The Utility of Serum Hepcidin as a Biomarker for Late-Onset
late-onset sepsis in very low birth weight infants.
Study design We compared the diagnostic performance of hepcidin with C-reactive protein from the serum con-
predictive accuracies were assessed from the areas under receiver operating characteristic curves and the cutoffs
that differentiated infants with and without sepsis were identified using classification and regression tree analysis.
Results Seventeen of the enrolled infants in this study were bacteremic and/or received antibiotics for neonatal
sepsis for $5 days (infants with sepsis). The concentrations of hepcidin were increased 4-fold in infants with com-
pared with infants without sepsis (P < .0001) and returned to similar levels following therapy. The areas under re-
ceiver operating characteristic curves of hepcidin was 0.93 compared with 0.83 for C-reactive protein, P = .06.
Hepcidin concentration >92.2 ng/mL correctly classified 91% of all infants (positive predictive value: 100%, nega-
tive predictive value: 87%, specificity: 100%, and sensitivity: 76%).
Conclusion Serum hepcidin concentration may be a useful adjunct test, in addition to blood culture and other
markers of infection, in the evaluation of late-onset sepsis in very low birth weight infants. (J Pediatr
hygieneandstrictadherencetoinsertionandmaintenanceprotocols forperipherally-inserted centralvenouscatheters.1Invery
low birth weight infants [(VLBW); birth weight <1500 g], the rates of late-onset infections are as high as 20%-25%.2The del-
eterious consequences of neonatal sepsis are particularly pronounced in VLBW infants, adversely impacting growth, neurode-
velopment, pulmonary function, and prolonging hospital stay.3Diagnosing late-onset neonatal sepsis is problematic because
the clinical signs are nonspecific, especially in low-birth weight infants. Reliance on blood culture as a ‘gold standard’ presents
several challenges in neonates because of long turn-around time for results and frequent falsely negative culture results second-
ary tolowinoculum ofbacteriain thesmallvolumeofbloodsample collections.4,5Consequently, variousadjunctivediagnostic
tests, including biochemical markers, hematological indices, and scoring systems are used to aid decision-making in antibiotic
therapy in VLBW infants suspected of sepsis.6,7
Hepcidin, a highly conserved antimicrobial peptide, is an acute-phase reactant that plays a critical role in inflammation and
iron homeostasis.8,9Hepcidin contributes to host defense by depriving microbes access to iron10and through direct antimi-
crobial activity against bacteria and viruses. The utility of hepcidin as an adjunct test for sepsis has not been assessed in preterm
neonates. The goal of this study is to compare the diagnostic performance of hepcidin with a well-established marker of
neonatal sepsis, C-reactive protein (CRP), in predicting late-onset neonatal sepsis in VLBW infants.
epsis is a major cause of morbidity and mortality in neonates. Adoption of intrapartum antibiotic prophylaxis against
Streptococcus agalactiae infection has resulted in 80% reduction in the incidence of early-onset neonatal bacterial sepsis
caused by group B Streptococcus. No similar preventive strategy exists for late-onset sepsis except for meticulous hand
All VLBW infants evaluated for late-onset neonatal sepsis (at >7 days of life) were eligible for the study. Infants born with
congenital anomalies, twin–twin transfusion syndrome, placenta abruption, or immediate postnatal hemoglobin level <10
g/dL on admission were excluded because of the important role of hepcidin in anemia.11Sepsis was defined post hoc as a pos-
itive blood culture and/or antibiotic therapy for 5 or more days in infants with
From the1Division of Neonatology, University of
Southern California Keck School of Medicine, Los
Angeles, CA;2Department of Pediatrics, Keio University
School of Medicine, Tokyo, Japan; Divisions of
3Biostatistics and Epidemiology, and4Neonatology,
Cincinnati Children’s Hospital Medical Center,
The authors declare no conflicts of interest.
0022-3476/$ - see front matter. Copyright ª 2013 Mosby Inc.
All rights reserved. http://dx.doi.org/10.1016/j.jpeds.2012.06.010
Classification and regression tree
Coefficient of variation
Negative predictive value
Positive predictive value
Very low birth weight
clinical signs of infection such as persistent cardiorespiratory
instability, neutropenia, immature-to-total granulocyte ratio
>0.2, and elevated CRP concentrations. Infants who were
treated for #3 days and with negative blood culture results
were classified as nonseptic. This study was conducted at
The University Hospital Neonatal Intensive Care Unit in
Cincinnati, Ohio, and was approved by the Institutional Re-
view Board of the hospital. The care providers were unaware
As per neonatal intensive care unit protocol, infants sus-
pected of sepsis are evaluated routinely with complete blood
antibiotic therapy. For this study, contemporaneous blood
specimens were assessed for hepcidin (designated acute spec-
imens). Blood specimens were also obtained 2 weeks
(denoted as convalescent specimens) after the initial evalua-
convalescent specimens. Hepcidin and CRP concentrations
were measured in healthy full-term infants to generate nor-
mative data for comparison with reference values reported
for adults. Duration of antibiotic therapy was at the discre-
tion of the providers who were not involved in the study.
