南方医科大学学报

南方医科大学学报 ›› 2022, Vol. 42 ›› Issue (4): 546-553.doi: 10.12122/j.issn.1673-4254.2022.04.10

• • 上一篇    下一篇

腹膜透析相关腹膜炎患者治疗失败预测模型的构建和验证：一项多中心临床研究

孟令飞，朱学研，杨立明，李忻阳，程思宇，郭师正，庄小花，邹洪斌，崔文鹏   

  1. 吉林大学第二医院肾病内科，吉林 长春 130041；吉林市中心医院肾病内科，吉林 长春 132011；吉林大学第一医院二部肾病内科，吉林 长春 130031
  • 出版日期:2022-04-20 发布日期:2022-04-29

Development and validation of a prediction model for treatment failure in peritoneal dialysis-associated peritonitis patients: a multicenter study

MENG Lingfei, ZHU Xueyan, YANG Liming, LI Xinyang, CHENG Siyu, GUO Shizheng, ZHUANG Xiaohua, ZOU Hongbin, CUI Wenpeng   

  1. Department of Nephrology, Second Hospital of Jilin University, Changchun 130041, China; Department of Nephrology, Jilin Central Hospital, Changchun 132011, China; Department of Nephrology, Second Division of First Hospital of Jilin University, Changchun 130031, China
  • Online:2022-04-20 Published:2022-04-29

RichHTML

27

PDF

633

摘要： 目的 构建和验证腹膜透析相关性腹膜炎（PDAP）患者治疗失败的风险预测模型。方法 对2013年1月1日~2019年12月31日在吉林省3个腹膜透析中心发生PDAP的腹膜透析（PD）患者进行了回顾性分析。收集入选者基线临床资料，主要研究终点为治疗失败。根据腹膜透析中心地域的不同，将数据分为训练集（吉林大学第二医院、吉林大学第一医院二部）和验证集（吉林市中心医院）。采用Logistic风险回归模型筛选影响PDAP治疗失败的危险因素，用Stata建立预测模型；用ROC曲线和校准曲线评估模型的区分度和准确性，并以DCA曲线评估列线图的临床有效性。结果 共纳入977例次PDAP，训练集中625例次，其中78例治疗失败，验证集中352例次，其中35例治疗失败。建模队列多因素Logistic回归分析结果显示，血清白蛋白、第5天腹透液白细胞计数、透析龄和致病微生物类型是治疗失败的独立危险因素，在训练集中的C统计量为0.827（95% CI：0.784–0.871）。在验证集中，C统计量为0.825（95%CI：0.743-0.908）。预测模型在训练和验证集的校准方面都表现良好。结论 基于血清白蛋白、第5天腹透液白细胞计数、透析龄和致病微生物类型构建了预测模型，性能良好。

关键词: 腹膜透析；腹膜透析相关性腹膜炎；列线图；治疗失败；预测模型

Abstract: Objective To develop and validate a risk prediction model of treatment failure in patients with peritoneal dialysis-associated peritonitis (PDAP). Methods We retrospectively analyzed the data of patients undergoing peritoneal dialysis (PD) in 3 dialysis centers in Jilin Province who developed PDAP between January 1, 2013 and December 31, 2019. The data collected from the Second Hospital of Jilin University and Second Division of First Hospital of Jilin University) were used as the training dataset and those from Jilin Central Hospital as the validation dataset. We developed a nomogram for predicting treatment failure using a logistic regression model with backward elimination. The performance of the nomogram was assessed by analyzing the C-statistic and the calibration plots. We also plotted decision curves to evaluate the clinical efficacy of the nomogram. Results A total of 977 episodes of PDAP were included in the analysis (625 episodes in the training dataset and 352 episodes in the validation dataset). During follow-up, 78 treatment failures occurred in the training dataset and 35 in the validation dataset. A multivariable logistic regression prediction model was established, and the predictors in the final nomogram model included serum albumin, peritoneal dialysate white cell count on day 5, PD duration, and type of causative organisms. The nomogram showed a good performance in predicting treatment failure, with a C-statistic of 0.827 (95% CI: 0.784-0.871) in the training dataset and of 0.825 (95% CI: 0.743-0.908) in the validation dataset. The nomogram also performed well in calibration in both the training and validation datasets. Conclusion The established nomogram has a good accuracy in estimating the risk of treatment failure in PDAP patients.

Key words: peritoneal dialysis; predictive model; peritoneal dialysis-associated peritonitis; treatment failure; nomogram