The diagnostic utility of procalcitonin, interleukin-6 and interleukin-8, and hyaluronic acid in the Norwegian consensus definition for early-onset neonatal sepsis (EONS).

INTRODUCTION: A key challenge in identifying serious bacterial infection in new born infants is the nonspecific clinical presentation of early-onset neonatal sepsis (EONS). Routinely used C-reactive protein, white blood cell count, and platelets are nonspecific. We assessed the diagnostic utility of single biomarkers or combinations of procalcitonin (PCT), interleukin (IL)-6, IL-8, and hyaluronic acid (HA) in newborn infant with EONS, and in human umbilical cord blood (HUCB) from deliveries with chorioamnionitis.
MATERIALS AND METHODS: Blood was collected from term infants with strictly defined EONS (group 1, n=15), healthy term infants (group 2, n=15), and the umbilical vein from pregnancies with suspected chorioamnionitis (group 3, n=8), and from healthy pregnancies with no signs of infection (group 4, n=15).
RESULTS: Neonatal plasma PCT and IL-8 showed good predictive value (90% and 83%) for EONS, and the combination of IL-6 or HA with PCT increased the predictability to 87% and 90%, respectively. PCT, IL-6, IL-8, and HA were 8.4-, 4.5-, 3.6-, and 1.9-fold higher when compared with plasma levels in noninfected neonates. PCT, IL-6, and IL-8 in HUCB predicted chorioamnionitis and fever in the delivering mother (89%, 83%, and 72%, respectively). HA was a poor predictor (59%), but its predictability increased in combination with PCT, IL-8, or IL-6. In HUCB from chorioamnionitic deliveries, IL-6, IL-8, and PCT were 23-, 14-, and 2.4-fold higher, respectively, when compared with HUCB from healthy deliveries. There was no correlation between C-reactive protein, white blood cell, and platelet count with PCT, IL-6, IL-8, or HA.
CONCLUSION: In neonates that fulfilled the Norwegian consensus definition of neonatal sepsis, PCT, IL-6, and IL-8, but not HA, have the potential to improve our management of neonates at risk. Except for PCT and IL-8, both with a predictability of >80% in neonatal plasma, combinations of biomarkers increased the predictability for EONS and chorioamnionitis.

Introduction

The diagnosis of early-onset neonatal sepsis (EONS) remains a challenge as clinical signs and biomarkers are nonspecific. It is often difficult to discriminate between neonatal bacterial infection, respiratory failure, and clinical deterioration from other causes, so there is a need for specific diagnostic tools or combinations of biomarkers. Thus far, no single laboratory test or combination of clinical signs and tests have ensured a rapid, reliable, and early identification of infected neonates. The fear of a fatal outcome often leads to overuse of antibiotics when clinical signs and symptoms are nonspecific and/or elevated markers of inflammation, such as C-reactive protein (CRP) are present.1 In Norway, 2.3% of liveborn term infants receive antibiotic treatment, but only 0.54 per 1000 liveborns have confirmed bacterial sepsis.2

Proposed biomarkers in EONS during recent years include procalcitonin (PCT).3 Changes in PCT values in cord and neonatal blood in proven histological chorioamnionitis suggest its production and release by fetal tissue.4 Another study in neonatal and umbilical cord blood was related to intrauterine infection and useful to predict infection in the newborn infant.5 In another study of umbilical cord blood samples, both interleukin (IL)-6 and PCT were reliable as a rule-in and a rule-out tool for neonatal sepsis, whereas IL-6 was useful in neonates.6 Dollner et al also observed increased IL-8 levels in the umbilical cord blood of infected preterm neonates.7 Others report that IL-6 and IL-8 are promising diagnostic tools for early diagnosis of neonatal sepsis and may emerge as superior to routine laboratory tests, such as CRP and white blood cell (WBC).6,8–10 Hyaluronic acid (HA), a glycosaminoglycan, is a potential biomarker for EONS. It decomposes from a high- to a low-molecular-weight form, which triggers inflammatory responses, while the high molecular form slows and inhibits ongoing inflammatory processes.8 Increased amounts of circulating HA have been found in human and experimental sepsis where it may signal through Toll-like receptor (TLR)-2 and TLR-4, and the leukocyte receptor CD44.8,11,12

The aim of this study was to assess the diagnostic utility of PCT, IL-6, IL-8, and HA in 1) term newborn infants with EONS and 2) umbilical cord blood from term pregnancies with maternal fever and chorioamnionitis. We also aimed to find if an increase in HA in infant or human umbilical cord blood (HUCB) plasma, when known risk factors are present, could be a predictor of EONS.

