Correspondence: Y.

Qiu, Department of Paediatrics, Nanjing Maternal and Child Health Hospital of Nanjing Medical University, Nanjing 210004, China.
E-mail: qiuyufang0298@126.com
(Received 22 December 2009; accepted 19 April 2010)
Introduction
Neonatal sepsis remains one of the leading causes of
neonatal admission, morbidity and mortality in devel-
oping and developed countries [1,2]. It affects the
neonatal period of an estimated 12 40.5 per 1000
live-births [3 5]. This condition has a gradual and
subtle onset, with non-specic symptoms that may
severely compromise the infant s clinical state if
untreated, and may lead to life-threatening conse-
quences [6]. The mortality rate varies among coun-
tries, ranging from 7% to 26% of affected infants
[7 9].
Early treatment of neonatal sepsis is vital to
improve outcome. In the absence of reliable infection
markers during the rst hours of life, neonatologists
often begin early antibiotic treatment in newborn
infants with risk factors for infection, exposing
many neonates to unnecessary treatment. This treat-
ment strategy is because the early diagnosis of sep-
sis is difcult; isolation of causative organisms from
microbiological cultures takes up to 72 h and does
not identify most infected infants [10].
Procalcitonin (PCT), a 116-amino acid peptide
involved as a precursor in calcium homeostasis, has
been studied as a marker to differentiate sepsis from
other non-infectious disease [11,12]. Early studies
were encouraging, and PCT has been proposed for
inclusion as a diagnostic marker in the international
denition of sepsis [13]. More recent studies, how-
ever, have produced conicting results [12,14 16]. At
the same time, it is well established that neonatal
PCT concentrations show a physiological increase
during the rst 2 days of life, which complicates the
interpretation of results during this period [17]. While
different hourly cut-off values have been proposed for
neonates, the diagnostic relevance of PCT for new-
born sepsis is still debated [18]. The aim of this review
was therefore to systematically and quantitatively
evaluate all published studies that have assessed the
diagnostic use of PCT for neonatal sepsis.
ORIGINAL ARTICLE
The accuracy of the procalcitonin test for the diagnosis of neonatal
sepsis: A meta-analysis
ZHANGBIN YU
1
, JIEBO LIU
2
, QING SUN
1
, YUFANG QIU
1
, SHUPING HAN
1

& XIRONG GUO
1

From the
1
Department of Paediatrics, Nanjing Maternal and Child Health Hospital of Nanjing Medical University, Nanjing,
China, and
2
Department of Paediatrics, The Fifth People s Hospital of Shenzhen, Shenzhen, China

Abstract
A meta-analysis was performed to assess the accuracy of the procalcitonin (PCT) test for diagnosing neonatal sepsis.
The major databases, MEDLINE, EMBASE and the Cochrane Library were searched for studies published between
January 1996 and May 2009 that evaluated PCT as a diagnostic marker for neonatal sepsis and provided sufcient
data to calculate sensitivity and specicity. Twenty-two studies were included in the analysis. Trials that evaluated the
PCT test for the diagnosis of early-onset neonatal sepsis at different time points (birth, 0 12 h, 12 24 h, and 24 48 h)
and late-onset neonatal sepsis (LONS) all showed moderate accuracy ( Q

