Hippocampal discharges during encoding predicted lower word recall in 37 patients, but post-hoc estimates weakened with stricter detection and list adjustment

Question: do hippocampal interictal epileptiform discharges (IEDs) during word encoding decrease later free recall? Data: OpenNeuro ds004789 FR1, session 0, bipolar EDF, hippocampal pairs by the 7b42f9b6 rule, word timing from the events.tsv sample column. 47 subjects were fetched and 37 were included. Method: an automatic detector (15-60 Hz envelope > 5 x median and a raw peak > 5 sigma), fixed in a hashed plan before data. A word is an IED word if a detection occurs 0-1.6 s after word onset. Primary result: recall adjusted for serial position was 3.8 points lower for IED words (95% CI -6.7 to -1.0; 21 of 35 non-zero subjects negative; sign-flip p = 0.014). Raw recall: -4.2 points (p = 0.0075). IEDs only before word onset had no effect (-0.3 points, p = 0.93, n = 29). There was no dose effect of the number of pairs with an IED. The detection rate did not increase after word onset. Post-hoc checks: the effect stayed when we removed zero-recall subjects or subjects with high detection rates. But it decreased to -2.5 to -2.7 points (p = 0.18 to 0.24) with stricter detector thresholds, and to -2.3 points (p = 0.074) when the residual also removed the mean recall of each list. Limits: no expert check of the detections, no seizure-onset labels, 47 of 155 eligible subjects, and a 2-subject code test before the full run.

hippocampus · interictal epileptiform discharges · verbal memory · free recall · intracranial EEG · OpenNeuro · ds004789 · preregistration