Table 2. Improvement in the Objective Function by Introducing Patients’ Age and Presence of Surgical Stimulation as Covariates for the Desflurane Concentration–Electroencephalographic Effect Relation in the Initial Models

The search for the final models started with an initial model for each electroencephalographic parameter, built by using the same parameter set for both groups, with and without surgical stimulation. The parameter set contained fixed and random (variances) effect parameters for k^e0, E⁰, EC⁵⁰, E^max, λ, and the error variance. The initial models served as the null hypothesis for a statistical comparison with models, including covariates. Presence or absence of surgery and patients’ age were introduced as covariates and kept in the models if they significantly improved the objective function. Models including presence of surgery as a covariate allowed a shift of the concentration–electroencephalographic effect curve toward higher concentrations by introducing the separate parameter EC⁵⁰and a greater steepness of the concentration–electroencephalographic effect curves by introducing the separate parameter λ for the patients during surgical stimulation. Age as a covariate was modeled using the EC⁵⁰in an age-dependent manner. The final estimated models’ parameters for spectral edge frequency 95 (SEF 95) are k^e0, η^{ke0, E0during surgery, Emax, EC50without surgery, EC50during surgery, γ without surgery, γ during surgery, age correction on EC50, ηEmax, ηEC50, ηγ, and ς2}. The final estimated models’ parameters for the median power frequency (MPF) are k^e0, η^{ke0, E0during surgery, Emax, EC50without surgery, EC50during surgery, γ, ηEmax, ηEC50, ηγ, and ς2}. The final estimated models’ parameters for the Bispectral Index (BIS) are k^e0, η^{ke0, EC50without surgery, EC50during surgery, γ without surgery, γ during surgery, ηEC50, ηγ, and ς2}. The minimum value of the objective function is minus twice the logarithm of the likelihood of the data, calculated by the computer program NONMEM.