Abstracts

Rationale: Hypsarrhythmia is the classic interictal electroencephalographic pattern associated with infantile spasms. It is generally considered an all-or-none phenomenon, and most definitions require (1) high voltage, (2) disorganization, and (3) multifocal independent epileptiform discharges. Given this seemingly simple definition, and that hypsarrhythmia is among the more dramatic electroencephalographic abnormalities in existence, one might expect excellent inter-rater reliability (IRR) in the clinical identification of this pattern. Alternatively, it may be argued that assessments of voltage and disorganization are fairly subjective, and thus quite challenging in borderline cases. We sought to test the IRR of hypsarrthymia assessments in a systematic fashion.Methods: We selected 25 EEGs from patients with infantile spasms. In an effort to represent routine clinical practice, we chose EEGs from patients at various stages of treatment, including untreated patients with a new diagnosis, as well as patients with varying degrees of success following treatment. Each EEG clip was 5 minutes in duration, included only wakefulness, and did not include seizures. No video was provided. Three experienced pediatric encephalographers independently reviewed all 25 EEGs in random order. Patients ages were provided to the readers, but they were otherwise blinded as to patient identity, diagnoses, and treatment status. For each EEG, readers were asked to determine if the study was normal/abnormal, whether hypsarrhythmia was present/absent, and to characterize relevant features: voltage, organization, slowing, epileptiform discharges, interictal attenuations, symmetry, and synchrony. Free-marginal multirater Kappa statistics were calculated for the assessment of hypsarrhythmia and each feature.Results: Although IRR was adequate in determining whether a study was normal or abnormal (kappa = 0.95), reliability was marginal for assessment of hypsarrhythmia (kappa = 0.63) and high voltage (kappa = 0.57), and 'poor' for the identification of disorganization (kappa = 0.34), multifocal epileptiform discharges (kappa = 0.47), interictal voltage attenuations (kappa = 0.41), slowing (kappa = -0.01), asymmetry (kappa = 0.25), and asynchrony (kappa = 0.36).Conclusions: This study contradicts the view that hypsarrhythmia assessment is straightforward. Even small variability in the identification of hypsarrhythmia has potentially deleterious consequences for clinical care, as its presence or absence impacts decisions to pursue high-risk and high-cost therapies. These inconsistencies may similarly confound studies in which abolition of hypsarrhythmia is an outcome measure. There is a great need for practical, reliable, and unbiased measures of hypsarrhythmia that can be applied in both research protocols and routine clinical care.