HUP SEEG data show no pre-onset heart-rate rise, but the pre-registered tachycardia rule did not replicate

This pre-registered out-of-sample replication applied the archived analysis code unchanged to HUP SEEG electrocardiogram data from OpenNeuro ds004100. Sample A included 9 patients not fetched in the target study (run-01); Sample B included run-02 from 32 patients. The pre-registered tachycardia rule required significant positive heart-rate differences in both the 0–30 s and 30–60 s windows. It did not replicate in either sample: Sample A increases were 5.5 and 9.0 bpm and were not significant; in Sample B, the 30–60 s increase was 23.5 bpm (95% CI 12.2–34.9; Holm p=0.001), while the 0–30 s increase was 10.9 bpm (Holm p=0.090). The absence of a group heart-rate rise during the 30 s before clinician-marked onset replicated in both samples (upper 95% CI 2.3 and 7.1 bpm). Across 28 patients with both runs, the 30–60 s heart-rate rise was moderately consistent (Spearman rho=0.38, one-sided permutation p=0.022). The pooled descriptive estimate was +24.9 bpm at 30–60 s and +13.0 bpm at 0–30 s. Sample A was small, Sample B reused patients from the target study, and run-02 is not confirmed to be the second seizure in time. The source data URL was an unversioned live mirror; recorded URLs and SHA-256 hashes identify the analyzed byte ranges, but the exact dataset release is uncertain.

epilepsy · heart rate · electrocardiography · seizure · SEEG · replication · OpenNeuro