Article Text
PDF
Abstract
Objective To compare necrotising enterocolitis (NEC), late-onset sepsis (LOS), focal intestinal perforation (FIP) and mortality in infants from a single neonatal unit before and after probiotic introduction.
Design Retrospective review of infants <32 weeks admitted January 2009–December 2012 (no probiotic) and January 2013–December 2017 (routine probiotics). Infants included were admitted before day 3, and not transferred out before day 3. NEC, LOS and FIP were defined with standard definitions.
Patients 1061 infants were included, 509 preprobiotic and 552 postprobiotic. Median gestation, birth weight and antenatal steroid use did not differ, and proportions of extremely low birthweight infants were similar (37% and 41%).
Results Overall unadjusted risk of NEC (9.2% (95% CI 7.1 to 12.1) vs 10.6% (95% CI 8.2 to 13.4), p=0.48), LOS (16.3% (95% CI 13.2 to 19.6) vs 14.1% (95% CI 11.5 to 17.4), p=0.37) and mortality (9.2% (95% CI 7.1 to 12.1) vs 9.7% (95% CI 7.6 to 12.6), p=0.76) did not differ, nor proportion of surgical NEC. In multiple logistic regression, accounting for gestation, birth weight, antenatal steroid, maternal milk, chorioamnionitis and sex, probiotic receipt was not significantly associated with NEC (adjusted OR (aOR) 1.08 (95% CI 0.71 to 1.68), p=0.73), LOS or mortality. In subgroup (645 infants) >28 weeks, aOR for NEC in the probiotic cohort was 0.42 (95% CI 0.2 to 0.99, p=0.047). FIP was more common in the probiotic cohort (OR 2.3 (95% CI 1.0 to 5.4), p=0.04), not significant in regression analysis (2.11 (95% CI 0.97 to 4.95), p=0.05).
Conclusions Probiotic use in this centre did not reduce overall mortality or rates of NEC, LOS or FIP but subgroup analysis identified NEC risk reduction in infants >28 weeks, and LOS reduction <28 weeks.
- neonatology
- gastroenterology
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Statistics from Altmetric.com
What is already known on this topic?
Current meta-analysis of randomised clinical trials supplementing probiotics to preterm infants suggests reduction in necrotising enterocolitis (NEC), and possibly late-onset sepsis (LOS) and mortality.
The efficacy of probiotics depends on which product is used, and the precise strain and species combination.
A recent single-centre UK observational study before and after routine probiotic introduction showed lower rates of NEC and LOS, but not mortality.
What this study adds?
Our centre identified no differences in overall rates of NEC, LOS or mortality after routine probiotic supplementation commenced, after adjustment for likely confounders.
Consistent with the meta-analysis, we demonstrate a lower rate of NEC in infants >28 weeks’ gestation receiving probiotics.
Further studies are required to determine optimal probiotic use in infants <28 weeks’ gestation.
Introduction
Despite advances in neonatal care over the past 30 years, necrotising enterocolitis (NEC) and late-onset sepsis (LOS) remain significant causes of mortality and morbidity in the preterm infant.1 Probiotic supplementation has been extensively studied and the most recent meta-analysis suggests reduction in NEC, and possibly late-onset infection and mortality2 but use in the UK remains uncommon,3 perhaps due to uncertainty around efficacy of commercially available products, and the lack of benefit in the UK Probiotics in Preterm babies (PiPs) trial.4 Probiotics are recommended by several professional organisations including the European Society for Pediatric Gastroenterology Hepatology and Nutrition,5 and strongly endorsed by parent-led advocacy organisations (NEC UK (https://www.necuk.org.uk/) and the US NEC Society (https://necsociety.org/)). In addition to reductions in NEC and LOS, probiotics have been shown to reduce time to full enteral feeds, time to regain birth weight and reduce episodes of feed intolerance.6 Data for infants <1000 g are least available and the American Academy of Paediatrics now recommends against routine administration, especially to these infants.7 A recent retrospective review from a single UK unit showed a >50% lower rate of NEC (from 7.5% to 3.1%) following introduction of probiotics, and an almost halving of LOS, but no impact on adjusted mortality.8 We report on a similar cohort of infants cared for in Newcastle (UK), exposed to the same probiotic products over a virtually identical time frame.
Methods and definitions
Eligible population: preterm infants <32 weeks’ gestation inborn (and not transferred out before day 3) or transferred in before day 3 between 1 January 2009 and 31 December 2017, until discharge from in-patient neonatal care. Infants transferred in with abdominal concerns were excluded.
