中国全科医学 ›› 2023, Vol. 26 ›› Issue (17): 2078-2088.DOI: 10.12114/j.issn.1007-9572.2022.0827

所属专题: 乳腺癌最新文章合集 肿瘤最新文章合集

• 论著 • 上一篇    下一篇

基于前瞻性队列研究的Meta分析构建乳腺癌相关淋巴水肿风险预测模型研究

沈傲梅1,2, 路潜1,*(), 符鑫1, 韦小夏1, 卞静如2, 张丽媛2, 强万敏2, 庞冬1   

  1. 1.100191 北京市,北京大学护理学院内外科教研室 北京大学循证护理研究中心2.300060 天津市,天津医科大学肿瘤医院护理部 国家肿瘤临床医学研究中心 天津市"肿瘤防治"重点实验室 天津市恶性肿瘤临床医学研究中心
  • 收稿日期:2022-08-08 修回日期:2022-12-10 出版日期:2023-06-15 发布日期:2023-01-18
  • 通讯作者: 路潜

  • 作者贡献:沈傲梅负责研究设计、文献检索、筛选、评价、资料提取分析及原稿写作;符鑫、韦小夏、卞静如、张丽媛负责文献检索、筛选、评价、资料提取及数据分析;路潜、强万敏、庞冬负责研究设计及方法、论文审阅及修订;所有作者确认论文终稿。
  • 基金资助:
    国家自然科学基金面上项目(72174011)——健康生态学视角下乳腺癌淋巴水肿风险预测与精准干预策略研究; 天津医科大学肿瘤医院院级课题(H2001)——基于Triangle模型的乳腺癌患者淋巴水肿分层支持管理标准的制定; '未名护理'领军人才科研创新孵化基金项目(LJRC21ZD03)——基于ITHBC的乳腺癌患者淋巴水肿自我管理行为状况及机制研究

Constructing a Risk Prediction Model of Breast Cancer-related Lymphedema Based on a Meta-analysis of Prospective Cohort Studies

SHEN Aomei1,2, LU Qian1,*(), FU Xin1, WEI Xiaoxia1, BIAN Jingru2, ZHANG Liyuan2, QIANG Wanmin2, PANG Dong1   

  1. 1. Division of Medical & Surgical Nursing, School of Nursing, Peking University/Evidence-based Nursing Center, Peking University Health Science Center, Beijing 100191, China
    2. Department of Nursing, Tianjin Medical University Cancer Institute & Hospital/National Clinical Research Center for Cancer/Key Laboratory of Cancer Prevention and Therapy, Tianjin/Tianjin's Clinical Research Center for Cancer, Tianjin 300060, China
  • Received:2022-08-08 Revised:2022-12-10 Published:2023-06-15 Online:2023-01-18
  • Contact: LU Qian

摘要: 背景 乳腺癌相关淋巴水肿(BCRL)是困扰乳腺癌患者术后常见的慢性并发症,早期评估和预测BCRL风险尤为重要,但目前仍缺乏权威公认、适宜推广的风险预测模型。目的 本研究拟基于Meta分析构建并验证乳腺癌患者BCRL风险预测模型。方法 计算机检索PubMed、Embase、CINAHL、Scopus、Web of Science、Cochrane Library、中国知网、中国生物医学数据库、万方数据知识服务平台自建库至2021年11月发表的有关BCRL危险因素的前瞻性队列研究。由2名经过系统培训的研究者分别独立筛选文献、提取资料,并采用纽卡斯尔-渥太华量表(NOS)进行质量评价。采用Stata 17.0软件进行Meta分析。根据Meta分析结果提取合并效应量具有显著性的风险因素及合并风险值,构建Logistic回归预测模型。基于回归系数及合并风险值构建Logistic和Additive评分模型。选取2017年4月至2018年12月在北京大学人民医院乳腺中心招募的486例乳腺癌术后患者为模型验证集,采用受试者工作特征(ROC)曲线下面积(AUC)、Hosmer-Lemeshow检验评价评分模型的预测性能,采用决策曲线分析评价预测模型的临床实用性。结果 共纳入49项前瞻性队列研究、32 543例乳腺癌患者。Meta分析结果显示,乳腺癌患者BCRL发生率为20.6%〔95%CI(17.9%,23.3%)〕。49项研究中报告次数>5次且Meta分析合并效应量结果具有显著性的危险因素共5个,分别为:体质指数(BMI)〔RR=1.777,95%CI(1.515,2.085)〕、乳腺手术类型〔RR=1.320,95%CI(1.125,1.549)〕、腋窝手术类型〔RR=3.058,95%CI(2.325,4.020)〕、放疗〔RR=1.620,95%CI(1.214,2.160)〕、术后并发症〔RR=2.373,95%CI(1.278,4.405)〕。Logistic及Additive评分模型总分分别为0~34分、5~11分。Logistic及Additive评分模型的AUC分别为0.748〔95%CI(0.701,0.794)〕、0.737〔95%CI(0.691,0.784)〕,Hosmer-Lemeshow检验P值分别为0.185、0.763。Logistic评分模型最佳截断值为18分,灵敏度为81.7%,特异度为43.1%;Additive评分模型最佳截断值为8.5分,灵敏度为80.9%,特异度为42.8%。当阈值概率在20%~30%时,预测模型具有较高的临床净获益。结论 基于Meta分析构建的BCRL风险预测模型具有较好的预测性能,可作为BCRL风险评估工具,指导BCRL的分层管理,但其预测性能和临床实用性仍有待进一步验证和优化。

