【摘要】 目的:根据黑加仑油软胶囊进行的多中心开放Ⅳ期临床试验数据,预测其对不同患者治疗高血脂的效果。方法:选入30～80岁符合西医高脂血症诊断标准的患者,服用黑加仑油软胶囊6周,对符合方案完成试验的2 154例患者分析血脂变化情况,将数据分为训练集和测试集,建立随机森林(RF)预测模型,采用内部交叉验证和外部数据进行验证。结果:筛选出的主要预测变量有治疗前的中医证候得分、血清总胆固醇(TC)、甘油三酯(TG)、体质量指数(BMI)、低密度脂蛋白胆固醇(LDL-C)、高密度脂蛋白胆固醇(HDL-C)、病程、体重、血压、运动情况等。使用这些变量预测服药是否"显效",交叉验证和外部验证的ROC曲线下面积(AUC)分别为0.827和0.857;按照外部验证数据最佳切点得到灵敏度和特异度分别为82.7%和72.8%,预测的正确率为77.9%。结论:黑加仑油软胶囊对高血脂证患者的治疗效果可以根据患者的病情、血脂水平、生命体征等情况进行预测,从而辅助临床决策。更多还原

【Abstract】 Objective:To predict the treatment effect for patients with hyperlipidemia based on the data from phase IV clinical trials of blackcurrant oil capsule in the hyperlipidemia treatment.Methods: Patients with age between 30 and 80 were enrolled in this study and were administrated with blackcurrant oil capsule for six weeks,then the lipids of 2154 patients before and after administration.These patients were divided into training sample and validation sample and construct random forest model with training sample to predict the discriminant effect with internal cross-validation and external validation sample.Results: The selected predictive variables included the traditional Chinese syndrome score,TC,TG,BMI,LDL-C,HDL-C,disease course,weight,blood pressure,exercises and so forth.The area under receiver operating characteristics(AUC)with cross-validation and external validation sample area 0.827 and 0.857,respectively.The sensitivity and specificity obtained from optimal threshold are 82.7% and 72.8%,the correct rate is 77.9%.Conclusions:The treatment effect for blackcurrant oil capsule can be predicted with factors,such as disease status,lipid level,vital signs and these results can assisted clinical decision-making.更多还原