0.79, 0.86, 0.81, 0.82, and 0.77, respec-

tively). The PCT test was more accurate than the C-reactive protein (CRP) test for the diagnosis of LONS. A sensitiv-
ity analysis found that differences in PCT assay producer, gestational age and severity of sepsis in the study population
may partially explain the between-studies heterogeneity. The PCT test showed moderate accuracy in diagnosing neo-
natal sepsis, regardless of differences in diagnostic criteria and time points for testing. For the diagnosis of LONS, the
PCT test showed better accuracy than the CRP test. PCT is a valuable additional tool for the diagnosis of neonatal
sepsis.
Scandinavian Journal of Infectious Diseases, 2010; 42: 723733
ISSN 0036-5548 print/ISSN 1651-1980 online 2010 Informa Healthcare
DOI: 10.3109/00365548.2010.489906
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
724 Z. Yu et al.
Methods
Identication of studies
A protocol was written before this study was under-
taken, as recommended by the Cochrane handbook
for diagnostic test accuracy reviews [19,20]. Two
investigators (ZY and JL) independently searched
MEDLINE, EMBASE and the Cochrane Library for
relevant citations from 1 January 1996 through 30
May 2009. Search terms included procalcitonin ,
PCT , neonate , newborn , infant , sepsis , septi-
caemia , and combinations of these terms. The refer-
ence lists of all known primary and review articles
were examined to identify cited articles not captured
by electronic searches. No language restriction was
used and all foreign language publications were
translated.
Study selection and data extraction
Studies that evaluated PCT levels as diagnostic
markers for neonatal sepsis were considered for
inclusion. The study selection was performed by 2
independent investigators (ZY and JL). Differences
between the 2 reviewers were resolved by consensus
or referred to a third reviewer (YQ) if necessary.
From each selected article, we extracted the fol-
lowing information on the study population and
methodology: timing of PCT test, study design, pro-
ducer of PCT assay, diagnostic cut-off points, meth-
odological quality, and diagnostic gold standard.
Data collected for this standard included verication
of sepsis diagnosis by microbiological (positive blood,
urine or cerebrospinal uid cultures) laboratory evi-
dence, and/or signicant clinical evidence. If we were
able identify enough studies ( n 2 articles) that
compared both PCT and C-reactive protein (CRP)
tests for the diagnosis of neonatal sepsis according to
the different diagnostic criteria, we planned to carry
out a meta-analysis. Accuracy data were used to con-
struct 2 2 contingency tables. Some studies reported
sensitivity and specicity data for many cut-off points.
In these cases, we chose the cut-off point with the
best efciency value, which was estimated by dividing
the sum of true-positive and true-negative cases by
the total number of cases. Where accuracy data were
not extractable, we contacted the corresponding
author by e-mail to seek his or her assistance in data
extraction.
Assessment of methodological quality
Study quality was assessed using the QUADAS tool,
based on the recommended methods of the Cochrane
handbook for diagnostic test accuracy reviews, with
each item scored as yes , no , or unclear [20,21].
Statistical analysis
Studies were grouped according to the diagnostic cri-
teria for sepsis and the timing of the PCT test. Diag-
noses were categorized as proven sepsis (culture-proven
sepsis) and proven or probable sepsis (culture-proven
sepsis or clinical sepsis). The time points of PCT
measurement for the diagnosis of early-onset neona-
tal sepsis (EONS) were birth, 0 12 h, 12 24 h, and
24 48 h; that for the diagnosis of late-onset neonatal
sepsis (LONS) was greater than 72 h of age. The
sensitivity, specicity and diagnostic odds ratio (OR)
and corresponding 95% condence intervals (CI)
were calculated for each study.
A pooled diagnostic OR was calculated for each
of the diagnostic tests evaluated. The diagnostic OR
expresses how much greater the odds of having sep-
sis are for newborns who have a positive test result,
relative to newborns who have a negative test result
[22]. Heterogeneity in the estimates of diagnostic OR
was assessed with use of the Cochran Q statistic [23].
We explored the reasons for heterogeneity by per-
forming subgroup analyses. We examined whether
characteristics of the patient population (e.g., disease
severity levels, group risk factors), different cut-off
values, and PCT assay producer affected the diag-
nostic accuracy of the PCT test.
We constructed summary receiver operator char-
acteristic (SROC) curves, which represented the
relationship between sensitivity and the proportion
of false-positives (1 specicity), to summarize the
study results. Q

values were calculated from the

SROC curves; this value represents the intersection
point of the SROC curve with a diagonal line at
which sensitivity equals specicity [24]. All analyses
were performed using Review Manager 5.0 (Oxford,
UK: The Cochrane Collaboration) and Meta-DiSc
version 1.4 [25].
Results
Literature identication and quality of the studies
The literature search identied 446 studies. Of these,
24 reports encompassing 22 studies met our inclu-
sion criteria. Figure 1 shows the literature search
leading to the selected nal 22 articles [14,18,26
45]. Descriptive information for each included study
is presented in Table I.
The methodology among the included studies
was variable. Figure 2 summarizes study quality.
The major limitations lay in the categories of:
blinded reference standard results; explanation of
withdrawals; acceptable delay between tests;
reporting of uninterpretable results; and blinding
of index test results. Strengths in the studies related
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Accuracy of the PCT test for diagnosing neonatal sepsis 725
to: preclusion of sponsoring; acceptable reference
standards; avoidance of partial or differential veri-
cation; representative spectrum; and avoidance of
incorporation.
Accuracy of the PCT test in the diagnosis of proven or
probable early-onset neonatal sepsis
Thirteen trials evaluating the use of the PCT test in
the diagnosis of proven or probable EONS, dened as
culture-proven sepsis or clinical sepsis in the rst 72
h after delivery, met the inclusion criteria [14,18,26
36]. Studies were grouped according to time points
for the measurement of PCT (at birth, 0 12 h, 12 24
h, and 24 48 h).
Among these trials, 3 collected umbilical cord
blood samples at birth [26 28]. The pooled sensitiv-
ity, specicity and diagnostic OR are shown in Figure
3 (part 3.1). The SROC curves for PCT markers are
plotted in Figure 4.
Eight trials evaluating the use of the PCT test in
the diagnosis of EONS collected blood samples
within 0 12 h [18,29 35]. The pooled sensitivity,
specicity and diagnostic OR are shown in Figure 3
(part 3.2). The SROC curves for PCT markers are
plotted in Figure 4.
Four trials evaluating the use of the PCT test in
the diagnosis of EONS collected blood samples
within 12 24 h [14,18,29,33]. The pooled sensitivity,
specicity and diagnostic OR are shown in Figure 3
(part 3.3). The SROC curves for PCT markers are
plotted in Figure 4.
Six trials evaluating the use of the PCT test in
the diagnosis of EONS collected blood samples
within 24 48 h [18,29,32 34,36]. The pooled sensi-
tivity, specicity and diagnostic OR are shown in
Figure 3 (part 3.4). The SROC curves for PCT
markers are plotted in Figure 4.
Accuracy of the PCT test in the diagnosis of late-onset
neonatal sepsis
Eight trials evaluating the use of the PCT test in the
diagnosis of proven LONS, dened as culture-proven
sepsis after 72 h, met the inclusion criteria [29,37
43]. All of these trials collected blood samples at time
points after 72 h. The pooled sensitivity, specicity
and diagnostic OR are shown in Figure 5 (part 5.1).
The SROC curves for PCT markers are plotted in
Figure 6.
Seven trials evaluating the use of the PCT test in
the diagnosis of proven or probable LONS met the
inclusion criteria [30,32,37,42,44 46]. The pooled
sensitivity, specicity and diagnostic OR are shown
in Figure 5 (part 5.2). The SROC curves for PCT
markers are plotted in Figure 6.
Potentially relevant publications identified
from electronic database searches
(n = 446)
Papers excluded on the basis of title and
abstract, clearly did not contain data on true-
positive and false-positive results (n = 412)
Papers retrieved for more detailed
evaluation (n = 32)
Studies with outcome data useful in
meta-analysis (n = 22)
Studies included in meta-analysis
(n = 24)
Studies assessed in duplicate for inclusion
(n = 2)
Study population including both newborns
and infants (n = 1)
No definite separating time point for PCT
measurement (n = 5)
Insufficient data to calculate sensitivity and
specificity (n = 2)
Figure 1. Summary of study assessment and inclusion in the meta-analysis of studies involving diagnosis of neonatal sepsis using a
procalcitonin test.
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
726 Z. Yu et al.
Table I. Characteristics of studies included in the meta-analysis of the diagnosis of neonatal sepsis using a procalcitonin test.
Study Population Study design
Patients
( n )
Sepsis
diagnosis
PCT
assay
a

