快速免疫测定法用于儿童急性心肺衰竭分子亚型分类的评估
回顾性多中心PICU队列纳入269例急性心肺衰竭伴高血糖的危重患儿(2012-2016)。比较快速免疫测定与常规多重组学平台的生物标志物测量,并将基于常规检测的简约分类器(IL-6、IL-8、sTNFR-1)应用于快速测定数据,AUROC为0.90(95%CI 0.85-0.95),但因快速平台系统性低估sTNFR-1导致校准差。随后基于快速测定数据重新推导并内部验证分类器,AUROC 0.90(95%CI 0.86-0.95),校准优秀,以概率阈值≥0.5分类与LCA分类一致241/269例(准确率89.6%)。快速测定亚型显示死亡率差异(高炎症33.3% vs 低炎症11.8%,p=0.009)及对强化胰岛素管理的差异反应(交互p=0.024)。
要点
- 在269例急性心肺衰竭伴高血糖的危重患儿中,快速免疫测定衍生的简约分类器以概率阈值≥0.5分类,与LCA分类一致241/269例(准确率89.6%)。
- 基于快速测定数据新推导的分类器AUROC为0.90(95%CI 0.86-0.95)且校准优秀;直接套用常规检测分类器AUROC亦为0.90(95%CI 0.85-0.95)但校准差。
- 快速平台系统性低估sTNFR-1浓度,是直接套用常规分类器校准不良的原因,需用快速测定数据重新训练分类器。
- 快速测定亚型保留预后意义:高炎症亚型死亡率33.3% vs 低炎症亚型11.8%(p=0.009)。
- 快速测定亚型保留预测意义:对强化胰岛素管理的反应存在亚型间差异(交互p=0.024)。
适用人群与场景
- 多中心PICU中急性心肺衰竭伴高血糖的危重患儿,年龄范围及具体纳入/排除标准原文未提及。
- 研究数据时段为2012-2016年,为回顾性队列。
诊断标准与关键阈值
- 分子亚型由既往基于13种血浆生物标志物的常规多重组学检测经潜在类别分析(LCA)衍生,分为高炎症类和低炎症类。
- 简约分类器使用IL-6、IL-8和sTNFR-1三个生物标志物预测LCA衍生亚型。
- 快速免疫测定衍生分类器以概率阈值≥0.5判定亚型分类。
处置流程与药物剂量
- 原文未提及具体处置流程或药物剂量;研究为观察性,无干预。
- 原文提及既往已发表强化胰岛素管理在高血糖伴心肺衰竭患儿中具有亚型特异性获益,但未给出具体剂量、目标值或方案。
红旗征与不能漏的情况
- 原文未提及
研究设计与关键结果
- 研究类型:回顾性队列研究;样本量269例危重患儿;多中心PICU;人群为急性心肺衰竭伴高血糖(2012-2016);无干预。
- 将基于常规检测的简约分类器(IL-6、IL-8、sTNFR-1)应用于快速免疫测定数据,AUROC为0.90(95%CI 0.85-0.95),但因快速平台系统性低估sTNFR-1导致校准差。
- 基于快速免疫测定数据重新推导并内部验证(bootstrapping)分类器,AUROC为0.90(95%CI 0.86-0.95),校准优秀;以概率阈值≥0.5分类与LCA分类一致241/269例(准确率89.6%)。
- 快速免疫测定亚型显示死亡率差异:高炎症亚型33.3% vs 低炎症亚型11.8%(p=0.009)。
- 快速免疫测定亚型显示对强化胰岛素管理的差异反应(交互p=0.024)。
证据类型、等级与局限
- 证据类型:回顾性多中心队列研究(原文标注Design: Retrospective cohort study)。
- 证据等级或推荐强度:原文未提及。
- 局限:回顾性设计;样本量269例;快速平台系统性低估sTNFR-1;新分类器仅内部验证(bootstrapping),未外部验证;原文未提及其他局限。
对床旁实践的改变
- 快速免疫测定平台不能直接套用基于常规多重组学检测训练的简约分类器,否则因sTNFR-1系统性低估导致校准差。
- 若要用快速免疫测定在床旁进行亚型分类,需使用基于快速测定数据重新训练的分类器(概率阈值≥0.5),其与LCA分类准确率89.6%。
- 快速免疫测定衍生的亚型保留了死亡风险分层(高炎症33.3% vs 低炎症11.8%)和对强化胰岛素管理的差异反应(交互p=0.024),但原文未给出可直接用于临床决策的阈值或治疗方案。
- 该分类器目前仅用于研究背景,原文结论为「有潜力为未来亚型导向的精准试验提供信息」,未推荐常规临床使用。
指南或共识出处
- 原文未提及指南或共识名称;文献来源为Crit Care Med。
以上提炼依据论文摘要(全文未开放获取,或许可不允许翻译转载);全文请点「阅读原文」。
英文原文(摘要)· 点击展开
OBJECTIVES: Molecular subphenotypes, identified through latent class analysis (LCA) of biomarker profiles, have the potential to guide targeted therapeutics in critical care. We have previously published that intensive insulin management has subphenotype-specific beneficial effects among children with hyperglycemia accompanying cardiorespiratory failure. However, the real-time application of subphenotype-based strategies in clinical settings remains challenging due to the