Medical records were abstracted for demographic data, he-
matocrit, reticulocyte counts, CRP, duration of antibiotic
therapy, and blood culture results.
Hepcidin concentrations were
enzyme-linked immunosorbent assay, which detects the
25-amino acid mature form of hepcidin. The intra-assay co-
efficient of variation (CV) was 5%-19% and the inter-assay
reproducibility had an average CV of 12%.12
assessed by capture
Thenatural logarithmic (ln)transformationsofhepcidinand
CRP wereused inorder tostabilize thevarianceofthese mea-
to statistical testing and hepcidin values on the original scale
when nonparametric tests were used or when transformed
back to the original scale. Continuous demographic and clin-
ical variables are reported as means with SD if normally dis-
tributed, means with CV if back transformed, or as medians
with IQRs if non-normally distributed. Correlations were
tested between continuous variables with Pearson or Spear-
man rank correlation coefficient. Differences in continuous
variables between infants with and without sepsis were tested
using an independent t test. An ANOVA with Tukey–
Kramer’s test for multiple comparisons was used to evaluate
differences in ln (hepcidin) between infants without sepsis,
infants with culture-negative sepsis, and infants with
culture-positive sepsis. Prediction models for neonatal sepsis
were developed using logistic regression. Hepcidin and CRP
were tested separately and in combination as predictors,
with the following potential covariates: gestational age,
birth weight, age at sample collection, hematocrit, mode
of delivery, sex, nutrition (enteral/parenteral), respiratory
support (mechanical ventilation/continuous positive airway
pressure/room air), and blood transfusion. Covariates with
P # .2 in bivariate analysis were retained in the regression
a parsimonious final model, considering Akaike’s informa-
ORwith 95% CI. In order to assess the utility of serum hepci-
din as a biomarker for late-onset neonatal sepsis, we com-
pared the areas under receiver operating characteristic
curves between hepcidin and CRP using a c2test. Cutoff
points of hepcidin concentrations that differentiated infants
with and without sepsis were determined using classification
and regression tree (CART) analysis. The inherent nonpara-
metric properties of CART render it an appropriate method
for determining cutoff points. To assess the pathophysiologic
relevance of hepcidin, we compared concentrations in acute
and convalescent sera using a paired t test. Statistical signifi-
cance was determined at a = 0.05. CART analysis was per-
formed using DTREG software v. 10.0.1 (Phillip Sherrod).
tute, Cary, North Carolina).
A total of 44 VLBW and 21 term infants were enrolled in this
study between October 2008 and September 2011. Mean
(SD) birth weight was 885 (245) g and mean gestational
age was 26.2 (1.7) weeks among VLBW infants and 3297
(348) g and 38.8 (1.2) weeks among term infants. Seventeen
of the VLBW infants had positive blood cultures and/or were
treated with antibiotics for >5 days. The mean SD age at
differences between infants with and without sepsis in the
study with respect to birth weight or Apgar scores. Bacteria
recovered from blood cultures included Escherichia coli
(n = 2), Streptococcus agalactiae (2), coagulase-negative
staphylococci (5), and Serratia marcencens (1). Two infants
in the septic group died of E. coli sepsis and necrotizing en-
terocolitis before completion of antibiotic therapy, and 1 in-
fant in the nonseptic group died of intractable respiratory
sis were 1 week older by post-menstrual age and blood sam-
ples were collected at an earlier age (Table). Otherwise, there
were no significant correlations between hepcidin and
gestational age, birth weight, or hematocrit, all P > .05.
There was a significant but weak negative correlation
between hepcidin and age at sample collection (s = ?0.34,
P = .02) and positive correlation between hepcidin and
CRP (s = 0.39, P = .008).
The range of hepcidin concentration was 5.3-89.8 ng/mL
in infants without sepsis and 26.8-67.7 ng/mL in healthy
term infants. There was a significant stepwise increment in
mean hepcidin concentration from infants without sepsis
(43.9 ng/mL, CV 13.7%) to infants with sepsis with negative
blood cultures (99.2 ng/mL, CV 22.4%) to infants with sepsis
with positive blood cultures (244.8 ng/mL, CV 10.7%;
P < .0001).