Data extraction: demographic data (gestation, birth weight, feed type, time to full enteral feeds (150 mL/kg/day maintained for 72 hours), gender, antenatal steroid receipt, delivery mode, mortality) was extracted from BadgerNet. We identified potential cases of abdominal pathology including NEC or focal intestinal perforation (FIP) and LOS using a combination of datasets: we searched BadgerNet for any baby with a diagnosis of NEC or suspected NEC, receiving metronidazole, undergoing laparotomy or dying. We cross-checked against a unit-specific research database where NEC and LOS are key primary outcomes and against hospital episode coding lists for NEC, abdominal surgery, FIP and LOS. Blood and cerebrospinal fluid culture results were cross-checked against laboratory records of positive cultures.
Definitions: infants with potential NEC, LOS, FIP or who died were reviewed by at least two clinicians independently. Surgical NEC (NECS) was defined as infants who underwent abdominal drainage or laparotomy, or died before undergoing planned laparotomy, where the surgical, histological and clinical diagnoses combined were most consistent with NEC (consistent with definitions applied by Robertson et al). Infants dying before laparotomy were classified as NEC if a postmortem confirmed NEC, or a diagnosis of NEC was made before death. Infants with clear alternate diagnoses at laparotomy, including FIP, were reclassified. Medical NEC (NECM) was defined as receiving 5 or more days of antibiotics including metronidazole and nil by mouth, with X-ray features suggestive of NEC (pneumatosis, portal venous gas, gasless or fixed dilated loops), and no alternate diagnosis. FIP was defined as infants who underwent abdominal drainage or laparotomy, or died before undergoing planned laparotomy, where the surgical, histological and clinical diagnoses combined were most consistent with FIP. LOS was defined by a positive blood culture of a single organism, or mixed growth with a clinically significant organism (excluding mixed coagulase-negative staphylococci species) after 72 hours of age, combined with clinical features suggestive of infection, accompanied by treatment (or intention) for ≥5 days with intravenous antibiotics.
LOS occurring in association with NEC was defined where infants had an enteric organism-associated bacteraemia while being actively managed for NEC or within 24 hours of a NEC diagnosis was made.
Survival was defined as live at discharge from neonatal in-patient services and causes of death categorised as either attributable to NEC, LOS, FIP or LOS and NEC co-occurring. Infants discharged to local neonatal units had complete analysis of their records to identify NEC/LOS/FIP or death before discharge. Diagnostic criteria were applied consistently across epochs and applied by the same clinicians, using a third to arbitrate occasionally as necessary (JB, CG, NE).9
Cohorts
Two epochs were characterised: from January 2009 to December 2012 (no probiotic use) and from January 2013 to December 2017 (routine probiotic use).
Probiotic use
We initially used Infloran (Lactobacillus acidophilus and Bifidobacterium bifidum) changing to Labinic (L. acidophilus, B. bifidum and Bifidobacterium longum spp infantis) in August 2016. In July and August 2016 no probiotic was available. We aimed to commence probiotics once minimal enteral feeds were tolerated and discontinue at 34 weeks corrected gestational age (CGA) or discharge.
Other relevant unit practices
A standard approach to feeding and antibiotic use was followed throughout both epochs (online supplemental material). In brief, mother’s own milk (MOM) was prioritised and encouraged, supplemented, if necessary, with preterm formula and/or cows’ milk-based fortifier; donor human milk was not used.
Supplemental material
Following results of the BOOST II trial (Benefits of Oxygen Saturation Targeting),10 we altered oxygen saturation targets from 81%–94% to 91%–95% in 2011. We also participated in several randomised controlled trials during these periods including enteral lactoferrin for very preterm infants,11 speed of increasing milk feeds trial12 and iodine supplementation study,13 none of which was shown to alter NEC rates. During 2015, bed closures resulted in admission being restricted to the smallest, sickest babies.
Statistics
Statistical analysis was undertaken using GraphPad Prism. We used Mann-Whitney U test to determine differences between non-parametric data, and χ2 test to analyse differences in frequencies between cohorts. Logistical binomial regression analysis was undertaken using a single multivariate model to adjust for multiple comparisons, with residual testing to ensure no significant uncontrolled variables remained.
Results
Population
There were 1061 eligible infants: 509 in the preprobiotic cohort and 552 in the probiotic cohort. There were no differences in median gestation, birth weight, delivery mode, antenatal steroid use, gender or death before 72 hours of age, but more infants in the probiotic cohort were exclusively MOM fed at full feeds (58% vs 48%, p < 0.001) (table 1).
Demographics of two cohorts
Probiotic receipt
Seventy per cent of infants who received probiotics were given Infloran, 30% Labinic. Median age at commencement across the whole probiotic cohort was 6 days (IQR 4–9). Of those in the probiotic cohort who died or developed disease, 79% received probiotics before onset of disease or death and rates did not vary significantly by year (75%–85%). Twenty-five per cent of infants with FIP had received probiotics before FIP.
