Conventional EEG (CEEG) in neonates is considered the gold standard for evaluating EEG background and detecting electrographic seizures. However, CEEG is expensive and cumbersome for long-term monitoring. A simplified method, amplitude-integrated EEG (AEEG) has been rapidly adopted to accomplish the same goals. The purpose of this study was to measure the agreement between the methods of classification in long-term EEG background assessments by CEEG and AEEG. Infants underwent CEEG monitoring after cardiac surgery and the background during four 12-hour epochs classified as “normal” or “mildly,” “moderately,” or “markedly” abnormal. CEEGs were converted to a single-channel AEEG and independently interpreted as “normal,” “moderately abnormal,” or “markedly abnormal” by standard amplitude criteria. The distributions of CEEG and AEEG interpretations were statistically compared, and the associations between CEEG and AEEG interpretations were measured. Generalized estimating equations were used to measure the effects of seizures and patient age on the agreement between AEEG and CEEG scores. Paired CEEGs and AEEGs were available for 637 epochs recorded from 179 infants. The distribution of CEEG backgrounds included 60% normal, 22% mildly abnormal, 13% moderately abnormal, and 5% markedly abnormal. The distribution of AEEG backgrounds was significantly different from CEEG and included 22% normal, 73% moderately abnormal, and 5% markedly abnormal. Nevertheless, the two techniques exhibited a significant, moderate positive association. Generalized estimating equations focusing on those with moderately abnormal AEEGs showed that younger patients with seizures were significantly more likely to have moderately or markedly abnormal CEEGs than older patients without seizures. Although there was overall significant moderate agreement between the two techniques, the distribution of backgrounds assigned by AEEG was significantly different from CEEG. Most moderately abnormal AEEGs were associated with normal or mildly abnormal CEEGs. However, the ability of moderately abnormal AEEGs to correctly predict moderately or markedly abnormal CEEG was significantly associated with the knowledge of the patient's age and the presence of seizures on CEEG.
From the *The Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, U.S.A.; †Harvard School of Public Health, Boston, Massachusetts, U.S.A.; and ‡George Washington University School of Medicine, Washington, DC, U.S.A.
Supported in part by grants from an American Heart Association National Grant-in-Aid (9950480), the Fannie E. Rippel Foundation, and an AATS Summer Intern Scholarship.
Address correspondence and reprint requests to Robert Clancy, Division of Neurology, Children's Hospital of Philadelphia, 34th Street and Civic Center Boulevard, Philadelphia, PA 19104, U.S.A.; e-mail: firstname.lastname@example.org.