Longitudinal dynamics of respiratory microbiome composition in infants after new tracheostomy placement

Abstract

Objectives

This prospective longitudinal study characterised respiratory microbiome dynamics following new tracheostomy placement among infants.

Setting

A tertiary care paediatric hospital system in the United States.

Participants

Fifteen infants ≤12 months of age contributed 84 tracheal aspirate samples collected from day 1 through 3–4 months post-placement.

Primary and secondary outcome measures

Bacterial composition, including abundance, from 16S rRNA gene sequencing; alpha and beta diversity measures over time.

Results

16S rRNA gene sequencing revealed immediate and sustained bacterial community shifts. Staphylococcus abundance increased and alpha diversity decreased in the first 30 days post-tracheostomy (p<0.05) before returning to baseline. Beta diversity demonstrated compositional changes immediately and with ongoing divergence through 3–4 months. Time and clinical factors (prematurity, ventilation and neurologic impairment) were significantly associated with microbiome structure (p=0.001).

Conclusions

This study provides novel evidence that new tracheostomy placement induces rapid and prolonged airway microbiome disruption in infants, highlighting a previously uncharacterised window of vulnerability with implications for respiratory health.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• Among infants, new tracheostomy placement leads to a transient but predictable increase in Staphylococcus abundance, followed by ongoing, continued disruptions in the infant respiratory microbiome community structure measured up to 3–4 months post-placement.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• In clinical practice, past tracheostomy aspirate samples may not be reflective of current microbiology or antibiotic needs in these infants. Future work should explore the clinical outcomes associated with ongoing microbiome community alterations, especially during the critical period of microbiome development in early life.

Introduction

There is emerging evidence for abnormal respiratory microbiomes among children with artificial airways, including both new and long-term tracheostomies and endotracheal tubes.^{1 2} However, the temporal dynamics of that dysbiosis, including onset and chronicity, have not been well-described. Defining these dynamics for infants with new tracheostomies and their causes and consequences could help identify strategies to promote microbiome health and clinical stability, especially during the critical period of microbiome development in early life.

In this study, we aimed to identify the temporal dynamics and clinical associations of tracheostomy aspirate (TA) microbiomes among a cohort of infants followed longitudinally after new tracheostomy placement.

Methods

This was a prospective study at Children’s Hospital Los Angeles from 2020 to 2021. We enrolled hospitalised infants ≤12 months old undergoing new tracheostomy placement and collected serial TA samples shortly following tracheostomy placement on day 1 or 2 (ie, postoperative day 0, 1 or 2), at first tracheostomy change (day 7) and approximately every 2 weeks thereafter during initial hospitalisation. Samples were collected immediately following tracheostomy changes and stored at −80 °C. Clinical characteristics were extracted from medical records. Institutional review board approval was obtained and parents provided written informed consent. Patients were not directly involved in formulating the research question, choosing the study design or in the interpretation of results. The authors fully support involving patients and members of the public in conducting medical research, and although no funding was available for this purpose in our study, the authors will pursue patient and public involvement in future studies.

DNA was extracted from samples using QIAamp PowerFecal Pro DNA Kits (Qiagen) via the Qiacube platform using manufacturer’s protocol. Bacterial DNA from the V4 region of 16S ribosomal gene was PCR-amplified using shared forward primer 806rB; each sample had unique reverse primers.^{3 4} Samples were pooled and sequenced using custom primers by paired-end Illumina MiSeq (2×150 bp). Reads were processed using DADA2 V.1.5.2 and taxonomies assigned using RDP naïve Bayes classifier.⁵ Diversity metrics were computed using genus taxa counts.

Taxon-level analyses were performed using linear mixed-effects models with log-transformed relative abundances (log₁₀[x+1e−5]) and clinical variables (chronic or home mechanical ventilation (HMV), preterm birth, neurologic impairment, upper airway obstruction and craniofacial anomaly). Models included time as a covariate and random intercept to account for repeated measures. A linear mixed-effects model was used to identify differences in each infant’s change from baseline in Shannon alpha diversity over time, including fixed effects for use of HMV (defined as ventilator use at discharge), gestational age category, sex, age at tracheostomy, presence of neurologic impairment, upper airway obstruction and craniofacial diagnoses, as well as a random intercept for infant. Principal coordinates analysis (PCoA) plots with a PERMANOVA model evaluated associations between time with tracheostomy and Bray-Curtis dissimilarity, accounting for key clinical factors from marginal testing and repeated measures with stratification. Analyses used ‘phyloseq’(V.1.20.0) and ‘vegan’(V.2.5-4) in R V.4.4.1.^6–8

Results

Among the 15 infants included, most were female (60%) and Latino/a/x/Hispanic (60%) or Black (33%, (online supplemental table1). Eight infants (53%) were preterm, and median gestational age was 36 weeks (IQR 32 to 38.5). Indications for tracheostomy included upper airway obstruction (40%), neurologic impairment (40%), craniofacial anomalies (20%) and bronchopulmonary dysplasia (20%), with some having multiple indications; 40% had other indication (eg, chronic respiratory failure). Median age at tracheostomy was 5 months (IQR 2.5–7). All infants were discharged home, 67% using HMV. Acute respiratory illnesses were rare (n=1) in this COVID-era cohort.