Materials and methods

Recruitment of patients

Blood samples from newborn infants and HUCB were collected at Akershus University Hospital in Norway in 2013–2014 after written consent from the parents or the pregnant woman. The study protocol was approved by the Regional Ethics Committee (REK no 2013/609) and the private protector officer at our hospital (PVO 13-045). Spoken and written informed consent was obtained from parents or pregnant women (HUCB).

Blood samples were analyzed according to 4 groups:

Term newborn infants (gestational age >37 weeks)

EONS (group 1)

Healthy controls (group 2)

Human umbilical cord venous blood (HUCB)

From term pregnancies with suspected chorioamnionitis and treated with peripartum antibiotics (group 3)

From healthy term pregnancies with no signs of infection (group 4)

Inclusion criteria for neonates

Group 1

Gestational age >37 weeks, appropriate for age birth weight with EONS as defined by the Norwegian Society of Pediatrics, ie, clinical signs of sepsis and a positive blood culture or clinical signs of sepsis, but negative blood culture, however, in combination with CRP >30 mg/L, and minimum 5 days of antibiotic treatment. Exclusion criteria included birth asphyxia or suspected syndrome disease or dysmorphism. Median Apgar scores and birth weight are given in Table 1. In addition, the National Institute for Health and Clinical Excellence (NICE) risk factors for sepsis was used13 (Table 2).

Table 1

Characteristics of newborns (groups 1 and 2) and mother/deliveries (groups 3 and 4)

Characteristics Mean (SD)	Group 1 (n=15)	Group 2 (n=15)	Group 3 (n=8)	Group 4 (n=15)
Female/male (%)	46/54	33/67	50/50	45/55
Birth weight (g), mean (SD)	3653 (496)	3618 (465)	3887 (657)	3560 (477)
GA (weeks), mean (SD)	40.3 (1.4)	40.6 (1.3)	40.7 (1.0)	40.4 (1.1)
Age* (h), total cohort, median (min–max)	22 (1–42)	48 (40–55)
Age (h) <24 h, median (min–max)	19 (1–24)
Age 24–72 h, median (min–max)	37 (29–42)
Age (years), median (min–max)			29 (22–36)	28 (24–37)
Vaginal/section (%)	92/8	78/22	38/62	90/10
APGAR median
1 min (min–max)	9 (3–10)	9	9 (2–9)	9
5 min	10 (4–10)	9	9.5 (4–10)	9
10 min	10 (6–10)	9	10 (6–10)	9
Top findings (%)a	Irritability 29	None	HRfetus ↑ 63	None
	Resp distr 33		NICU 38
	Hypotonic 13		Antibiotics 100
	Hypoxemia13		Fever 100
Meconium SA (%)	42	28	75	0
Temp >38°C (%)
Newborn	12	0
Mother			100	0
CRP (mg/L) infant
At start treatment	62.4 (30.1)	ND	ND	ND
Highest level	75.7 (35.4)	ND	ND	ND
Hb (mg/L) infant	16.9 (2.1)	ND	ND	ND
Leukocytes (109/L) infant	17.8 (6.6)	ND	ND	ND
Platelets (109/L) infant	259.0 (117.7)	ND	ND	ND
Blood culture infant + (%)	27	ND	ND	ND
PROM >18 h (%)	29	6	25	ND
Blood sampling (median) age (h) postnatal (min)	22*	48**	0–15***	0–15***

Notes:

Blood sampling in septic babies (group 1) at admission to NICU for sepsis.

Always taken at about age 48 h (group 2).

Cord blood harvested within 15 min after birth. Inclusion criteria in groups 2 and 4 were Apgar 9-10/9-10/9-10.

Top clinical findings from medical records.

