Performance of propofol target-controlled infusion models in the obese: pharmacokinetic and pharmacodynamic analysis.

Abstract

Obesity is associated with important physiologic changes that can potentially affect the pharmacokinetic (PK) and pharmacodynamic (PD) profile of anesthetic drugs. We designed this study to assess the predictive performance of 5 currently available propofol PK models in morbidly obese patients and to characterize the Bispectral Index (BIS) response in this population. Twenty obese patients (body mass index >35 kg/m), aged 20 to 60 years, scheduled for laparoscopic bariatric surgery, were studied. Anesthesia was administered using propofol by target-controlled infusion and remifentanil by manually controlled infusion. BIS data and propofol infusion schemes were recorded. Arterial blood samples to measure propofol were collected during induction, maintenance, and the first 2 postoperative hours. Median performance errors (MDPEs) and median absolute performance errors (MDAPEs) were calculated to measure model performance. A PKPD model was developed using NONMEM to characterize the propofol concentration-BIS dynamic relationship in the presence of remifentanil. We studied 20 obese adults (mean weight: 106 kg, range: 85-141 kg; mean age: 33.7 years, range: 21-53 years; mean body mass index: 41.4 kg/m, range: 35-52 kg/m). We obtained 294 arterial samples and analyzed 1431 measured BIS values. When total body weight (TBW) was used as input of patient weight, the Eleveld allometric model showed the best (P < 0.0001) performance with MDPE = 18.2% and MDAPE = 27.5%. The 5 tested PK models, however, showed a tendency to underestimate propofol concentrations. The use of an adjusted body weight with the Schnider and Marsh models improved the performance of both models achieving the lowest predictive errors (MDPE = <10% and MDAPE = <25%; all P < 0.0001). A 3-compartment PK model linked to a sigmoidal inhibitory Emax PD model by a first-order rate constant (ke0) adequately described the propofol concentration-BIS data. A lag time parameter of 0.44 minutes (SE = 0.04 minutes) to account for the delay in BIS response improved the fit. A simulated effect-site target of 3.2 μg/mL (SE = 0.17 μg/mL) was estimated to obtain BIS of 50, in the presence of remifentanil, for a typical patient in our study. The Eleveld allometric PK model proved to be superior to all other tested models using TBW. All models, however, showed a trend to underestimate propofol concentrations. The use of adjusted body weight instead of TBW with the traditional Schnider and Marsh models markedly improved their performance achieving the lowest predictive errors of all tested models. Our results suggest no relevant effect of obesity on both the time profile of BIS response and the propofol concentration-BIS relationship.