文献速递呼吸道感染呼吸危重症与机械通气

非HIV肺孢子菌肺炎肺泡损伤与免疫衰竭的生物学表型预测死亡率:西班牙多中心队列研究

Biological phenotypes of alveolar injury and immune failure predict mortality in non-HIV Pneumocystis jirovecii pneumonia: a Spanish multicentre cohort study.
PubMed 文献 · Chest · 2026-09-28

该回顾性多中心队列纳入78例非HIV肺孢子菌肺炎(PJP)成人患者,根据入院LDH(≥400 U/L)和淋巴细胞计数(<700 cells/μL)定义四种表型(F1-F4)。院内死亡率为50.0%,30天累积死亡率随表型单调升高(F1 12.5%至F4 64.0%,Gray's test p=0.008),F4表型独立与死亡率相关(subdistribution hazard ratio 6.15;95% CI 1.27-29.84),模型区分度良好(乐观校正AUC 0.88)。

要点

  • 入院LDH≥400 U/L且淋巴细胞<700 cells/μL的F4表型院内死亡风险显著增高。
  • 四种表型30天累积死亡率呈单调梯度,F1至F4分别为12.5%、28.6%、47.1%、64.0%。
  • 基于LDH和淋巴细胞计数的表型可用于非HIV PJP早期床旁风险分层。

核心结论

  • 两个常规入院标志物定义的四种表型呈现有序的死亡率梯度,F4为高风险聚合表型。
  • F4表型独立预测院内死亡,支持早期床旁风险分层和表型导向的辅助治疗研究。

研究设计与样本

  • 回顾性多中心队列研究,纳入三家医院2019-2024年收治的78例EORTC/MSGERC probable或proven非HIV PJP成人患者。
  • 以入院LDH(≥400 U/L)和淋巴细胞计数(<700 cells/μL)定义四种表型(F1-F4)。
  • 主要结局为院内全因死亡,采用竞争风险框架(活体出院为竞争事件)分析,主要分析为Fine-Gray subdistribution hazards模型,调整CURB-65和实体器官恶性肿瘤。
  • 次要分析为30天Kaplan-Meier/Cox分析,区分度和校准通过bootstrap内部验证。

主要发现

  • 78例患者中院内死亡39例(50.0%)。
  • 30天累积死亡率随表型单调升高:F1 12.5%、F2 28.6%、F3 47.1%、F4 64.0%(Gray's test p=0.008)。
  • F4表型独立与死亡率相关(subdistribution hazard ratio 6.15;95% CI 1.27-29.84),CURB-65和实体器官恶性肿瘤亦独立相关。
  • 中间表型方向一致但精度不足。
  • 多变量预测模型保留良好区分度(乐观校正AUC 0.88)和校准。

关键数值与推荐

  • 入院LDH≥400 U/L和淋巴细胞计数<700 cells/μL定义四种表型(F1-F4)。
  • 主要结局为院内全因死亡,活体出院为竞争事件。
  • 主要分析采用Fine-Gray subdistribution hazards模型,调整CURB-65和实体器官恶性肿瘤。
  • F4表型subdistribution hazard ratio 6.15(95% CI 1.27-29.84)。
  • 模型乐观校正AUC 0.88。

临床怎么用

  • 入院时检测LDH和淋巴细胞计数可快速识别非HIV PJP高危表型。
  • F4表型患者院内死亡风险高,应加强监护和早期干预。
  • 表型分层可用于辅助治疗研究的入组筛选。

局限性

  • 回顾性设计。
  • 样本量较小(78例)。
  • 中间表型关联精度不足。
  • 内部验证采用bootstrap,需外部验证。

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: Non-HIV Pneumocystis jirovecii pneumonia (PJP) carries high in-hospital mortality, yet bedside risk stratification remains limited. Admission lactate dehydrogenase (LDH) and lymphocyte count are universally available and may capture two distinct pathophysiological axes (cytolytic tissue injury and failure of adaptive immune control). RESEARCH QUESTION: Do admission phenotypes defined by LDH and lymphocyte count identify distinct in-hospital mortality risk in non-HIV PJP? STUDY DESIGN AND METHODS: Retrospective multicentre cohort of adults with EORTC/MSGERC probable or proven non-HIV PJP admitted to three hospitals (2019-2024). Admission LDH (≥400 U/L) and lymphocyte count (<700 cells/μL) defined four phenotypes (F1-F4). The primary outcome was in-hospital, all-cause mortality analysed within a competing-risks framework (live discharge as the competing event); the primary analysis was a Fine-Gray subdistribution hazards model adjusted for CURB-65 and solid-organ malignancy, with 30-day Kaplan-Meier/Cox analyses as secondary. Discrimination and calibration were internally validated by bootstrap. RESULTS: Seventy-eight patients were included; in-hospital mortality was 39/78 (50.0%). The 30-day cumulative incidence of death rose monotonically across phenotypes (F1 12.5%, F2 28.6%, F3 47.1%, F4 64.0%; Gray's test p=0.008). The convergent cytolytic-immunodepressed phenotype (F4) was independently associated with mortality (subdistribution hazard ratio 6.15; 95% CI 1.27-29.84), as were CURB-65 and solid-organ malignancy; intermediate phenotypes were directional but imprecise. The multivariable prediction model retained good discrimination (optimism-corrected AUC 0.88) and calibration. INTERPRETATION: Two universally available admission markers define an ordered mortality gradient and a high-risk convergent phenotype in non-HIV PJP, supporting early bedside risk stratification and phenotype-stratified investigation of adjunctive therapies.

来源:Chest,查看原文。版权归原作者所有。

引用

  1. Martínez de Victoria Carazo J, Arrabal EG, Zafra LG, et al. Biological phenotypes of alveolar injury and immune failure predict mortality in non-HIV Pneumocystis jirovecii pneumonia: a Spanish multicentre cohort study. Chest. 2026. PMID: 42805328. doi:10.1016/j.chest.2026.09.011

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。