Abbreviations: Antibiotics, antibiotic treatment; Appr, approximately; CRP, C-reactive protein; Fever, Fever in the mother; HRfetus ↑, tachycardia in the fetus; ND, No data; NICU, newly born admitted to the intensive care unit; PROM, premature rupture of membranes; Resp failure, respiratory failure; Resp distress, respiratory distress; SA, Stained amniotic fluid; Vag/sect, vaginal delivery/cesarean section; GA, gestational age.

Table 2

Risk factors for early-onset neonatal infection (EOS), including NICE red flags

Peripartum risk factors for EOS	Red flag
Neutrophil count (109/L, median) in the delivering mother	ND; ND; 15.6; ND
Group 1; Group 2; Group 3; Group 4
CRP in the delivering mother	ND; ND; 90; ND
Group 1; Group 2; Group 3; Group 4
Invasive GBS infection in a previous baby (%)	4; 0; 13; 3
Group 1; Group 2; Group 3; Group 4
Maternal GBS colonization, bacteriuria, or infection in the current pregnancy (%)	4; 0; 13; 0
Group 1; Group 2; Group 3; Group 4
Prelabor rupture of membranes (%)	8; 5; 25; 3
Group 1; Group 2; Group 3; Group 4
Preterm birth following spontaneous labor (before 37 weeks’ gestation)	None, inclusion criteria GA >37 weeks
Suspected or confirmed rupture of membranes for more than 18 h in a preterm birth	None, inclusion criteria GA >37 weeks
Group 1; Group 2; Group 3; Group 4
Intrapartum fever higher than 38°C, or confirmed or suspected chorioamnionitis (%)	ND; 0; 100; 0
Group 1; Group 2; Group 3; Group 4
Parenteral antibiotic treatment given to the woman for confirmed or suspected invasive bacterial infection (such as septicemia) at any time during labor, or in the 24-h period before and after the birth	ND; 0; 100; 0
Group 1; Group 2; Group 3; Group 4
Suspected or confirmed infection in another baby in the case of a multiple pregnancy	0, only singlets

Note: Data from https://www.nice.org.uk/guidance/cg149/chapter/1-guidance#risk-factors-for-infection-and-clinical-indicators-of-possible-infection-2.13

Abbreviations: CRP, C-reactive protein; GA, gestational age; GBS, group B streptococcal; ND, No data; NICE, National Institute for Health and Clinical Excellence.

Group 2

Healthy term neonates with no clinical signs of infection and recruited in the maternity ward at about 48 h of age when the routine sample for newborn metabolic screening was drawn.

Inclusion criteria for HUCB samples

Group 3

HUCB from term pregnancies where the mother had signs and symptoms of chorioamnionitis, including fever (rectal temperature >38°C), peripartum antibiotic treatment, irrespective of the infant developing suspected EONS or not.

Group 4

HUCB from term healthy pregnancies of nonsmoking mothers with no signs of preeclampsia or infection.

Characteristics of all 4 groups are presented in Table 1.

Blood sampling

Newborn infants

Venous blood (1.5 mL) was collected in lithium heparin-coated tubes along with routine blood tests for suspected neonatal infection (according to Norwegian guidelines and consensus for EONS), or when blood was routinely sampled for newborn metabolic screening at age 48 h in healthy infants.

Human umbilical cord blood

Puncture of the umbilical vein was done immediately after delivery after the cord had been cut, and blood collected directly into lithium heparin-containing vacutainers (BD Bio-sciences, San Jose, CA, USA). We did not delay cord clamping and the cord was not manually expressed. The preparations for cytokine analyses were performed within an hour.

Cytokine analysis

Plasma was prepared from blood that was centrifuged at 2500 rpm for 10 min at room temperature and aliquoted into sterile polystyrene tubes for later analyses of soluble biomarkers. Concentrations of IL-6, IL-8, PCT, and HA were assessed by Quantikine immunoassays from R&D Systems (Minneapolis, MN, USA). Plasma samples were analyzed following the manufacturer’s instructions. Spectrophotometric measurements were carried out at 450 nm, with wavelength correction set to 540 nm on the Varioskan instrument (Thermo Fisher Scientific Inc, Waltham, MA, USA) and using SkanIt Software version 2.4.3. Plasma samples were diluted before analyses if necessary.

Statistical analysis

A power analysis indicated the need for 14 individuals in order to achieve a power of 80%.8

Differences between cytokine concentrations were estimated using the Wilcoxon signed-rank test for paired samples (exact method, 2-tailed), since they were not normally distributed. The p-values <0.05 were considered statistically significant.

