Objective To conduct a systematic review of neonatal necrotising enterocolitis (NEC) rates in high-income countries published in peer-reviewed journals.
Methods We searched MEDLINE, Embase and PubMed databases for observational studies published in peer-reviewed journals. We selected studies reporting national, regional or multicentre rates of NEC in 34 Organisation for Economic Co-operation and Development countries. Two investigators independently screened studies against predetermined criteria. For included studies, we extracted country, year of publication in peer-reviewed journal, study time period, study population inclusion and exclusion criteria, case definition, gestation or birth weight-specific NEC and mortality rates.
Results Of the 1888 references identified, 120 full manuscripts were reviewed, 33 studies met inclusion criteria, 14 studies with the most recent data from 12 countries were included in the final analysis. We identified an almost fourfold difference, from 2% to 7%, in the rate of NEC among babies born <32 weeks’ gestation and an almost fivefold difference, from 5% to 22%, among those with a birth weight <1000 g but few studies covered the entire at-risk population. The most commonly applied definition was Bell’s stage ≥2, which was used in seven studies. Other definitions included Bell’s stage 1–3, definitions from the Centers for Disease Control and Prevention, International Classification for Diseases and combinations of clinical and radiological signs as specified by study authors.
Conclusion The reasons for international variation in NEC incidence are an important area for future research. Reliable inferences require clarity in defining population coverage and consistency in the case definition applied.
PROSPERO International prospective register of systematic reviews registration number CRD42015030046.
- necrotising enterocolitis
Statistics from Altmetric.com
What is already known on this topic?
The number of babies at risk of developing neonatal necrotising enterocolitis (NEC) in high-resource settings is increasing with improving survival of preterm babies.
Little is known of international variation in disease burden, and no study has systematically reviewed the published literature for NEC rates in high-income countries.
What this study adds?
Internationally, there is a fourfold to fivefold reported difference in rates of NEC between high-income countries.
A standardised approach to reporting population coverage, study inclusion criteria and NEC definition would enable more accurate international comparisons.
The early survival of preterm babies has increased in high-resource settings, and as a consequence, the number at risk of serious complications that commonly occur in the neonatal period, such as necrotising enterocolitis (NEC), is increasing. NEC is now one of the leading causes of mortality and morbidity in neonatal intensive care.1 2 The poor understanding of its aetiology and pathophysiology3 and absence of a non-invasive diagnostic test have resulted in a lack of clarity of what constitutes ‘NEC’. Recent reports from the USA describe a fall in NEC incidence despite a rising preterm term birth rate.4 Understanding variation between different populations and over time in the incidence of this serious gastrointestinal disease is important in identifying determinants, designing preventive trials and implementing quality improvement programmes. Little is known of international variation in disease burden. The aim of this study was to review systematically and summarise published data that most closely describe the national incidence of NEC in high-income countries.
We included countries in the Organisation for Economic Co-operation and Development (OECD) as these have broadly comparable resources, rates of survival of very preterm babies and the necessary infrastructure to capture and report national-level data.
Data sources and search strategy
The systematic review was registered prospectively on PROSPERO (registration no CRD42015030046).5 Methods were developed according to recommendations from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).6 Two review authors independently performed the search. The first systematic search was performed by CB using MEDLINE, between 1946 and 5 December 2015, Embase, between 1974 and 15 December 2015 and PubMed between 1979 and 28 April 2016, respectively, using ‘Necrotising Enterocolitis’ (MesH) and ‘Country’. TS repeated the search between 30 April and 3 May 2016. We restricted the search to the 34 countries in the OECD7 and did not apply language restrictions. These countries comprised Australia, Austria, Belgium, Canada, Chile, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Japan, Korea, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Turkey, UK and the USA.