关键词: 乳腺癌淋巴水肿, 乳腺肿瘤, 危险因素, 风险预测模型, Meta分析, 前瞻性队列研究

Abstract: Background Lymphedema is a common chronic complication bothering breast cancer patients. Early assessment and prediction of the risk for developing breast cancer-related lymphedema (BCRL) is particularly important. However, there is still a lack of an authoritatively recognized and suitably promoted risk prediction model.Objective To construct and validate a risk prediction model for BCRL based on the results of a meta-analysis.Methods Electronic databases including PubMed, Embase, CINAHL, Scopus, Web of Science, Cochrane Library, CNKI, CBM, and Wanfang Data, were searched for prospective cohort studies on risk factors of BCRL from inception to November 2021. Two systematically trained researchers independently screened the literature, extracted data, and assessed the study quality using the Newcastle-Ottawa Scale. Stata 17.0 was used for meta-analysis. The risk factors with significant pooled effect size and their combined risk value were extracted to construct the Logistic risk prediction model. The Logistic and additive risk scoring models were constructed based on regression coefficients and pooled risk values, respectively. The data of 486 breast cancer patients recruited in the breast cancer research center of Peking University People's Hospital, from April 2017 to December 2018, were selected as the validation set. The area under the ROC curve (AUC) and the Hosmer-Lemeshow test were used to evaluate the prediction performance of the risk scoring model. Decision curve analysis was used to evaluate the clinical practicability.Results A total of 49 prospective cohort studies involving 32 543 breast cancer patients were included. Meta-analysis showed that the incidence of BCRL was 20.6%〔95%CI (17.9%, 23.3%) 〕. Among 49 studies, five risk factors with significant pooled effect sizes were reported more than five times, including body mass index (BMI) 〔RR=1.777, 95%CI (1.515, 2.085) 〕, type of breast surgery〔RR=1.320, 95%CI (1.125, 1.549) 〕, type of axillary surgery〔RR=3.058, 95%CI (2.325, 4.020) 〕, radiotherapy〔RR=1.620, 95%CI (1.214, 2.160) 〕, and postoperative complications〔RR=2.373, 95%CI (1.278, 4.405) 〕. The total score for the Logistic risk scoring model ranged from 0 to 34, and that for the additive risk scoring model ranged from 5 to11. The AUCs of Logistic and additive risk scoring models were 0.748〔95%CI (0.701, 0.794) 〕and 0.737〔95%CI (0.691, 0.784) 〕, respectively. The values of Hosmer-Lemeshow test were 0.185 and 0.763, respectively. The optimal cut-off value of the Logistic risk scoring model was 18, with a sensitivity of 81.7%, and a specificity of 43.1%. The optimal cut-off value of the additive risk scoring model was 8.5, the sensitivity was 80.9%, and the specificity was 42.8%. When the probability threshold ranged from 20% to 30%, the model achieved higher net clinical benefit. Conclusion The BCRL risk prediction model based on this meta-analysis has good predictive performance. It can be used as a risk assessment tool for BCRL to guide the hierarchical monitoring and management of BCRL. However, prediction performance and clinical practicability of the model still needs to be validated and optimized in future research.

Key words: Breast cancer lymphedema, Breast neoplasms, Risk factors, Prediction model, Meta-analysis, Prospective cohort studies