To investigate how well the biomarker could predict sepsis, we used a quadratic discriminant analysis with a leave-one-out cross-validation test that was performed by taking 1 data point from the data set and use the remaining to adapt the model. Then we used this model to predict which group (EONS or healthy newborn [group 1 vs 2], infection vs not in the mother [group 3 vs 4]) the omitted observation belongs to. This way, we obtained a realistic indication of how good the markers were to predict sepsis, while also taking into account the uncertainty related to the limited amount of data. Log-transformed markers were used in all analyses. As CRP, in addition to clinical suspicion of sepsis, was included in the inclusion criteria for sepsis, it was not tested for predictability. The approval by the regional ethics committee did not allow for analyzing CRP, or other laboratory routine analyses, in those infants and deliveries that appeared healthy.

The area under the receiver operating characteristic (ROC) curve (AUC) with 95% CI was calculated for each biomarker that was different between septic and non-septic infants and HUCB samples. We also estimated the cutoff value for each parameter that would give the test at least 70% sensitivity for detecting EONS and report the corresponding specificity for that value. Spearman’s correlations were calculated for the different biomarkers. Statistical analyses were performed with IBM SPSS 22 for Mac (IBM Corporation, Armonk, NY, USA).

Results

In the sepsis (group 1), there were 4 culture-confirmed cases of EONS (27%) – 2 Escherichia coli and 2 Staphylococcus epididermidis – and 11 cases were classified as culture-negative EONS (85%), according to Norwegian guidelines. Empiric therapy consisted of an aminoglycoside and aminopenicillin in newborns and aminopenicillin in the delivering mother. All infants survived and were discharged without complications caused by the infection (data not presented in detail).

Overall response in soluble biomarkers

Correlations

We found positive correlations in neonatal plasma between PCT and IL-8 (r=0.70, p<0.001) between PCT and IL-6 (r=0.40, p=0.007), but not between PCT and CRP or CRP and WBC at initiation of antibiotic treatment (r=0.26, p=0.22 and r=0.02, p=0.95, respectively). Regression analysis to show the extent to which a routinely used predictor (CRP, WBC, and platelet count) was correlated with an alternative biomarker did not find any significant correlation between the predictor and PCT, IL-8, IL-6, or HA.

Procalcitonin

The predictability for EONS with PCT was high (90%) (Table 3) and did not increase when this biomarker was combined with IL-6, IL-8, or HA. PCT in septic newborns was increased to 8.4-fold compared with healthy newborns (age 48 h, Table 4). EONS at age <24 h was 9.3 and age 24–72 h 6.7-fold higher compared with group 2 (age 48 h). The EONS subgroup of positive blood cultures (n=4) had a median PCT of 2015 pg/mL, and those without confirmed blood culture (n=10) had a median PCT of 2112 pg/mL, ie, no difference between positive and negative blood culture subgroups. PCT in HUCB was predictable for chorioamnionitis (89%, n=8) and was 2.4-fold higher compared with HUCB of deliveries without any risk factors or signs of infection (NICE).13 A ROC curve provided an AUC of 0.87 (0.70–1.00) for PCT in separating septic from healthy infants with a cutoff value of 321 pg/mL to give the test a sensitivity of 88% and a specificity of 93%; correspondingly, in cord blood, a ROC curve provided an AUC of 0.87 (0.69–1.00) separating chorioamniotic deliveries (n=8) from controls with a cutoff value of 54 pg/mL and the best sensitivity of 88% and specificity of 60%.

Table 3

Percentage of correctly classified patients of single and combinations of biomarkers in neonatal sepsis and in human umbilical cord plasma of perinatal infection in the delivering mother

Biomarker	Correctly classified to group 1	Correctly classified to group 2	Predictability groups 1 and 2 (%)	Correctly classified to group 3	Correctly classified to group 4	Predictability groups 3 and 4 (%)
PCT	13/15	14/15	90	6/8	10/15	89
IL-6	9/15	13/15	73	5/8	10/15	83
IL-8	12/15	13/15	83	5/8	8/15	72
HA	6/15	7/15	43	6/8	7/15	72
PCT–IL6	13/15	13/15	87	5/8	10/15	83
PCT–IL8	11/15	14/15	83	5/8	10/15	83
PCT–HA	13/15	14/15	90	6/8	10/15	89
IL6–IL8	11/15	14/15	83	5/8	10/15	83
IL6–HA	8/15	11/15	63	6/8	10/15	89
IL8–HA	12/15	11/15	77	5/8	9/15	78

