Neonatology

Changes in the intestinal microbiome of the preterm baby associated with stopping non-invasive pressure support: a prospective cohort study

Abstract

Background Intestinal dysbiosis is implicated in the pathogenesis of necrotising enterocolitis and late-onset sepsis in preterm babies. The provision of non-invasive positive pressure ventilation is a common clinical intervention in preterm babies, and may be hypothesised to adversely affect intestinal bacterial growth, through increased aerophagia and induction of a hyperoxic intestinal environment; however this relationship has not been previously well characterised.

Methodology In this prospectively recruited cohort study, high-throughput 16S rRNA gene sequencing was combined with contemporaneous clinical data collection, to assess within-subject changes in microbiome development around the time of transitioning from non-invasive positive pressure respiratory support to unsupported spontaneous breathing.

Results In a group of 14 preterm infants, bacterial diversity was seen to increase by 0.34 units/week (inverse Simpson index) at the point of transitioning off non-invasive positive pressure respiratory support. Correspondingly, a significant increase in anaerobic genera (Bifidobacteria spp, Veillonella spp), and a non-significant fall in Enterobacteriaceae was also seen at this time.

Conclusions Provision of non-invasive positive pressure ventilation is associated with suppression of both diversity accrual and obligate anaerobic growth in the preterm intestine. This has clinical implications in view of the widespread use of non-invasive positive pressure ventilation in preterm neonatal care (and wider adult use), and demonstrates the need for potential strategies (eg, probiotic support; reduced aerophagia) to support the development of a healthy gut microbiome during this time.

What is already known on this topic

  • Intestinal dysbiosis is implicated in the pathogenesis of important neonatal conditions such as necrotising enterocolitis and late-onset sepsis.

  • Iatrogenic interventions such as antibiotic administration are known to induce intestinal dysbiosis.

What this study adds

  • The use of non-invasive positive pressure respiratory support is associated with impaired development of microbial diversity in the preterm intestine, and suppression of anaerobic colonisation, through induction of a theorised hyperoxic intestinal environment.

How this study might affect research, practice or policy

  • This may encourage the use of supplemental probiotic interventions to support eubiosis during times of non-invasive positive pressure respiratory support; or encourage the development of respiratory support interfaces which reduce the risk of aerophagia and intestinal hyperoxia.

Introduction

In view of the association between intestinal dysbiosis and the development of important preterm neonatal conditions such as necrotising enterocolitis (NEC) and late-onset sepsis (LOS), efforts have been undertaken to assess the impact that different components of clinical care may have on the developing preterm intestinal microbiome. Understandably, the main focus of these works has been on assessing the impact of more microbiologically plausible interventions, such as prebiotics, probiotics and antibiotics. However, there are other interventions which may also be theorised to affect preterm intestinal microbiome development.

One such component of clinical care which is less well reported in the literature is the relationship between preterm neonatal respiratory support, and the developing microbiome. While the respiratory and gastrointestinal systems may seem only tangentially related, respiratory interventions can impact on the intestinal intraluminal environment, as evidenced by the concept of ‘CPAP-belly’,1 whereby inhaled gases enter the gastrointestinal tract via aerophagia, potentially changing the environment from its oxygen-depleted natural state.2 Swallowed gas is one of the factors that influences intraintestinal gas content, and it can be reasonably hypothesised that if aerophagia is increased (as occurs in continuous positive airway pressure (CPAP) use), then the oxygen fraction in intraintestinal gas contents will similarly increase,3 and would be compounded through the use of hyperoxic inhaled gas (as frequently occurs in preterm babies).

Where the relationship between respiratory support and microbiome development has been examined, an increased relative abundance of aerobic and facultative anaerobic bacteria is seen in association with respiratory support provision.3–5 However, these studies either do not discriminate between invasive and non-invasive respiratory modalities, with their profoundly different effects on intestinal gas delivery; or have not differentiated the purported dysbiotic effects of mechanical ventilation from the recognised and concurrent dysbiotic effects of preterm life6–9 and the medical interventions associated therewith.10 A recent clinical trial in adults examining the effect of CPAP on intestinal dysbiosis (in obstructive sleep apnoea patients) failed to recruit sufficient numbers.11

With the advent of respiratory management strategies which advocate the use of non-invasive positive pressure ventilation in more extremely preterm infants and earlier in life,12 understanding the effects of this on the developing intestinal microbiome is increasingly important, in these infants more at-risk of the serious complications of preterm life.

This study aims to assess the potential changes in microbiome development following the introduction of discrete changes in respiratory management, in isolation (as much as is feasible) from other known or potential modulators of bacterial community development, through a high-fidelity assessment of changes in the intestinal microbial community at the time of respiratory support transition.

