (2025). GLP-1RA-induced delays in gastrointestinal motility: Predicted effects on coadministered drug absorption by PBPK analysis. Pharmacotherapy. PMID 39989027
  7. [7] AbuAlrob MA, Hussein A, Abdellatif R, et al. (2026). Seizure recurrence after GLP-1 receptor agonist initiation in adults with epilepsy. Epilepsia. PMID 41251033
  8. [8] Son H, Hwang I, Lee SW, et al. (2026). Safety and Efficacy of GLP-1 Receptor Agonists in Adults With Epilepsy, Obesity, and Type 2 Diabetes. Ann Clin Transl Neurol. PMID 42563258
  9. [9] Cheng CY, Lo SC, Huang CN, et al. (2026). Association Between GLP-1 Receptor Agonist Use and Epilepsy Risk in Type 2 Diabetes. Neurology. PMID 41370744
  10. [10] Eun Y, Bong S, Koh HY, et al. (2026). Semaglutide and Risk of Adult-Onset Seizure: A Target Trial Emulation. Neurology. PMID 42308439
  11. [11] Sindhu U, Sharma A, Zawar I, et al. (2024). Newer glucose-lowering drugs reduce the risk of late-onset seizure and epilepsy: A meta-analysis. Epilepsia Open. PMID 39487832
  12. [12] Chochoł P, Arturo N, Łajczak PM, et al. (2025). Effects of Antiseizure Medications on Lipid Profile and Weight in Patients with Epilepsy: A Systematic Review with Meta-analysis. CNS Drugs. PMID 41032224
  13. [13] AbbVie Inc. (2026). DEPAKOTE (divalproex sodium) delayed-release tablets — Adverse Reactions 6.1 DailyMed, U.S. National Library of Medicine. Source
  14. [14] Janssen Pharmaceuticals, Inc. (2026). TOPAMAX (topiramate) tablets and sprinkle capsules — Adverse Reactions 6.1 DailyMed, U.S. National Library of Medicine. Source

More in Safety