Jackknife分析法在环孢素AUC_(0-12)预测模型中的应用

Application of jackknife analysis in AUC_(0-12) predictive models of cyclosporine

目的用jackknife分析法对环孢素AUC0-12预测模型进行验证和评价,为临床提供简便实用的AUC0-12预测模型。方法用酶多种免疫测定技术(EMIT)测定16例肝、肾移植患者服用环孢素后的经时血药浓度,建立AUC0-12与各血药质量浓度效应变量间的回归模型。用jackknife分析法验证和评价这些模型预测AUC0-12的准确性和精密性。结果ρ0、ρ2、ρ6、ρ9与AUC0-12相关性极为显著(P<0.01),ρ0+ρ2与AUC0-12相关性高于任一单点。各验证模型的预测值与实测值没有差异(P>0.05)。由验证参数判别各模型预测的准确性和精密性顺序为:三点模型、二点模型和单点模型。结论从实际操作和成本的角度考虑,二点模型更为实用,尤其是ρ0+ρ2简化模型。

AIM To validate and evaluate AUC0 predictive models of cyclosporine with jackknifeanalysis, and so as to provide simple and practical AUC0 12 predictive models for clinical work. METHODS The whole blood cyclosporine concentration of 16 patients with liver and kidney transplantation after taking cyclosporine by enzyme multiplied immunoassay technique （EMIT） were determined, and regression models between AUC0 12 and various points whole blood cyclosporine concentration were founded. Accuracy and precision of these model that predicted AUC0_~2 were validated and evaluated with jackknife analysis. RESULTS Correlation between po, p2, p6, ,09 and AUC0_j2 was very significant （P 〈 0.01 ）. The correlation between po ＋ p2 and AUC0-12 was higher than any single point. The predicted and observed values of various validation models had no difference （P 〉 0.05）. The predicted accuracy and precision of these models were discriminated by validation parameters and the order was three point predictive model, two point predictive model and single point predictive model. CONCLUSION Considering actual operation and cost, two point predictive model was more practical, especially simplified p0 ＋ p2 predictive model.