Neonatal sepsis: within and beyond China

Incidences and pathogen profiles

Intrapartum antibiotic prophylaxis (IAP) for EOS

Maternal post-partum transmission of GBS is underestimated

LOS still accounts for about one-third of neonatal GBS sepsis, and the case fatality rate of GBS LOS was higher than that of GBS EOS (7.7% vs. 5.2%).^[20] Mother-to-infant transmission may constitute an important source of GBS LOS. A longitudinal study was carried out in 160 mother-baby pairs from Italy to assess postnatal colonization with GBS and the impact of IAP.^[26] Specimens from the rectum, vagina, and milk of mothers were collected from the time of delivery to 8 weeks post-partum. Women were grouped into culture-positive carriers (n = 83), culture-negative carriers (n = 26), and non-carriers (n = 51) at discharge from hospital. A total of 35 (22%) neonates were colonized by GBS from at least one body site, and the majority of them (30/35) were born to culture-positive carriers. Infants of culture-positive carriers exposed to IAP were less likely to be colonized (15/57 vs. 15/26, P = 0.01), or heavily colonized (7/57 vs. 1/26, P = 0.04). However, neonates exposed to IAP and discharged GBS-free from hospital often became subsequently colonized (12/57 vs. 1/26, P = 0.09), and six out of 83 culture-positive carrier mothers showed positive milk cultures with heavy colonization of GBS in their babies. Molecular typing analysis confirmed identical GBS strains in all mother-baby pairs. These data call the attention of physicians to maternal transmission after delivery as an underestimated source of neonatal LOS.

GNB are associated with severe neonatal LOS

Despite being less common than Gram-positive bacteria to cause neonatal LOS, GNB are associated with more severe clinical manifestations such as meningitis, and have a higher mortality, especially in VLBW infants.^[27]E. coli is the most common cause of bacterial meningitis in preterm infants and the second most common after GBS in term neonates in the UK.^[28] In France, the average number of LBs each year was around 800,000 between 2001 and 2013, of whom 6.6% were born preterm.^[29] In 2001, the Pediatric Infectious Group of the French Pediatric Society (GPIP/ACTIV) established an active bacterial pediatric meningitis surveillance network including 233 pediatric wards and 168 microbiology laboratories. From 2001 to 2013, 325 infants were prospectively diagnosed with E. coli meningitis, which was seven-fold more frequent in preterm than in term infants.^[30] The median age at diagnosis was 14 days, with two peaks of infection onset at 0 to 3 days of age (EOS in mostly preterm neonates), and 11 to 15 days of age (LOS in mostly term infants). Severe clinical manifestation was reported in 51.9% of patients, and 9.2% died. Death was associated with very preterm birth (odds ratio [OR]: 7.3, 95% CI 2.7–20.9; P < 0.001), cerebrospinal fluid (CSF) to blood glucose ratio <0.1 (OR: 15.3, 95% CI 1.8–128.3; P = 0.012), uncommon strains like O7 serogroup (P = 0.034), and PapGII adhesin (OR: 2.3, 95% CI 1.2–4.5; P = 0.015). Translocation of GNB into the systemic circulation via an interrupted gastrointestinal barrier has been speculated to be a key mechanism of GNB LOS.^[27,31]

Pathogen profiles of community- and hospital-acquired neonatal LOS

An emerging body of evidence has demonstrated completely different microbiological findings between various types of sepsis. This is of critical importance to guide empirical antibiotic therapy strategy, which should not only be adapted to the timing (early vs. late), but also to circumstances (hospital- vs. community-acquired).

The incidence of overall viral sepsis

Viruses are common but largely underappreciated pathogens of neonatal sepsis. In a population-based pregnancy surveillance at five sites in Bangladesh, India, and Pakistan from 2011 to 2014, babies with illnesses meeting the World Health Organization definition of possible serious bacterial infections (pSBI) were referred to study physicians for culture and molecular assays of blood and respiratory samples.^[8] Among 63,114 babies, 6022 pSBI episodes were identified (95.4 per 1000 LBs), with causes specified in 28% of episodes (16% bacterial and 12% viral). The mean incidence of bacterial and viral infections was 13.2 (95% credible interval 11.2–15.6) and 10.1 (95% credible interval 9.4–11.6) per 1000 LBs, respectively. The leading viral pathogen was respiratory syncytial virus (5.4 per 1000 LBs). This study was among the first to provide valuable population-based data on neonatal viral sepsis with universal detection of pathogen species. So far, species-specific data on neonatal viral sepsis are limited.

