Incidence of necrotising enterocolitis before and after introducing routine prophylactic Lactobacillus and Bifidobacterium probiotics

Objective To compare rates of necrotising enterocolitis (NEC), late-onset sepsis, and mortality in 5-year epochs before and after implementation of routine daily multistrain probiotics administration in high-risk neonates. Design Single-centre retrospective observational study over the 10-year period from 1 January 2008 to 31 December 2017. Setting Level 3 neonatal intensive care unit (NICU) of the Norfolk and Norwich University Hospital, UK. Patients Preterm neonates at high risk of NEC: admitted to NICU within 3 days of birth at <32 weeks’ gestation or at 32–36 weeks’ gestation and of birth weight <1500 g. Intervention Prior to 1 January 2013 probiotics were not used. Thereafter, dual-species Lactobacillus acidophilus and Bifidobacterium bifidum combination probiotics were routinely administered daily to high-risk neonates; from April 2016 triple-species probiotics (L. acidophilus, B. bifidum, and B. longum subspecies infantis) were used. Main outcome measures Incidence of NEC (modified Bell’s stage 2a or greater), late-onset sepsis, and mortality. Results Rates of NEC fell from 7.5% (35/469 neonates) in the pre-implementation epoch to 3.1% (16/513 neonates) in the routine probiotics epoch (adjusted sub-hazard ratio=0.44, 95% CI 0.23 to 0.85, p=0.014). The more than halving of NEC rates after probiotics introduction was independent of any measured covariates, including breast milk feeding rates. Cases of late-onset sepsis fell from 106/469 (22.6%) to 59/513 (11.5%) (p<0.0001), and there was no episode of sepsis due to Lactobacillus or Bifidobacterium. All-cause mortality also fell in the routine probiotics epoch, from 67/469 (14.3%) to 47/513 (9.2%), although this was not statistically significant after multivariable adjustment (adjusted sub-hazard ratio=0.74, 95% CI 0.49 to 1.12, p=0.155). Conclusions Administration of multispecies Lactobacillus and Bifidobacterium probiotics has been associated with a significantly decreased risk of NEC and late-onset sepsis in our neonatal unit, and no safety issues. Our data are consistent with routine use of Lactobacillus and Bifidobacterium combination probiotics having a beneficial effect on NEC prevention in very preterm neonates.


What is already known on this topic?
► Necrotising enterocolitis (NEC) remains a leading cause of mortality and morbidity in premature and very low birthweight infants. ► The aetiology of NEC is multifactorial, but the development of an abnormal gut microbiota is an important predisposing risk factor. ► Multiple meta-analyses and observational studies have shown that ameliorating gut bacterial colonisation through the enteral administration of live probiotic bacteria significantly reduces NEC incidence.

What this study adds?
► Routine daily dosing with Lactobacillus and Bifidobacterium probiotics was a cheap, simple intervention associated with significantly decreased NEC, surgical NEC and late-onset sepsis in our neonatal unit. ► No episode of sepsis due to Lactobacillus or Bifidobacterium occurred in the routine probiotic supplementation epoch. ► The apparent impact of reduced NEC incidence with probiotics was particularly pronounced in the first 2 weeks postnatal, implying that achieving early probiotic bacterial gut colonisation is crucial.

InTRODuCTIOn
Necrotising enterocolitis (NEC) is a leading cause of mortality and morbidity in premature very low birthweight (VLBW; <1500 g) babies. 1 2 Prevention of NEC is a top UK research priority. 3 NEC aetiology is multifactorial, 4 with abnormal gut microbiota increasingly recognised as central. 5 Probiotics are 'live microorganisms that, when administered in adequate amounts, confer a health benefit on the host'. 6 Multiple randomised controlled trials (RCTs), systematic reviews and meta-analyses of RCTs and observational studies have shown that prophylactic probiotics prevent NEC in preterm babies. [7][8][9] A 2012 meta-analysis, 10 updated in 2017 to include 25 RCTs and >7000 neonates, 11 showed strong evidence for using multispecies probiotics to reduce NEC incidence (pooled OR=0.36, 95% CI  12 Uncertainties about effectiveness may arise from RCTs which have not shown benefit, heterogeneity, multiplicity, and dosage of probiotic microorganisms, differing inclusion criteria, treatment durations and NEC definitions used to assess benefit. 13 There are no published reports on the safety or efficacy of routine probiotics used in any UK centres. We therefore aimed to assess whether routine probiotic supplementation has been associated with a decreased incidence of NEC, sepsis, and all-cause mortality in our centre.