Methodology

Subject recruitment, sample collection and handling, library preparation, sequencing and bioinformatics processing were conducted as described in our previous study,10 and are included as online supplemental file 1. The wet-lab and analysis protocols are similarly attached as a online supplemental files 2 and 3, respectively.

Assessment of the contemporaneous clinical data revealed the many ways in which respiratory support modalities could change:

  1. Parameter changes within invasive respiratory support modes.

  2. Changes between invasive respiratory support modes.

  3. Extubation from invasive to non-invasive respiratory support modalities.

  4. Parameter changes within non-invasive respiratory support modes.

  5. Changes between non-invasive respiratory support modes.

  6. Changing from non-invasive respiratory support modes, to self-ventilation (without pressure support, but +/− supplementary oxygen).

A pragmatic decision was made to assess changes in microbiome progression around the point of conversion from non-invasive positive pressure respiratory support (ie, bi-level positive airway pressure (BIPAP); CPAP; or heated, humidified high-flow nasal cannulae (HHFNC)) to spontaneous breathing (either with or without supplemental oxygen). This particular transition was chosen as it occurred in a significant number of subjects (maximising the power to detect any differences); and as it took place at times of clinical stability (hence the decision to progress to lesser levels of support), there was a reduced incidence of concurrent microbiome modulators (eg, antibiotics), which would have confounded the findings. This clinical event also represented the key hypothesised change in the intestinal environment: the transition from a potentially hyperoxic environment in the presence of non-invasive positive pressure respiratory support, to the increasingly anaerobic environment of the gastrointestinal tract in a spontaneously breathing subject.

Episodes were identified using the following criteria:

  1. The transition must be from a period of non-invasive positive pressure respiratory support (eg, BIPAP/CPAP/HHFNC), to a period of self-ventilation.

  2. Sufficient (at least three) samples must be available from the periods prior to and after the transition from non-invasive positive pressure respiratory support to self- ventilation, in order to accurately reflect the microbiome progression during these periods.

  3. The period under analysis must occur at least 5 days clear of preceding antibiotic use, to reduce the risk of microbiome changes related to recent use and/or cessation of antibiotics being attributed to the change in respiratory support. The length of ‘washout’ period was pragmatically based on knowledge of antibiotic half-lives in this patient cohort, and our understanding of the effects of antibiotics on preterm intestinal development.10

  4. The periods under analysis must be within 10 days either side of the respiratory support transition point.

Where antibiotics had been used within 5 days of the 10-day period prior to the respiratory support transition, the episode could still be assessed provided at least three samples were available from the pretransition period, which were taken greater than 5 days after antibiotics ceased.

Parameters of microbiome progression were studied in the pretransition and post-transition period. Longitudinal diversity progression and longitudinal progression of Enterobacteriaceae relative abundance were seen to follow roughly linear trajectories, and are quantitatively described. Using the regression coefficients of these parameters as the summary statistic for each individual’s progression prerespiratory and post respiratory support transition, a paired analysis (using t-test or a Wilcoxon signed-rank test, as appropriate) was performed to assess whether the rate of diversity progression (units/week) and rate of Enterobacteriaceae relative abundance progression (%/week) differed prerespiratory and post-respiratory support transition (see figure 1).

Figure 1
Figure 1

Idealised diagram demonstrating the changing rate of diversity progression after a clinical intervention. Black dots represent individual sample alpha diversities. Black lines represent regression lines through those points. Grey dashed lines represent translation of regression lines to demonstrate adjacent comparison. Difference in rates of diversity progression is represented by red angle.

Ordination analyses (using adonis and permutest functions within the vegan package in R13) were undertaken to assess for changes in the microbiome composition around the time of transitioning off non-invasive positive pressure support. Representative paired samples from the pretransition and post-transition periods were identified from each subject; an NMDS (non-metric multidimensional scaling) plot constructed, and the DESeq2 R package14 used to identify taxa which differed significantly between the two periods.

Patient and public involvement

Patients or the public were not involved in the design, or conduct, or reporting or dissemination plans of our research. However, research questions which this work contributes to were identified as priorities 5 and 7 in a recent paper, which heavily incorporated patient and public involvement, addressing research priorities in neonatology.15

Results

Study population and samples

Of 158 infants admitted to the neonatal unit during the period under assessment, 76 were successfully recruited after informed consent was obtained. Of these 76, there were 5 subjects who did not remain in the study long enough to produce a stool sample, so microbial community data were available from 71 infants (see figure 2). Of these infants, 14 met the eligibility criteria for inclusion for assessment of microbiome development around the time of cessation of positive pressure respiratory support.

Figure 2
Figure 2

Recruitment diagram.

