SOZ-label coverage did not add clear outcome information beyond treated contact fraction in public SEEG data

In a metadata-only analysis of public SEEG datasets, SOZ-label coverage did not show significant outcome information beyond the treated fraction of all contacts in the primary set (52 patients; 28 seizure-free). The adjusted odds ratio for seizure freedom per SD of SOZ coverage was 1.34 (95% CI 0.69–2.71; permutation p=0.356; Holm-adjusted p=0.859). The data therefore do not show added information, but cannot exclude a moderate effect. In a descriptive sensitivity analysis that added 38 patients with mixed SEEG/ECoG implants and counted only SEEG contacts, SOZ coverage predicted seizure freedom without adjustment (AUC 0.625; p=0.016; Holm p=0.049), but not after adjustment for treated contact fraction (OR 1.41, 95% CI 0.80–2.66; p=0.197). Small sample size, different label definitions and follow-up times, use of one channels file per subject, and a ceiling effect limit interpretation; SEEG contacts from mixed implants may not sample the brain like the full implant.

epilepsy · stereo-EEG · seizure-onset zone · surgical outcome · metadata analysis