Is there inter-observer variation in the interpretation of SSEPs in comatose cardiac arrest survivors? Further considerations following the Italian multicenter ProNeCa study

Background: Bilateral absence of N20 peak in median nerve Somatosensory Evoked Potentials (SSEPs) is considered the most valid predictor of poor outcome in comatose survivors after cardiopulmonary resuscitation. We investigated the consistency in interpreting SSEP recordings in a multicentre study. Methods: 44 SSEP recordings randomly extracted from 600 recordings of 392 patients included in the "Prognostication of Neurological outcome after Cardiac Arrest (ProNeCa) study" were blindly read by three expert neurophysiologists. Agreement between raters, and individual agreement of each rater vs. reference standard (RS), were calculated using Kappa Coefficients. Inter-rater reliability was calculated with Intra-class Correlation Coefficient (ICC). Results: When raters had to evaluate the presence of N20 with normal amplitude, the inter-rater agreement was very high (Kappa = 0.84). In the case of N20 absence the agreement was good (Kappa = 0.66), but when N20 amplitude was low, the agreement decreased to moderate (Kappa = 0.579) becoming even weaker when it was "Non Assessable" (Kappa = 0.107). The agreement of each rater with the RS had a range from moderate to very good; rater1 Kappa = 0.589 (95%CI 0.397-0.781; p < 0.001), rater2 Kappa = 0.644 (95%CI 0.460-0.828; p < 0.001), rater3 Kappa = 0.859 (95%CI 0.698-1.000; p < 0.001). The ICC was barely good, 0.682 (95%CI 0.539-0.798; p = 0.0075). Conclusion: Different health professionals, using different equipment in a multicentre study, had very good inter-rater agreement in interpreting SSEP records. The interpretation of "Non Assessable" SEPPs, mainly in relation to noise level, is still a crucial issue because it increases rater uncertainty. For this reason, it is important to focus on improving recording quality and interpretation of records.