Transition off non-invasive respiratory support occurred at a median of 34 days postbirth (range: 14–62), at a mean corrected gestational age of 33/40 (29+3–35+5)—see table 1. Stool samples obtained around this period of transition allowed analyses to be conducted on 143 samples (median 10 samples/subject)—see figures 3 and 4.

Figure 3
Figure 3

Individual profiles of longitudinal diversity progression versus postnatal age, over the period of transition from non-invasive respiratory support to self-ventilation. Vertical hashed red line indicates point of respiratory support transition.

Figure 4
Figure 4

Longitudinal taxonomic profiles of the majority taxa in babies around the period of transitioning from non-invasive respiratory support to self-ventilation. Red hashed line represents point of respiratory support transition.

Table 1
|
- Demographics and clinical data of study population

Following the cessation of non-invasive positive pressure respiratory support, the rate of diversity progression was seen to increase by a median of 0.34 units (as measured by the inverse Simpson index) per week (95% CI: 0.04 to 1.88, p=0.04)—see figure 5.

Figure 5
Figure 5

Composite plot demonstrating the changing rate of diversity progression in the period prior to and after transition from non-invasive respiratory support to self-ventilation. Thin grey lines represent changes in individuals’ rate of diversity progression; thick black line represents median change in diversity progression (dotted lines represent 95% CI).

Correspondingly, the rate of Enterobacteriaceae accumulation was seen to fall by 16.5%/week (95% CI −3.7–36.6), although this did not meet statistical significance (p=0.1) – —see figure 6.

Figure 6
Figure 6

Composite plot demonstrating the changing rate of Enterobacteriaceae relative abundance progression in the period prior to and after transition from non-invasive respiratory support to self-ventilation. Thin grey lines represent changes in individuals’ rate of Enterobacteriaceae relative abundance progression; thick black line represents mean change in diversity progression (dotted lines represent 95% CI).

No significant difference was seen in microbiome composition around the time of transitioning off non-invasive positive pressure respiratory support (PERMANOVA (Permutational multivariate analysis of variance) test p=0.74; dispersion p=0.74)—see figure 7. However, differences were seen in some individual taxa, with amplicon sequence variants corresponding to Bifidobacteria spp and Veillonellaceae spp. significantly increased in abundance following transition to self-ventilation.

Figure 7
Figure 7

Ordination plot demonstrating the changing composition of the microbiome after transitioning from non-invasive respiratory support to self-ventilation.

Discussion

In this study, it has been shown that following the transition from non-invasive positive pressure respiratory support to unsupported spontaneous breathing, there is an associated increase in diversity accrual by 0.34 units/week (p=0.04); a fall in the relative abundance of Enterobacteriaceae (not meeting statistical significance), and a significant increase in the relative abundance of some anaerobic species (ie, Bifidobacteria spp and Veillonella spp). These findings support the hypothesis that aerophagia and induction of a hyperoxic intraluminal intestinal environment, secondary to provision of non-invasive positive pressure support, suppresses the growth of obligate anaerobic bacteria in the gut, in favour of continued colonisation with aerobic and facultative anaerobic taxa. This represents a deviation from the natural development of the preterm intestinal microbome,10 with the potential loss of the symbiotic action and positive effects on gut health conferred by these commensal anaerobic organisms.16

This is a significant finding, as while it may not influence the decision to progress to non-invasive respiratory support (in view of the clear holistic advantages of this in the care of preterm infants), it may prompt more targeted support of probiotic bacterial communities (eg, with facultative anaerobic probiotic bacteria, or greater abundances of anaerobic organisms); influence the rapidity with which the non-invasive respiratory support is weaned, or affect the mode of delivering non-invasive respiratory support, in favour of strategies or technologies which reduce aerophagia. These microbiome signatures of the theorised hyperoxic distal gut align with the dysbiotic origins of important neonatal conditions such as NEC. Increased relative abundance of Enterobacteriaceae is consistently associated with the subsequent development of NEC17–19 and is plausibly linked to proposed pathophysiological mechanisms underlying this condition.10 Similarly, an increased relative abundance of obligate anaerobes (of which Bifidobacteria and Veillonella are representative taxa) is associated with a healthy preterm gut biosis.19 20 While bifidobacterial species are widely employed in neonatal probiotics, species within Veillonellaceae have not been proposed as a component of neonatal probiotic formulations. The recurrent detection of this taxon within apparently healthy gut microbiota7 21 (and its susceptibility to diminution through clinical interventions such as antibiotic use10 22 and non-invasive positive pressure ventilation), may venture the exploration of the potential role and mechanism of action of Veillonellaceae within neonatal probiotic supplements. Veillonella spp in adult studies have been characteristically associated with the oral microbiome,23 but can be seen to penetrate deeper into the gastrointestinal tract through diminution of the chemical barrier within the stomach (through medical24 or surgical25 means); the naturally less acidic nature of the preterm infant gastric fluid26 may similarly permit this. Veillonella spp have been shown to exhibit activity that may be considered probiotic in nature, through the production of short-chain fatty acids,27 which are associated with gut health, acting as a nutritional source for enterocytes28 and promoting intestinal growth—this may explain their presence in the microbiome of preterm infants with a healthy gut state.