Time of PCT
test: cut-off
(ng/ml)
Guibourdenche, 2002 [14] Full-term/premature neonates
suspected of sepsis
Prospective 120 Culture;
clinical
qILMA 12 24 h: 2.5
Santuz, 2008 [18] Newborns at risk of early-onset
neonatal sepsis
Retrospective 149 Culture;
clinical
qCLIA At birth: 2.5
24 h: 34
48 h: 15
Kordek, 2008 [26] Cases: newborns with intrauterine
bacterial infection; controls:
non-infected newborns
Prospective 286 Culture;
clinical
qILMA At birth: 1.22
Kordek, 2003 [27] Cases: infected full-term/premature
neonates with perinatal risk
factors; controls: non-infected
full-term and premature neonates
Prospective 187 Culture;
clinical
qILMA At birth: 1.2
Joram, 2006 [28] Cases: full-term/premature neonates
with suspected maternofetal
infection; controls: full-term/
non-infected premature neonates
Prospective 197 Culture;
clinical
sqICA At birth: 0.5
Lopez Sastre, 2007 [29] Consecutive neonates of neonatal
services
Prospective 205 Culture;
clinical
qILMA At birth: 0.55
12 24 h: 4.7
36 48 h: 1.7
72 h: 0.59
Ucar, 2008 [30] Cases: NICU patients suspected of
sepsis; controls: non-infected
newborns
Prospective 72 Culture;
clinical
qILMA At birth: 0.8
72 h: 0.8
Pastor Peidro, 2007 [31] Consecutive neonates of neonatal
services
Prospective 123 Culture;
clinical
sqICA 0 12 h: 2
Cetinkaya, 2009 [32] Cases: NICU preterm neonates
suspected of sepsis; controls:
non-infected preterm neonates
Prospective 163 Culture;
clinical
qILMA At birth: 0.55
48 h: 4.7
72 h: 1.7
Chiesa, 2003 [33] Consecutive critically ill NICU
newborns
Retrospective 134 Culture;
clinical
sqICA At birth: 1
24 h: 100
48 h: 50
Maire, 1999 [34] Consecutive neonates of neonatal
services
Prospective 102 Culture;
clinical
qILMA At birth: 1.5
24 48 h: 10
Koksal, 2007 [35] Cases: NICU patients suspected of
sepsis; controls: non-infected
newborns
Prospective 55 Culture;
clinical
sqICA At birth: 0.34
Perez Solis, 2006 [36] Cases: NICU patients suspected of
sepsis; controls: non-infected
newborns
Prospective 69 Culture;
clinical
qILMA 24 48 h: 3.3
72 h: 0.65
Isidor, 2007 [37] NICU patients suspected of sepsis Prospective 176 Culture;
clinical
sqICA 72 h: 0.5
Fendler, 2008 [38] NICU preterm neonates suspected
of sepsis
Retrospective 78 Culture sqICA 72 h: 0.99
Sherwin, 2008 [39] NICU patients suspected of sepsis Prospective 164 Culture sqICA 72 h: 98
Boo, 2008 [40] NICU patients suspected of sepsis Prospective 87 Culture sqICA 72 h: 2
Sakha, 2008 [41] NICU patients suspected of sepsis Prospective 117 Culture qCLIA 72 h: 2
Vazzalwar, 2005 [42] VLBW infants suspected of sepsis Prospective 67 Culture;
clinical
qILMA 72 h: 0.5
Chiesa, 1998 [43] Cases: full-term and premature
neonates suspected of sepsis;
controls: non-infected neonates
Prospective 115 Culture;
clinical
qILMA 72 h: 2
Turner, 2006 [44] NICU patients suspected of sepsis Prospective 85 Culture;
clinical
qILMA 72 h: 2.3
Ma, 2004 [45] Cases: NICU patients suspected of
sepsis; controls: non-infected
full-term and premature neonates
Retrospective 85 Culture;
clinical
sqICA 72: 2
NICU, neonatal intensive care unit; PCT, procalcitonin test; VLBW, very low birth weight.

