文献速递流行病学与生物统计基础与转化研究

Meta回归重新解读开颅与颅骨修补术后手术部位感染危险因素的效应修饰与异质性

Effect moderation and heterogeneity in meta-analysis: a meta-regression-based reinterpretation of surgical site infection risk factors after craniotomy and cranioplasty.
PubMed 文献 · J Clin Epidemiol · 2026-07-27

该研究从四项Meta分析纳入的研究中提取18项参数,用改良GRADE工具评估感染发生率估计的确定性,并通过气泡图检验参数与手术部位感染的单变量关联。结果发现研究人群规模是效应修饰因子,人群越小效应估计越大,按人群规模分层后四项潜在危险因素的森林图解读发生改变。作者认为Meta回归可识别效应修饰因子并通过分层分析改变Meta分析结论。

要点

  • 研究人群规模是手术部位感染效应估计的效应修饰因子,人群越小效应估计越大。
  • 按人群规模分层(阈值500、750、1000)后,四项潜在危险因素的森林图解读发生改变。
  • 男性与植入物使用变为与感染增加相关,创伤比例增加与感染发生率降低相关。

核心结论

  • Meta回归可识别效应修饰因子并通过分层分析改变Meta分析得出的结论。

研究设计与样本

  • 从四项Meta分析纳入的研究中提取18项参数,用改良GRADE工具评估各研究感染发生率估计的确定性。

主要发现

  • 研究人群规模为效应修饰因子,研究人群越小效应估计越大,多变量建模敏感性分析证实。
  • 按人群规模分层阈值500、750、1000后,四项潜在危险因素的森林图解读发生改变。
  • 男性与植入物使用变为与感染增加相关。
  • 创伤病例比例增加与感染发生率降低相关。
  • 急诊手术从有关联变为无关联。

关键数值与推荐

  • 分层阈值采用500、750、1000。

临床怎么用

  • 在解读手术部位感染危险因素的Meta分析时,应考虑按研究人群规模分层分析以识别效应修饰。

局限性

  • 本研究为探索性研究,仅基于四项Meta分析纳入的研究数据。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

BACKGROUND AND OBJECTIVES: The presence of effect moderation can compromise the interpretation of meta-analyses. Hence, this characteristic has potential clinical relevance. One tool to identify effect moderation is meta-regression; however, this approach is seldom used. Meta-regression incorporates study-level characteristics to explore whether such characteristics are associated with variation in effect estimates. We applied meta-regression to examine a significant problem in neurosurgery: surgical site infection after craniotomy and cranioplasty. Numerous meta-analyses have attempted to synthesize data from individual studies that have sought to identify infection risk factors. Nevertheless, the results have been equivocal, and these meta-analyses never applied meta-regression. In this exploratory study, we hypothesized that meta-regression might provide insight into the interpretation of prior meta-analyses. METHODS: We extracted data on 18 parameters from studies included in four meta-analyses. A modified GRADE tool was used to assess the certainty of the estimates of infection incidence in each study. We examined bubble plots to assess univariate associations between these parameters and surgical site infection. RESULTS: Population size was found to be an effect moderator; the smaller the study population, the greater the effect estimate (confirmed in a sensitivity analysis using multivariable modeling). Therefore, we determined whether the interpretation of forest plots in published meta-analyses would change if analyses were stratified by population size. We used stratification thresholds of 500, 750, and 1000. The interpretation of forest plots changed for four potential risk factors at different stratification thresholds. Male sex and implant use became associated with increased infection. An increase in the proportion of trauma cases was associated with a reduced incidence of infection, whereas emergency surgery shifted from an association to no association. CONCLUSION: Meta-regression provides a complementary interpretive framework that can alter conclusions drawn from meta-analyses by identifying effect moderators and by applying stratified analysis.

来源:J Clin Epidemiol,查看原文。版权归原作者所有。

引用

  1. Wang B, Sun M, Ma C, et al. Effect moderation and heterogeneity in meta-analysis: a meta-regression-based reinterpretation of surgical site infection risk factors after craniotomy and cranioplasty. J Clin Epidemiol. 2026. PMID: 42508585. doi:10.1016/j.jclinepi.2026.112428

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。