Notes: Biomarkers in infant plasma (groups 1 and 2) and human umbilical cord plasma (groups 3 and 4). Predictability = the likelihood of the biomarker testing correctly for septic or healthy infant (groups 1 and 2) and infection in the mother (groups 3 and 4). Bold indicates the percentage of correctly classified patients (≥83%). PCT (pg/L) in the total sample in groups 1–4, in group 1 irrespective of age after birth, in group 2 age approximately 48h after birth. IL-6 (pg/L) and IL-8 (pg/L). HA (ng/L).

Abbreviations: HA, hyaluronic acid; IL, interleukin; PCT, procalcitonin.

Table 4

Biomarkers in neonatal plasma in groups 1 and 2 and in HUCB plasma in groups 3 and 4

Soluble biomarkers	PCT (pg/mL) Mean/SD (median)	IL-6 (pg/mL) Mean/SD (median)	IL-8 (pg/mL) Mean/SD (median)	HA (ng/mL) Mean/SD (median)
Group 1 (n=15)	2213/689 (1922)	102/136 (31)	104/71 (88)	224/188 (176)
Age <24 h (n=10)	2442/719 (2244)	104 (62)	148 (134)	276 (179)
Age >24 h (n=5)	1756/320 (1863)	37 (20)	80 (64)	219 (202)
Group 2 (n=15)	262/223 (229)	23/41 (10)	29/14 (27)	116/45 (111)
Group 1 vs 2 mean (median)	8.4 (8.4)	4.5 (3.1)	3.6 (3.2)	1.9 (1.6)
Group 3 (n=8)	127/72 (110)	244/220 (269)	271/406 (136)	77/107 (43)
Group 4 (n=15)	52/16 (49)	12/9 (10)	21/17 (13)	136/86 (115)
Group 3 vs 4 mean (median)	2.4 (2.3)	23 (27)	14 (11)	0.6 (0.4)

Abbreviations: HA, hyaluronic acid; HUCB, human umbilical cord blood; IL, interleukin; PCT, procalcitonin; h, hour.

Interleukin-6

The predictability of IL-6 in EONS was poor (73%), using the quadratic discriminant analysis with a leave-one-out cross-validation test (Table 3), whereas the combination of IL-6 with PCT or IL-8 increased the predictability to 87% and 83%, respectively. IL-6 was 4.5-fold higher in group 1 vs 2. In the subgroup of positive blood cultures, IL-6 was 91 pg/L (median, n=4) and 4-fold higher than group 2. The predictability for chorioamnionitis was high (89%) and 23-fold higher (Table 4) compared with healthy deliveries. Predictability did not increase when IL-6 was combined with IL-8 or PCT, but increased slightly to 89% with HA. A ROC curve provided an AUC of 0.84 (0.69–0.98) for IL-6, separating septic from control infants. A cutoff value of 20 pg/mL gave the test a sensitivity of 80% and a specificity of 88% ; correspondingly, the ROC curve for IL-6 separating HUCB in chorioamniotic from healthy deliveries provided an AUC of 0.88 (0.72–1.00) with a cutoff value for IL-6 of 19 pg/mL to give the test a sensitivity of 88% and a specificity of 80%.

Interleukin-8

In newborn plasma, the predictability of IL-8 was 83% and did not increase in combination with PCT, IL-6, or HA (Table 3). The corresponding predictabilities of IL-8 in HUCB was 72%, which increased to 83% with both PCT and IL-6. A lower predictability was seen with HA. IL-8 was 3.6-fold higher in EONS compared with controls and 15-fold higher in chorioamnionitic deliveries (n=8) vs control group 4. The subgroup of positive blood culture consisting of only 4 patients had higher IL-8 levels (median 184 pg/L) than the total group 1. A ROC curve for IL-8 separating septic from healthy deliveries provided an AUC of 0.88 (0.74–1.00) with a cutoff value of 43 pg/mL to give a sensitivity of 83% and a specificity of 88%; correspondingly, the ROC curve separating HUCB from chorioamniotic deliveries from healthy provided an AUC of 0.81 (0.59–1.00) with a cutoff value of 32 pg/mL to give the test a sensitivity of 75% and a specificity of 87%.