Logistic regression models adjusted for gestational age
demonstrated that individually, hepcidin and CRP were
THE JOURNAL OF PEDIATRICS
Vol. 162, No. 1
Wu et al
significant predictors of neonatal sepsis (Figure 1). Increases
andincreasesinln(CRP), adjusted OR1.4(95% CI:1.1,1.7).
The areas under receiver operating characteristic curves for
logistic regression models were 0.93 (95% CI: 0.83, 1.0) and
0.83 (95% CI: 0.71, 0.95) for ln (hepcidin) and ln (CRP)
respectively, P = .06 (Figure 2). Models combining
hepcidin and CRP did not perform better than hepcidin
alone (data not shown). CART analysis established hepcidin
values > 92.2 ng/mL as a cutoff point, which correctly
classified 91% of all infants, with positive predictive value
(PPV) 100%, negative predictive value (NPV) 87%,
specificity 100%, and sensitivity 76%. When analysis was
restricted to infants with positive blood cultures, the PPV,
NPV, specificity, and sensitivity of hepcidin were 75%,
97%, 91%, and 90% respectively. In comparison, CART
analysis determined CRP values > 7.95 mg/dL as the cutoff
point, with PPV 89%, NPV 74%, specificity 96%, and
sensitivity 47%. Pairwise comparisons revealed a greater
decrease in hepcidin between acute hepcidin (median
226.5) and convalescent sera (median 42.1) in infants with
sepsis (mean decrease of 1.4 ng/mL ln (hepcidin),
P < .0001) compared with infants without sepsis (mean
decrease 0.2 ng/mL ln (hepcidin), P = .04). Concentrations
of hepcidin in convalescent sera of infants with sepsis did
not significantly differ from hepcidin samples in infants
without sepsis (P = .11).
Timely diagnosis and prompt institution of antimicrobial
therapy are essential in order to mitigate the high case fatality
and to avert morbidity associated with late-onset neonatal
sepsis. The vexing challenge in managing neonatal sepsis is
the inability to differentiate infants who are truly septic, in
whom antimicrobial therapy could be life-saving, from those
infants who have a low probability of sepsis in whom antibi-
otics can be safely withheld or discontinued.13The human
Table. Demographic and clinical characteristics of
VLBW infants and concentrations of serum markers of
(n = 27)
(n = 17)
Gestational age (wk), mean (SD)
Birth weight (g), mean (SD)
Age at sampling (d), mean (SD)
Mode of delivery, Cesarean
delivery, n (%)
Male, n (%)
Nutritional support, (TPN), n (%)
(mechanical ventilator/CPAP), n (%)
Blood transfusions, n (%)
Acute ln (hepcidin), mean (SD)
Acute ln (CRP), mean (SD)
Hematocrit as %, mean (SD)
Convalescent hepcidin, mean (SD)
Convalescent CRP, mean (SD)
833.6 (228.8) 965.4 (255.9)
CPAP, continuous positive airway pressure; TPN, total parenteral nutrition.
*Fisher exact test.
Figure 1. Predicted probabilities (curve) with 95% confidence limits (shaded region) for sepsis at the mean gestational age (26.2
weeks) based on hepcidin concentration (left panel) compared with CRP concentration (right panel). The observed values are
represented by circles for infants with negative cultures and by plus signs for infants with positive cultures.
The Utility of Serum Hepcidin as a Biomarker for Late-Onset Neonatal Sepsis
immune system consists of an armamentarium of defensive
strategies that militate againstthe invasionof harmfulmicro-
organisms or ameliorate the deleterious consequences of
sepsis. Assessment of the concentrations of proteins that
are induced in response to sepsis can serve as adjuncts to
blood culture. Our data affirmed the diagnostic utility of
CRP. We also showed that serum concentrations of hepcidin
were significantly increased in infants who were bacteremic
or who were treated for sepsis. The biologic relevance of hep-
ing completion of therapy, hepcidin concentrations in
convalescent sera from infants with sepsis were comparable
with levels in infants without sepsis or healthy term infants.
In addition, we report that hepcidin was comparable with
CRP, a well-established marker, in identifying neonates
with sepsis. The NPV of hepcidin is sufficiently high that uti-
lization of hepcidin concentrations may minimize needless
exposure of VLBW infants to empiric antimicrobials when
the probability of sepsis is low.
