【作者】 张骥；

【导师】 卢国元；

【作者基本信息】 苏州大学 ， 内科学（肾脏病学）（专业学位）， 2019， 博士

【摘要】 背景:膜性肾病(membranous nephropathy,MN)是一个通过肾脏病理形态学诊断的疾病,其病理特征性是肾小球毛细血管袢上皮侧大量免疫复合物沉积,肾小球基底膜破坏;它也是成人肾病综合征常见的原因之一,临床的主要特征为常表现为显著的蛋白尿、低蛋白血症和高脂血症。目前我国的流行病学调查显示MN发病率呈现逐年增高趋势,已经引起肾脏病医师的广泛的重视。MN按照致病原因可以分为2类,第一类为继发性膜性肾病,约占总MN的20%,由系统性疾病或暴露于某些因素导致;第二类为特发性膜性肾病(idiopathic membranous nephropathy,IMN),也称为原发性膜性肾病,约占80%,其病变仅局限于肾脏。相对于继发性膜性肾病,IMN的治疗更为棘手。虽然既往的研究认为IMN是一种“良性疾病”,但是仍有30%-40%的IMN患者,特别是那些表现为持续性大量蛋白尿且伴有肾功能受损的患者,可能在5-15年内进入终末期肾病。目前IMN的治疗仍有争议,尽管已经有多种免疫抑制药物可以用来治疗IMN,KDIGO指南给出了免疫抑制治疗的适应证和禁忌证,但是由于免疫抑制剂存在显著的副作用,需要谨慎使用免疫抑制治疗,并且KDIGO指南也指出需要寻找合适的临床、病理和生物学标志物来预测IMN的预后情况,以筛选合适的患者,给予积极治疗,从而改善IMN预后及减少免疫抑制治疗带来的副反应。然而早期预测IMN的肾脏预后仍具有较大的挑战性,需要更加广泛和深入的研究和验证。第一部分特发性模型肾病不良肾脏预后的危险因素分析目的:分析IMN患者肾活检时的年龄、24小时尿蛋白定量、肾功能水平、血清抗PLA2R抗体滴度和血尿酸水平的分布特征及其和肾脏预后的关系,寻找能够早期预测IMN患者不良肾脏预后的预测指标。对象与方法:回顾性收集2009到2017年在温州医科大学附属第一医院行肾活检的患者8000余例,详细记录这些患者的人口学数据、肾活检时的临床指标、病理数据、随访过程中的实验室检查结果和治疗方案。按照本研究预先制定的纳入和排除标准筛选符合标准的IMN患者。对相关临床指标的分布进行展示并比较性别亚组的分布。采用Pearson′s卡方检验和矫正的标准化卡方值分析临床指标间的关系。受试者工作曲线评价相关临床指标对于鉴别不良肾脏预后的能力。时间事件分析用来评价这些临床指标和不良肾脏预后之间的关系。首要终点事件定义为不良肾脏预后,指在随访期间估计的肾小球滤过率较基线水平下降50%以上或进展至终末期肾病。使用R统计软件(版本号3.5.2)完成统计分析和相关的作图。结果:在2009-2017年肾活检的8000余例患者中,病理诊断为膜性肾病患者989例,约占12.4%,其中572名IMN患者符合筛选标准,最终纳入到本研究中。在纳入的队列中,中位随访时间为18个月,其中45名(7.9%)患者进展到首要终点。虽然肾活检时的年龄、24h小时尿蛋白定量和肾活检时的估计肾小球清除率和IMN患者不良肾脏预后存在相关性,但是多因素的Cox回归显示这三项临床指标并不是IMN患者的独立危险因素。另外本研究采用受试者工作曲线评价肾活检时的血清尿酸水平和随访过程中时间平均尿酸水对于鉴别IMN患者肾脏转归的效能,结果显示两者之间的效能差异并不显著,其中基线血清尿酸的曲线下面积(area under curve,AUC)=0.66,95%置信区间(confidence interval,CI)=0.58-0.74,时间平均血清尿酸的AUC=0.69,95%CI=0.60-0.77,两者之间的P值=0.6。多因素的Cox回归分析进一步显示基线血清尿酸水平是IMN患者不良肾脏预后的独立危险因子,而且亚组分析显示血清尿酸水平预测效能存在性别差异。结论:IMN患者肾活检时的年龄、24小时尿蛋白水平、e GFR水平和血尿酸水平均和不良肾脏预后相关,经多因素Cox回归分析显示基线血清尿酸水平是IMN患者不良肾脏预后的独立预测因子。第二部分特发性膜性肾病肾脏预后的早期预测模型构建目的:通过第一部分的分析显示IMN患者肾活检时的临床指标和不良肾脏预后存在密切相关,但是单因素预测IMN患者不良肾脏预后并不稳健。因此本研究拟结合既往成熟的IMN预后预测模型,采用稳健的统计学方法构建一个基于多因素的预测模型从而早期预测不良肾脏预后。对象与方法:筛选出2009年1月至2017年12月在我院经肾活检确诊为IMN的患者。收集患者的人口学数据,活检时的临床指标、病理数据及随访过程中的实验室检查结果和治疗方案。利用本研究收集的数据集验证Cattran等构建的IMN患者风险评估(Toronto Risk Score)模型,评价模型性能。利用多因素Cox回归模型构建一个新的基于肾活检时临床和病理数据的IMN预后预测模型,并用AIC原则筛选变量优化模型,然后采用Therneau-Grambsch方法对模型内所有变量进行诊断,排除不符合模型假设的变量。最后采用净重分类改进指数(net reclassification improvement,NRI)评估本研究新构建的早期预测模型和Toronto Risk Score模型之间的分类性能差异。不良肾脏预后定义包括首要终点事件和次要终点事件,首要终点事件定义为随访过程中估计肾小球滤过率(estimated glomerular filtration rate,e GFR)下降超过基线值的30%或e GFR<30ml/min/1.73m~2,次要终点事件定义为随访过程中进展至慢性肾功能不全(chronic renal insufficiency,CRI)即e GFR<60ml/min/1.73m~2。所有统计及数据处理均采用R统计软件(版本号3.5.2)完成。结果:共纳入符合标准的患者572例,其中进展至首要终点事件的患者118(20.6%)例,而进展至次要终点事件的患者98(17.1%)例。比较发生首要终点事件的患者和未发生终点事件的患者,两者肾活检时的临床指标存在显著差异,尤其是年龄、性别、血白蛋白水平、估计肾小球滤过率、血肌酐水平、血纤维蛋白原水平、24h尿蛋白定量和收缩压。本研究数据显示Toronto Risk Score模型对于预测首要终点事件的敏感性和特异性分别为28%和96.4%。新构建的多因素Cox回归模型经AIC原则筛选后仅保留年龄、血白蛋白、血尿酸、血甘油三脂和肾小管损害程度等变量,提示肾活检时这些变量和远期肾脏预后密切相关。然而Therneau-Grambsch分析显示血白蛋白水平不符合等比例风险假设,在模型中剔除。最终构建的模型纳入年龄、血尿酸水平、血甘油三脂水平和肾小管损害程度。模型预测不良肾脏预后的敏感性和特异性分别为68.8%和71.5%。和Toronto Risk Score模型比较显示NRI为0.174(P值=0.003),提示本研究构建的早期预测模型在分类性能上有了显著提高。结论:本研究构建了一个基于多因素Cox回归模型的早期预测IMN肾脏预后的评估模型。模型的分类性能较Toronto Risk Score模型有了显著的提高,但由于数据限制,本研究结果仍需要外部数据的验证。更多还原

【Abstract】 Background:Membranous nephropathy(MN)with the characteristics of massive proteinuria and hypoalbuminemia is a common cause of nephrotic syndrome in adults.The special pathological feature of MN is the formation of immune complexes in the epithelial side of the capillary loop of the glomerulus.The incidence of membranous nephropathy in our country is increasing gradually,which has attracted extensive attention from nephrologists.The membranous nephropathy was classed by the cause of the disease named secondary and idiopathic membranous nephropathy.Secondary membranous nephropathy,which accounts for about 20%of total membranous nephropathy,is caused by systemic disease or exposure to certain factors.Idiopathic membranous nephropathy(IMN),also known as primary