Hyaluronic acid

The predictability of HA was poor; however, it increased in combination with PCT in neonates from 43% to 90% and in HUCB from 72% to 89% (Table 3). When combining neonatal plasma HA with IL-6 or IL-8, the predictability was still low, 63% and 77%, respectively. The AUC for the ROC curve for HA separating septic infants from controls was 0.77 (0.57–0.97). A cutoff value of 131 pg/mL gave the best sensitivity of 79% and specitivity of 80%; correspondingly, in cord blood, the ROC curve provided an AUC of 0.18 (0.69–1.00) with low sensitivity and specificity. HA increased 1.9-fold in group 1 vs 2 (Table 4) and decreased in group 3 vs 4. The 4 septic newborns with a positive blood culture had 2.7-fold higher HA values (median 302 pg/L) compared with group 2.

As CRP, leukocyte, or platelet counts were not analyzed in groups 2 and 4, we could not estimate the predictability for sepsis based on CRP, leukocyte, or platelet counts, neither with combinations of CRP, leukocyte or platelet counts and biomarkers PCT, IL-8 or HA. In neonatal sepsis, PCT tended to predict a high CRP (regression coefficient [RC]=15.3, p=0.06). However, when antibiotic treatment was initiated, PCT did not correlate with CRP (RC=14.3, p=0.22), platelet count (RC=−5.8, p=0.14), or WBC (RC=−51, p=0.31), nor did corresponding values for IL-6, IL-8, or HA (all p-values >0.08).

Discussion

In this cohort of strictly defined EONS, according to Norwegian consensus criteria, the PCT predictability was superior to IL-8 and IL-6, and did not increase when PCT was combined with IL-6, IL-8, or HA. HA was a poor predictor of EONS and chorionamnionitis, unless it was combined with any of the other biomarkers. PCT and IL-6 in umbilical venous cord plasma (HUCB) predicted chorioamnionitis. PCT had a probability of 90% discrimination between healthy infants and infants with EONS, and was overall 8.4-fold higher in septic neonates compared with healthy newborns; correspondingly, it was 2.7-fold higher in HUCB in chorioamniotic deliveries. Furthermore, with optimal cutoff values based on ROC curves and logistic regression analysis, these biomarkers could improve the sensitivity for the diagnosis of EONS and increased to sensitivity 93%, which is higher than previous studies on biomarkers accuracy. This may, at least partly, be due to the strict criteria for the diagnosis of neonatal sepsis in this cohort.14

The most common routine blood test used to detect infection is CRP. CRP is nonspecific for diagnosis of EONS but has a high negative predictive value.15–18 Neither CRP nor WBC was correlated with any of the biomarkers PCT, IL-8, IL-6, or HA, or combinations of them, indicating that these routine clinical biomarkers are not good diagnostic tools for EONS. This is in contrast to other studies that found serum levels of PCT, CRP, and WBC to be higher in sepsis than in non-septic neonates.10 In spite of the low number of subjects (Table 2), the risk factors in mothers with chorioamnionitis13 led us to examine these deliveries, and high predictability and increased HUCB plasma concentrations of PCT and IL-6 were found.

Blood culture results are often inconclusive and late, and the drawn blood volume is sometimes too small. If presumed to be neonatal sepsis, treatment with antibiotics normally starts before the blood culture result is reported. Therefore, many researchers have looked for early biomarkers that are sensitive and specific for EONS and can help clinicians decide which newborns should be given antibiotics. In this study, only 4 of 15 clinically septic patients had a positive blood culture and a low percentage, but in accordance with other studies.19–21