a
sqICA, semi-quantitative immunochromatographic assay (Brahms, Berlin, Germany); qILMA, quantitative immunoluminometric assay
(Brahms, Berlin, Germany); qCLIA, quantitative chemiluminescent assay (Diasorin, Germany).
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Accuracy of the PCT test for diagnosing neonatal sepsis 727
Analysis of heterogeneity and sensitivity
Heterogeneity among the estimates of diagnostic OR
was assessed with the Cochran Q statistic. We found
considerable heterogeneity among the studies evalu-
ating the use of the PCT test in the diagnosis of
EONS that collected umbilical cord blood samples
at birth ( I
2
86%). Sensitivity analyses (Table II)
indicated that the differences in diagnostic cut-off
points may partially explain the between-studies het-
erogeneity. Excepting the study of Joram et al. [28],
with a different cut-off point, the remaining studies
showed homogeneity ( I
2
0%).
Heterogeneity was also found among studies eval-
uating the use of the PCT test in the diagnosis of
EONS that collected blood samples within 0 12 h
( I
2
76%). Subgroup analyses found that the study
population explains the between-studies heterogene-
ity. Three trials analyzed all consecutive newborns in
neonatal services, who were considered to lack risk
factors for sepsis [29,31,34]. Four trials enrolled new-
borns at risk of infection, admitted to neonatal inten-
sive care units [18,32,33,35]. The pooled diagnostic
ORs for the diagnosis of sepsis in low-risk neonates
and no-risk neonates were 135.8 and 8.2, respec-
tively; the diagnostic accuracy of PCT for the diag-
nosis of sepsis in low-risk neonates was signicantly
higher than in no-risk neonates ( p 0.05).
No obvious heterogeneity was found in those trials
evaluating the use of the PCT test in the diagnosis of
EONS that collected blood samples within 12 24 h
( p 0.05). We found considerable heterogeneity
among studies evaluating the use of the PCT test in
the diagnosis of EONS that collected blood samples
within 24 48 h ( I
2
75%). Sensitivity analyses
(Table II) indicated that the difference in the severity
of sepsis in the study population may partially explain
the between-studies heterogeneity. Excepting the
study of Chiesa et al. [33], with a study population
of consecutive critically ill neonatal intensive care
unit (NICU) newborns, homogeneity was found
among studies ( I
2
48%).
Heterogeneity was also found among studies
evaluating the use of the PCT test in the diagnosis
of proven LONS ( I
2
82%). Sensitivity analyses
were performed. When the studies of Vazzalwar et al.
[42] and Fen and Piotrowski [38] (difference in ges-
tational age), Chiesa et al. [43] (severity of sepsis),
and Sakha et al. [41] (different PCT assay producer:
quantitative chemiluminescent assay, Diasorin) were
removed, homogeneity was found among studies
( I
2
47%). However the results were not interpre-
table because 50% of the studies were excluded (this
is not methodologically acceptable).
Comparison of accuracy between the PCT and CRP
tests for LONS
Four trials evaluating the use of both the PCT and
CRP tests for proven LONS met the inclusion crite-
ria [37,39,40,42]. The pooled sensitivity for the PCT
test was 72% (95% CI 63 81%), compared with 55%
(95% CI 45 65%) for the CRP test. There was a
Figure 2. Assessment of study quality: review authors judgements about each methodological quality item are presented as percentages
across all included studies.
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
728 Z. Yu et al.
statistically signicant difference between the sensi-
tivities ( p 0.05). Pooled specicity for PCT mark-
ers was lower than for the CRP test (77% (95% CI
72 81%) vs 85% (95% CI 81 88%)), although this
difference was not statistically signicant ( p 0.05).
The pooled diagnostic OR for the PCT test was
higher than for the CRP test (11.6 (95% CI 5.2 26.0)
vs 8.6 (95% CI 3.5 21.0)), although this difference
was also not statistically signicant ( p 0.05). The
SROC curves for the PCT and CRP tests were plot-
ted; the Q

value was slightly higher for the PCT test

than for the CRP test (0.77 vs 0.75; p 0.05).
Five trials evaluating the use of both the PCT
and CRP tests for proven or probable LONS met
the inclusion criteria [36,37,42,44,45]. The pooled
sensitivity for the PCT test was higher than for the
CRP test (81% (95% CI 74 86%) vs 77% (95%
CI 70 83%)), although this difference was not
statistically signicant ( p 0.05). The pooled speci-
city for the PCT test was 92% (95% CI 88 95%),
compared with 79% (95% CI 74 84%) for CRP test.
This difference was statistically signicant ( p 0.05).
The pooled diagnostic OR for the PCT test was sig-
nicantly higher than for the CRP test (56.7 (95%
CI 28.0 114.9) vs 17.9 (95% CI 4.5 70.4); p
0.05). The Q