MeThODS
Single-centre retrospective review of neonates admitted to our tertiary-level neonatal intensive care unit (NICU) which caters to ~6000 deliveries annually and is a regional neonatal surgical centre.

eligible population
'Eligible high-risk neonates' comprised all those inborn at <32 weeks' gestation, plus 32-36 weeks' gestation VLBW infants. Outborn babies were included if transferred in within 72 hours of birth, providing no abdominal concerns at referral.

Study period
Eligible neonates were reviewed in two consecutive 5-year epochs, allocated by date of birth: 1 January 2008 to 31 December 2012 (pre-probiotics epoch) and 1 January 2013 to 31 December 2017 (routine probiotic epoch).

enteral feeding practices
Our unit adopted a standardised regional enteral feeding guideline for preterm neonates in January 2011. Ideally, human milk feeds commenced on day 1. From May 2013, donor breast milk (DBM) was available to supplement shortfalls in mother's own breast milk supply before full feeds. Cow's milk-based fortifier was added to breast milk between full enteral feeds (≥150 mL/ kg/day) and discharge-a consistent policy across epochs. Milk type used in the period until full feeds was categorised as exclusive maternal breast milk, exclusive DBM, formula, or mixed (ie, any combination of own maternal milk, DBM, and formula).

Probiotic exposure and compliance
We introduced routine daily combination Lactobacillus and Bifidobacterium probiotics for prophylaxis of NEC in high-risk babies resident in our NICU on 1 January 2013, and for all eligible infants subsequently admitted. 14 Written parental information was provided. 15 We administered first probiotic dose with first colostrum feeds, ideally on postnatal day 1. Probiotics continued until ~34 weeks postmenstrual age. We initially used the dual-species probiotic Infloran capsules (Desma Healthcare, Chiasso, Switzerland), containing Lactobacillus acidophilus and Bifidobacterium bifidum, half a capsule (125 mg) twice daily (1×10 9 colony-forming units (CFU) of each bacterial species daily). From April 2016, we used triple-species Labinic Drops (Biofloratech, Walton-on-Thames, UK), four drops once daily (~0.5×10 9 CFU dosage each of L. acidophilus, B. bifidum and B. longum subspecies infantis daily). Probiotic batches were subject to quality control (LJH's laboratory) for species and CFU confirmation alongside our prospective microbiota study. 16 Probiotics were prescribed on drug charts in accordance with our NICU's guideline, and compliance was good. 17

Oxygen saturation limits
Between 2008 and May 2011, we targeted oxygen saturation (SaO 2 ) range 84%-92% for babies aged <32 weeks. From 31 Original research Figure 1 The cumulative incidence of necrotising enterocolitis (NEC) from date of birth stratified by epoch. Risk table shows the number of cases still at risk with each passing week from birth, as well as the cumulative numbers of those diagnosed with NEC, who died without a diagnosis of NEC, or who were discharged home without a diagnosis of NEC. One NEC case was censored in this time-to-event analysis, a baby born in the month before 1 January 2013 who developed NEC after that date, leaving 34 cases in the pre-probiotics epoch and 16 cases in the routine probiotics epoch included.

Identification and classification of neC cases
From BadgerNet electronic patient records (CleverMed, UK), we identified all cases of definite, suspected, and perforated NEC. All eligible infants were screened between admission and final discharge home (or death if earlier) from either our NICU/paediatric ward, or step-down NICU/other hospital ward if transferred out. Electronic clinical records, including X-ray request forms, abdominal radiographs, radiologist reports, blood results, surgical notes, histopathology and autopsy reports, and death certificates were independently reviewed by two unblinded clinicians (CR, PC) to determine whether they met NEC case definition according to each of three diagnostic systems: modified Bell's staging criteria, 19 Vermont-Oxford Network, 20 and Battersby et al. 1 We defined 'surgical NEC' as any baby with definite NEC in whom an abdominal drain was inserted or who underwent laparotomy. Potential surgical NEC cases whose operative findings and/or histopathology reports were uncertain or required expert interpretation, and also those with uncertainty regarding staging, were additionally reviewed and adjudicated by a consultant paediatric surgeon (AM) blinded to epoch.
We imputed as either NEC Bell's stage 2b/3a cases babies who had acute clinical deterioration with Bell's stage 2 or 3 systemic and gastrointestinal signs (eg, blood in stool, abdominal distension, tenderness, acidosis), but without classic stage 2/3 radiographic findings of pneumatosis intestinalis, portal venous gas or pneumoperitoneum, provided treated clinically as presumed NEC with ≥10 days' antibiotics and nil enterally and with fixed distended loops, definite ascites or gasless abdomen on abdominal X-ray and no alternative diagnosis (eg, sepsis) for the episode. These included babies who underwent surgical drain placement who died before planned laparotomy without autopsy, but with NEC recorded on their death certificate.
Isolated pneumoperitoneum without other clinical/radiographic evidence of NEC was diagnosed as spontaneous intestinal perforation, not NEC. Similarly, any case who satisfied any staging/clinical-radiological definition of NEC but who at proximate surgery or postmortem examination for that episode had no NEC findings was excluded as NEC.

