Day-21 gut microbiota community state types are associated with bronchopulmonary dysplasia classification in preterm infants: a pilot shotgun metagenomic study.

Introduction: Bronchopulmonary dysplasia (BPD) is a major complication in preterm infants, and its clinical classification remains strongly influenced by gestational maturity and the evolving respiratory course. In this exploratory study, we investigated whether early-life gut microbiota configurations at postnatal day 21 are associated with subsequent formal BPD classification at 36 weeks postmenstrual age and whether they provide ecological information relevant to preterm infant microbiome development. Methods: In a prospective cohort of 23 preterm infants with gestational age <32 weeks or birth weight <1,500 g, shotgun metagenomic sequencing of day-21 fecal samples was performed. Community state types (CSTs) were identified using unsupervised clustering, and their taxonomic, functional, and exploratory discrimination patterns were assessed in relation to subsequent BPD classification. Results: Two CSTs were identified. CST1 was dominated by commensal taxa and exhibited functional enrichment in metabolic homeostasis pathways. CST2 was characterized by pathobionts, fragmented taxon-pathway association networks, and enrichment in virulence-related pathways. BPD was observed in 1 of 11 CST1 infants and 7 of 12 CST2 infants (9.1% vs. 58.3%; two-sided Fisher's exact test, p = 0.027), although this association was based on very small cell counts. In exploratory discrimination analysis, a model combining CST status with gestational age showed an apparent AUC of 0.892; however, leave-one-out cross-validation yielded a lower AUC of 0.800, indicating likely optimism in the apparent model performance. Discussion: These preliminary, observational findings suggest that day-21 gut microbiota profiles and CST classification may provide ecological information associated with formal BPD classification. However, this analysis should be interpreted as exploratory discrimination rather than validation of a clinically useful prediction model. Establishing causality or clinical utility requires validation in larger cohorts that systematically track longitudinal confounders such as gestational age, feeding mode, antibiotics, and probiotics.