value was slightly higher for the PCT

test than for the CRP test (0.88 vs 0.81; p 0.05),
suggesting that in terms of overall accuracy, the PCT
test is better than the CRP test for the diagnosis of
proven or probable LONS.
Discussion
Early identication of neonatal sepsis would help
reduce the high mortality and morbidity associated
with this disorder [46]. The rst clinical signs of
Figure 3. Sensitivity and specicity of the procalcitonin test for the diagnosis of early-onset neonatal sepsis. FN, false-negative; FP, false-
positive; TN, true-negative; TP, true-positive.
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Accuracy of the PCT test for diagnosing neonatal sepsis 729
Figure 4. Summary receiver operating characteristic (SROC) curve of the procalcitonin test for the diagnosis of early-onset neonatal sepsis.
Each point contributing to the SROC curve represents one study.
Figure 5. Sensitivity and specicity of the procalcitonin test for the diagnosis of late-onset neonatal sepsis. FN, false-negative; FP, false-
positive; TN, true-negative; TP, true-positive.
neonatal sepsis are non-specic; laboratory indica-
tors include complete blood cell count and ratio of
immature to total neutrophils. These markers do not
have high sensitivity, especially if measured early in
the course of sepsis [47]. Therefore, a specic marker
for neonatal sepsis is desired. In this meta-analysis
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
730 Z. Yu et al.
we observed that the PCT test shows moderate
accuracy for the diagnosis of neonatal sepsis.
Ideally, tests would be both sensitive and specic;
in practice, there is often a choice between a more
sensitive test and a more specic test. A useful mne-
monic for determining the use of a specic test vs a
sensitive test is SpIn and SnOut . SpIn stands for
Specic tests rule In the condition when they re
positive . SnOut stands for Sensitive tests rule Out
the condition when they re negative [48]. The meta-
analysis suggests that the PCT test has a higher value
of specicity than sensitivity (SpIn/SnOut 1) in
both the diagnosis of EONS and LONS; this means
that the PCT test is a valuable additional tool for the
conrmatory diagnosis of neonatal sepsis.
Like other meta-analyses of diagnostic tests, our
work has limitations. The studies included in the
meta-analysis were subjected to a cut-off value
dened by differences in instructions for the PCT
test and operators. The cut-off values in 2 studies
[18,33] at 24 h, ranging from about 30 to 100 ng/
ml, were very much higher than those found in other
studies. We believe that these studies used a binary
model of cut-off value to differentiate infected from
non-infected newborns, which is inappropriate to
clinical reality. Feinstein [49] proposed a 3-zone
partition (a grey zone), including a middle uncer-
tain, indeterminate, inconclusive zone of intermedi-
ate values, corresponding to a prediction not precise
enough for diagnostic decision . The indeterminate
value (a grey zone) rather than a single value may
better represent the physiological variation of disease
markers. This was also recognized in one of the ref-
erenced studies [18]. Further diagnostic studies on
the accuracy of the PCT test, with inclusion dened
by the same testing instruments and the grey zone of
cut-off value, are recommended [50].
Publication bias (lower probability of publication
of negative results) may be more difcult to avoid in
studies such as the one presented here than in meta-
analyses of randomized controlled trials [51]. In each
study, blinding of reference standard results and the
blindness of study design were often not fully reported;
the expertise of the test reader was never reported; and
the details of the analytical validation of the test were
not reported. Several studies did not explicitly explain
withdrawals, dene the acceptable delay between tests,
or report uninterpretable results. By pooling studies
dealing with various samples, clinical settings, and
study groups, we may have introduced excessive het-
erogeneity, since diagnostic tests may have different
accuracies during various phases of a disease.
Figure 6. Summary receiver operating characteristic (SROC) curve of the procalcitonin test for the diagnosis of late-onset neonatal sepsis.
Each point contributing to the SROC curve represents one study.
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Accuracy of the PCT test for diagnosing neonatal sepsis 731
In order to explore the reasons for heterogeneity,
we performed subgroup analyses. We observed dif-
ferences in diagnostic cut-off points, PCT assay pro-
ducer, gestational age and severity of sepsis in the
study population, which may partially explain the
between-studies heterogeneity. The considerable het-
erogeneity limits the feasibility of this meta-analysis;
there is a need for prospective studies of good meth-
odological quality (taking into account the very low
incidence of proven EONS, and powered to include