Identification and classification of late-onset sepsis
Late-onset sepsis was defined as clinical signs of sepsis and a concomitant positive blood or cerebrospinal fluid culture occurring >72 hours after birth and before 46 weeks' postmenstrual age or discharge/transfer from our NICU if sooner. We reviewed all positive blood and CSF cultures from our local microbiology laboratory database for the whole study period. Repeated growth of the same isolate in successive samples obtained within 7 days was considered the same episode.

Outcome measures
NEC was coded as a binary outcome, and NEC case definition used for primary analyses was modified Bell's criteria ≥2a. 19 Mortality from any cause was included as a secondary outcome.

Statistical analysis
Full details of statistical methods are provided in online supplementary figure S1. In short, time-to-event analysis was used to estimate the difference in NEC rate between epochs, controlling for potentially confounding variables, and with death and discharge considered competing risks. Cumulative NEC incidence is reported, with sub-hazard ratios (sub-HR: ratio of cumulative incidence rates) reflecting the association between each factor and cumulative incidence. Effect of epoch appeared to vary significantly over the risk period, thus we conducted a second analysis including three separate coefficients for the effect of epoch on NEC in the first, second and beyond the second week after birth. A regression discontinuity analysis 21 assessed whether the effect of epoch would be better explained by a linear trend over time or by a step change coinciding with introduction of probiotics. A time-toevent analysis estimated the effect of epoch on all-cause mortality. We used χ 2 and Fisher's exact tests to compare mortality rates following NEC diagnosis and other between-epoch proportions.

Descriptive analysis
The sample included 982 eligible neonates: 469 born in the pre-probiotics epoch and 513 in the routine probiotics epoch (online supplementary figure S2, patient flow). Table 1 shows comparative distributions of covariates between epochs. The only covariate to differ substantially between epochs was mode of milk feeding in the period up to full feeds: exclusive breast Original research Figure 3 The rate of necrotising enterocolitis (NEC) in each year (solid bars with error bars), and linear trends estimated within each era (lines with error bands). Each estimate is shown with a 95% bootstrapped confidence interval. milk feeding was more common in the routine probiotics epoch (table 1). Table 1 shows NEC incidence in the pre-probiotics and routine probiotics epochs using the three NEC classifications. NEC Bell's stage ≥2a occurred in 35 (7.5%) babies in the preprobiotics epoch and 16 (3.1%) in the routine probiotics epoch. Imputed cases (without classical X-ray signs) were n=5 in the pre-probiotics epoch and n=1 in the routine probiotics epoch. In the latter era, three babies who developed NEC during the second postnatal week had not received prior probiotics; dosing had been overlooked in two cases. NEC rates within the routine probiotics epoch were similar irrespective of product used (10 cases during 39 months using Infloran; 6 cases during 21 months using Labinic). More NEC cases of Bell's stage ≥3a occurred in the pre-probiotics era, 29 (6.2%) vs 14 (2.7%) (p=0.017), but comprised a similar proportion of overall NEC cases within epochs, 29/35 (83%) vs 14/16 (88%) (p=0.70). Surgical NEC occurred in 30 infants in the pre-probiotics epoch (6.4% of eligible neonates) vs 15 (2.9%) in the routine probiotics epoch (p=0.014). Figure 1 shows the cumulative incidence of NEC from day of birth until final discharge home (or earlier death). NEC incidence was strongly associated with birth gestational age (GA) (figure 2). Irrespective of GA, all cases of NEC occurred within 40 days of birth. NEC rate was lower in the routine probiotic-use era for all GA subgroups, with differences particularly pronounced in the first 2 weeks after birth.

