Prognostic significance of BALF mNGS-detected Candida in ICU patients with severe pneumonia: a multicenter retrospective cohort study

Background The clinical interpretation of Candida detected in bronchoalveolar lavage fluid (BALF) by metagenomic next-generation sequencing (mNGS) remains challenging in ICU patients with severe pneumonia. Respiratory Candida detection usually reflects airway colonization rather than invasive candidiasis, and BALF mNGS alone cannot establish organism viability, tissue invasion, or causality. This study investigated the clinical characteristics, co-detected organism patterns, and prognostic significance of BALF mNGS-detected Candida in ICU patients with severe pneumonia. Methods We conducted a regional multicenter retrospective cohort study of adult ICU patients with severe pneumonia who underwent BALF mNGS in 11 hospitals in Zhejiang and Henan provinces, China, between December 26, 2018 and November 9, 2023. Patients were classified according to whether at least one reportable Candida species was included in the first eligible clinically issued BALF mNGS report. BALF mNGS-detected Candida was analyzed as a microbiological exposure variable rather than as evidence of invasive candidiasis. Baseline characteristics, co-detected organisms, and 28-day ICU mortality were compared between groups. Multivariable logistic regression, Cox proportional hazards regression, and propensity score matching were used to assess the association between BALF mNGS-detected Candida and mortality. Results Among 1,737 included patients, 583 patients (33.6%) had BALF mNGS-detected Candida . Candida albicans was the most frequently reported species, followed by Candida tropicalis and Candida glabrata . Before matching, patients with BALF mNGS-detected Candida had higher frequencies of viral co-detection, non- Candida fungal co-detection, and mixed microbial detection, whereas bacterial co-detection rates were similar between groups. The 28-day ICU mortality was higher in patients with BALF mNGS-detected Candida . In the full cohort, BALF mNGS-detected Candida remained associated with 28-day ICU mortality after multivariable adjustment in logistic regression and Cox regression analyses. After 1:1 propensity score matching, 583 matched pairs were analyzed. The mortality association persisted, whereas differences in viral, non- Candida fungal, and mixed microbial detection were substantially attenuated. Conclusions In ICU patients with severe pneumonia, BALF mNGS-detected Candida was associated with higher 28-day ICU mortality after adjustment for baseline confounders. This finding should be interpreted as a prognostic association rather than evidence that Candida caused pneumonia or contributed directly to death. BALF mNGS-detected Candida should not be considered direct evidence of invasive candidiasis or Candida pneumonia, nor should it be used as a stand-alone indication for antifungal therapy. Instead, it may identify a subgroup of critically ill patients with greater host vulnerability, airway microbial disturbance, and increased short-term clinical risk. Prospective studies incorporating standardized BALF sampling, longitudinal microbiological assessment, fungal biomarkers, antifungal treatment data, and adjudication of invasive candidiasis are needed to clarify its clinical relevance. Clinical trial number Not applicable.