Fig. 3. Prediction probabilities (PKs) of different predictors and predicted variables. The PKvalues according to Smith et al. 12were estimated from 1,000 random samples of 263 data points from 44 patients, with 1 data point per patient in each sample. The data are presented as box plots with median values depicted as horizontal lines and the interquartile range (lower and upper limits of the boxes ). The error bars represent the 10 and 90 percentiles, and the black circles the 5 and 95 percentiles. Pairwise comparison of the PKvalues by comparing the median PKof one predictor with the 2.5–97.5 percentile range of another predictor (corresponding to the 95% confidence interval). If the median PKof the first is outside this percentile range, it is considered significantly different. A : observer assessment of alertness and sedation score (OAAS), B : loss of eyelash reflex, C : loss of motor response tetanic stimulation, D : bispectral index (BIS), and E : acoustic evoked potential index (AAI).*= Significantly different compared with noxious stimulation response index (NSRI), probability to tolerate laryngoscopy (P TOL), and effect-site remifentanil concentration (Remi); # = Significantly different compared with effect-site propofol concentration (Prop), BIS, AAI, and Remi.