CB and TS independently reviewed all abstracts. Reference lists were further hand-searched for relevant studies omitted in the initial electronic searches. We excluded reviews, viewpoints, editorials and assessments of interventions including randomised controlled trials. Full manuscripts for observational studies reporting the incidence of NEC in the country of interest were retrieved. Following review of the final manuscript, exclusion criteria were single centre, missing numerators and denominators and not generalisable, for example, study population included only small for gestational age (GA) babies or singleton births. A study was eligible for inclusion if the rate of NEC was provided at national, regional or multicentre level. Of the eligible studies, those most representative of the population were included (in order of preference: national, regional and multicentre). In order to minimise risk of double counting babies when multiple studies from the same country overlapped in time, only the most recent was included unless the studies presented the data differently in more detail (e.g. birth weight categories) or relevant ways (e.g. one by GA and the other by birth weight) in which case both were used. Any disagreements between TS and CB over article inclusion, exclusion and/or data extraction were resolved through consensus.
Data extraction and synthesis
Extracted data were stored in Microsoft Excel file format. The following information was extracted from each study: country, year of publication, study time period, study population inclusion and exclusion criteria, case definition, gestation or birth weight-specific NEC rates (with or without laparotomy) and mortality if available (with or without laparotomy). Where raw numbers were available, we present specific NEC rates; where only graphs were available in the published report, we sought raw numbers from the authors.
We further present country rates of NEC by the case definition applied and by widely used GA and birth weight categories (<28 weeks’ GA, 28–31 weeks’ GA and <32 weeks’ GA; birth weight <1000 g, 1001–1499 g and <1500 g).
Study quality assessment
We assessed the risk of bias of each study using a modified version of the8 tool including 8 of the 10 appropriate parameters addressing internal and external validity8 (online supplementary table 2). We omitted the two parameters non-response bias and whether data were obtained directly from the subjects as these were not applicable to our study population. Each parameter was assessed as having either low or high risk of bias. Unclear or data unavailable to make a judgement were regarded as having a high risk of bias. The overall risk of bias was then scored according to the number of high risk of bias parameters per study: low (≤2), moderate3 4 and high (≥5).
Supplementary file 1
The PRISMA flow chart of search results is shown in figure 1. We identified 1888 publications, in which 1633 studies remained after removing duplicates. After screening titles and abstracts, 1513 articles were excluded because they did not report NEC in any population; we reviewed the full manuscripts of the 120 remaining articles. Thirty-three articles were eligible for inclusion, and 14 studies presented the most recent data (table 1).
Characteristics of included studies
The country, time period, study inclusion and exclusion criteria, design, case definition applied and rates of NEC by GA and/or birth weight categories of the included studies are shown in table 2. Eligible but excluded studies are reported in online supplementary table S1. The 14 included studies represented data from 12 of the 34 OECD countries across Europe, Asia and the USA (table 2). One study included data from both Australia and New Zealand.9 We included two studies from Canada: one provided the most recent data from 2008 to 2012 for babies born less than 29 weeks10 and the other included the most recent data from 2003 to 2011 for babies born 29–32 weeks’ GA.11 We included two studies from the USA with overlapping time periods: one from 511 US Vermont Oxford Network Centres in 2005–2006 across birth weight categories ≤1500 g12 and a regional study with more recent data from California in 2005–2011 by GA categories.13 We included one study from each of the following countries: Finland, Germany, Italy, Japan, Korea, Poland, Spain, Sweden and Switzerland. Four studies were multicentre,14–17 two studies were regional13 18 and eight studies were national.9–12 19–22 Most study populations comprised babies admitted to a neonatal unit; two studies included live births obtained from birth registry data, as the denominator.13 19 Inclusion criteria were based on birth weight in seven studies, GA in four studies, GA and birth weight in two studies and no GA or birth weight restriction in one national linkage study. Some studies restricted the population to babies that survived beyond 12–72 hours after birth and excluded babies with congenital anomalies.11 21 With exception of the study from Finland conducted in 1996,19 all studies included data from the last 15 years. The study from Sweden covered the longest period (1987–2009) using data linkage between national registers.
Case definitions for NEC
The most commonly applied definition was Bell’s stage ≥2, which was used in seven studies.9–11 17 18 21 22 Other definitions included Bell’s stage 1–3,19 the definition from the Centers for Disease Control and Prevention (CDC)14 and the International Classification for Diseases (ICD), which consisted of a code rather than a definition.20 The remaining four studies used a combination of clinical and radiological signs as specified by study authors.12 13 15 16
Risk of bias
Assessments using the modified risk of bias tool8 showed that of the 14 included studies, the overall risk of bias was low for 1, moderate for 11 and high for 2 studies (table 3). We were unable to find any validation studies for the data sources used. Eight studies were prospectively, and six studies were retrospectively designed.
