METHODS: PubMed was searched for IFX PopPK models in children. Selected models were rebuilt and analysed using R. Model performance was assessed through goodness-of-fit-plots, residuals against time, prediction error and prediction-corrected visual predictive checks. The validation cohort consisted of 73 children with IBD who were treated with IFX in our centre between 2017 and 2023 (340 IFX measurements).
RESULTS: We identified 9 PopPK models. Model bias for individual predicted values ranged from -9.29% to 8.01% compared to bias for population predicted values. The model by Vande Casteele et al. demonstrated superior performance (individual predicted bias 2.13, population predicted bias -6.11); upon Bayesian estimation, it predicted induction trough levels with median error of 12.95% but had a median error of -69% predicting maintenance concentrations.
CONCLUSIONS: The model by Vande Casteele et al. displayed superior performance in initial evaluations but had a high error in estimating next IFX levels and can only be used in practice to predict induction levels.
方法:在儿童中搜索PubMed的IFXPopPK模型。使用R重建和分析选定的模型。通过拟合优度图评估模型性能,相对于时间的残差,预测误差和预测校正的视觉预测检查。验证队列包括2017年至2023年在我们中心接受IFX治疗的73例IBD儿童(340例IFX测量)。
结果:我们确定了9个PopPK模型。与人群预测值的偏差相比,个体预测值的模型偏差范围为-9.29%至8.01%。VandeCasteele等人的模型。表现优异(个体预测偏差2.13,群体预测偏差-6.11);经贝叶斯估计,它预测诱导谷水平的中位误差为12.95%,但预测维持浓度的中位误差为-69%.
结论:VandeCasteele等人的模型。在初始评估中表现出优异的性能,但在估计下一个IFX水平时误差很大,只能在实践中用于预测诱导水平。