狼疮肾炎患者肾脏预后预测列线图的建立

A nomogram to predict the renal prognosis of patients with lupus nephritis

目的探讨利用狼疮肾炎患者的临床、病理指标构建列线图，用于预测患者的肾脏预后。方法回顾性分析2005年1月1日至2015年10月1日在中山大学附属第一医院确诊为狼疮肾炎并有肾活检资料的患者数据。肾脏终点定义为：终末期肾脏病（ESRD）[eGFR〈15ml·min^-1·（1．73m^2）^-1、维持性透析或肾移植]或血肌酐翻倍。以单因素COX回归对患者的人口学资料、临床及实验室资料、肾脏病理资料进行分析；根据多因素COX回归分析结果、赤池信息量准则及连续性重分类改善指数（cNRI）筛选模型并构建预测患者5年及8年肾脏预后的列线图，并以一致性指数（C—index）、标准校正曲线评价模型的效能。结果共纳入513例狼疮肾炎患者，中位随访时间为48（24，71）个月，44例（8．58％）到达。肾脏终点。患者1年、5年、8年无肾脏终点事件率分别为97．57％、92．89％、79．89％。拟合最优COX回归模型纳入变量：eGFR小于30ml·min^-1·（1．73m^2）。（HR=4．44，95％CI2．16～9．13，P〈0．01）、肾脏活检病理中球性硬化。肾小球比例高（HR=12．28，95％C13．58～42．13，P〈0．01）和诱导治疗6个月未达到部分缓解（HR=9．16，95％CI4．71～17．82，P〈0．01）。预测5年、8年无终点事件肾脏存活率C-index分别为0．88（95％C10．78～0．92）和0．76（95％C10．68—0．85），校正曲线接近45。参考线。结论基于患者基线肾小球滤过率、肾活检中肾小球硬化比例及完成6个月诱导治疗时对治疗的反应构建的狼疮肾炎患者肾脏预后的预测列线图，有助于临床实践中预测患者的肾脏预后。

Objective To develop a nomogram for the use of predicting renal outcomes of Chinese lupus nephritis （LN） patients. Methods From January 1, 2005 to October 1, 2015, 513 patients with biopsy-proven LN in the First Affiliated Hospital of Sun Yat-Sen University were enrolled into this study. Renal outcomes were defined as end- stage renal disease or doubling of serum creatinine. Demographic characteristics, laboratory data, and pathologic data were recorded and included for analysis, Nomograms were designed using multivariate Cox proportional hazards regression to predict the non-outcome renal survival in 5 and 8 year according to the Akaike information criterion （AIC） and continuous reclassification net improvement （cNRI）. Predictive accuracy and discriminative ability of the models were determined by concordance index （C-index） and calibration curve. Results During a median follow up of 48 （24,71） months, 44 patients （8.58%） reached the endpoint. 1-year, 5- year and 8-year non-outcome renal survival were 97.57%, 92.89%, 79.89% respectively. According to multivariate Cox regression, four nomograms including index for baseline renal function, pathologic severity, and response to treatment were designed. The best model, within which included eGFR was lower than 30 ml·min^-1·（1.73 m^2）^-1（HR=4.44, 95% CI 2.16- 9.13, P 〈 0.01）, percentage of global glomerulosclerosis was higher （HR=12.28, 95%CI 3.58-42.13, P 〈 0.01） and partial remission occurred after 6 - month induction treatment （HR=9.16, 95% CI 4.71 - 17.82, P 〈 0.01） demonstrated good discrimination to predict 5-year and 8-year non-outcome renal survival [C-index, 0.80（95%CI 0.81- 0.91）, 0.76（95%CI 0.68-0.85）, respectively]. The nomogram based on above model also performed good calibration. Conclusion The nomogram based on patients＇ baseline eGFR, percentage of global glomerulosclerosis, and treatment reaction after 6- month induction therapy can accurately predict 5- year and 8-year non-outcome renal survival in Chinese lupus nephritis patients.