Herpes simplex virus (HSV)

HSV may cause potentially devastating infection in neonates. However, large-scale assessment of its frequency in potentially infected infants has not been performed, apart from one retrospective cross-sectional study in 23 North American pediatric emergency departments enrolling infants ≤60 days old with CSF samples for HSV polymerase chain reaction (PCR) or culture.^[43] Of 26,533 eligible encounters, 112 infants had HSV identified (0.4%), corresponding to 1.2% of all infants tested for HSV. Among them, 90 (80.4%) occurred in weeks 1 to 4, with the median age of HSV-infected infants being 14 days (interquartile range 9–24 days). Notably, 68 of 112 test-positive infants had central nervous system (CNS) or disseminated HSV disease. The proportion of infants subjected to an HSV test (35%, range 14%–72%) and to whom acyclovir was administered (23%, range 4%–53%) varied widely across centers. Thus, both pediatric emergency physicians and neonatologists should keep in mind this rare but deadly cause of neonatal sepsis, which requires urgent acyclovir administration. A careful review of the maternal history and close communication with obstetricians are critical for a timely diagnosis and treatment of neonatal HSV disease.

Enteroviruses

Enteroviruses have been increasingly recognized to be causes of meningitis, sepsis-like disease, and fever without source in neonates and older children. A prospective multicenter study performed at 35 French pediatric and emergency departments has shown that positive blood PCR results for enterovirus was common.^[44] Between 2015 and 2016, a total of 71 newborns were subjected to enterovirus testing, the detection of enterovirus was more frequent in blood than in CSF (70/71 vs. 62/71, P = 0.011). As such, PCR for enterovirus should be part of the sepsis workup in neonatal sepsis, and empiric antibiotic therapy should be promptly discontinued in confirmed cases of viral infection. So far, there is a lack of population-based data on neonatal infection due to enterovirus in China. In a hospital-based prospective study in a level-3 NICU in East China, enterovirus infection among febrile neonates with an admission temperature >38°C was diagnosed by PCR testing of stool and/or CSF samples.^[45] One-hundred thirty one (39.2%) of 334 febrile neonates had PCR-confirmed enterovirus infection, with PCR results being positive in 130 (99.2%) of stool and in 58 (44.3%) of CSF samples, respectively. Among 131 infected neonates, mostly term infants in their 2 to 3 postnatal weeks, 69 (52.7%) had diarrhoea, 48 (36.6%) respiratory symptoms, 22 (16.8%) poor feeding, 34 (26.0%) rash, and 18 (13.7%) lower platelet counts <150 × 10⁹/L.^[45] These data indicate that enterovirus infection may be common in febrile neonates and stool samples should be included in diagnostic work-up. Although all infected patients received supportive treatment with none developing fulminant disease course, the long-term outcome of neonates infected with enterovirus may be worrying. Another case series of 12 Chinese neonates with a GA of 35 to 39 weeks showed that enterovirus can cause severe encephalitis associated with white matter damage in the neonatal period, and the severity of the imaging abnormalities might correlate with later neurodevelopmental outcome.^[46]

Culture-dependent and independent tests

Sufficient amount (1–2 mL) of blood should be taken to optimize the yield of blood culture,^[47] besides the volume needed for a full blood cell count, venous pH, and lactate, and other parameters to assess the disease status [Table 2]. Urine sample obtained via aseptic bladder catheterization and CSF via lumbar puncture (LP) should be considered in a respiratory- and hemodynamically-stable infant.^[48–50] A positive bacterial culture after 24 to 36 h confirm the type of bacterial pathogen, and allows antimicrobial susceptibility testing.^[42] Biological markers, such as procalcitonin (PCT), C-reactive protein (CRP), white blood cell, and absolute neutrophil count (ANC), are not specific of bacterial infection. However, the Lab-score based on PCT levels (<0.5, ≥0.5, and ≥2 ng/mL), CRP levels (<40, 40–99, and ≥100 mg/L), and urine dipstick (negative or positive) has been shown to be highly predictive of a urinary tract infection (UTI).^[51] Nonetheless, UTI should be defined by bacterial counts of uropathogen (mostly E. coli), with the minimum concentration ranging between 10⁵ and 10⁶ colony-forming units/mL.^[52] Recently, two prediction rules to identify young febrile infants at risk of SBI in the pediatric emergency department have been validated. The “Step-by-Step” approach developed in Europe assesses abnormal clinical appearance at first, followed by young age (<3 weeks), significant leukocyturia and bacteruria confirmed by high-power field microscope, PCT >0.5 ng/mL, CRP >20 mg/L, or ANC >10,000/mm³ to define low-risk and high-risk of SBI.^[53] The prediction rule of the PECARN group in USA identifies febrile infants 60 days or younger at low risk of SBI using the urinalysis, ANC, and PCT levels.^[54]

