Fig. 1. Estimated probability of electroencephalographic seizure activity as a function of glucose concentration at T7, 90 min after the cessation of circulatory arrest (CA) or low-flow cardiopulmonary bypass (LF). Logistic regression curves are shown by treatment group; the P value adjusts for treatment group and diagnosis of ventricular septal defect or intact ventricular septum. The area under the receiver–operator characteristic curve corresponding to this model is 0.764 and quantifies how well the model is able to discriminate between patients who did and did not experience electroencephalographic seizures; an area of 0.5 indicates a model with no predictive power, whereas a model that perfectly predicts outcome every time has an area of 1.0. Point estimates and exact 95% confidence intervals for the probability of seizures are plotted for the mean of each quartile of glucose concentration.