The following summarizes some additional highlights from the 11th Congress of the European Academy of Neurology, Helsinki, Finland, June 21-24, 2025.
Genetics affect MS risk, not clinical course
GFAP increases with normal aging
Midbrain lesions associated with migraine in MS
When to switch off anti-CGRP MAbs for migraine
Serial sNfL monitoring needed for correct interpretation
Specialists outperform Chat GPT
Genetics affect MS risk, not clinical course
A group in The Netherlands examined the effect of genetic risk factors on the clinical course of MS in 415 patients with clinically isolated syndrome (CIS) (Corsten et al. EAN 2025;EPR-273). Family history, HLA-DRB1*15:01 status, genetic variants related to vitamin D and body-mass index, and levels of Epstein-Barr virus nuclear antigen (EBNA)-1 antibodies were obtained at baseline. Overall, 19.5% of CIS patients had a family history of MS; 66.7% with first-degree relatives were HLA-DRB1*15:01 carriers. Weighted genetic risk scores and anti-EBNA-1 titres were higher in subjects with familial MS. However, the early clinical course did not differ in patients with and without familial MS, suggesting that distinct pathophysiological processes mediate MS risk and the MS clinical course.
GFAP increases with normal aging
Levels of glial fibrillary acidic protein (GFAP) become highly variable in older individuals, complicating its use as a biomarker of astrocytosis in neurological conditions, according to a new analysis of GFAP in healthy community-living individuals (Demjaha et al. EAN 2025;OPR-030). GFAP levels were analysed from 316 healthy persons over a 5.6 year follow-up period; mean age was 64.5 years (range 38-82 years). Mean sGFAP increased from 73.1 pg/mL in persons aged <50 years to 86.8 pg/mL for ages 50-60, to 136.9 pg/mL for age 60-70 years to 154.6 pg/mL for those aged >70 years. GFAP showed an increasing variability with aging. sGFAP increases were larger in women. The authors noted that there is a need to establish normative values based on Z-scores.
Midbrain lesions associated with migraine in MS
A prospective study of 96 MS patients has concluded that the occurrence of migraine may be determined in part by lesion location (Gklinos et al. EAN 2025;EPO-463). Mean age was 42 years; 77.1% had RMS and 22.9% had PMS; and mean EDSS score was 3.1. Midbrain (periaqueductal gray) lesions were significantly associated with migraine (odds ratio 4.7). Thalamic lesions (OR 7.2) and cortical lesions (OR 9.1) were also associated with a higher prevalence of migraine. Lesion location did not appear to be associated with tension-type headache.
When to switch off anti-CGRP MAbs for migraine
An estimated 72% of chronic migraine patients will respond to anti-CGRP monoclonal antibody therapy over the first year of treatment but the probability of a response in non-responders declines after three months, according to a new analysis (Jaimes et al. EAN 2025;OPR-007). The study evaluated 462 migraine patients; median age was 48 years. Response was defined as a >50% reduction in headache frequency. The probability of response declined from 36.1% after one month to 14.9% after three months and <10% from month 4 onward. There was a higher likelihood of response in patients with hemicranial pain (hazard ratio 1.31) and photophobia (HR 1.64). The probability of response was lower in males (HR 0.68). The authors stated that while later responses are possible, consideration should be given to changing therapies in non-responders after a three-month trial period.
Serial sNfL monitoring needed for correct interpretation
A retrospective study of 162 MS patients reported that the timing of serum NfL sampling affected the prognostic value (Martinez-Serrat et al. EAN 2025;OPR-079). A median of seven serum samples were obtained over a 10-year period. sNfL Z-scores were significantly elevated in patients who experienced disease activity (relapse, EDSS worsening, new MRI activity) over the subsequent year, but only if the samples were obtained during a period of remission. sNfL levels were not predictive of future disease activity beyond one year. The authors also noted that sNfL Z-scores increased at the time of relapse and remained elevated for up to nine months thereafter.
Specialists outperform Chat GPT
The latest generation of artificial intelligence is the generative pre-trained transformer 4 omni (Chat GPT-4o), and researchers in Italy evaluated its diagnostic skills (De Lorenzo et al. EAN 2025;OPR-023). Chat GPT-4o was fed data from 100 polyneuropathy cases and asked for its leading diagnosis, differential diagnoses and a confirmatory test. Responses were compared to those from 26 specialist and 12 non-specialist neurologists. Diagnostic accuracy was higher with specialists compared to Chat GPT (73.9% vs. 65.5%), although AI outperformed non-specialist neurologists (54.4%). The rankings were the same when differential diagnoses were included, with specialist neurologists outperforming Chat GP and non-specialist neurologists (88.1% vs. 82.0% vs. 68.5%). GPT-4o was shown to be overly reliant on laboratory findings and past history, overlooked clinical information, offered vague conclusions and demonstrated limited internal knowledge. It also offered reasonable but inaccurate responses. However, its accuracy in recommending diagnostic tests was comparable to that of specialists (68.0% vs. 67.3%) and superior to that of non-specialists (45.3%).