Although hepcidin is reputed for its central role in anemia,
the biologic plausibility of our data is supported by the
munity.14,15Lactoferrin is an iron chelator that exhibits bac-
tericidal and antiviral activity through immunoregulatory
effects on T-helper cells.16Transferrin, an iron transporter,
is required early in T-cell differentiation.17Ferritin is synthe-
sized by macrophages and T-lymphocytes.18Hepcidin plays
a dual function in innate immunity: it enhances intracellular
sequestration of iron by degrading the membrane iron trans-
tial mineral,19and it has direct antimicrobial activity.20The
levels of the pro-hormone, pro-hepcidin, has been reported
inpregnant mothers andtheir offspring.21However, thecon-
hepcidin, the active form of the protein. To our knowledge,
the current study is the first to assess the pathophysiologic
relevance of hepcidin in late-onset sepsis in neonates.
The concentrations of hepcidin were dramatically elevated
in infants who were bacteremic or who were treated for pre-
sumed sepsis. Consistent with the report that synthesis of
hepcidin is up-regulated by lipopolysaccharide and interleu-
kin-6,8our data suggest that infection is a potent inducer of
reliably detected and quantified in the urine.22This will ob-
viate the need for frequent blood draws for serial measure-
ments of hepcidin during evaluations for sepsis. These
properties make hepcidin a potential biomarker for sepsis.
The gold standard for diagnosing neonatal sepsis remains
the blood culture. However, negative blood cultures are dif-
ficult to interpret because bacteremia can be intermittent in
neonates with sepsis. Therefore, biomarkers with high sensi-
tivity and high negative predictive accuracies are desirable in
order to identify infants who are truly septic or to minimize
inappropriate exposure of those infants with low probability
of sepsis to antimicrobials. Most of the well-established
markers (eg, CRP, elastase a-1 protease inhibitor, CD64,
and procalcitonin), are useful in excluding infection because
they have high negative predictive accuracies. However, they
hepcidin reported in the current study suggest that hepcidin
the diagnosis of neonatal sepsis.
Figure 2. Receiver operating characteristic curves for hepcidin (left panel) compared with CRP (right panel) as predictors of
sepsis in VLBW infants suspected of late-onset sepsis. AUC, areas under receiver operating characteristic curves.
THE JOURNAL OF PEDIATRICS
Vol. 162, No. 1
Wu et al
There are several limitations to our study that are worthy
of note. We acknowledge that the definition of sepsis used
ogens were recovered from the blood culture. However,
treatment bias on the part of the providers is highly unlikely
the analysis focusing on the sepsis status was performed post
hoc. Also, the comparison between the diagnostic perfor-
mance of hepcidin and CRP is confounded by the timing
of specimen collection because hepcidin is reported to peak
at 6 hours after interleukin-6 injection in humans, and
CRP peaks at 24-48 hours after an inflammatory stimulus.
Therefore, hepcidin concentration and CRP levels very early
in thecourseof late-onset sepsis asreported in our study may
have underestimated the predictive ability of CRP as a bio-
marker for sepsis. Our study assesses the utility of hepcidin
in the setting of late-onset sepsis. It is unclear if our conclu-
sion can be extrapolated to early-onset neonatal sepsis. The
sample size in this study is relatively small, and the data
need to be validated in a larger cohort of VLBW infants. Al-
though hepcidin plays a central role in iron homeostasis and
to expectation, there were no correlations between hepcidin
and the hemoglobin, reticulocyte count, or ferritin concen-
quency of red blood cell transfusions in this patient
infants evaluated for late-onset sepsis. In addition to blood
culture and other markers of infection, hepcidin concentra-
tion may be a useful adjunct test in the evaluation of late-
onset sepsis in VLBW infants. Before adoption into clinical
tings, including early-onset neonatal sepsis, other pediatric
age groups, and in a larger cohort of infants are warranted. n
The authors gratefully acknowledge Dr Steve Hoath for thoughtful cri-
tique of the early draft of the manuscript and Dr Jareen Meinzen-Derr
for assistance with statistical analysis.
Submitted for publication Jan 13, 2012; last revision received May 22, 2012;
accepted Jun 5, 2012.
Reprint requests: Henry Akinbi, MD, Divisions of Pulmonary Biology and
Neonatology, Cincinnati Children’s Hospital Medical Center, 3333 Burnet Ave,
Cincinnati, OH 45229-3030. E-mail: firstname.lastname@example.org
1. Borghesi A, Stronati M. Strategies for the prevention of hospital-
acquired infections in the neonatal intensive care unit. J Hosp Infect
2. Stoll BJ, Hansen N. Infections in VLBW infants: studies from the
NICHD Neonatal Research Network. Semin Perinatol 2003;27:293-301.