very large cohorts of newborns).
The studies were grouped according to 2 diag-
nostic criteria for sepsis and 5 time points for the
PCT test. Homogeneity was found among groups
that collected blood samples within 12 24 h and at
older than 72 h. The quality and quantity of
included studies affects the bias in the nal conclu-
sions and limits the feasibility of the meta-analysis.
We excluded these different studies and assured the
homogeneity of the included studies in a meta-
analysis. Only 3 trials evaluating the use of the PCT
test in the diagnosis of EONS collected umbilical
cord blood samples at birth in the meta-analysis.
This could induce a bias in the conclusions; an
adequately powered, high quality trial is needed to
investigate these ndings.
CRP is an acute-phase protein released by the
liver after the onset of inammation or tissue dam-
age [52]. The increase in serum concentration of
CRP is rather slow during the rst 12 24 h of infec-
tion, and this may negatively affect the sensitivity of
the test. CRP is neither highly specic nor sensitive
for bacterial infection since it can remain at low con-
centrations in bacterial infections and can increase
signicantly in viral infections [53]. The sensitivity
of CRP in initial determinations for the detection of
EONS has been reported to vary from 35% to 89%,
and specicity to vary at the initial determination
from 70% to 90% [54]. In the meta-analysis, PCT
sensitivity in the diagnosis of EONS was found to be
69 80%, while the specicity was 75 97%, and stud-
ies that compared accuracy between the PCT and
CRP tests for LONS showed that the PCT test was
more accurate, both for proven sepsis and proven or
probable sepsis. However, CRP was used to classify
the newborns into the infected or not infected group
Table II. Heterogeneity and sensitivity analyses of the use of a procalcitonin test for the diagnosis of neonatal sepsis.
Study
Sensitivity
(95% CI)
Specicity
(95% CI)
Pooled OR
(95% CI) Cochrans Q p-Value I
2
(%)
Effects
model
PCT test in the diagnosis of proven or probable early-onset neonatal sepsis
Umbilical cord blood
samples at birth
78% (6886%) 83% (8086%) 28.7 (5.3154.3) 14.7 0.01 86 Random
Studies using a cut-off
of 0.5 [26,27]
76% (6585%) 75% (7180%) 10.0 (5.617.8) 0.02 0.89 0 Fixed
Blood samples within
012 h
77% (7281%) 87% (8490%) 39.8 (13.6116.1) 29.6 0.01 76 Random
Neonates with no risk
factors for sepsis
[29,31,34]
76% (6782%) 76% (7181%) 8.2 (4.714.4) 2.2 0.34 8 Fixed
Neonates with low
risk factors
[18,32,33,35]
80% (7485%) 97% (9499%) 135.8 (48.4381.2) 3.0 0.39 1 Fixed
Blood samples within
1224 h
77% (68 84%) 89% (8591%) 30.0 (9.0 99.2) 10.5 0.06 52 Fixed
Blood samples within
2448 h
70% (6475%) 88% (8591%) 22.4 (10.150.0) 12.0 0.01 75 Random
Removal of the study
of Chiesa et al. with
consecutive
critically ill NICU
newborns [33]
73% (6481%) 86% (8290%) 16.7 (7.437.3) 3.8 0.15 48 Fixed
PCT test in the diagnosis of late-onset neonatal sepsis
Proven or probable
sepsis
62% (5767%) 92% (8995%) 49.3 (27.588.2) 5.1 0.53 0 Fixed
Proven sepsis 0.82% (0.770.86%) 77% (7380%) 18.5 (6.255.2) 38.5 0.01 82 Random
Removal of the study
of different
characteristics
[38,4143]
74% (66 81%) 82% (7886%) 13.2 (6.427.3) 5.6 0.13 47 Fixed
CI, condence interval; NICU, neonatal intensive care unit; OR, odds ratio; PCT, procalcitonin test.
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
732 Z. Yu et al.
and there were different criteria used to determine
the infectious status of newborns; the design of these
studies was not conducive to a PCT/CRP compari-
son, and this was a major bias that limited the feasi-
bility of this conclusion.
In conclusion, the use of PCT as a marker of
neonatal sepsis showed moderate accuracy regardless
of differences in diagnostic criteria for sepsis and
time points for PCT testing. The PCT test was also
shown to have better accuracy than the CRP test in
the diagnosis of LONS, and is a valuable additional
tool for the diagnosis of neonatal sepsis. Because of
the relatively poor quality and quantity of included
studies and the possibility of publication bias in the
meta-analysis, an adequately powered, high quality
trial is needed to investigate these ndings further.
Declaration of interest: None of the authors have
commercial relationships or other associations that
might pose a conict of interest (e.g., pharmaceutical
stock ownership, consultancy, advisory board mem-
bership, relevant patents, or research funding).
References
Qazi SA, Stoll BJ. Neonatal sepsis: a major global public [1]
health challenge. Pediatr Infect Dis J 2009;28:S1 2.
Lusyati S, van den Broek P, Sauer PJ. Neonatal sepsis in a [2]
neonatal intensive care unit in Indonesia. J Hosp Infect 2009;
71:383 5.
Shin YJ, Ki M, Foxman B. Epidemiology of neonatal sepsis [3]
in South Korea. Pediatr Int 2009;51:225 32.
Sundaram V, Kumar P, Dutta S, Mukhopadhyay K, Ray P, [4]
Gautam V, et al. Blood culture conrmed bacterial sepsis in
