Abstract / Summary
Among patients with migraine without aura (MO) and 0–3 monthly migraine days (MMDs) at baseline, neurophysiological correlates of subsequent frequency increase remain unclear. We characterized interictal source-space resting-state electroencephalography (EEG) features and evaluated their associations with short-term frequency increase and prognostic contribution. This single-center prospective cohort study was nested within the Neuroelectrophysiology Core of the China Migraine Registry (CHIME). Among 697 patients screened, 216 patients with MO and 0–3 baseline MMDs were included; 94 healthy controls (HCs) were included separately. Short-term frequency increase was defined as ≥ 4 MMDs during the third month after baseline. Patients were chronologically assigned to development ( n = 126) and temporal validation ( n = 90) cohorts. Periodic alpha power, aperiodic spectral features, alpha-band functional connectivity, and network topology were assessed across seven canonical networks. Clinical, EEG, and combined elastic-net models were developed using repeated nested cross-validation and applied unchanged to the validation cohort. Ninety patients (41.7%) met the outcome criterion. Compared with HCs, patients with MO had lower periodic alpha power in the visual network (VIS) and limbic networks, lower regional and global aperiodic offset, lower within-VIS alpha-band connectivity and system segregation, and a higher mean participation coefficient. In the development cohort, frequency increase was associated with lower baseline within-VIS connectivity, global aperiodic offset, and system segregation and with a higher mean participation coefficient. The combined model achieved areas under the receiver operating characteristic curve (AUROCs) of 0.871 (95% confidence interval [CI], 0.811–0.931) in development and 0.821 (95% CI, 0.732–0.910) in validation. The combined model significantly outperformed the clinical model in the development cohort, whereas this improvement was not statistically established in the temporal validation cohort. Validation sensitivity was 0.553, specificity was 0.981, and the calibration slope was 0.604. Excluding boundary cases yielded broadly similar principal patterns and model performance. Interictal source-space EEG identified spectral and network differences in MO, while lower baseline alpha-band connectivity within the VIS was associated with short-term frequency increase. The combined model retained discrimination in temporal validation, but its incremental value over clinical variables was not established. Independent multicenter validation is required before clinical application. Not applicable.