Abstract
OBJECTIVE: Quantitative calculation models for the ratio of daily area under the concentration-time curve (AUC(24)) to the minimum inhibitory concentration (MIC) (i.e., AUC(24)/MIC) and the amount of time that concentration stays above the MIC during a dosing interval (i.e., T(>MIC)%) in regular intermittent i.v. infusion (RIIVI) are currently absent. This work set out to construct the models of AUC(24)/MIC, T(>MIC)% and matching daily dosage (D(d)) in RIIVI, and further examine their performance by comparing with the documented models currently used widely, and concomitantly create a closed loop for evaluating the original scheme's effectiveness and developing the personalized dosing regimen using these established models. METHODS: 1-compartment model was used to construct the AUC(24)/MIC, T(>MIC)% and matching D(d) models. 20 designed individuals with different renal functions in different clinical scenarios were employed to examine the models. Bland-Altman plots and Bootstrap analysis were applied to assess the consistency, and the prediction reliability and accuracy of the models in calculating AUC(24)/MIC and T(>MIC)%, respectively. Tornado method based on global sensitivity analysis was used to perform the sensitivity analysis of the models to examine the effect of parameter variation on predictions. Combining the AUC(24)/MIC or T(>MIC)% model-based efficacy assessment with the D(d) model-based regimen optimization to creates a closed loop consisting of efficacy assessment and regimen optimization. RESULTS: The AUC(24)/MIC, T(>MIC)% and D(d) models in RIIVI were developed. Bland-Altman plots and Bootstrap analysis indicated that the established and the documented models had no consistency and the established models had better prediction reliability and accuracy in calculating AUC(24)/MIC and T(>MIC)%. Sensitivity analysis suggested that MIC was an important factor on AUC(24)/MIC and T(>MIC)% variation. Cooperative application of the AUC(24)/MIC, T(>MIC)% and D(d) model created a closed loop consisting of efficacy assessment and regimen optimization for creation of customized antibiotic regimens. CONCLUSIONS: The established AUC(24)/MIC and T(>MIC)% models displayed better performance relative to the documented models. Cooperative application of these models and the corresponding D(d) model can create a fully closed loop for evaluating the original scheme's effectiveness and developing the optimization regimen, and thus construct a basic framework for the creation of customized antibiotic regimens.