PCT, IL-6, and IL-8 have been shown to be early and specific markers of neonatal sepsis, confirming these as possible tools that may confirm or exclude infection shortly after birth.22 They rise earlier than CRP, and thus, we anticipated that they would have a higher sensitivity than CRP early in the course of an infection. There are few studies of upper limits for CRP in the healthy newborn, and these are based on relatively small sample sizes with wide-ranging postnatal ages.23–25 Data pertaining to reference intervals for PCT and cytokines are also limited. The present study describes cutoff values for some of the most potential biomarkers to define EONS. A recent paper showed higher PCT (AUC=0.791; p<0.001) in septic patients compared with controls, but with lower AUC values than in the neonate cohort reported in this paper.26 Montaldo et al27 identified an AUC of 0.83 (95% CI, 0.72–0.92) in newborns who were likely to have sepsis. Another study also examined patients with neonatal sepsis compared with non-sepsis and evaluated the diagnostic value of several biomarkers, including PCT, CRP, and WBC. These were significantly higher in the sepsis group than in the non-sepsis group (p<0.001), and they provided the recommended biomarker cutoff levels.10 In non-surviving infectious sick children, the PCT level has been reported to be significantly higher than in surviving children.28 A prospective study of serum data during the first week of life in infants receiving therapeutic hypothermia found that the peak of serum levels of IL-6, PCT, and CRP was on postnatal days 1, 2, and 4, respectively. The authors speculated that the peak value of CRP might reflect the early production and secretion of IL-6 and concluded that serial measurements of IL-6 might help to avoid invasive sepsis workup and unnecessary change of antibiotics in infants.29

When evaluating the impact of potential biomarkers, it is important to consider kinetics and the pattern(s) of response in healthy neonates. PCT increases steadily in term healthy infants from birth to reach peak values at about age 24 h and then decreases gradually.30 This immediate postnatal increase of PCT in healthy infants probably represents endogenous synthesis.31 In a study of neonatal sepsis, PCT measurements at birth were initially normal but, during the first 24 h of life, increased and became a more sensitive marker of infection than CRP.32 Increased PCT, IL-6, and IL-8 concentrations have also been reported for newborns presenting with apparently noninfectious conditions.15–17 These studies, however, may lack enough specificity and have arbitrarily cutoff values to distinguish between infectious and noninfectious clinical conditions,16 and the definition of EONS was not as precisely defined as in this paper. Here, we report septic infants with a diverse range of ages, but only up to age 72 h, and PCT was always higher in EONS compared with healthy controls. We found slight differences of PCT values in infants aged <24 h or older (2242 [n=5] vs 1756 [n=10]), and in both subgroups, PCT was higher (>6.7-fold) than in healthy infants aged 48 h. Blood sampling in healthy newborns was performed at about age 48 h when routine screening for metabolic disorders was performed in accordance with approval of the regional ethics committee. Neonatal plasma values for IL-6, IL-8, or HA was also higher in the youngest subgroup at signs/symptom debut, but as we report a small number of infants, no significant conclusions may be drawn with regard to differences between age subgroups.

In a recent review on biomarkers in sepsis, the authors concluded that biomarkers like PCT, IL-6, and IL-8 can be the key to personalized targeted treatment in the future clinical management of sepsis.33 Memar et al34 proposed that IL-6 is the most potent marker for evaluation of EONS prognosis, whereas PCT and CRP seem to be appropriate indicators for the detection and monitoring of antibiotics therapy. In their meta-analyses and systematic review, they concluded that a panel of sepsis biomarkers along with present routine tests can make early identification easier, and appropriate management and improved outcome may be more efficient than with a single indicator. This is in accordance with conclusions in this paper. PCT determination may also be helpful in detecting meningitis, bacteremia, and sepsis in children with fever without an apparent source,35 but PCT, like other biomarker cytokines, also holds promise to identify diverse medical conditions beyond those associated with acute bacterial infection.36