operational aspects of biomarker assays. Our study had three objectives: 1) to compare biomarker measurements from rapid immunoassay and conventional multiplex platforms; 2) to evaluate the cross-platform transportability of a conventional assay-based parsimonious classifier for LCA-derived subphenotypes and compare it with a classifier trained directly on rapid immunoassay data; and 3) to assess the prognostic and predictive significance of rapid immunoassay-based subphenotypes.
DESIGN: Retrospective cohort study.
SETTING: Multicenter PICUs.
PATIENTS: Two hundred sixty-nine critically ill children with acute cardiorespiratory failure and hyperglycemia (2012-2016).
INTERVENTIONS: None.
MEASUREMENTS AND MAIN RESULTS: LCA was previously used to derive hyperinflammatory and hypoinflammatory classes using a conventional multiplex assay of 13 plasma biomarkers. After feature selection, a parsimonious classifier using interleukin (IL)-6, IL-8, and soluble tumor necrosis factor receptor 1 (sTNFR-1) was developed to predict LCA-derived subphenotypes. We applied this classifier to rapid immunoassay biomarker measurements, yielding an area under the receiver operating characteristic curve (AUROC) of 0.90 (95% CI, 0.85-0.95). However, calibration was poor due to systematic underestimation of sTNFR-1 concentrations by the rapid platform. We then derived and internally validated via bootstrapping a de novo classifier using rapid immunoassay data, achieving an AUROC of 0.90 (95% CI, 0.86-0.95) and excellent calibration. Using a probability threshold of ≥ 0.5, the de novo classifier matched LCA-derived classifications in 241 of 269 cases (accuracy 89.6%). Rapid immunoassay-based subphenotypes demonstrated differences in mortality (33.3% in hyperinflammatory vs. 11.8% in hypoinflammatory; p = 0.009) and differential response to intensive insulin management (interaction p = 0.024).
CONCLUSIONS: A parsimonious, rapid immunoassay-based classifier can approximate LCA-derived molecular subphenotypes while preserving their prognostic and predictive significance, with the potential to inform future subphenotype-based precision trials.
参考来源
- Sallee CJ, Taylor CS, Zinter MS, et al. Evaluation of a Rapid Immunoassay for Molecular Subphenotype Classification in Pediatric Acute Cardiorespiratory Failure. Crit Care Med. 2026. PMID: 42704284. doi:10.1097/CCM.0000000000007339
供执业医师抢救与诊疗参考,不替代临床判断;剂量请结合体重、肝肾功能和本院药品规格复核,以指南原文为准。