neonates in a North Indian tertiary care center: changes over
the last decade. Jpn J Infect Dis 2009;62:46 50.
Hervas JA, Alomar A, Salva F, Reina J, Bened VJ. Neonatal [5]
sepsis and meningitis in Mallorca, Spain, 1977 1991. Clin
Infect Dis 1993;16:719 24.
Lin YJ. The changing characteristics of neonatal sepsis in the [6]
neonatal intensive care unit: a never-ending challenge. Pedi-
atr Neonatol 2009;50:83 4.
Wu JH, Chen CY, Tsao PN, Hsieh WS, Chou HC. Neonatal [7]
sepsis: a 6-year analysis in a neonatal care unit in Taiwan.
Pediatr Neonatol 2009;50:88 95.
Seale AC, Mwaniki M, Newton CR, Berkley JA. Maternal and [8]
early onset neonatal bacterial sepsis: burden and strategies for
prevention in sub-Saharan Africa. Lancet Infect Dis 2009;
9:428 38.
Makhoul IR, Sujov P, Smolkin T, Lusky A, Reichman B; [9]
Israel Neonatal Network. Pathogen-specic early mortality
in very low birth weight infants with late-onset sepsis: a
national survey. Clin Infect Dis 2005;40:218 24.
Reier-Nilsen T, Farstad T, Nakstad B, Lauvrak V, Steinbakk [10]
M. Comparison of broad range 16S rDNA PCR and con-
ventional blood culture for diagnosis of sepsis in the new-
born: a case control study. BMC Pediatr 2009;9:5.
Uzzan B, Cohen R, Nicolas P, Cucherat M, Perret GY. Pro- [11]
calcitonin as a diagnostic test for sepsis in critically ill adults
and after surgery or trauma: a systematic review and meta-
analysis. Crit Care Med 2006;34:1996 2003.
Muller B, Christ-Crain M, Nylen ES, Snider R, Becker KL. [12]
Limits to the use of the procalcitonin level as a diagnostic
marker. Clin Infect Dis 2004;39:1867 8.
Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, [13]
Cook D, et al. 2001 SCCM/ESICM/ACCP/ATS/SIS inter-
national sepsis denitions conference. Crit Care Med 2003;
31:1250 6.
Guibourdenche J, Bedu A, Petzold L, Marchand M, Mariani- [14]
Kurdjian P, Hurtaud-Roux MF, et al. Biochemical markers
of neonatal sepsis: value of procalcitonin in the emergency
setting. Ann Clin Biochem 2002;39:130 5.
Lopez Sastre JB, Perez Solis D, Roques Serradilla V, Fern n- [15]
dez Colomer B, Coto Cotallo GD, Krauel Vidal X, et al.
Procalcitonin is not sufciently reliable to be the sole marker
of neonatal sepsis of nosocomial origin. BMC Pediatr 2006;
6:16.
Monneret G, Labaune JM, Isaac C, Bienvenu F, Putet G, [16]
Bienvenu J. Increased serum procalcitonin levels are not
specic to sepsis in neonates. Clin Infect Dis 1998;27:
1559 61.
Sachse C, Dressler F, Henkel E. Increased serum procalci- [17]
tonin in newborn infants without infection. Clin Chem 1998;
44:1343 4.
Santuz P, Sofati M, Dorizzi RM, Benedetti M, Zaglia F, [18]
Biban P. Procalcitonin for the diagnosis of early-onset neo-
natal sepsis: a multilevel probabilistic approach. Clin Bio-
chem 2008;41:1150 5.
Virgili G, Conti AA, Murro V, Gensini GF, Gusinu R. Sys- [19]
tematic reviews of diagnostic test accuracy and the Cochrane
collaboration. Intern Emerg Med 2009;4:255 8.
Deeks JJ, Bossuyt PM, Gatsonis C. Cochrane handbook for [20]
systematic reviews of diagnostic test accuracy; version 1.0.0.
Oxford, UK: The Cochrane Collaboration; 2009. Available
at: http://srdta.cochrane.org/en/authors.html (accessed).
Whiting P, Rutjes AW, Reitsma JB, Bossuyt PM, Kleijnen J. [21]
The development of QUADAS: a tool for the quality assess-
ment of studies of diagnostic accuracy included in systematic
reviews. BMC Med Res Methodol 2003;3:25.
Glas AS, Lijmer JG, Prins MH, Bonsel GJ, Bossuyt PM. The [22]
diagnostic odds ratio: a single indicator of test performance.
J Clin Epidemiol 2003;56:1129 35.
Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring [23]
inconsistency in meta-analyses. BMJ 2003;327:557 60.
Walter SD. Properties of the summary receiver operating [24]
characteristic (SROC) curve for diagnostic test data. Stat
Med 2002;21:1237 56.
Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. [25]
Meta-DiSc: a software for meta-analysis of test accuracy
data. BMC Med Res Methodol 2006;6:31.
Kordek A, Halasa M, Podraza W. Early detection of an early [26]
onset infection in the neonate based on measurements of
procalcitonin and C-reactive protein concentrations in cord
blood. Clin Chem Lab Med 2008;46:1143 8.
Kordek A, Giedrys-Kalemba S, Pawlus B, Podraza W, [27]
Czajka R. Umbilical cord blood serum procalcitonin concen-
tration in the diagnosis of early neonatal infection. J Perina-
tol 2003;23:148 53.
Joram N, Boscher C, Denizot S, Loubersac V, Winer N, Roze [28]
JC, et al. Umbilical cord blood procalcitonin and C reactive
protein concentrations as markers for early diagnosis of very
early onset neonatal infection. Arch Dis Child Fetal Neona-
tal Ed 2006;91:F65 6.
Lopez Sastre JB, Solis DP, Serradilla VR, Colomer BF, [29]
Cotallo GD; Grupo de Hospitales Castrillo. Evaluation of
procalcitonin for diagnosis of neonatal sepsis of vertical
transmission. BMC Pediatr 2007;7:9.
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Accuracy of the PCT test for diagnosing neonatal sepsis 733