Cumulative incidence of neC from birth
The overall sub-HR for being born in the routine probiotics compared with pre-probiotics epoch on NEC rate was 0.43 (95% CI 0.23 to 0.77, p=0.005). When all other covariates were included in the model (table 2), the estimated association between epoch and NEC rate remained very similar (sub-HR=0.44, 95% CI 0.23 to 0.85, p=0.014). However, as reflected in the cumulative incidence curve, the difference between epochs was greatest in the first week after birth, and there was no difference after the second week (test for proportional hazard vs difference in sub-HR over time since birth: p=0.013). Estimates for the association between NEC and probiotics stratified by time since birth are shown in online supplementary table S3. In the first week after birth, the sub-HR for routine versus no probiotics was 0.08 (95% CI 0.01 to 0.59, p=0.014), in the second week was 0.43 (95% CI 0.13 to 1.4, p=0.17) and thereafter was 0.85 (95% CI 0.36 to 2.0, p=0.72).
NEC was also independently linked to GA, and appeared less common before initiation of enteral feeds. There was no evidence that birth weight (once adjusted for GA) or milk type used for feeding were significantly associated with NEC rates, although our study was not designed to test these associations.

All-cause mortality and neC-associated rate of death within epochs
There were 67 deaths from all causes before discharge in the pre-probiotics epoch (14.3% of eligible neonates) and 47 in the routine probiotics epoch (9.2%). The cumulative incidence of death among eligible included babies is shown in figure 2, stratified by epoch and GA. All-cause mortality reduced between epochs (sub-HR=0.62, 95% CI 0.42 to 0.90, p=0.013). This was attenuated and no longer statistically significant after adjusting for all covariates (online supplementary table S4). However, as with NEC, the proportional hazard assumption was not met and on stratifying by time since birth there was a significantly lower death rate beyond 2 weeks during the routine probiotics epoch compared with in the pre-probiotics epoch (sub-HR=0.34, 95% CI 0.17 to 0.68, p=0.003; online supplementary table S5). Other factors independently associated with mortality were lower GA, never having received any enteral milk feeds, and non-reception of antenatal steroids.

Regression discontinuity analysis
Within individual years across the whole study period, the proportion of NEC cases as a proportion of all eligible admissions is shown in figure 3, along with estimates of the trend across years within each epoch. After adjusting for calendar year as a continuous variable in the time-to-event analysis, the effect of epoch was largely unchanged and remained statistically significant in the first week after birth (online supplementary table S3). This suggests that our results are not explained by an underlying linear improvement in NEC rates with time, but rather reflect a step change in incidence around the introduction of routine probiotics administration. Conversely, the improvement in death rate between epochs is better explained by a linear trend over time (online supplementary table S5).

Sensitivity analyses
Estimates were similar when postmenstrual age was used as a time-scale instead of days since birth (data not shown). Since milk type used between first and full feeds was the only other recorded variable to change significantly between epochs, we also stratified plots by mode of milk feed (among those who initiated enteral feeds) (figure 2); this showed a clear reduction in NEC rate in the routine probiotics epoch irrespective of milk feed type. Replacing 'birth weight' with 'small for gestational age' in the multivariable analysis did not affect findings. Repeating analyses using the Vermont-Oxford Network NEC case definition 20 did not significantly affect findings either: the estimate of effect of epoch was sub-HR=0.47 (95% CI 0.24 to 0.92, p=0.027), and the stratified analysis was also largely consistent. The only difference was in the regression discontinuity analysis, where the evidence was less strong that the trend was due to probiotics introduction as opposed to a linear trend (in the first week after birth, adjusted sub-HR=0.09, 95% CI 0.00 to 1.62, p=0.10), although this involved a very small number and was still consistent with a big drop in NEC rate in the first week after birth.

Incidence of late-onset sepsis
Babies having one or more episodes of late-onset sepsis were significantly fewer in the routine probiotics epoch compared with the previous epoch (p<0.0001), and there were fewer episodes of coagulase-negative staphylococcal sepsis (p<0.0001) (table 1).
No case of any probiotic-organism bacterial sepsis was observed in either epoch.