Origins of data
With the exception of Finland,19 Sweden20 and Italy,18 data were extracted from established networks that have maintained registers or databases, restricted to very low birth weight (VLBW) (<1500 g) or very preterm (<32 weeks’ GA) babies. Population coverage differed, even among national studies, with some only including tertiary neonatal units, for example, Canadian Neonatal Network, which includes 75% of VLBW babies in Canada, and the Swiss Neonatal Network, which includes level 2 and 3 neonatal units, representing 90% of all VLBW babies born in Switzerland.22
Comparison of rate of NEC by case definition, GA and birth weight categories
We present the rate of NEC by case definition, GA and birth weight categories (table 4). Comparing by GA, the rate of NEC was highest among the most preterm but higher among babies born at 24 weeks’ GA compared with 23 weeks’ GA.18 19 For studies using Bell’s staging ≥2, the rate of NEC was lowest in Japan across all GA categories. The rate of NEC among preterm babies born <28 weeks’ GA ranged from 2% in Japan,21 4% in Switzerland,22 to 7%–9% in Australia, Canada and Italy.9 11 18 Among babies born 28–31 weeks, the rate of NEC ranged from 0.2% in Japan to 2%–3% in the other countries. Overall, the rate of NEC for all babies <32 weeks’ GA ranged from 2% in Japan to 3%–4% in the other countries. In studies using other definitions, the rate of NEC among <28 weeks ranged from 5% in Sweden20 to 10% in USA.13 Studies using the Vermont Oxford Network (VON) definition, requiring one clinical sign and one radiological finding reported higher incidences of NEC, around 7% for babies <32 weeks’ GA, compared with 2%–4% in other studies.
Among babies <1500 g, studies using Bell’s staging ≥2 reported an NEC rate ranging from 2% in Japan, to 3% in Italy18 and to 6%–7% in Korea and Spain.15 17 Studies using other definitions, reported an NEC rate ranging from 3% in Sweden and Germany,14 20 to 6%–7% in the USA and to 9% in Poland.16 The rate among babies born <1000 g was highest in Finland (22.0%), which used Bell’s stage 1–3 as the case definition.19
Laparotomy and mortality for NEC
Only four of the studies reported mortality of babies with NEC; for all NEC mortality, this ranged from 21.9% to 38%.12 15 17 23
We identified an almost fourfold difference, from 2% to 7%, in the rate of NEC among babies born <32 weeks’ gestation in reports from 12 high-income countries. We also identified an almost fivefold difference from 5% to 22% in the rate of NEC among those born <1000 g. However, caution is warranted in interpreting this variation. Studies varied in NEC case definition, quality, risk of bias and population coverage. Bell’s stage 2–3 was most commonly used to ascertain cases, but growing recognition of the need for a more suitable case definition has led to investigators applying different combinations of radiological and clinical signs. The inconsistencies in definitions used make it difficult to make international comparisons. A further important observation is that population data are necessary for accurate estimation of burden of disease, but there was variable coverage among the national studies reported with very few studies truly covering the entire at-risk population. Some studies described as population based were in fact limited to tertiary centres, introducing selection bias as babies that might have died in lower level units prior to transfer were excluded; for example, the Neonatal Research Network of Japan represented only 45% of VLBW babies and the VON centres only two-thirds at the time of the study. Differences in inclusion criteria also limit comparability, with some studies including all neonatal unit admissions and others including only babies that survived a certain length of time. These limitations detract from the ability to make meaningful comparisons and argue for establishing standard reporting criteria for NEC studies such as have been produced in other disease areas.24
Variation among studies with similar NEC definitions and inclusion criteria may indicate differences in clinical practices, such as criteria for offering intensive care to babies at the limits of viability.25 26 Countries with a more active approach may have higher NEC rates because of the higher survival of the extreme preterm babies most at risk of developing NEC. Variations in feeding practices may also influence the risk of NEC. Japan has the lowest reported rate, attributed to early and ‘aggressive’ enteral feeding and high use of unpasteurised human donor milk, though these practices have not been investigated with rigour.