Median hospital stay and sampling duration was 67 days (IQR 29 to 118.5); 84 samples were of sufficient quality for analysis, with a median of 6 samples per infant (range 3–11, (online supplemental figure 1). Among samples, 12 (14%) were collected on day 1 after tracheostomy placement, 13 (16%) at Week 1, 17 (20%) at Weeks 2–4, 22 (26%) at Month 2, 8 (10%) at Month 3 and 12 (14%) at Months 4–6.

The most abundant genera identified included Staphylococcus, Streptococcus, Klebsiella, Neisseria, Pseudomonas and Stenotrophomonas (figure 1). Only three infants had samples with >1% relative abundance of Pseudomonas, all with HMV. Staphylococcus relative abundance increased following tracheostomy, peaking at 40 days and mean relative abundance of 27%, followed by a subsequent decrease; this was more pronounced among infants without HMV (57% vs 20% peak, figure 2). In mixed-effects models, HMV was associated with 3.7-fold lower mean Staphylococcus abundance (p=0.059), while preterm birth was associated with substantially higher Klebsiella abundance (p<0.001, figure 1 and online supplemental figure 2). Neurologic impairment was associated with 10.1-fold higher Neisseria abundance (p=0.0096). Mean Pseudomonas abundance had no clinical association. Longitudinal analysis of alpha (within-sample) diversity using Shannon index identified a significant decrease in median alpha diversity from baseline in the first 30 days (p=0.017 day 7–13; p=0.040 day 13–30, figure 3). By 61–90 days, median alpha diversity returned to baseline (p=0.11 day 31–60; p=0.55 day 61–90).

Figure 1. Bacterial composition of tracheostomy aspirate samples. (a) Stacked bar plot of mean relative abundance of the top 20 bacterial genera across time categories relative to tracheostomy placement. Sample size (n) and number of contributing participants (N) are shown beneath each category label, (b) Table summarising the same top 20 genera ranked by overall mean relative abundance across the full dataset, along with their prevalence (ie, proportion of samples in which each genus is detected) and (c) Forest plot showing effect estimates (β) from mixed-effects models of log-transformed relative abundance with key clinical factors, accounting for repeated measures and time since tracheostomy. Points represent model estimates and horizontal lines indicate 95% CIs.

Figure 2. Staphylococcus relative abundance in tracheostomy aspirates over time following tracheostomy. Lines represent differences in Staphylococcus relative abundance in TA samples over time for individual infants (light lines) and for overall cohorts (dark lines). Grey shading denotes the 95% CI of the Overall Cohort Median. Staphylococcus relative abundance increased following tracheostomy placement, peaking at 40 days with a mean relative abundance of 27%, followed by a subsequent decrease; this increase was more pronounced among infants with no home mechanical ventilation (peak 57% No HMV vs. peak 20% Using HMV). Online supplemental figure 2 presents all modelled associations, including those not reaching statistical significance. HMV, home mechanical ventilation; TA, tracheostomy aspirate.

Figure 3. Difference in Alpha Diversity (Shannon) from baseline sample following tracheostomy. Lines represent differences in tracheostomy aspirate Shannon index alpha diversity relative to baseline samples for individual infants. Positive values reflect increased diversity compared to baseline, while negative values indicate decreased diversity. Grey shading denotes the 95% CI of the Overall Cohort Median. There was a statistically significant decrease in alpha diversity in the first 30 days following tracheostomy (p=0.017 for time category Day 7–13, p=0.040 for category Day 13–30). This decrease in alpha diversity largely returned to baseline by 60–90 days, and these later time points were not significantly different from baseline.

PCoA plots of beta (between-sample) diversity using Bray-Curtis dissimilarity demonstrated that samples collected immediately following tracheostomy clustered together (Day 1–2, figure 4), indicating relatively similar microbiome compositions, followed by divergence by Day 14–30 and some return of clustering by Day 61–90. PERMANOVA showed time post-tracheostomy was significantly associated with community structure when accounting for clinical variables (p=0.001, figure 4). Biplots indicated Day 14–30 samples had higher abundances in Staphylococcus. Gestational age, HMV, neurologic impairment and upper airway obstruction remained significantly associated with microbiome composition (ie, beta diversity) via PERMANOVA (all p=0.001). Over time, microbiome communities rapidly shifted into communities dissimilar to baseline and continued to have ongoing composition changes with each successive sampling timepoint post-tracheostomy without stabilisation or return to baseline (figure 5).