Summary of findings
In this sample of 982 high-risk neonates admitted to our NICU over a 10-year period, NEC rate fell by more than half (from 7.5% to 3.1%) in the 5-year period following the introduction of a policy of routine multispecies probiotics administration from birth (~55% relative risk reduction). Our apparent more-than halving of NEC risk with probiotics mirrors remarkably the highly significant risk reductions shown in large meta-analyses of RCTs and observational studies, [7][8][9][10][11] including that of a very large German observational study which also used Infloran. 22 While some quality improvements occurred in our NICU over the study period, the difference observed was not attributable to changes in the main important covariates assessed between epochs, including exposure to breast milk. The improvement affected all high-risk neonates irrespective of GA. The impact appeared greatest in the first 2 weeks after birth, suggesting that aiming for very early postnatal probiotic bacterial gut colonisation is crucial. Late-onset sepsis rates also halved in the routine probiotics epoch compared with in the previous epoch (from 22.6% to 11.5%). There was a fall in the all-cause mortality rate between epochs, although this was more consistent with a continuous trend over the study period.

Strengths and limitations
This is the first UK study to evaluate the potential impact of routine probiotics use on NEC rates. We used multispecies Lactobacillus and Bifidobacterium probiotics, shown to be most effective for NEC prophylaxis in large meta-analysis. 11 These were routinely available, and commercially produced to the same accredited Good Manufacturing Practice standards required for drug production. We had a high rate of compliance and early probiotics administration in eligible babies. We applied three different NEC classification systems; all showed similar reductions in NEC rates across epochs. A further study strength is the variety of rigorous statistical analyses applied.
Being retrospective and observational, our findings only support an association and cannot conclusively determine that probiotics caused the observed markedly reduced NEC incidence; unmeasured confounders might underlie observed differences between epochs. The only measured covariate observed to change between epochs was milk-type used, but this did not explain the improved NEC rate. Furthermore, a regression discontinuity analysis suggested that the drop in NEC rate was better explained by a step-change at the introduction of probiotics, and no other significant policy changes occurred between epochs that could have affected NEC incidence so markedly.
Importantly, colonisation and development of the early life microbiota, including Bifidobacterium and Lactobacillus species, occurs immediately postbirth. Previous studies (clinical and preclinical) indicate that optimal colonisation by supplemented species in a 'naïve' gut environment enables more effective persistence, linking to improved short-term and long-term host health. [23][24][25] Choice of bacterial species and strains is a crucial consideration. Many previous studies already highlighted that strains of species B. bifidum, B. longum subspecies infantis, and L. acidophilus were efficacious in other NICUs globally, key to our selecting these combinations for routine use.
Mechanistically, B. bifidum and B. infantis strains are pioneer colonisers of the infant gut, and their innate ability to digest components of breast milk preferentially, for example, human milk oligosaccharides (HMOs), augments their establishment within the infant gut. 26 27 Numerous studies have shown how these species may provide resistance to potentially pathogenic microbes, so-called 'colonisation resistance', and also aid development of the mucosal and systemic immune systems, key beneficial traits to enhance development of the preterm gut and prevent NEC and late-onset sepsis. 25 28 Furthermore, a multispecies approach may promote better colonisation because Lactobacillus may serve to reduce the oxygen content in the neonatal gastrointestinal tract, facilitating colonisation and persistence of anaerobic (ie, oxygen sensitive) Bifidobacterium species. 24 However, it is critical to note that there are significant differences between individual Bifidobacterium and Lactobacillus species and also huge strain variability, not least in ability to digest dietary components such as HMOs and in their immune and infection modulatory traits. Such complexities may explain conflicting results of some previous RCTs, such as the large UK multicentre Probiotics in Preterm infants Study (the PiPS trial) which-using the single-strain probiotic B. breve BBG-001-did not reduce NEC or late-onset sepsis. 29 Those findings may link to poor colonisation ability of the selected B. breve BBG-001 strain, 30 compounded by low dosage used, 31 and inability of this strain to digest key early life dietary components. B. breve strains frequently lack the genes required to assimilate HMO molecules, and consequently have limited ability to assimilate HMOs compared with B. infantis and B. bifidum. 32 Further studies are needed to determine the generalisability of the underlying potential benefits of multispecies probiotics in this at-risk cohort.

COnCLuSIOn
Administration of multispecies Lactobacillus and Bifidobacterium combination probiotics has been associated with a significantly decreased risk of NEC and late-onset sepsis in our NICU. This evaluation supports our routine use of multi-species Lactobacillus and Bifidobacterium combination probiotics for preventing NEC in very preterm neonates.