Since undertaking this systematic review, we have completed a 2-year prospective whole population study of NEC in England27 using the Neonatal Dataset extracted from the National Neonatal Research Database (NNRD), a repository containing data from admissions to all neonatal units in England. Since this study, the NNRD now also includes all admissions from Scotland and Wales. We found that the incidence of severe NEC, defined as that confirmed at laparotomy/postmortem, or resulting in death, was 3% in babies born <32 weeks’ GA.
The strengths of our study are the broad search criteria aimed at maximising the likelihood that all relevant studies worldwide with data from more than one centre would be identified. We also acknowledge limitations. By restricting the scope of this review to NEC rates published in peer-reviewed journals, we excluded annual data published by some networks on their websites, for example, Australia and New Zealand Neonatal Network reports.28 Furthermore, due to the large volume of literature on NEC, we accept that despite the broad criteria we may still have missed relevant and more recent publications.
Although we set predefined inclusion and exclusion criteria, these were difficult to apply because of the heterogeneity of the studies. We justified the exclusion of studies published by research groups such as National Institute of Child Health and Development2 29–38 based on the restriction to academic units, representing 5% of the population of VLBW born in the USA.2 For some countries with multiple studies, selecting the most recent and nationally representative data was not straightforward. For example, for Italy, we excluded a study that had a greater number of neonatal units and geographical spread39 but was older, presenting data for the period 1995–1996, which may explain their lower NEC incidence despite applying a more liberal NEC definition when compared with the more recent study presenting data for the period 1999–2002.18 For Australia and New Zealand, we excluded studies conducted over 15 years ago from the NSW and Australian Capital Territory.40 41 For Canada, we included two studies that contained most recent data, although their primary focus was not NEC. One included data for babies born up to 29 weeks, and the other for babies born ≥29 weeks.10 11
The wide range of definitions used for case ascertainment reflect the disparate purposes for which studies were conducted, some primarily for research, others for benchmarking and quality improvement exercises, but they limit the extent to which the data can reliably be compared and pooled nationally and internationally. Bell’s criteria were compiled to assist in management after the diagnosis of NEC was made, and not as a case definition, yet are widely used for this latter purpose. Additionally, different components of Bell’s staging are commonly selected by authors to define NEC. Studies mandated different numbers of clinical and radiological signs for NEC case ascertainment. Germany and Poland requires the presence of one radiological sign and at least two clinical signs, making it a more stringent criterion than the VON definition, which requires one radiological and one clinical sign. The Swedish study used ICD codes to identify NEC cases and included all cases without further specification, including ‘suspected NEC’.20 ICD codes are not a definition but are assigned after the diagnosis has been made. We have recently published evidence-based gestation-specific criteria for NEC case ascertainment that include abdominal X-ray findings and clinical signs.42
In conclusion, in this review, we highlight the limited information on population incidence of NEC internationally, and the challenges in achieving complete population coverage and applying a consistent case definition. A number of preventive and therapeutic approaches to reduce the incidence and impact of NEC are under current and planned investigation. We recommend that to improve comparisons and generalisability of conclusions, international consensus is sought for the case definition for NEC, and criteria for reporting on the population covered.
We wish to thank Luigi Gagliardi, Riccardo Pfister and Fei Chen for providing the raw numbers corresponding to the graphs in their publications.
Contributors CB conceptualised and designed the study and data collection forms, performed the initial searches, extracted the data, drafted the initial manuscript and approved the final manuscript as submitted. TS carried out the initial searches and extracted the data independently, reviewed and revised the manuscript and approved the final manuscript as submitted. KC critically reviewed the manuscript, contributed to each draft and approved the final manuscript as submitted. NM critically reviewed the manuscript and approved the final manuscript as submitted.
Funding This paper represents independent research funded by the National Institute for Health Research (NIHR) under its Programme Grants for Applied Research (Grant Reference Number RP-PG-0707-10010).
Disclaimer The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.