Antibodies to glutamic acid decarboxylase define a form of limbic encephalitis




Antibodies to glutamic acid decarboxylase (GAD) have been described in a few patients with temporal lobe epilepsies consistent with limbic encephalitis (LE). We studied a cohort of patients with recent-onset temporal lobe epilepsy caused by LE to test for GAD antibody positivity and response to immunotherapies.


Over a period of 3.75 years, 138 patients aged ≥18 years investigated at the Department of Epileptology, University of Bonn, for recent-onset epilepsy were prospectively collected and studied for cliniconeuroradiological features of LE, autoantibodies, and treatment responses.


Fifty-three adult patients fulfilled the criteria for LE: (1) limbic signs and symptoms for ≤5 years and (2) brain MRI revealing mediotemporal encephalitis (T2/fluid attenuated inversion recovery hyperintensity without atrophy). Nine had high-titer GAD antibodies; 10 had voltage-gated potassium channel (VGKC) antibodies. Patients with GAD antibodies were younger (median age, 23 years; range, 17-66 years) (p = 0.003) and presented with seizures only, whereas polymorphic limbic features were more common in the VGKC antibody-positive group (p < 0.001). None had tumors. Patients with GAD antibodies more frequently had cerebrospinal fluid oligoclonal bands (p = 0.009) and intrathecal secretion of the specific antibody (p = 0.01). Following monthly intravenous methylprednisolone pulses, GAD antibodies remained highly elevated in 6/6 patients, whereas VGKC antibodies normalized in 6/9 patients (p = 0.03). Despite more intense anticonvulsive treatment in the GAD antibody-positive group (p = 0.01), none of these patients became seizure free, unlike all of the patients with VGKC antibodies (p < 0.001).


High-titer GAD antibodies define a form of nonparaneoplastic LE. This is a chronic, nonremitting disorder and should be included in the differential diagnosis of patients with TLE and mediotemporal encephalitis. Therapeutic trials of other immunotherapies should be undertaken. ANN NEUROL 2010;67:470–478