Figure 4. Tracheostomy aspirate (TA) microbiome communities diverge in the first month after tracheostomy placement. (a) Principal coordinates analysis (PCoA) with Biplot vectors of TA microbial communities using the Bray-Curtis dissimilarity index. Points represent individual sample microbiota; colour represents days since tracheostomy. Ellipses indicate 95% CIs for group centroids. Vectors represent bacterial genera that are significantly associated with microbial community composition (p<0.05). Beta-diversity measures over time demonstrated clustering immediately following tracheostomy (Day 1–2, purple, inclusive of 12 samples), followed by notable shifts by Day 14–30 (green, 17 samples) and again by Day 60–90 (yellow, 8 samples). Biplots (vector arrows) indicated that Day 1–2 communities had abundance of Streptococcus while Day 14–30 communities moved transiently to higher abundance of Staphylococcus and Klebsiella. (b) A PERMANOVA analysis based on Bray-Curtis dissimilarity index showed that time post-tracheostomy was significantly associated with bacterial community structure when accounting for other clinical variables (p=0.001). Prematurity, home mechanical ventilation, neurologic impairment and upper airway obstruction were also statistically significantly associated with bacterial composition (all p=0.001). Bronchopulmonary dysplasia diagnosis was not included in the model as it was collinear with neurologic impairment. The model included key clinical factors from marginal testing and accounted for repeated measures with stratification. The by =‘terms’ argument was used to sequentially test each clinical predictor factor in the model, providing marginal effects for each predictor.

Figure 5. Tracheostomy aspirate (TA) microbiome communities have ongoing changes with each sampling time point during the first months after tracheostomy placement. Differences in Bray-Curtis dissimilarity over time following tracheostomy. Lines represent differences in Bray-Curtis dissimilarity for individual infants from baseline samples (a) and between successive samples (b). A difference value of 0 indicates complete microbiome community similarity between an individual’s samples, while 1.0 indicates complete dissimilarity between samples. Grey shading denotes the 95% CI of the Overall Cohort Median. (a and b) Within the first week post-tracheostomy, TA microbiome communities rapidly shifted into >75% dissimilar compositions from baseline, as evidenced by Difference in Bray-Curtis values >0.75. (a) After initial microbiome composition shifts post-tracheostomy, TA microbiome communities did not return to compositions similar to those of baseline, but rather remained dissimilar throughout the sampling period. (b) Infants had ongoing community changes with each successive sampling time point, without achieving compositional similarity.

Discussion

We identified significant shifts in respiratory sample microbiomes in infants within the first weeks after new tracheostomy, including a transient bloom in Staphylococcus abundance. Although bacterial richness and evenness largely returned to baseline, there were ongoing microbiome community composition changes through 3–4 months. These patterns occurred across gestational age and HMV status. Together, our findings suggest that tracheostomy placement leads to immediate and sustained microbiome community disruptions in the infant airway.

Among healthy infants, the oropharyngeal and lung microbiomes are dominated by Streptococcus, Neisseria, Haemophilus and Veillonella after the first week of life.^9–11 Therefore, the transient increase in Staphylococcus and declines in diversity that we observed likely reflect effects of tracheostomy tube placement, including entry of skin flora via the stoma, perioperative antibiotics, and altered mucociliary clearance. A similar increase in Staphylococcus was also reported in children following new tracheostomy placement and following intubation among infants with pneumonia.^{1 2} Blooms of other resident bacteria have also been described among children with tracheostomies during viral infection.¹² Differential diversity recovery among gestational age and HMV groups suggests that these factors influence microbiological trajectories.

It is not known if the continual shifts in microbial communities we observed at 3 months persist beyond this timepoint. Powell et al also documented shifts in beta diversity during the 3 months following new tracheostomy, with some dysbiosis noted even prior to placement.¹ If dysbiosis does persist after 3 months, that could have lasting implications for airway health, and as such could influence the occurrence and severity of respiratory illness given the respiratory microbiome’s role in shaping respiratory immunity and inflammatory responses.^{1 13} Understanding the role of these dynamic community structures in resilience or vulnerability will be critical for predicting disease trajectory and guiding individualised, long-term management strategies.

This study has several limitations, including a small patient cohort and single-centre design, lack of control group, lack of paired bacterial culture data which may limit clinical interpretation of microbiome profiles and potential residual confounding from unmeasured variables. Although our findings demonstrate a clear early disruption in airway microbial communities following tracheostomy, later timepoint observations should be interpreted with caution, as they reflect a smaller and potentially non-representative subset of participants due to differential lengths of hospitalisation, potentially introducing bias and limiting interpretation of longer-term trends. These limitations are balanced by notable strengths: a longitudinal design, dense sampling and low respiratory illness prevalence. Notably, our individual-level analysis contrasts with previous group-level studies and underscores the importance of analysing microbiome temporal dynamics at the individual level.^{1 14} The racial and ethnic composition of this cohort reflects the patient population served by their institution, which predominantly includes children from historically underrepresented groups and should not be interpreted as evidence of differential access to tracheostomy care.

In summary, tracheostomy placement leads to both short- and longer-term disruptions in the infant respiratory microbiome. Future work should explore the clinical outcomes associated with microbiome community alterations, potentially leading to targeted treatment approaches.

Data availability statement

Data are available upon reasonable request.