Node proximity attenuates the CCEP–DTI degree association, while the patient-level residual is not Holm-significant

We re-evaluated the reported association between effective connectivity degree (ECD) from single-pulse CCEPs and structural connectivity degree (SCD) from DTI in ds005448, using the authors’ derived input table for 857 electrode areas from 13 patients (5 ECoG, 8 SEEG). Within each patient, we calculated Spearman correlations and partial correlations controlling for node proximity (NP), then used exact sign-flip tests across patients and Holm correction across seven tests within each cohort. The unadjusted SCD–ECD association was positive in 12 of 13 patients (mean rho 0.32; authors’ 11-patient cohort p = 0.005, Holm p = 0.029). ECD and NP were strongly correlated (mean rho -0.62). After control for NP, mean partial rho fell to 0.12; it was not significant after Holm correction in the authors’ cohort (7/11 positive, p = 0.061, Holm p = 0.18), all 13 patients (9/13 positive, p = 0.029, Holm p = 0.087), or SEEG-only cohort (6/8 positive, p = 0.047, Holm p = 0.125). Fixed-effects OLS recovered the authors’ raw-unit ECD coefficient (0.125 versus 0.126), but its patient-cluster bootstrap 95% CI included zero (-0.047 to 0.295); a within-patient z-score model had a CI excluding zero (0.013 to 0.303), so significance depended on method. The reported direction was reproduced, while much of the association was consistent with shared dependence on proximity. We did not recompute tractography, CCEP N1 detection, or NP. Rank residualization assumes linear control, non-linear effects may remain, and 13 patients provide low power; a small partial association (rho about 0.1) cannot be excluded.

focal epilepsy · effective connectivity · structural connectivity · CCEP · DTI · node proximity · partial correlation · patient-level inference · sign-flip test