Word-locked hippocampal IEDs were associated with lower recall in an expanded sample, but the preregistered test was inconclusive

In ds004789 FR1 session 0, a prior analysis of 38 SEEG patients reported no recall change for words with a hippocampal interictal epileptiform discharge (IED; +0.5 percentage points, p = 0.78), but its epoch times used events.tsv onset as EDF time. In this dataset onset starts at session start, whereas EDF time is sample/fs; the recorded shift ranged from -124 s to +3246 s (median +98.7 s). We reran the same detector, inclusion rules, and statistics with word time = sample/fs. The preregistered fetch ended at 53 of 64 subjects; 33 met the primary inclusion rules. Their recall was 4.1 percentage points lower for IED words (95% CI -8.2 to +0.1; sign-flip p = 0.068; 23/33 negative; Mantel-Haenszel OR 0.85, 95% CI 0.70-1.03), so the preregistered test did not support H1 at alpha 0.05. After this result, the remaining 11 subjects were fetched. In the non-preregistered expanded analysis, 43 included patients showed a -5.2-point difference (95% CI -8.9 to -1.6; p = 0.009; OR 0.80, 95% CI 0.67-0.95). In the same 38 patients as the prior report, the corrected-timing estimate was -5.5 points (p = 0.012), versus +0.5 with onset timing; paired change was -6.0 points (p = 0.022). Pre-word IEDs showed a smaller, nonsignificant decrease. The two runs’ word-level IED labels agreed weakly (phi = 0.16). These results are correlational and do not establish that IEDs cause lower recall. The detector had no expert review; seizure-onset-zone labels were unavailable; only session 0 was analyzed; and no correction was made for the sensitivity analyses. Dataset: OpenNeuro ds004789, snapshot 3.1.0.

hippocampus · interictal epileptiform discharges · word encoding · free recall · stereo-EEG · memory · event timing