Abstracts

Rationale: Epilepsy is a heterogeneous disorder characterized by alterations in white matter (WM) microstructure within multiple brain networks, likely influenced by syndrome-specific, demographic, and clinical characteristics. ENIGMA-Epilepsy was formed to develop the largest quantitative imaging dataset available, aggregating data across centers worldwide to investigate patterns of network abnormalities in common epilepsy syndromes, including temporal lobe epilepsy with hippocampal sclerosis (TLE-HS), nonlesional TLE (TLE-NL), genetic generalized epilepsy (GGE), and nonlesional extratemporal epilepsy (ExE). Our primary goal was to rank the most robust WM microstructural differences across and within syndromes in a large sample of patients across 21 centers from North and South America, Europe and Australia. Methods: Diffusion-weighted imaging (dMRI) data were available for 1,249 patients and 1,069 age- and sex-matched healthy controls: TLE-HS (N=319 left; N=280 right), TLE-NL (N=162 left; N=113 right), GGE (N=182) and ExE (N=193). DMRI data were processed using the same image processing pipeline at each center. Tract-based spatial statistics was used to derive maps of fractional anisotropy (FA) and mean diffusivity (MD) for each patient, and fiber tracts were segmented using a WM atlas. Data were harmonized to correct for scanner-specific variation in average tract FA and MD using a batch-effect correction tool (ComBat). Analyses of covariance, controlling for age, age2, and sex were performed to examine differences between each epilepsy syndrome and healthy controls for each WM tract (Bonferroni-corrected at p<0.001).  Results: ComBat harmonization demonstrated strong alignment of FA/MD values across scanners while maintaining biological variance expected with age (Fig 1). An across syndrome comparison revealed lower FA in all patients with epilepsy compared to controls in most fiber tracts (Fig 2) with small to medium effect sizes (ES), especially in the genu and body of the corpus callosum (CC), cingulum (CING) and external capsule (EC). Less robust effects were seen with MD. However, even within WM pathways affected across syndromes, evidence of syndrome-specific differences in pathology emerged. The most pronounced FA/MD differences were observed in patients with TLE-HS, with large ES differences in the parahippocampal cingulum and EC ipsilateral to the seizure focus, and small to medium ES across most other association, projection, and commissural tracts. TLE-NL showed a similar pattern to TLE-HS, but with smaller ES differences. Patients with ExE demonstrated the most pronounced effects in the body and genu of the CC for FA, and anterior corona radiata bilaterally for FA/MD (small to medium ES).  GGE patients showed diffuse, small effects for FA throughout the brain. Earlier age of seizure onset and longer disease duration were associated with loss of WM microstructure in right and left TLE-HS for many tracts, with stronger association in right TLE-HS. Conclusions: In the largest epilepsy dMRI study to date, we demonstrate microstructural injury across major association, commissural, and projection fibers. Whereas all patients showed WM injury in antero-midline fibers, the magnitude of change varied across syndromes. These data further inform our understanding of epilepsy as a WM network disorder with shared and syndrome-specific features, providing new insights into pathological substrates in epilepsy which could guide treatment or genetic studies. Funding: NIH/NINDS (R01NS065838; R21NS107739)