membranous nephropathy,accounts for about 80%of total membranous nephropathy and is limited to the kidney.Against the secondary MN,the treatment of IMN is more complicated.Although IMN is often considered a"benign disease",30%-40%of IMN patients,especially those with persistent proteinuria and impaired kidney function,may develop the end-stage renal disease within 5-15 years.At present,the treatment of IMN is still controversial.The treatment of IMN is still contentious.Even though immunosuppressive therapies can be selected for IMN,KDIGO guidelines give the indications and contraindications for immunosuppressive treatment.For the significant side effects of immunosuppressive therapies,it should be selected cautiously.The KDIGO guidelines also pointed out that it is necessary to find appropriate clinical,pathological,and biological markers to predict the prognosis of IMN to screen suitable patients and give active treatment for improving the prognosis of IMN and reducing the side effects of immunosuppressive therapy.However,early prediction of the renal prognosis of IMN is still very challenging and requires more extensive and in-depth research and verification.Part I.Risk factors of renal prognosis in idiopathic membranous nephropathyPurpose:The distribution characteristics of age,24-hour urinary protein level,renal function level,serum anti-PLA2R antibody titer,and serum uric acid(UA)level at the time of renal biopsy in IMN patients were analyzed,and their relationship with renal prognosis was analyzed so as to screen for early predictors of poor renal prognosis in IMN patients.Methods:We identified patients with a renal biopsy-confirmed diagnosis of IMN from2009 to 2017 in Wenzhou medical university’s first affiliate.The demographic,clinical data and renal histopathological reports were collected at the renal biopsy.Cases were filtered by the including and exclusion criteria.The receiver operating characteristic(ROC)analysis was used for evaluating the identification power of risk factors,including serum UA for the poor renal outcomes.The time-event analysis was used to assess the impact of risk factors for poor renal outcomes in patients with IMN.The primary endpoint was the poor renal outcome,which was defined as a decrease in the estimated glomerular filtration rate to 50%of the baseline level or progression to end-stage renal disease during the follow-up.R(version 3.5.2,R Core Team)and it′s packages were used to perform the statistical analyses and plot the figures.Results:Selected 989 cases of MN from 8000 renal biopsy cases,and 572 eligible patients were included.During a median of 18 months of follow-up,45(7.9%)patients progressed to the primary endpoint.Although the age at biopsy,24-hour proteinuria and estimated glomerular clearance(e GFR)at biopsy were associated with poor renal outcomes in IMN patients,multivariate Cox regression showed these three clinical indicators were not independent risk factors for IMN patients.Both baseline serum UA and time-averaged UA levels could be used for discrimination of renal outcomes,but the difference was not significant(p-value=0.6).Our multivariate Cox regression analysis further demonstrated that baseline serum UA was an independent predictor of poor renal outcome in IMN patients,and subgroup analysis revealed a gender difference in the predictive effect of serum UA.Conclusions:Our study demonstrated that age 24-hour urinary protein level,EGFR level,and serum UA at the time of renal biopsy were associated with poor renal outcomes in IMN patients.Furthermore,our multivariate Cox regression analysis indicated that serum UA was an independent predictor for poor renal outcomes in patients with IMN.Part II.Establishment of an early prognostic model for idiopathic membranous nephropathyPurpose:The first part showed that