3. Stoll BJ, Hansen NI, Adams-Chapman I. Neurodevelopmental and
growth impairment among extremely low-birth-weight infants with
neonatal infection. JAMA 2004;292:2357-65.
4. Connell TG, Rele M, Cowley D, Buttery JP, Curtis N. How reliable is
a negative blood culture result? Volume of blood submitted for culture
in routine practice in a childre’s hospital. Pediatrics 2007;119:891-6.
5. Kellogg JA, Ferrentino FL, Goodstein MH, Liss J, Shapiro SL,
Bankert DA. Frequency of low level bacteremia in infants from birth
to two months of age. Pediatr Infect Dis J 1997;16:381-5.
6. Edgar JD, Gabriel V, Gallimore JR, McMillan SA, Grant J. A prospective
studyof thesensitivity, specificityand diagnosticperformanceof soluble
intercellular adhesion molecule 1, highly sensitive C-reactive protein,
soluble E-selectin and serum amyloid A in the diagnosis of neonatal in-
fection. BMC Pediatr 2010;10:22.
7. Newman TB, Puopolo KM, Wi S, Draper D, Escobar GJ. Interpreting
8. Kemna E, Pickkers P, Nemeth E, van der Hoeven H, Swinkels D. Time-
course analysis of hepcidin, serum iron, and plasma cytokine levels in
humans injected with LPS. Blood 2005;106:1864-6.
9. Nemeth E, Rivera S, Gabayan V, Keller C, Taudorf S, Pedersen BK, et al.
IL-6 mediates hypoferremia of inflammation by inducing the synthesis
of the iron regulatory hormone hepcidin. J Clin Invest 2004;113:1271-6.
10. Patruta SI, Horl WH. Iron and infection. Kidney Int Suppl 1999;69:
11. M€ ullerKF,LorenzL,PoetsCF,WestermanM,FranzAR.HepcidinCon-
centrationsin Serum and UrineCorrelatewith Iron Homeostasis inPre-
term Infants. J Pediatr 2012.
human serum hepcidin. Blood 2008;112:4292-7.
13. Polin RA, Randis TM. Biomarkers for late-onset neonatal sepsis. Ge-
nome Med 2010;2:58.
14. Vyoral D, Petrak J. Hepcidin: a direct link between iron metabolism and
immunity. Int J Biochem Cell Biol 2005;37:1768-73.
15. Ong ST, Ho JZ, Ho B, Ding JL. Iron-withholding strategy in innate im-
munity. Immunobiology 2006;211:295-314.
16. Valenti P, Berlutti F, Conte MP, Longhi C, Seganti L. Lactoferrin func-
tions: current status and perspectives. J Clin Gastroenterol 2004;38:
17. Macedo MF, de Sousa M, Ned RM, Mascarenhas C, Andrews NC, Cor-
reia-NevesM. Transferrinis required forearly T-celldifferentiation. Im-
18. D€ orner MH, Silverstone A, Nishiya K, de Sostoa A, Munn G, de
Sousa M. Ferritin synthesis by human T lymphocytes. Science 1980;
19. Nemeth E, Tuttle MS, Powelson J, Vaughn MB, Donovan A, Ward DM,
inducing its internalization. Science 2004;306:2090-3.
20. Cuesta A, Meseguer J, Esteban MA. The antimicrobial peptide hepcidin
exerts an important role in the innate immunity against bacteria in the
bony fish gilthead seabream. Mol Immunol 2008;45:2333-42.
21. Ervasti M, Sankilampi U, Luukkonen S, Heinonen S, Punnonen K. Ma-
ternal pro-hepcidin at term correlates with cord blood pro-hepcidin at
birth. Eur J Obstet Gynecol Reprod Biol 2009;147:161-5.
22. Bansal SS, Halket JM, Bomford A, Simpson RJ, Vasavda N, Thein SL,
chromatography-mass spectrometry. Anal Biochem 2009;384:245-53.
23. Verboon-Maciolek MA, Thijsen SF, Hemels MA, Menses M, van
Loon AM, Krediet TG, et al. Inflammatory mediators for the diag-
nosis and treatment of sepsis in early infancy. Pediatr Res 2006;59:
The Utility of Serum Hepcidin as a Biomarker for Late-Onset Neonatal Sepsis