Reliability in interpretation of median somatosensory evoked potentials in the setting of coma: factors and implications.

BACKGROUND: Median nerve short-latency somatosensory evoked potentials (SSEPs) are useful in determining prognosis for awakening after coma following hypoxic ischemic encephalopathy, but reliability of interpretation is unclear.
OBJECTIVE: To measure inter- and intra-observer reliability of determining presence or absence of SSEPs in comatose patients following hypoxic ischemic encephalopathy and to identify factors that enhance reliability.
DESIGN: Retrospective review by four readers (experience ranging from 2-27 years) of SSEP recordings in 95 comatose patients. Twenty waveforms were presented twice and five waveforms were presented pre- and post-neuromuscular junction (NMJ) blockade.
SETTING: Academic teaching hospital and level 1 trauma center. PATIENTS: Inpatients in coma due to hypoxic ischemic encephalopathy.
MEASUREMENTS AND MAIN RESULTS: Kappa values among the four readers, reflecting inter-observer reliability, ranged from 0.39 to 0.79 (mean 0.60, SD 0.16), a level of agreement interpreted as "moderate." Better inter-observer reliability was seen when: cortical amplitudes were ≥ 0.7 μV, baseline noise was smaller, and when experience level between reviewers was most similar. In cases examined both pre- and post-NMJ blockade, average kappa values of inter-observer reliability increased significantly from 0.03 pre-NMJ blockade to 0.41 post-NMJ blockade. Intra-observer reliability was higher than that for inter-observer, with a "substantial" average kappa of 0.84 (range 0.79-0.89, SD .06).
CONCLUSIONS: Inter- and intra-observer reliability of SSEP interpretation in comatose patients varies from moderate to substantial, respectively. In order to reliably interpret the presence of small cortical responses, NMJ blockade should be used when baseline noise is excessive.