After de-duplication, CCEP N1 latency increased with distance in 17 archived SEEG patients

A re-analysis counted each patient once in archived feature tables from OpenNeuro datasets ds005448, ds004696, and ds004457. Three recordings in ds004457 were byte-identical to recordings in ds004696, so these patients had been counted twice in the original 20-patient analysis. Using the pre-written, hashed plan and unchanged original pooling functions, the primary analysis removed the three duplicate rows from ds004457 and merged the Mayo datasets as one source. The pooled N1 latency-distance slope remained positive across 17 patients (16 positive; 0.455 ms/mm, 95% CI 0.311–0.599; Holm-adjusted sign-flip p = 0.020). This p-value is about five times the original 0.0022. No difference was detected between Mayo and Utrecht (Holm-adjusted permutation p = 0.18), though the comparison had low power (9 Mayo and 8 Utrecht patients). Sensitivity analyses retained a positive pooled slope. In an exploratory comparison of archived tables for the three duplicate patients, one pair of slopes differed by |z| > 1.96; the tables used different recording contacts and pipelines, so this is not a pipeline test-retest. The analysis used archived features and did not re-detect N1 from raw recordings. Duplicate detection found only byte-identical recordings, and the assumption that these were all shared patients may be false.

CCEP · N1 latency · SEEG · duplicate recordings · patient de-duplication · latency-distance slope · intracranial EEG · re-analysis