The comparison of predictive performance in bispectral index prediction during target effect-site controlled infusion of propofol using different blood effect-site equilibration rate constants in the same pharmacokinetic model

BACKGROUND: Blood-brain equilibration rate constant (k(e0)) is derived from either pharmacokinetic and pharmacodynamic modeling (k(e0_model)) or a model-independent observed time to peak effect (k(e0_tpeak)). Performance in bispectral index (BIS) prediction was compared between k(e0_model) and k(e0_tpeak) for microemulsion or long chain triglyceride (LCT) propofol. METHODS: Time to peak effect (t(peak), time to a maximally reduced BIS value) of microemulsion propofol after an intravenous bolus (1 mg/kg) was measured in 100 patients (group A(micro)). An observed t(peak) of 1.6 min for LCT propofol was obtained from an earlier study. Another 40 patients received a target controlled infusions of microemulsion propofol (k(e0_model) = 0.187/min, group B(micro) = 20) or LCT propofol (k(e0_model) = 0.26/min, group B(LCT) = 20) and remifentanil. The k(e0_tpeak)'s in group B(micro) and B(LCT) were calculated using the observed t(peak) value obtained from group A(micro) and 1.6 min, respectively. Effect-site concentrations of propofol were recalculated using the amounts of propofol infused over time and k(e0_tpeak)'s. Predicted BIS values calculated by sigmoid Emax equations with k(e0_model) and k(e0_tpeak) were compared with observed BIS values during induction and emergence for both formulations of propofol. RESULTS: Observed t(peak) of microemulsion propofol was 1.68 min. The median performance errors of BIS in group B(micro) were -1.83% (-24.8 to 18.9, k(e0_model)) and -2.42% (-26.1 to 36.2, k(e0_tpeak)), while 8.01% (-20.5 to 30.1, k(e0_model)) and 7.37% (-27.0 to 49.1, k(e0_tpeak)) in group B(LCT). The median absolute performance errors of BIS in group B(micro) were 11.87% (2.2-31.1k(e0_model)) and 14.38% (-0.6 to 44.6, k(e0_tpeak)), while 17.31% (5.54-36.0, k(e0_model)) and 18.28% (-0.1 to 56.0, k(e0_tpeak)) in group B(LCT). CONCLUSIONS: The k(e0_model) showed better performance in BIS prediction than the k(e0_tpeak).