GLP 1 multiple sclerosis records study links the drugs to fewer relapses and optic neuritis

A GLP 1 multiple sclerosis records analysis presented at MSToronto 2026, the joint meeting of the ECTRIMS and ACTRIMS societies, found fewer steroid treated relapses, optic neuritis episodes and demyelinating events among people with MS who had used GLP 1 drugs. The study drew on 339,092 patients in a US health records network. People who used the drugs were heavier and far more often had diabetes, and the result is an association, not proof that the drugs protect nerves. A randomised trial is now under way.

What the TriNetX analysis compared

Mahsa Khayat-Khoei of the University of California Riverside presented the work, according to MedPage Today. Her team searched the TriNetX research network, which pools electronic health records from many US health systems, for people coded with multiple sclerosis. Of 339,092 patients, 14,300 had been exposed to a GLP 1 receptor agonist, and 82% of those were women. The exposed group differed sharply from the rest. Mean body mass index was 36.7 against 29.0, type 2 diabetes was recorded in 49% against 4%, and obesity in 58% against 6%. That is expected, because the drugs are prescribed for diabetes and obesity. Khayat-Khoei told MedPage Today that patients were propensity score matched by age, sex and other illnesses before outcomes were compared.

The main window covered 12 years of records, and the team said analyses over 21 years gave similar results. A separate look at people treated with ocrelizumab, a B cell therapy available since 2017, used a 10 year window and included 30,934 patients, 1,953 of them exposed to a GLP 1 drug. Outcomes were drawn from diagnosis and treatment codes in the records: relapses treated with steroids, optic neuritis, demyelinating events, gait problems, fatigue and depression. Codes like these show what clinicians recorded and billed. They do not come from scheduled neurological exams or regular MRI scans, which are the usual tools for judging MS activity in trials. The presentation has not been published as a peer reviewed paper.

GLP 1 multiple sclerosis relapse figures

Over 12 years, steroid treated relapses were recorded in 1.85% of exposed patients against 4.84% of unexposed patients. Optic neuritis appeared in 1.46% against 2.56%, and demyelinating events in 3.48% against 7.01%. Gait abnormalities, fatigue and depression were also less common in the exposed group, at 10.46% against 16.15%, 11.87% against 17.55%, and 19.34% against 26.38%. All of these differences were highly significant. The way the risk ratios were reported is easy to misread. A figure of 2.62 for relapses means the unexposed group had about 2.6 times the rate of the exposed group, not that the drugs raised risk. Put the other way, exposed patients had roughly 62% fewer recorded steroid treated relapses.

Put the other way, exposed patients had roughly 62% fewer recorded steroid treated relapses.

In the ocrelizumab subgroup the pattern was similar and the gaps were even larger, with a risk ratio of 5.52 for steroid treated relapses and 2.37 for optic neuritis. Khayat-Khoei noted that GLP 1 drugs may affect inflammation in the nervous system through the gut and brain axis, and that animal models of MS, Alzheimer's disease, Parkinson's disease and stroke point the same way. She said prospective studies are needed to learn whether the drugs have disease modifying or neuroprotective effects in MS. Paul Edison of Imperial College London, who was not involved, called the data "encouraging" and said the drugs act on insulin signalling, mitochondria, blood vessels and inflammation, giving "a credible biological basis" for testing them in neurology.

How the spring and autumn numbers differ

The same research group, with Aayma Irfan and Lynsey Lakin of UC Riverside, submitted an abstract on this cohort to the American Academy of Neurology 2026 meeting earlier in the year. It reported the same 339,092 patients and 14,300 exposed, but different outcome figures, as the table shows. That version matched groups only by age and sex, and it did not state a time window. Differences between the two releases probably reflect changes in matching and follow up, but neither abstract gives enough detail to say exactly why. Readers should treat the precise GLP 1 multiple sclerosis percentages as provisional until a full paper sets out the methods, the matched sample sizes and how missing data were handled.

Both versions share the same central weakness. People who receive GLP 1 drugs differ from those who do not in weight, diabetes, contact with doctors and probably access to care, and matching can only balance what the records contain. MS disease modifying treatment was not used to build the cohorts in the spring abstract, so differences in MS therapy could also shape relapse rates. Steroid treated relapses depend on someone seeking and receiving treatment, which can vary with insurance and habits. Neither abstract lists funding or disclosures. Similar records signals for other conditions, such as a recent ankylosing spondylitis signal on this site, show how often these databases produce associations that still need independent confirmation.

What a trial would need to show

A direct test has started. TAG-MS, a phase 2 trial led by Johns Hopkins University with support from the National Multiple Sclerosis Society, began enrolling on 28 April 2026 and aims to randomise about 120 people to a GLP 1 receptor agonist or placebo, with participants, care providers, investigators and outcome assessors all masked. Its main measure is the change in brain volume on MRI over about 96 weeks, with retinal thickness and disability scores as secondary outcomes. Primary completion is estimated for January 2030, according to the TAG-MS trial registry. MedPage Today reports that the study drug is NLY01, a long acting form of exenatide, and that the trial focuses on progressive MS.

Edison also pointed to the mixed record of these drugs in brain disease. Oral semaglutide did not slow decline in the EVOKE trials, covered on this site when semaglutide failed in Alzheimer's, while liraglutide showed some benefit on cognition and brain volume in the smaller ELAD trial, and exenatide's positive phase 2 result in Parkinson's disease was not sustained in phase 3. He argued that GLP 1 drugs differ in structure, brain reach and receptor action, so one failure should not close the field. That is a reasonable point, and it also means a records signal for one class cannot predict how any single drug will perform in a controlled trial.

Conclusion

This GLP 1 multiple sclerosis records study adds a large and consistent association between the drugs and fewer coded relapses, optic neuritis and demyelinating events. It cannot separate the drugs from the very different people who take them, its figures have shifted between two meeting abstracts, and it is not yet peer reviewed. TAG-MS and similar randomised trials will have to show whether any protective effect is real.

Source Khayat-Khoei M and colleagues, University of California Riverside. Presentation at MSToronto 2026 (ECTRIMS and ACTRIMS), reported by MedPage Today; earlier abstract: Irfan A, Lakin L, Khayat-Khoei M, AAN 2026 abstract 005882. Not peer reviewed.

Amino Weekly reports research. It is not medical advice and never gives doses.