Particularities of LP

LP is mandated in young febrile infants if sepsis is highly suspected and the patient's condition allows.^[53–55] The previously recommended practice of performing brain imaging before a LP has been shown to significantly delay antibiotic treatment and increase the likelihood of a poor neurologic outcome, both in adults and children with proven bacterial meningitis.^[56,57] As a traumatic LP can complicate the interpretation of CSF, it is helpful to evaluate the mean spinal canal depth (MSCD) using either the following rule MSCD (mm) = 0.4 × weight (kg) + 20 or direct ultrasound imaging.^[58] Direct examination of CSF sample under microscope is mandatory to determine the bacterial type and number per high-power field, providing valuable guidance for empirical antibiotic therapy. The presence of Gram-positive cocci in long chains under microscope may suggest GBS meningitis, and Gram-negative bacilli meningitis caused by E. coli or even extended-spectrum beta-lactamase (ESBL) Enterobacteriaceae, depending on microbial surveillance results at the local institutions. In a multicenter study in Asia which included 453 documented episodes of neonatal sepsis, the most common pathogens for neonatal meningitis were Klebsiella spp. (27/76), followed by CoNS (11/76) and E. coli (10/76), with GBS identified only in one case.^[15] One recent province-level multicenter study in South China including 838 cases of neonatal meningitis from 12 hospitals (2011–2016) showed that the first three causes were GBS (88/249), E. coli (55/249), and CoNS (32/249).^[59]

Timing and regimen of antibiotics

Empirical antibiotic treatment should be initiated as soon as neonatal sepsis is suspected, in fear of negative outcomes associated with missed septic cases. As streptococci are often sensitive to penicillin/ampicillin and about 50% of E. coli, noticeably K1, are resistant to ampicillin/amoxicillin, most recommendations of treating neonatal EOS are now to combine amoxicillin/ampicillin with gentamicin.^[5,60] If culture confirms GBS, amoxicillin/ampicillin is recommended, and gentamicin should be discontinued after 48 h to prevent cumulative nephro- and oto-toxicity.^[61,62] In case of ampicillin-resistant E. coli infection, a third-generation cephalosporin like cefotaxime should be used (100 mg/kg per day), and at a higher dosage (200 mg/kg per day) when meningitis is suspected. In case of hospital-acquired LOS which is often caused by CoNS like Staphylococcus epidermidis,^[42] vancomycin by intravenous injection (or continuous infusion through a CVC if it is infected) is recommended at first,^[63,64] as many of CoNS are indeed methicillin-resistant. When culture results are positive for GNB, aminoglycoside is preferred given its narrow spectrum and reduced risk of resistance.^[5] However, escalation to third-generation cephalosporin and even carbapenem should be considered, depending on local epidemiology, clinical presentation especially with CNS involvement, and whether there is evidence of ESBL Enterobacteriaceae.^[5] As community-acquired LOS is often related to respiratory viral infections,^[65] antibiotic treatment must be rapidly discontinued when viral infections is confirmed by multiplex PCR of nasopharyngeal aspirates.

The duration of antibiotic treatment

Recently, two statements regarding the management of neonates born at <35 and >35 weeks’ gestation with suspected or proven early-onset bacterial sepsis have been released in the US.^[66,67] Suspected EOS should always be confirmed by cultures of blood, urine or CSF, and antibiotic therapy should be discontinued if cultures are sterile.^[66,67] This rule also applies to the management of LOS. Furthermore, non-specific biological markers (eg, blood cell count, PCT, CRP) alone do not justify prolonged use of antibiotics.^[66,67] A fruitful dialogue should be maintained between neonatologists and microbiologists to optimize the duration of antibiotic treatment, which relies often upon empirical rules: 7 to 10 days for bacteremia, 14 days for Gram-positive meningitis, and 21 days for Gram-negative meningitis,^[3,68] Given the general trend to shorten antibiotic therapy and minimize the selection of multi-resistant bacteria,^[69] PCT as a highly sensitive biomarker of sepsis has been investigated to guide the cessation of antibiotic therapy.^[70]

