2023 年2 期 第31 卷
专题研究![HTML下载 HTML下载](/Areas/201911120101/Content/img/icon_html.png)
![PDF下载 PDF下载](/Areas/201911120101/Content/img/icon_pdf.png)
2型糖尿病患者发生心力衰竭风险预测模型的系统评价
Risk Prediction Models for Heart Failure in Patients with Type 2 Diabetes Mellitus: a Systematic Review
作者:杨玉涵,刘岩,袁如月,胡超越,张晔,张力,杨晓晖
- 单位:
- 1.北京中医药大学东直门医院肾病内分泌二区2.北京中医药大学东方医院科研处
- Units:
- 1.Department of Nephrology and Endocrinology, Dongzhimen Hospital, Beijing University of Chinese Medicine, Beijing 100700, China 2.Scientific Research Department, Dongfang Hospital, Beijing University of Chinese Medicine, Beijing 100078, China
- 关键词:
- 2型糖尿病; 心力衰竭; 预测模型; 系统评价;
- Keywords:
- Type 2 diabetes mellitus; Heart failure; Predictive model; Systematic review
- CLC:
- R 587.1 R 541.6
- DOI:
- 10.12114/j.issn.1008-5971.2023.00.045
- Funds:
- 国家自然科学基金资助项目(81974541);北京市自然科学基金资助项目(7212176)
摘要:
目的 系统评价2型糖尿病(T2DM)患者发生心力衰竭的风险预测模型,以期为临床医生选择合适的预测模型提供参考。方法 计算机检索PubMed、Embase、Cochrane Library、中国知网、万方数据知识服务平台、维普网及中国生物医学文献数据库中有关T2DM患者发生心力衰竭风险预测模型的文献,检索时间从建库至2022-04-30。根据CHARMS清单提取文献资料,应用预测模型偏倚风险和适用性评估工具(PROBAST)评估模型的偏倚风险和适用性。结果 最终纳入12篇文献,共构建了14个风险预测模型。仅1个模型未提及AUC,13个模型的AUC为0.72~0.87;仅1个模型未提及校准方法,13个模型报告了校准方法;8个模型采用Bootstrap法进行内部验证,5个模型采用分割样本法进行内部验证,1个模型采用交叉验证法进行内部验证;8个模型是作者或其他研究者进行外部验证,6个模型未进行外部验证;14个模型包含3~16个模型变量,其中最常见的模型变量是年龄(8个模型)、糖化血红蛋白(HbA1c)(8个模型)及BMI(7个模型);模型最常见的呈现形式为评分分级(6个模型),其次为方程(5个模型)。14个模型整体均存在高偏倚风险,但整体适用性高。结论 目前构建的T2DM患者发生心力衰竭风险预测模型的AUC为0.72~0.87,具有一定区分度,但部分模型缺乏外部验证,且所有模型存在高偏倚风险。
Abstract:
Objective To systematically evaluate the risk prediction models for heart failure in patients with type 2 diabetes mellitus (T2DM) , in order to provide a reference for clinicians to choose the appropriate prediction model. Methods PubMed, Embase, Cochrane Library, CNKI, Wanfang Data, VIP and CBM were searched to collect the literature on the risk prediction model of heart failure in patients with T2DM. The retrieval time was from the establishment of the database to 2022- 04-30. The literature data was extracted according to the CHARMS checklist, and the risk of bias and applicability of the model were evaluated by the Predictive Model Risk of Bias and Suitability Assessment Tool (PROBAST) . Results Finally, 12 literature were included, and 14 risk prediction models were constructed. Only 1 model did not mention the AUC, and the AUC of the other 13 models was 0.72-0.87; only 1 model did not mention the calibration method, and 13 models reported the calibration method; 8 models were internally validated by Bootstrap method, 5 models were internally validated by split-sample method, and 1 model was internally validated by cross-validation method; 8 models were externally validated by authors or other researchers, and 6 models were not externally validated; 14 models had 3-16 model variables, the most common model variables were age (8 models) , glycosylated hemoglobin (HbA1c) (8 models) and BMI (7 models) ; the most common presentation of the models was score grading (6 models) , followed by equations (5 models) . Fourteen models had high risk of bias and high applicability. Conclusion The AUC of the current risk prediction model for heart failure in T2DM patients was 0.72-0.87, with a certain degree of discrimination, but some models lack external validation and all models have a high risk of bias
ReferenceList: