Introduction

Cronobacter spp. (Enterobacter sakazakii) is a ubiquitous agent causing infrequent infections in immunologically incompetent patients of all age groups. In 2008, E. sakazakii was reclassified as the new genus Cronobacter spp. including the six species C. sakazakii, C. turicensis, C. muytjensii, C. malonaticus, C. dublinensis and Cronobacter genomospecies I [1]. The limited epidemiological information about invasive neonatal Cronobacter infections warrants a summary of the current knowledge regarding case numbers, frequency, fatality rates and risk factors of neonatal meningitis, bacteraemia and necrotising enterocolitis (NEC). Globally, about 120–150 cases have been reported in the high-risk group of infants up to 2 months of age [2, 3]. The main clinical features of Cronobacter infections in neonates are meningitis, septicaemia and NEC [4–6]. The relevance for public health is due to the high rate of fatal outcome and neurological complications of neonatal Cronobacter infections. Diarrhoea, urinary tract infections and conjunctivitis [4, 7] occur even less frequently in neonates. In extensive outbreaks, more than ten neonates may be affected simultaneously [7–9]. In 2–12-month-old infants, some single Cronobacter infections (bacteraemia, urinary tract infections) have been described [10–12]. Recently, suspicious cases of Cronobacter meningitis have been reported in infants >2 month of age [13]. A further 18 cases of (meningitis or) bacteraemia in infants aged 1–11 months have been reported by the laboratory summary of the United Kingdom 1997–2007 to the FAO/WHO [14]. Hence, Cronobacter infections in neonates are very rare acute diseases, reliable incidence rates are not available. In the USA, incidences of one Cronobacter infection per 100,000 infants, 8.7 per 100,000 low-birth-weight neonates [8] and one Cronobacter infection per 10,660 very-low-birth-weight neonates [15] have been reported. In addition to powdered infant formula (PIF) as the main source of neonatal Cronobacter infections, contamination of the preparation and cleaning equipment and the hospital environment should be considered [16]. Some reports describe the problem of contaminated infant formula [17–20] and Cronobacter infections [21–23] in less developed countries with complicated climatic and hygienic conditions. In young children, six cases of bacteraemia [4, 14, 24, 25] and a case of an infected intradural dermoid cyst [24] have been known. Further 27 clinical Cronobacter isolates from young children aged 1–4 years have been reported in England and Wales to the FAO/WHO in 2008 [14]. In general, Cronobacter infections in infants are associated with intensive medical interventions or immunodeficiency syndrome. In 2008, the FAO/WHO summarised the current knowledge about invasive Cronobacter infections in infants and young children in a detailed meeting report. The main purpose of this expert group was to examine the risk of infection by follow-up-formula in infants >6 months of age [14]. Because of the less severe nature of the illness, little attention is focused on Cronobacter infections in adults. Recently, the FAO/WHO cited interesting cases of colonisation and infection with Cronobacter spp. in the elderly population [26, 27] and mentioned the necessity of systematic reviews of Cronobacter infections in adults [13]. The majority of Cronobacter infections in infants [28] and adults [11, 29, 30] is health care-related. In the European Union, the Rapid Alert System for Food and Feed (RASFF) and the Early Warning and Response System (EWRS) are tools to promote the prevention and control of communicable diseases, including (risk factors for) Cronobacter infections. Microbiological criteria for food and process hygiene, relevant for the European Economic Area (EEA), consider Cronobacter spp. as a possible risk factor in PIF and dried dietary foods for special medical purposes for infants <6 month of age [31].

Reports of microbiologically confirmed Cronobacter infections in neonates, published between 2000 and 2008, have been included in this literature and data analysis. More than 100 cases of neonatal Cronobacter infections have been reported in this period. The overall lethality of the 67 invasive cases with known outcome data was 26.9%. The lethality of Cronobacter meningitis, bacteraemia and NEC was 41.9% (P < 0.0001), <10% and 19.0% (P < 0.05), respectively.

Logistic regression models (P < 0.0001) revealed a higher gestational age at birth and parentage not from Europe as significant factors for a higher reporting probability of neonatal Cronobacter meningitis. Neonates with parentage from North America have a lower probability (P < 0.01) of dying from Cronobacter meningitis. These epidemiological measures of lethality and risk factors are based on limited numbers of reported cases and on different case recruitment strategies. Further consistent and sufficiently informative data of neonatal Cronobacter infections should be recorded in a centralised reporting system.

Materials and methods

Data of microbiologically confirmed neonatal Cronobacter spp. (E. sakazakii) infections have been extracted from scientific publications and from international (FAO, WHO), European (RASFF, EWRS) and national epidemiological reports or reporting systems. PubMed, ISI (Web of Science) and Scopus abstract and citation databases, as well as the alert systems of big publishers (ASM, Blackwell, Elsevier, Oxford and Springer), were searched using the following keywords: E. sakazakii, Cronobacter, neonate, infant, newborn, NICU; described anywhere in the text; without limitation of language; published between 2000 and 2008 (updated 31 March 2009). Cases of haemorrhagic colitis have been coded as cases of NEC for statistical modelling. Bacteraemia, sepsis and septicaemia have been used synonymously. Colonisations without clinical signs of infection have not been considered as cases of infection. Information about continental parentage, kind and outcome of infection was available for 67 invasive Cronobacter infections in neonates (aggregated data sets). Information about further risk factors like gestational age at birth, birth wight, onset of infection and gender has been available for 59 cases (complete data sets). Calculations have been carried out with SAS 9.1.—Enterprise Guide 2.1. Logistic regression models are based on the complete data set for dependent and effect variables of invasive neonatal Cronobacter infections during the period 2000–2008. Logistic regression models have been performed for the dependent variables ‘kind of infection’ (i.e. meningitis, sepsis or NEC) and ‘outcome of infection’ (i.e. death), respectively. Birth weight, day of onset, gender, gestational age at birth, kind of infection and parentage have been investigated as effect (or independent) variables in the relevant univariate logistic regression models, respectively. In the following model steps, significant variables have been included in multivariate logistic regression models (forward selection procedure), respectively. P-values ≥ 0.05 were considered as statistically not significant (n.s.).

Results

Epidemiology of neonatal Cronobacter infections published 2000–2008

Microbiologically confirmed cases of Cronobacter infections in neonates published 2000–2008 are summarised in Table 1. The overall case fatality rate of invasive neonatal Cronobacter infections was 26.9%. The lethality of Cronobacter meningitis, Cronobacter septicaemia and Cronobacter NEC were calculated to be 41.9% (P < 0.0001), <10% (n.s.) and 19.0% (P < 0.05), respectively.

Table 1 Neonatal Cronobacter infections published 2000–2008
Full size table

Significant logistic regression models for all kinds of invasive neonatal Cronobacter infections in the years 2000–2008 have been summarised in Table 2 according to the goodness of fit. The model with the best adaptation to the data comprises the bivariate model no. 5 with the significant risk factors of gestational age at birth and parentage not from the EU for the outcome of meningitis. Neonates with meningitis not originating from North America have the highest risk for lethal outcome (model no. 9).

Table 2 Risk factors of neonatal Cronobacter infections in 2000–2008
Full size table

Significant logistic regression models based on the data of neonatal Cronobacter meningitis cases in 2000–2008 have been summarised in Table 3. Among the reported cases of Cronobacter meningitis, parentage from North America (model no. 12), and parentage from the USA (model no. 13) is associated with a lower probability of death, respectively. The gestational age of cases of neonatal Cronobacter meningitis from South America and other continents (Asia, Australia and Oceania) does not differ from the gestational age at birth of the remaining cases of neonatal Cronobacter meningitis. For illustrating previous modelling, the corresponding aggregated case numbers and fatality rates of invasive neonatal Cronobacter infections have been stratified by continents as shown in Table 4.

Table 3 Risk factors for death in neonatal Cronobacter meningitis in 2000–2008
Full size table
Table 4 Case fatality rates of neonatal Cronobacter infections stratified by continents
Full size table

Reporting of Cronobacter-contaminated powdered infant formula in Europe

RASFF alerts concerning Cronobacter in PIF are summarised in Table 5. During the reporting period 2002–2008, Cronobacter infection and colonisation had been associated with three of the 11 reported contaminated PIF products. Two of the three products associated with diseases have been intended as dietary products for special medical purposes, i.e. one product of PIF for premature neonates and one product of hypoallergenic PIF, respectively.

Table 5 Notification of Cronobacter spp. in the European Rapid Alert System for Food and Feed (RASFF)
Full size table

Cronobacter infections in infants not related with powdered formulae

Between 2000 and 2008, 11 cases of invasive Cronobacter infections in neonates and infants not fed with powdered infant or follow-up formula have been reported (Table 6). Nine of these 11 cases are health care-related infections. Four of five Cronobacter infections occurred in infants treated in intensive medical units. Further cases of Cronobacter infections in infants with other environmental sources of infections, e.g. drinking water installations [14, 32] and intensive medical interventions [12] have been described.

Table 6 Cronobacter infections in infants not fed with powdered infant formula (PIF) published in 2000–2008
Full size table

Discussion

The lethality of neonatal Cronobacter infections, especially of neonatal Cronobacter meningitis, remains at a high level. So far, reliable rates of incidence, neurological complications and lethality of neonatal Cronobacter infections could not be reported due to missing or different reporting criteria. Most of the available data have derived from case and outbreak descriptions. Some data of sporadic cases have been communicated by passive reporting (systems) of clinical or laboratory data. Further submerged information have been retrieved actively by organised calls for data [33] or sporadical personal requests. After the occurrence of clustered cases, passive reporting systems have been established, e.g. for invasive Cronobacter infections in infants up to 12 months of age in the USA [8] and for Cronobacter meningitis in New Zealand [34]. In Brazil and Hungary, Cronobacter infections are mandatorily notifiable diseases [14]. Cases of bacteraemia (and meningitis) with the related age (groups) have been reported in England and Wales [35]. In Canada, adverse symptoms associated with the consumption of infant formulae should be reported to local food inspection agencies [36]. National epidemiological bulletins have reported current Cronobacter infections, e.g. in the Netherlands [37] and in France [7], as well as surveys of Cronobacter infections [38]. In the Norwegian national register for nosocomial infections, Cronobacter spp. is registered as a causative agent [39]. Laboratory data-based Cronobacter infections have been reported, e.g. from England, Wales and Northern Ireland [40] and the Philippines [14]. In most countries, foodborne diseases or outbreaks should be reported to local authorities. In notifications of the European RASFF about contaminated PIF, three associated cases (Table 1) have been known. “Affected persons associated with the dangerous subject” is a notification parameter of the RASFF. Even in 2009, notifications of contaminated infant formulae have appeared [41]. Via the European EWRS, outbreaks will be communicated to the competent public health authorities [42].

The main source of neonatal Cronobacter infections, contaminated PIF, has been ascertained in various outbreaks [7, 9, 43, 44] and single cases [4, 45]. The environment in PIF processing facilities has to be regarded as potentially contaminated. Manufacturers of powdered formulae for infants and young children should control the microbiological hazards in the raw materials during the whole processing chain and of the final products according to international recommendations and European legislation [31, 46, 47]. Atypical Cronobacter infections like conjunctivitis and urinary tract infections are unlikely to be directly related with contaminated PIF. The hospital environment, e.g. water outlets, medical equipment, surfaces and interpersonal contacts, may act as sources of infection, since Cronobacter spp. has been isolated from, e.g. incubators for newborn infants [48], the stethoscope of a physician [49], the sinks of a maternity ward, as well as from contaminated infusion [32] and blood culture bottles [49].

The present statistical analyses are based on spontaneously published case reports and on additionally actively reported data of neonatal Cronobacter infections in response to the call for data. Consequently, interpretations of statistical models refer to this database. Bowen and Braden found similar results of a higher gestational age and birth weight in cases of Cronobacter meningitis in comparison to Cronobacter bacteraemia in infants, respectively [28]. Continental differences in case numbers and fatality rates in neonatal Cronobacter meningitis are noticeable. They may depend on different health care-related, therapeutic and infection control procedures, on seasonal and climatic differences, as well as genetic aspects of cases. The kind and the direction of a possible publication and reporting bias, e.g. concerning the number of related cases, kind of outcome, location of acquisition, case reporting and data recruitment strategies, and case numbers from less developed countries, should be considered. Other feasible causes like differences in the microbiological agent have not been indicated. While the virulence factors among the Cronobacter genus remains to be described, each species has to be considered as virulent. The generality of present conclusions should be confirmed by models of comprehensive data, which are, so far, not available. For a reliable risk assessment, a consistent reporting system for Cronobacter infections in neonates and infants should be established.