Ucar B, Yildiz B, Aksit MA, Yarar C, Colak O, Akbay Y, et al. [30]
Serum amyloid A, procalcitonin, tumor necrosis factor-alpha,
and interleukin-1beta levels in neonatal late-onset sepsis.
Mediators Inamm 2008;2008:737141.
Pastor Peidro JA, Gonzalez de Dios J, Uran Moreno MM, [31]
Garc a Avil s B, de la Morena Campillo A, Moya Benavent
M. Usefulness of procalcitonin as an early diagnostic test of
neonatal sepsis in newborns with risk factors for infection.
An Pediatr (Barc) 2007;67:530 5.
Cetinkaya M, Ozkan H, Koksal N, Celebi S, Hacimustafao g lu [32]
M. Comparison of serum amyloid A concentrations with
those of C-reactive protein and procalcitonin in diagnosis
and follow-up of neonatal sepsis in premature infants. J Peri-
natol 2009;29:225 31.
Chiesa C, Pellegrini G, Panero A, Osborn JF, Signore F, [33]
Assumma M, et al. C-reactive protein, interleukin-6,
and procalcitonin in the immediate postnatal period:
inuence of illness severity, risk status, antenatal and
perinatal complications, and infection. Clin Chem 2003;49:
60 8.
Maire F, Heraud MC, Loriette Y, Normand B, B gue RJ, [34]
Labb A. The value of procalcitonin in neonatal infections.
Arch Pediatr 1999;6:503 9.
Koksal N, Harmanci R, Cetinkaya M, Hacimustafaog lu M. [35]
Role of procalcitonin and CRP in diagnosis and follow-up of
neonatal sepsis. Turk J Pediatr 2007;49:21 9.
Perez Solis D, Lopez Sastre JB, Coto Cotallo GD, [36]
Di guez Junquera MA, Deschamps Mosquera EM, Crespo
Hern ndez M. Procalcitonin for the diagnosis of neonatal
sepsis of vertical transmission. An Pediatr (Barc) 2006;64:
341 8.
Isidor B, Caillaux G, Gilquin V, Loubersac V, Caillon J, Roze [37]
JC, et al. The use of procalcitonin in the diagnosis of late-
onset infection in neonatal intensive care unit patients. Scand
J Infect Dis 2007;39:1063 6.
Fendler WM, Piotrowski AJ. Procalcitonin in the early diag- [38]
nosis of nosocomial sepsis in preterm neonates. J Paediatr
Child Health 2008;44:114 8.
Sherwin C, Broadbent R, Young S, Worth J, McCaffrey F, [39]
Medlicott NJ, et al. Utility of interleukin-12 and interleukin-10
in comparison with other cytokines and acute-phase reactants
in the diagnosis of neonatal sepsis. Am J Perinatol 2008;25:
629 36.
Boo NY, Nor Azlina AA, Rohana J. Usefulness of a semi- [40]
quantitative procalcitonin test kit for early diagnosis of neo-
natal sepsis. Singapore Med J 2008;49:204 8.
Sakha K, Husseini MB, Seyyedsadri N. The role of the pro- [41]
calcitonin in diagnosis of neonatal sepsis and correlation
between procalcitonin and C-reactive protein in these
patients. Pak J Biol Sci 2008;11:1785 90.
Vazzalwar R, Pina-Rodrigues E, Puppala BL, Angst DB, Sch- [42]
weig L. Procalcitonin as a screening test for late-onset sepsis
in preterm very low birth weight infants. J Perinatol 2005;25:
397 402.
Chiesa C, Panero A, Rossi N, Stegagno M, De Giusti M, [43]
Osborn JF, et al. Reliability of procalcitonin concentrations
for the diagnosis of sepsis in critically ill neonates. Clin Infect
Dis 1998;26:664 72.
Turner D, Hammerman C, Rudensky B, Schlesinger Y, [44]
Schimmel MS. The role of procalcitonin as a predictor of
nosocomial sepsis in preterm infants. Acta Paediatr 2006;95:
1571 6.
Ma L, Liu CQ, Liu ZH, Liu SZ, Jia XQ, Li WJ, et al. Sig- [45]
nicance of soluble intercellular adhesion molecule-1 and
procalcitonin in diagnosis of neonatal septicemia. Zhonghua
Er Ke Za Zhi 2004;42:654 8.
Gerdes JS, Polin R. Early diagnosis and treatment of neona- [46]
tal sepsis. Indian J Pediatr 1998;65:63 78.
Chiesa C, Panero A, Osborn JF, Simonetti AF, Pacico L. [47]
Diagnosis of neonatal sepsis: a clinical and laboratory chal-
lenge. Clin Chem 2004;50:279 87.
Coste J, Pouchot J. A grey zone for quantitative diagnostic [48]
and screening tests. Int J Epidemiol 2003;32:304 13.
Nylen E, Muller B, Becker KL, Snider R. The future diag- [49]
nostic role of procalcitonin levels: the need for improved
sensitivity. Clin Infect Dis 2003;36:823 4.
Feinstein AR. The inadequacy of binary models for the clin- [50]
ical reality of three-zone diagnostic decisions. J Clin Epide-
miol 1990;43:109 13.
Knudsen TB, Kristiansen TB. Issues pertaining to data [51]
extraction and classication and publication bias in meta-
analysis of the diagnostic accuracy of markers for bacterial
infection. Clin Infect Dis 2005;40:1372 3.
Hawk M. C-reactive protein in neonatal sepsis. Neonatal [52]
Netw 2008;27:117 20.
Caksen H, Kurtoglu S, Hallac IK, Ustunbas HB, Uzum K, [53]
Kilic H. The relationship between scoring systems and
cytokine levels in neonatal sepsis. Ann Acad Med Singapore
2003;32:418 20.
Benitz WE, Han MY, Madan A, Ramachandra P. Serial [54]
serum C-reactive protein levels in the diagnosis of neonatal
infection. Pediatrics 1998;102:E41.
S
c
a
n
d

J

I
n
f
e
c
t

D
i
s

D
o
w
n
l
o
a
d
e
d

f
r
o
m

i
n
f
o
r
m
a
h
e
a
l
t
h
c
a
r
e
.
c
o
m

b
y

N
y
u

M
e
d
i
c
a
l

C
e
n
t
e
r

o
n

0
6
/
1
3
/
1
2
F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.