Sasakawa Tomoki, Masui Kenichi, Kazama Tomiei, Iwasaki Hiroshi
Department of Anesthesiology, Asahikawa Medical University, Midorigaoka Higashi 2-1-1-1, Asahikawa, Hokkaido, Japan.
Department of Anesthesiology, National Defense Medical College, Namiki 3-2, Tokorozawa, Saitama, Japan.
J Anesth. 2016 Aug;30(4):620-7. doi: 10.1007/s00540-016-2174-5. Epub 2016 Apr 20.
Rocuronium concentration prediction using pharmacokinetic (PK) models would be useful for controlling rocuronium effects because neuromuscular monitoring throughout anesthesia can be difficult. This study assessed whether six different compartmental PK models developed from data obtained after bolus administration only could predict the measured plasma concentration (Cp) values of rocuronium delivered by bolus followed by continuous infusion.
Rocuronium Cp values from 19 healthy subjects who received a bolus dose followed by continuous infusion in a phase III multicenter trial in Japan were used retrospectively as evaluation datasets. Six different compartmental PK models of rocuronium were used to simulate rocuronium Cp time course values, which were compared with measured Cp values. Prediction error (PE) derivatives of median absolute PE (MDAPE), median PE (MDPE), wobble, divergence absolute PE, and divergence PE were used to assess inaccuracy, bias, intra-individual variability, and time-related trends in APE and PE values.
MDAPE and MDPE values were acceptable only for the Magorian and Kleijn models. The divergence PE value for the Kleijn model was lower than -10 %/h, indicating unstable prediction over time. The Szenohradszky model had the lowest divergence PE (-2.7 %/h) and wobble (5.4 %) values with negative bias (MDPE = -25.9 %). These three models were developed using the mixed-effects modeling approach. The Magorian model showed the best PE derivatives among the models assessed.
A PK model developed from data obtained after single-bolus dosing can predict Cp values during bolus and continuous infusion. Thus, a mixed-effects modeling approach may be preferable in extrapolating such data.
使用药代动力学（PK）模型预测罗库溴铵浓度，对于控制罗库溴铵的效应将很有用，因为在整个麻醉过程中进行神经肌肉监测可能很困难。本研究评估了仅根据单次静脉推注给药后获得的数据开发的六种不同房室PK模型，能否预测在静脉推注后持续输注罗库溴铵时测得的血浆浓度（Cp）值。
回顾性地将来自19名健康受试者的罗库溴铵Cp值用作评估数据集，这些受试者在日本的一项III期多中心试验中接受了单次静脉推注剂量后持续输注。使用六种不同的罗库溴铵房室PK模型来模拟罗库溴铵Cp随时间变化的值，并与测得的Cp值进行比较。使用中位绝对预测误差（MDAPE）、中位预测误差（MDPE）、摆动度、离散绝对预测误差和离散预测误差的预测误差（PE）导数，来评估预测误差（APE）和PE值的不准确程度、偏差、个体内变异性以及与时间相关的趋势。
仅Magorian模型和Kleijn模型的MDAPE和MDPE值是可接受的。Kleijn模型的离散预测误差值低于-10%/小时，表明随时间的预测不稳定。Szenohradszky模型的离散预测误差值最低（-2.7%/小时），摆动度值（5.4%）且具有负偏差（MDPE = -25.9%）。这三个模型是使用混合效应建模方法开发的。在评估的模型中，Magorian模型显示出最佳的PE导数。
根据单次静脉推注给药后获得的数据开发的PK模型，可以预测静脉推注和持续输注期间的Cp值。因此，在推断此类数据时，混合效应建模方法可能更可取。
