Identifying subgroup of severe community-acquired pneumonia based on clinical metagenomics, a multicenter retrospective cohort study

Mingqiang Wang, Yue Jin, Wenxiao Zhang, Ling Ye, Huanzhang ShaoFrontiers in Cellular and Infection Microbiology 2025 · DOI

TL;DR

Researchers have identified two distinct "phenotypes" of severe community-acquired pneumonia (sCAP) by combining routine clinical data with lung metagenomics. The study reveals that while corticosteroids significantly reduce mortality in one group (Class B), they offer no benefit to the other (Class A), providing a potential roadmap for personalized treatment in the ICU.

What Problem Does This Paper Address?

Severe community-acquired pneumonia (sCAP) is a leading cause of death in intensive care units (ICUs). For decades, clinicians have debated the use of corticosteroids (like methylprednisolone) to dampen the life-threatening inflammation associated with the disease. While some trials show steroids save lives, others show no benefit or even harm.

The challenge is that sCAP is "heterogeneous"—it isn't one single disease state. Different patients have different pathogens and different immune responses. This paper seeks to move past a "one-size-fits-all" approach by using clinical metagenomics (mNGS) to see if the specific microbial landscape of a patient's lungs can help predict whether steroids will actually work.

Methods

The researchers conducted a multicenter retrospective cohort study across 17 medical centers in China. They focused on immunocompetent patients with sCAP who required mechanical ventilation and had undergone metagenomic testing of their bronchoalveolar lavage fluid (BALF).

The core of their approach was Latent Class Analysis (LCA). Instead of just looking at one variable, LCA allows researchers to group patients based on patterns across multiple parameters, including:

  • Demographics: Age and gender.
  • Inflammatory Markers: White blood cell counts, lymphocytes, C-reactive protein (CRP), and Procalcitonin (PCT).
  • Metagenomics: Presence of specific pathogens (bacteria, fungi, and viruses) detected via mNGS.

Datasets

The study initially screened 1,897 patients, ultimately narrowing the cohort down to 569 patients who met strict inclusion criteria (e.g., excluding those with prior immunosuppression or those who didn't require invasive ventilation). The mNGS data included reports on over 100 microbial species, though the researchers focused on those appearing in more than 10% of the patient population to ensure statistical robustness.

Key Insights and Ideas

  • Two Distinct "Classes" of Pneumonia: The LCA identified two optimally fitting models, Class A and Class B, which look very different clinically.
  • Class A (The "Hyper-inflammatory" Group): These patients were generally younger but had much higher inflammatory markers (CRP and PCT) and higher rates of organ failure (liver, kidney, and circulatory). They were more likely to have Streptococcus infections.
  • Class B (The "Viral/Fungal" Group): These patients were older and had lower systemic inflammation markers but showed higher rates of HSV-1 and Candida species in their lungs.
  • The Steroid Paradox: The most striking insight was the lopsided response to treatment. Corticosteroids were associated with a significant reduction in mortality for Class B, but had absolutely no impact on the mortality of Class A.

Key Results

  • Mortality Benefit: In Class B, corticosteroid use was associated with a reduced mortality rate (adjusted Hazard Ratio [HR] = 0.719; p=0.04).
  • Drug Specificity: Within Class B, methylprednisolone was specifically linked to better survival (adjusted HR = 0.61; p=0.005), whereas dexamethasone showed no significant benefit.
  • Dose Matters: Interestingly, higher doses of corticosteroids in Class B were actually associated with increased mortality (adjusted HR = 1.01; p=0.005), suggesting a "sweet spot" for dosing.
  • Class A Non-Response: In Class A, the unadjusted HR for corticosteroid treatment was 0.988 (p=0.959), indicating the treatment was essentially ineffective for this subgroup.

Limitations and Future Work

The authors acknowledge several limitations:

  1. Retrospective Nature: As a retrospective study, it can show associations but not definitive causation. Prospective clinical trials are needed to validate these phenotypes.
  2. DNA vs. RNA: The mNGS used was DNA-based, which might overlook the impact of certain RNA viruses (like specific strains of influenza) unless specifically targeted.
  3. Generalizability: The study excluded immunosuppressed patients, meaning these findings might not apply to transplant recipients or patients with HIV/cancer.
  4. Future Directions: The authors suggest that future models should incorporate "absolute pathogen load" and other inflammatory markers like IL-6 to further refine these subgroups.

By proving that metagenomics can help "sort" patients into responders and non-responders, this study marks a significant step toward precision medicine for the most critically ill pneumonia patients.

Summary updated

Abstract

We identified two classes based on clinical metagenomics and clinical features. Class B exhibited a better response to corticosteroid compared to class A. The rapid identification of these phenotypes could facilitate the screening of sCAP patients responsive to corticosteroid in future prospective clinical trials.
BibTeX
@inproceedings{wang2025identifying,
  author = {Mingqiang Wang and Yue Jin and Wenxiao Zhang and Ling Ye and Huanzhang Shao},
  title = {{Identifying subgroup of severe community-acquired pneumonia based on clinical metagenomics, a multicenter retrospective cohort study}},
  booktitle = {Frontiers in Cellular and Infection Microbiology},
  year = {2025},
  doi = {10.3389/fcimb.2024.1516620},
}
Figure