We anticipated that HA could be a useful biomarker in sepsis. In addition, the methodology for quantification of HA seemed easy, quick, and feasible in a routine hospital setting, but we found a poor predictability. Serum HA is reported to be increased in inflammatory diseases and elevated in patients with septic conditions associated with a poor prognosis.11,37,38 Infants in this cohort were not seriously ill, and discharged home and survived without any complication (data not presented). In recent years, the role of HA has been summarized in several good reviews,39–42 and evidence has accumulated showing that HA serves important biological roles beyond its function as a structural component of interstitial and connective tissues. HA is a mediator of inflammation,18,43 and this is both HA dose and molecular size dependent. For instance, HA has been shown to promote cell motility and migration possibly because of the loose, hydrated, and porous nature of the HA “coat” bound to cells.44–47 Many of these biological activities are mediated by HA through the receptor CD44, but HA is also an endogenous mediator through TLR4 ligand. Fragments of HA can stimulate a chemokine or cytokine production in macrophages by signaling through TLR4 and TLR2.48 In a sub-study, we did analyze the expression of CD44 and TLRs in groups 1–4 by flow cytometry, but we found no difference between the groups (results not shown). Furthermore, in a previous paper, we provided evidence that HA was unchanged after in vitro lipopolysaccharide stimulation of cord blood leukocytes.8 Another paper reported decreased HA levels during group B streptococcal (GBS) pneumonia in neonatal piglets, and interestingly GBS virulence factors included specific secretory enzymes, such as hyaluronidase, an enzyme that breaks down HA.49

Strengths and limitations

A strength of this study would be the strict criteria for EONS according to national consensus diagnostic coding guidelines set by the Norwegian Pediatric Association (http://legeforeningen.no/PageFiles/25877/Neonatale%20diagnosekoder%20i%20ICD-10.pdf). We analyzed only term babies in the first 72 h of life to omit the problem of gestational and age-specific differences that may confound the interpretation of what consists a positive value for the biomarkers. When we subdivided the EONS group according to age at initiation of antibiotic therapy, in spite of the small number of infants (n=5–10), we persistently found higher cytokine levels in the septic group compared with the control group for PCT, IL-8, IL-6, and HA. A limitation is the small number of samples in group 3. In spite of repetitive information provided to the maternity unit in meetings and mails, it was challenging to recruit these samples of cord blood at inconvenient hours during the day and night. Cho-rioamnionitis in the delivering mother was defined as fever (temperature >38°C) with clinical signs of chorioamnionitis and an indication for antibiotic treatment. We do not refer any pathological placental findings.50 Only 4 neonates (27%) in group 1 had a positive blood culture. The majority of the clinically presumed neonatal infection episodes could not be confirmed by a positive blood culture, so we cannot confirm that these are true infections, and we may not draw conclusions with regard to safety of the proposed biomarkers. This may be due to insufficient sample volumes for blood cultures and possibly intermittent or low-density bacteremia very early after birth.19 In a recent Norwegian report, including all neonatal intensive care units (NICUs) in Norway, only 5.9% of those diagnosed with sepsis were blood culture positive.2 Antibiotic treatment prior to blood culture may, if this occasionally happens, reduce the diagnostic performance of blood culture, but in this study, no infant received antibiotic treatment prior to blood culture. Data were collected over a long time span and the enrollment frequency varied, but samples for infectious and control groups were similarly spread over the year, and we aimed to collect samples from controls in the same time period and season as the infectious groups. Furthermore, cytokine analyses were done for all with the same kits and batches on the same days and including all groups. The cohort was small but powered to indicate a need for 14 individuals per group in order to estimate the predictability of the biomarkers for neonatal sepsis. We only succeeded in recruiting 8 deliveries with clinical chorioamnionitis, which is a major drawback. Therefore, we could not perform subanalyses, which may have produced more solid conclusions. A larger cohort would have enabled us to analyze subgroups.

Conclusion

We examined neonates admitted to our NICU that fulfilled the Norwegian consensus definition of neonatal sepsis, which includes elevated CRP at the start of antibiotics. We found high predictability for EONS with PCT and IL-8 (90% and 83%) as we did for IL-6 combined with IL-8 or PCT (87% and 83%). These biomarkers and combinations may have the potential to improve our management of neonates at risk. There were corresponding results in HUCB comparing delivering mothers with chorioamnionitis with healthy delivering mothers. HA was a poor predictor of EONS and chorioamnionitis. At start of antibiotic therapy, biomarkers did not correlate with CRP and WBC, indicating that the routinely used markers are inappropriate for early workup and act as a tool to decide on antibiotic treatment or not. With the help of optimal cutoff values based on ROC curves and logistic regression analysis of single or combinations of these biomarkers, we can improve the sensitivity and specificity in the diagnosis of suspected EONS. We conclude that these biomarkers, especially PCT single or in combination, are better diagnostic tools for neonatal sepsis compared with routine blood analyses like CRP, WBC, or platelet counts.