Rates of NEC, LOS, FIP and mortality
Overall mortality was similar between cohorts (9.2% vs 9.7%), unadjusted NEC rates did not differ (9.2% vs 10.6%), nor the proportion of NECS (4.1% vs 5.3%). In unadjusted analysis, FIP was significantly more common in the probiotic epoch. Rates of LOS or causative organisms did not change overall (table 2). No differences were seen in rates of NEC, LOS, FIP or mortality with either probiotic used (data not shown).
Morbidity and mortality associated with each time epoch
Binomial logistic regression demonstrated no impact of cohort on NEC, LOS, FIP or mortality adjusting for gestation, gender, feed type, chorioamnionitis and birth weight. Because NEC developed before full feeds in some infants, additional sensitivity analysis was undertaken showing no difference using feed type before NEC or day of initial milk feeds to replace feed type at full feeds. Increasing gestational age was associated with reduced NEC, LOS and FIP, and female gender was associated with reduced NEC and LOS (table 3).
Regression analysis showing adjusted aORs for NEC, LOS, FIP and mortality of key factors
NEC, LOS and FIP cases
Infants with NEC were significantly less mature and lighter in the probiotic cohort with more infants with NEC born <28 weeks’ and <26 weeks’ gestation (table 4). Day of life of NEC onset, proportion undergoing surgery, timing of surgery in relation to onset and NEC-related mortality were not different across cohorts. In both cohorts, around half of the NEC infants had not reached full feeds before NEC. In the probiotic cohort, a higher proportion received exclusive MOM at full feeds and in the 3 days before LOS. Infants who developed LOS had a higher rate of chorioamnionitis in the probiotic cohort (table 4). There were no demographic differences in FIP infants between cohorts.
Demographic and outcome data for infants developing NEC, LOS or FIP
Discussion
Summary of findings
We identified no change in crude or adjusted rates overall or in subgroups of NEC or LOS rates, overall mortality or NEC-associated or LOS-associated mortality between cohorts exposed and not exposed to probiotics. The exceptions were a statistically significant lower rate of NEC in infants >28 weeks’ gestation and LOS in infants <28 weeks. We observed wide variation in annual NEC rates (6.5%–11.7% and 7.5%–18.3%) that has also been shown by others.14–16 Although cohorts did not differ in median gestation or birth weight and the contribution to each cohort of the babies considered to be highest risk (<28 weeks and <1000 g) was similar (37% vs 41% and 37% vs 42%, respectively), regression analysis was undertaken to account for any impact of the different characteristics of infants across cohorts. In this, impacts of gestation and sex were seen, but no difference in NEC risk was identified between cohorts meaning there was no overall impact of probiotic receipt after adjustment for these possible confounders. Infants with NEC were more premature and had a lower birth weight in the probiotic cohort–in keeping with a lower NEC rate in infants >28 weeks’ gestation, and consistent with meta-analysis of randomised controlled trials that demonstrate an advantage of probiotic use in infants >1000 g but not below this.2 We also noted an increased rate of FIP, no longer significant after adjustment for confounders, and in only one in four did this occur after probiotics had commenced.
Comparison with other studies
Our findings are strikingly different to Robertson et al,8 who identified reduced rates of NEC and LOS, despite almost identical product use (including change in probiotic preparation from Infloran to Labinic), time frames, cohort size and median gestations and birth weights. Interestingly, despite the reductions in NEC and LOS, two of the most important contributors to mortality, Robertson et al saw no impact on adjusted mortality, and nor did we. We note an approximately 10% higher incidence of LOS in infants in the study by Robertson et al, and very low rates of NECM (<1%) suggesting potentially different thresholds for diagnosing NECM/LOS across the two studies. In contrast to our study, most observational studies report intention to treat with probiotics rather than actual administration. The intent was to commence probiotics on day 1 of life in Norwich, earlier than our actual median start date of day 6. We note the approximately double rate of FIP in our probiotic cohort compared with the study by Robertson et al. FIP was surgically and histologically confirmed in our cohort, and occurs in the smallest sickest infants, also at greatest risk of NEC; this may suggest an underlying difference in infants cared for at the two units.
In the network meta-analysis by van den Akker et al,17 the strain combination contained in Infloran was identified as reducing mortality (two studies, combined n=494) but not NEC or LOS, whereas strains contained in LB2 were not identified as impacting on mortality, NEC or LOS.
