验证并比较华法林稳定剂量预测模型对中国心脏瓣膜置换术后患者预测准确性

目的 验证并比较12个华法林稳定剂量预测模型及目前国内常用固定剂量给药方案(2.5 mg/d)对本院心脏瓣膜置换术后患者的预测准确性。方法 收集已达华法林稳定剂量的804例人工心脏瓣膜置换术后患者完整临床资料,采集每位患者外周血 2 mL,焦磷酸测序法检测其VKORC1-1639G>A和CYP2C9^*3基因型。Pubmed检索出符合筛选条件的12个华法林稳定剂量预测模型。采用绝对误差均值(MAE)和预测百分比两个指标评价并比较各模型预测准确性。结果 共有9个预测模型及固定剂量方案计算的MAE小于±1.0 mg/d,其中MAE最低的三个模型分别为Gage、Wen和Ohno。共有8个模型及固定剂量方案计算理想预测百分比大于40%,其中理想预测百分比最高的三个模型分别为Wen、Huang和Gage。敏感性分析表明预测模型对中剂量(稳定剂量>1.88,且<4.38 mg/d)和高剂量(稳定剂量≥4.38 mg/d)患者预测准确性较好,但对低剂量患者(稳定剂量≤1.88 mg/d)预测较差;固定剂量方案对高、低剂量患者预测准确性均为0。结论 基因导向的华法林稳定剂量预测模型可以较好预测大部分患者,具有临床应用价值。

Validation and comparison of warfarin stable dosage prediction algorithms in Chinese patients undergoing heart valve replacement

AIM: To validate and compare 12 previously published warfarin stable dosage prediction algorithms and the fixed dosing regimen (2.5 mg/d) as applied to our patients undergoing heart valve replacement (HVR). METHODS: 804 Chinese HVR patients on stable maintenance dose of warfarin were enrolled from our hospital. Comprehensive clinical data were collected. 2 mL blood sample was drawn from each patient for VKORC1-1639G>A and CYP2C9^*3 genotype analyses by pyrosequencing method. 12 previously published algorithms were selected from Pubmed database for validation and comparison. The performance of all the algorithms was determined by mean absolute error (MAE) and percentage of ideal/under/over prediction.RESULTS: 9 algorithms and the fixed dosing method showed a MAE less than±1.0 mg/d, of which the Gage, Wen and Ohno algorithms had the lowest MAE. 8 algorithms and the fixed dosing regimen showed more than 40% of ideal prediction, of which the Wen, Huang and Gage algorithms showed the highest percentage of ideal prediction. Sensitivity analysis demonstrated that these algorithms had better prediction for the medium-dose (>1.88 and <4.38 mg/d) and high-dose (≥4.38 mg/d) patients but showed poor performance for the low-dose (≤1.88 mg/d) patients.The ideal prediction of the fixed dosing regimen for both low-dose and high-dose patients was 0.CONCLUSION: Genotype-guided warfarin dosage prediction algorithms could accurately predict most of our patients, therefore these algorithms may be potentially useful in clinical practice.