基于双能量CT成像建立预测痛风患者尿酸盐沉积的列线图模型

Establishment of the Nomogram Model for Predicting Urate Deposition in Patients with Gout Based on Dual-energy CT Imaging

目的:探究痛风患者发生尿酸盐沉积(uratedeposition,UD)的危险因素及风险列线图模型的建立。方法:将246例痛风患者根据双能量CT成像结果分为发生UD组及未发生UD组,采用单因素及多因素logistic回归分析筛选独立危险因素,然后利用列线图建立风险模型,使用C-index及ROC曲线评价预测模型的准确度,并使用自举验证评估内部验证。结果:通过对两组患者一般临床资料进行Logistic回归分析可知,年龄、饮酒、高血脂、肾结石、痛风病程、血尿酸水平为痛风患者发生尿酸盐沉积的独立危险因素,基于筛选出的6项独立危险因素,建立预测痛风患者发生尿酸盐沉积风险的列线图模型。通过对该模型进行验证表明:预测值和观察值基本一致,说明本研究的列线图预测模型具有较好的预测能力,同时本研究使用Bootstrap内部验证法对痛风患者发生尿酸盐沉积的列线图模型进行验证,C-index指数高达0.810(95%CI:0.720~0.900),说明本研究列线图模型具有良好的精准度和区分度。结论:对痛风患者及时地考虑年龄、饮酒、高血脂、肾结石、痛风病程、血尿酸水平等因素综合评估患者发生尿酸盐沉积的发生率,具有较高的临床应用价值,值得进一步推广使用。

Objective: To explore the risk factors of urate deposition(UD) in patients with gout and the establishment of the risk nomogram model. Methods: 246 patients with gout are divided into UD group and non-UD group according to results of the dualenergy CT imaging. Independent risk factors are screened by single-factor and multi-factor logistic regression analysis, and then the risk model is established by nomogram. The accuracy of the prediction model is evaluated by C-index and ROC curve, and the internal validation is evaluated by Bootstrap validation. Results: The Logistic regression analysis of the general clinical data of the two groups of patients shows that age, alcohol consumption, hyperlipidemia, kidney stones, course of gout and serum uric acid level are independent risk factors for UD in patients with gout. Based on the 6 independent risk factors screened out, the nomogram model is established to predict the risk of UD in patients with gout. The validation of the model shows that the predicted value and the observed value are basically consistent, which indicates that the prediction model of nomogram in this study has good prediction ability. At the same time, Bootstrap internal validation method is used in this study to verify the nomogram model of UD in patients with gout. C-index is as high as 0.810(95% CI: 0.720~0.900), indicating that the nomogram model in this study has good accuracy and discriminability. Conclusion: The timely consideration of age, alcohol consumption, hyperlipidemia, kidney stones, course of gout, serum uric acid level and other factors in patients with gout to comprehensively evaluate the incidence of UD in them has high clinical application value and is worthy of further promotion and use.