Three duplicate recordings reduce the pooled SEEG sample to 17 people, while the CCEP latency-distance slope remains positive

An audit of archived pair-level tables from three OpenNeuro datasets examined a pooled analysis that reported 20 SEEG patients. Independent comparison of anonymous S3 listings found matching ETag and size for every MEF channel file in three subject pairs: ds004457 sub-1/sub-2/sub-4 and ds004696 sub-04/sub-01/sub-05. These are strong evidence that the 20 entries represent 17 people, although they do not prove identity in clinical records. Re-running the original code and inputs reproduced the reported slope (0.47 ms/mm, 19 of 20 positive). Removing the ds004457 copies and applying the same rules gave a pooled slope of 0.45 ms/mm (95% CI 0.31–0.60), with 16 of 17 people positive (sign-flip p = 0.010; Holm p = 0.020). Alternative deduplication choices also gave positive slopes (0.47–0.53 ms/mm). The apparent speed remained within 2–5 m/s. After deduplication, the ds004457 center contained only two independent people, limiting the center comparison. Only archived tables were analyzed; no raw signals were read. The conclusion is limited by uncertain clinical identity, the small number of duplicated people, and the descriptive comparison of two pipelines that used mostly different recording contacts.

stereo-EEG · SEEG · CCEP · N1 latency · duplicate patients · OpenNeuro · replication · random effects