the clinical indicators of renal biopsy in IMN patients were closely related to adverse renal outcomes.Still,the single factor prediction of poor renal outcomes in IMN patients was not robust.Therefore,this study intends to combine the mature prognosis prediction model of IMN with robust statistical methods to construct a multi-factor prediction model for early prediction of poor renal prognosis.Methods:Patients with a renal biopsy-confirmed diagnosis of IMN between 2009 and2017 in our department were identified.The demographic,clinical data,and renal pathology recorded at the time of renal biopsy as well as follow-up data were collected.The data set of this study was used to verify the performance of a risk score model for evaluating the renal outcomes among IMN patients constructed by Cattran et al.,which is called the Toronto Risk Score model.A new IMN prognostic model based on clinical and pathological data at the time of renal biopsy was established by using a multivariate Cox regression model.A variable optimization model was selected by the AIC principle.Then,all variables in the model were diagnosed by the Therneau-Grambsch method,and variables that did not fit the model hypothesis were excluded.Finally,net reclassification improvement(NRI)was used to evaluate the classification performance difference between the newly constructed early prediction model in this study and the Toronto Risk Score model.Poor renal outcomes included primary endpoint events defined as a decrease of the estimated glomerular filtration rate(e GFR)by 30%of the baseline level or e GFR<30ml/min/1.73m~2,and secondary endpoint events defined as progression to CRI(chronic renal insufficiency),which means e GFR<60 ml/min/1.73m~2 during the follow-up.Finally,we carried out an internal validation of the model with our data set.R(version 3.5.2,R Core Team)and it′s packages were used to perform the statistical analyses and create the figures.Result:A total of 572 eligible patients were included,including 118(20.6%)patients who progressed to primary endpoint events and 98(17.1%)patients who progressed to secondary endpoint events.There were significant differences in the parameters at the time of renal biopsy between patients with or without primary endpoint events,especially in the age,gender,serum albumin levels,e GFR,serum creatinine levels,serum fibrinogen levels,24-hour urinary protein quantification,and systolic blood pressure.Data from this study showed that the Toronto Risk Score model had a sensitivity and specificity of 28%and 96.4%for predicting primary endpoint events,respectively.In the newly constructed multivariate Cox regression model,only age,serum albumin,serum uric acid,blood triglyceride,and tubulointerstitial injury were retained after screening by the AIC principle,suggesting that these variables were closely related to long-term renal prognosis at renal biopsy.However,the Therneau-Grambsch analysis showed that serum albumin levels did not conform to the proportional risk hypothesis and were eliminated from the model.The final model included age,blood uric acid level,blood triglyceride level,and renal tubular damage.The model’s sensitivity and specificity for predicting poor renal outcomes were 68.8%and 71.5%,respectively.Compared with the Toronto Risk Score model,the NRI was 0.174(P=0.003),indicating that this study’s early prediction model has significantly improved in the classification performance.Conclusion:In this study,an evaluation model based on the multivariate Cox regression model for early prediction of renal prognosis of IMN was established.The classification performance of the model has been significantly improved compared with the Toronto Risk Score model.However,due to data limitations,the results of this study still need to be verified by external data.更多还原