Automatic hippocampal IEDs showed a nonsignificant 4.4-point recall decrease during word encoding in 21 patients

This reanalysis tested whether automatically detected hippocampal interictal epileptiform discharges (IEDs) during word encoding predict later recall, and whether IEDs just before a word predict recall. In OpenNeuro ds004789 (RAM FR1, ses-0, bipolar SEEG), a pre-registered detector and analysis included 21 subjects for the primary comparison. Recall was 4.4 percentage points lower for words with an IED in the 0–1.6 s encoding window than for words without one (95% CI −8.6 to +0.1 points; sign-flip p = 0.065; Holm-adjusted p = 0.194; 15/21 subject differences were negative). Adjustment for serial position and list number gave a −3.8-point difference (p = 0.081). Pre-word IEDs did not predict recall (+0.7 points; p = 0.824; n = 17). Sensitivity thresholds gave −2.5 points (p = 0.23; n = 24) and −4.2 points (p = 0.066; n = 20). An exploratory analysis restricted to high-amplitude events showed no effect. The results show a weak trend in the direction of the earlier report, not confirmation: no pre-registered test was significant after Holm correction, and the primary confidence interval includes zero. No expert reviewed the detected events, which may include artifacts or high-amplitude beta/gamma bursts. Only 29 of 155 eligible subjects were saved under a fixed run-time limit; analyses used bipolar hippocampal pairs in ses-0 and did not separate hemispheres or seizure-onset sides.

hippocampus · interictal epileptiform discharges · word encoding · free recall · stereo-EEG · automatic event detection · reproducibility