In the 30 preceding observational studies of infants <34 weeks’ gestation or <1500 g birth weight identified,18 19 varying results have been seen with use of Infloran and LB2. Fourteen individually demonstrated efficacy against NEC, six using Infloran,20–25 (one also with LB28). Five studies using Infloran reported no impact on NEC reduction.26–30 Of 20 reporting impact on LOS, 5 individually demonstrated efficacy (1 with Infloran/LB2,8 1 just Infloran).27 31–33 Six studies using Infloran (one with additional species30) reported no impact on LOS21 22 28 30 34 35 and one reported increased LOS.24 Twenty-six studies reported unadjusted mortality, nine individually identifying efficacy, five with Infloran (one with additional,30 one with LB28 21 29 30 35). Six studies using Infloran20 22 25–28 and one with LB28 reported no impact on unadjusted mortality.
The lack of apparent efficacy in our study could relate to product type or lower exposure before disease (median start day 6), although 79% of infants who died or developed disease were exposed before onset, making this unlikely. In keeping with our data, other studies have shown static or increasing incidence of NEC over time, independent of probiotic use.36–38
Strengths and weaknesses
We present a large cohort of similar size to Robertson et al, with detailed independent review by experienced clinicians using standardised definitions,9 39 followed to discharge home and adjusted our analysis for multiple potential confounders. We carefully cross-checked details of NEC, LOS and death against a research database and hospital coding records, including cases back to 2009 only, before which further review of radiographs was not practical. There are inherent challenges with historical cohort comparisons using existing datasets and the inability to adjust for unknown confounders. While practices in neonatology have changed over this time frame, other interventions (such as higher oxygen saturation limits and a greater emphasis on MOM exposure) would be expected to result in lower risk of NEC. In our cohort, without a reduction in NEC rates over time, it is less likely that a benefit of probiotics is artificially masked by other factors. However, it remains possible that other unidentified confounders or external influential events altering physician thresholds for investigations and NEC treatments may have affected findings. Although NEC typically occurs sporadically, there is some literature on spatiotemporal nosocomial clustering of cases among high-risk infants.40 While it is possible that such clustering could have led to an increase in cases in 2015, with a reduction in lower acuity babies, this is more likely due to the increased risk of NEC or LOS in these smallest, sickest babies. In spite of these limitations, we do not think it likely that other interventions or exposures such as oxygen saturation limits, or greater use of MOM, will have obscured a true reduction in serious morbidity due to the use of probiotics.
Meaning and future studies
Interpretation of observational data is always challenging but in >1000 infants our results are consistent with the network meta-analysis data for the products we used and show no reduction in NEC, LOS or mortality. The specific product, dose and timing of exposure all potentially determine probiotic efficacy, and are widely debated, but potentially explain differences between other observational studies and ours. The importance of patient population, patient setting, specific colonising neonatal intensive care unit (NICU) flora and NICU antibiotic practices are relatively underexplored, but potentially key modulators. Infants developing NEC have very different gut microbial community structures between different NICUs41 42: a successful probiotic in one unit may not work elsewhere. Large, randomised studies including head-to-head trials remain key to choosing and refining probiotic products.
The differences in observed NEC, LOS and FIP rates between our study and the study by Robertson et al underline the importance of continuing to refine how cases are defined and ensuring consistency across studies.9 While all-cause mortality may be considered the optimal outcome measures for probiotic efficacy in large studies, documenting differences in NEC and LOS are key to clinical and mechanistic understanding, yet are notoriously difficult to standardise. NEC, LOS or FIP may all be miscoded as each other, as shown in validation studies,43 44 meaning detailed data on all three (rates, timing, organisms, interventions) should be included in any study reporting any of these outcomes.
Our data along with that of others suggest that currently studied probiotic products may have greatest benefit against NEC in infants >28 weeks or >1000 g. This may be because less mature infants take longer time to establish enteral tolerance and hence receive probiotics later or due to other factors such as greater antibiotic exposure or reduced exposure to MOM, that reduce efficacy. Despite the many probiotic studies undertaken to date, optimal probiotic use in the most immature infants remains a key area of uncertainty, and where need for further trials remains.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
Ethics approval
This retrospective practice review did not require ethical approval. All infants received clinical care from the reviewing team.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @clairelgranger, @neonatalbiobank
Contributors JB was responsible for the initial study concept and methodology. JB, NE and CG were responsible for data collection and analysis. CG and JB wrote the initial manuscript. LCB, CG and JB undertook statistical and regression analysis.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests NE and JB declare institutional research funding from Prolacta Biosciences and Danone Early Life Nutrition, and honoraria from Danone Early Life Nutrition, and Nestle Nutrition Institute. The other authors declare that they have no competing interests.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.