华法林稳定剂量预测模型对新疆肺栓塞患者预测准确性的验证与评价

Validation and evaluation of the predictive accuracy of warfarin stable dose prediction algorithm for patients with pulmonary embolism in Xinjiang

目的验证与评价12个华法林稳定剂量预测模型对肺栓塞患者的预测准确性。方法收集已达到华法林稳定剂量的131例肺栓塞患者完整临床资料,采集每位患者外周血2 mL,用聚合酶链反应-基因芯片杂交技术检测其细胞色素P450酶2C9(CYP2C9)和维生素K环氧化物还原酶复合体亚单位1(VKORC1)基因型,并对患者基因型分布特征和实际稳定剂量与不同基因型之间的差异进行分析。挑选出符合标准的12个华法林稳定剂量预测模型,用绝对误差均值和预测百分比评价并比较各模型预测准确性。结果 131例患者的VKORC1 AA型、GA型及GG型华法林实际稳定剂量分别为(2.76±0.78),(3.57±0.78)和(4.67±0.74)mg·d^(-1),组间比较差异有统计学意义(P<0.001)。除去CYP2C9中只有1例患者的基因型,在129例患者的实际稳定剂量中,CYP2C9*1/*1型患者高于*1/*3型患者,差异有统计学意义(P<0.05)。在12个华法林稳定剂量预测模型中,MAE最低的3个模型是都丽萍、SCONCE和吴超君,理想百分比最高的3个模型是都丽萍、吴超君和OHNO。敏感性分析显示,在8个模型的理想预测百分比大于40%中,稳定剂量预测模型对高、中、低3个剂量患者预测准确性依次下降。结论对高、中2个剂量的肺栓塞患者,都丽萍模型有很好的预测准确性;而吴超君模型相较于其他模型,对低剂量患者有较好的预测准确性。

Objective To validate and evaluate the predictive accuracy of 12 warfarin stable dose prediction algorithms for pulmonary embolism patients in the First Affiliated Hospital of Xinjiang Medical University. Methods The complete clinical data of 131 patients with stable warfarin dose were collected. Two mL of peripheral blood was collected from each patient, and the cytochrome P450 2 C9( CYP2 C9),vitamin K epoxide reductase complex 1(VKORC1)genotypes were detected by polymerase chain reation-microarray hybridization technology. Genotypes distribution characteristics and actual stable dose with differences between different genotypes were analyzed.Twelve stable dose prediction algorithms of warfarin were selected through previous publications. Mean absolute error(MAE) and prediction percentage were used to evaluate and compare the prediction accuracy of each algorithm. Results Among the 131 patients,the actual stable dose of warfarin of VKORC1 AA,GA and GG were( 2. 76 ± 0. 78),( 3. 57 ± 0. 78)and( 4. 67 ± 0. 74) mg · d^(-1),with statistically significant difference between groups( P < 0. 001). Remove the genotype of only one patient in CYP2C9,among the 129 patients,the actual stable dose of warfarin in CYP2C9* 1/* 1 genotype was higher than that in * 1/* 3 genotype,with significant differences( P < 0. 05). Among the 12 warfarin stable dose prediction algorithms,the MAE of DU Li-ping,SCONCE and WU Chao-jun in algorithms were lowest,and the ideal percentage of DU Li-ping,WU Chao-jun and OHNO in algorithms were highest. The ideal prediction percentage of 8 algorithms was greater than 40%,and the sensitivity analysis showed that the accuracy of stable dose prediction model for high,medium and low dose patients declined in turn. Conclusion DU Li-ping model has a good prediction accuracy for patients with pulmonary embolism in Xinjiang with high and medium doses,and WU Chao-jun model has better predictive accuracy than other models for patients with low doses.