At the conception of this work, it was envisaged that changes in the developing microbiome would be examined at the transition from invasive respiratory support (where there is minimal additional passage of ventilated gases into the gastrointestinal tract), to non-invasive respiratory support—that is, at extubation. This typically occurs within the higher risk period for the development of LOS and NEC, and thus changes in the intestinal microbiome would have been of especial interest, in light of the proposed dysbiotic origins to NEC. However, in recognition of previous work demonstrating the profound impact which antibiotic administration had on the developing microbiome,10 it was felt that to detect the ‘signal’ of microbiome changes truly related to changing respiratory support, this assessment would need to take place separate from the ‘noise’ of antibiotic-mediated dysbiosis. Correspondingly, as many extubations occur early in postnatal life, in accordance with the generally preferred strategy in neonatology of an early transition to non-invasive respiratory support, samples from around this period were frequently devoid of detectable bacterial DNA; were obtained in close proximity, or in the midst of, the postnatal antibiotic course; or there were insufficient numbers of samples prior to the respiratory transition (due to initial delays in the passage of stool), to allow such assessments.

Consequently, the converse transition was studied: the progression from non-invasive positive pressure support to self-ventilation; that is, the progression from a theoretically aerated bowel environment, to a more natural anaerobic state, with the intention that these findings could be extrapolated to the transposed situation of extubation to non-invasive respiratory support. This allowed an increased number of subjects and clinical episodes to be potentially assessed and as clinical stability is generally required prior to the cessation of non-invasive respiratory support, fewer subjects were excluded on the basis of recent or concurrent antibiotic use.

All modes of non-invasive positive pressure respiratory support were included (and analysed as one intervention) and self-ventilation was regarded as a uniform status, regardless of supplementary oxygen, in view of the proposed pressure-based mechanism of increased air±oxygen in the intestine, rather than due to the fraction of inspired oxygen. Further studies may demonstrate differences between types of positive pressure modalities, or allow detection of a difference in magnitude of effect in a dose-dependent relationship depending on pressure levels or flow rates. It may also be considered that positive pressure and oxygen delivery may be separately and independently associated with changes in microbiome development. Due to the considerable overlap between these two aspects of non-invasive respiratory support, large cohorts would be required to analyse these factors in isolation.

While efforts were undertaken to assess the effect on microbiome development of changing ventilation modality in isolation from other known modulators of the microbiome (eg, antibiotics), it is possible that other changes in clinical care were similarly influencing microbiome development over this time period (eg, feeding route). Post-hoc assessment of medications received during the studied periods identified no consistent occurrences of likely microbiome modulators (being largely limited to standard electrolyte and vitamin supplementation), and no subjects within the group transitioned to breastfeeding from tube feeding during this period. It was noted that during the period of assessment, 7/14 subjects were in the process of either establishing full enteral feeds, or re-establishing full enteral feeds, following a period of feed cessation. However, on post-hoc assessment, no significant difference was noted in the change in diversity progression between infants whose transition in ventilation modality occurred during a stable feeding pattern, and those in whom it had taken place during a dynamic feeding progression. Nevertheless, further investigations on larger patient cohorts should allow the study of such ventilation transitions in isolation from these potentially confounding feeding changes, although elimination of all theorised (and unsuspected) possible confounding microbiome modulators is likely to be prohibitively difficult. Additionally, larger patient cohorts may necessarily require multisite involvement, which would help to negate the susceptibility that single-site studies such as this have in describing microbiome changes that might reflect local practice and colonisation patterns.

Conclusion

This study has demonstrated the feasibility and strength of using an individual-subject approach and paired-analysis (with each subject as their own control) to determine the effects that changes in respiratory support may have on the developing gut microbiome in preterm infants. In parallel with the use of strict criteria to allow the study of this transition in as much isolation from other microbiome modulators as is clinically feasible, these approaches have allowed the detection of more subtle changes, which may be concealed by the more profound effects of other clinical interventions (ie, antibiotics).

The most notable finding has been of a dysbiotic effect seen in association with the use of non-invasive positive pressure ventilation—this has not previously been demonstrated independently from other clinical factors in preterm infants. This is of particular importance, considering the widespread use of this form of respiratory support in this patient population, more so during the highest risk period for the development of NEC and LOS. This should prompt further studies to assess the extent to which this phenomenon is seen with different degrees and modes of non-invasive ventilation, and to propose strategies to combat it.