Heartbeat-locked bipolar SEEG activity exceeded a registered pseudo-beat null after linear ECG regression in 18 patients

This registered replication tested whether heartbeat-locked stereo-electroencephalography (SEEG) activity exceeds a pseudo-beat control in an independent multicenter cohort, OpenNeuro ds003876. We analyzed the first 180 s of one interictal run from each of 30 candidate subjects, using same-shaft bipolar SEEG pairs and the first 40 Hz of signal. Eighteen subjects met inclusion and quality criteria: 10 NIH, 6 JHH, 1 RNS, and 1 UMF. For each pair, heartbeat-locked averages around real R peaks and a pseudo-beat train were compared with 100 random-time surrogates. The registered primary outcomes were the per-subject fractions of pairs with p < 0.05 in the 200–600 ms window, before and after regression on the ECG average and its derivative. The median fraction was higher for real than pseudo beats both before regression (0.123 vs 0.081; 15 of 18 subjects; one-sided Wilcoxon, Holm-adjusted p = 0.008) and after regression (0.150 vs 0.040; 14 of 18; Holm-adjusted p = 0.005). Secondary tests found the pseudo-beat fraction above 0.05 (median 0.081, p = 0.043), excess in the QRS window (p = 0.018), and lower heartbeat-locked z values for seizure-onset-zone pairs than other pairs (median difference −0.51, p = 0.034, uncorrected). The result shows heartbeat-locked activity that a linear ECG model does not explain; it does not establish a neural heartbeat-evoked response because pulse-related artifact may remain. The analysis used only 180 s per subject and 100 surrogates, and 16 of 18 included subjects came from two centers.

stereo-electroencephalography · heartbeat-locked activity · pseudo-beat control · electrocardiography regression · epilepsy · registered replication