Overuse and misuse of antibiotics contribute to microbial resistance

Both the inappropriate overuse of broad-spectrum antibiotics, such as third-generation cephalosporins and carbapenem, and inadequacy of antibiotic stewardship have led to the emergence of multi-resistant bacteria, mostly ESBL Enterobacteriaceae, in neonatal sepsis in China.^[71] In a retrospective study reporting the temporal trend of antimicrobial resistance over a period of 25 years in the largest neonatal unit in Southwest China, a dramatic six- and two-fold increase was found in the resistance rate of ceftazidime and imipenem respectively, in Klebsiella spp. isolated from neonates with sepsis.^[38] A recent meta-analysis of Chinese literature (2009–2014) on neonatal sepsis revealed that more than 50% of E. coli and Klebsiella spp. were resistant to third-generation cephalosporin, and ampicillin resistance of E. coli was almost 80%,^[37] similar to that in the US (85%).^[11] Unsurprisingly, the resistance rate of amikacin was relatively low (around 10%–15%) in E. coli and Klebsiella spp.,^[37] probably due to the fact that aminoglycoside is not routinely used to treat neonatal sepsis in China. Carbapenem-resistance was found in 9% of E. coli and 6% of Klebsiella spp., respectively.^[37] Similar antibiotic susceptibility results of GNB were corroborated by other systematic reviews and web-based surveys.^[72,73] As for Gram-positive bacteria, although resistance to glycopeptide including vancomycin and linezolid seem to be rare in China, methicillin resistance was reported for 60% to 70% of Staphylococcus spp.^[37,38,40]

Antimicrobial resistance in developing countries, such as India^[74] and Egypt,^[75] is alarming. A very recent report from a multinational, multicenter online database showed that third-generation cephalosporin resistance rates among Gram-negative isolates ranged from 26% to 84%, and carbapenem resistance rates ranged from 0% to 81% in 12 low- and middle-income countries.^[73] Glycopeptide resistance rates among Gram-positive bacteria ranged from 0% to 45%.^[73] Staff to patient ratio and antibiotic availability were considered to be key factors affecting antimicrobial resistance rate in these settings.^[73] The rising antibiotic resistance very much complicates both the choice of initial empirical antibiotic treatment and secondary antibiotic strategy, as last-resource antibiotics are not readily accessible in many settings.^[76]

To minimize the selection of multi-resistant bacteria, there is an urgent need to restrict antibiotic therapy to proven neonatal bacterial infection and shorten antibiotic course. Furthermore, surveillance and assessment of antibiotic consumption has been demonstrated to positively influence antibiotic stewardship strategy and reduce antibiotic use.^[77] Screening for GNB on neonatal body sites may have a predictive value in neonatal LOS,^[78] and decolonization has been investigated as an approach to prevent sepsis and the spread of multi-resistant GNB.^[79] However, the prognostic value of routine screening of neonatal microbial colonization remains controversial, due to a lack of high-quality evidence, especially in VLBW infants.^[31,78]

Supportive treatment of neonatal sepsis: uncertainties and certainties?

Neonatal sepsis can progress to septic shock, which may requires volume loading and vasopressors in order to maintain organ function. However, after the publication of the paradoxical results of the fluid expansion as supportive treatment (FEAST) trial comparing repeated boluses of 20 mL/kg of either saline or 5% albumin with no bolus in 3141 African children with severe infection,^[80] there is now a trend to restrict unnecessary volume loading in order to stabilize hemodynamics while preventing the occurrence of brain and lung edema, especially in resource-limited settings.^[81]

For years, intravenous immunoglobulins (IVIG) have been used in severe neonatal sepsis, because newborns are relatively deficient in endogenous immunoglobulins. However, a large phase III clinical trial of 3493 infants at 113 hospitals comparing two iv infusions of 500 mg (10 mL) per kg body weight 48 h apart with placebo failed to demonstrate a significant difference in the incidence of subsequent sepsis episodes and the rates of major or non-major disability at 2 years of follow-up.^[82] Since the large clinical trial included neonates from both developing and developed countries, the negative results should be interpreted as high-level evidence to abandon IVIG administration among septic neonates in China.