PEDIATRICS/ORIGINAL RESEARCH
Comparison of the Test Characteristics of Procalcitonin to
C-Reactive Protein and Leukocytosis for the Detection of Serious
Bacterial Infections in Children Presenting With Fever Without
Source: A Systematic Review and Meta-analysis
Chia-Hung Yo, MD, Pei-Shan Hsieh, BPH, Si-Huei Lee, MD, Jiunn-Yih Wu, MD, Shy-Shin Chang, MD,
Kuang-Chau Tasi, MD, MSc, Chien-Chang Lee, MD, MSc
From the Department of Emergency Medicine, Far Eastern Memorial Hospital, Taipei, Taiwan (Yo, Tsai); the Department of Rehabilitation and
Physical Medicine, Taipei Veteran General Hospital, Taipei, Taiwan (S-H Lee); the Department of Family Medicine, Chang Gung Memorial Hospital,
Taoyuan, Taiwan (Chang); the Graduate Institute of Clinical Medical Sciences, College of Medicine, Chang Gung University, Taoyuan, Taiwan
(Chang); the Department of Emergency Medicine, Chang Gung Memorial Hospital, Keelung, Taiwan (Wu); and the Department of Emergency
Medicine, National Taiwan University Hospital Yunlin Branch, Douliou, Taiwan (Hsieh, C-C Lee).
Study objective: We determine the usefulness of the procalcitonin for early identification of young children at
risk for severe bacterial infection among those presenting with fever without source.
Methods: The design was a systematic review and meta-analysis of diagnostic studies. Data sources were
searches of MEDLINE and EMBASE in April 2011. Included were diagnostic studies that evaluated the
diagnostic value of procalcitonin alone or compared with other laboratory markers, such as C-reactive protein or
leukocyte count, to detect severe bacterial infection in children with fever without source who were aged
between 7 days and 36 months.
Results: Eight studies were included (1,883 patients) for procalcitonin analysis, 6 (1,265 patients) for C-reactive
protein analysis, and 7 (1,649 patients) for leukocyte analysis. The markers differed in their ability to predict
serious bacterial infection: procalcitonin (odds ratio [OR] 10.6; 95% confidence interval [CI] 6.9 to 16.0), Creactive protein (OR 9.83; 95% CI 7.05 to 13.7), and leukocytosis (OR 4.26; 95% CI 3.22 to 5.63). The randomeffect model was used for procalcitonin analysis because heterogeneity across studies existed. Overall
sensitivity was 0.83 (95% CI 0.70 to 0.91) for procalcitonin, 0.74 (95% CI 0.65 to 0.82) for C-reactive protein,
and 0.58 (95% CI 0.49 to 0.67) for leukocyte count. Overall specificity was 0.69 (95% CI 0.59 to 0.85) for
procalcitonin, 0.76 (95% CI 0.70 to 0.81) for C-reactive protein, and 0.73 (95% CI 0.67 to 0.77) for leukocyte
count.
Conclusion: Procalcitonin performs better than leukocyte count and C-reactive protein for detecting serious
bacterial infection among children with fever without source. Considering the poor pooled positive likelihood ratio
and acceptable pooled negative likelihood ratio, procalcitonin is better for ruling out serious bacterial infection
than for ruling it in. Existing studies do not define how best to combine procalcitonin with other clinical
information. [Ann Emerg Med. 2012;60:591-600.]
Please see page 592 for the Editor’s Capsule Summary of this article.
A feedback survey is available with each research article published on the Web at www.annemergmed.com.
A podcast for this article is available at www.annemergmed.com.
0196-0644/$-see front matter
Copyright © 2012 by the American College of Emergency Physicians.
http://dx.doi.org/10.1016/j.annemergmed.2012.05.027
SEE EDITORIAL, P. 601.
INTRODUCTION
Fever is a common reason for pediatric visits to the
emergency department (ED). Although the majority of patients
have minor bacterial or viral infections, it is important to
Volume , .  : November 
recognize those having serious bacterial infections to provide
appropriate care with antibiotics and early hospitalization.1,2
After history-taking and physical examination, it is estimated
that 20% of febrile infants and young children receive a
diagnosis of fever without an apparent source of infection.3 Of
these, about 20% may have severe bacterial infection, such as
lobar pneumonia, bacteremia, bacterial meningitis,
Annals of Emergency Medicine 591
Tests for Sources of Serious Bacterial Infections in Children
Editor’s Capsule Summary
What is already known on this topic
Only a small subset of febrile children younger than
3 years has a serious bacterial infection, but we lack
accurate markers to identify that subset.
What question this study addressed
A meta-analysis included studies of children
younger than 3 years with fever without source.
Eight studies with 1,887 cases evaluated
procalcitonin as a marker of serious bacterial
infection, including bacteremia, pneumonia, and
urinary tract infection.
What this study adds to our knowledge
Procalcitonin performed better than leukocytosis or
C-reactive protein at identifying serious bacterial
infection, with sensitivity 83% and specificity 69%.
How this is relevant to clinical practice
Procalcitonin may have some utility in identifying
serious bacterial infection, but it is not clear how
many of these infections could be identified with
other tests such as chest radiograph or urinalysis, or
how procalcitonin should be combined with other
clinical information.
pyelonephritis, or urinary tract infection.4-19 After the
introduction of an effective Hib and PCV7 vaccine, the rate of
severe bacterial infection decreased dramatically, with occult
bacteremia rates now ranging from 0.02% to 0.7%.20 However,
given the serious outcomes of missed diagnoses, this is still a
great diagnostic challenge for clinicians. The risk is greatest
among infants and children younger than 36 months, making
proper diagnosis and management paramount. For decades,
investigators have attempted to find clinical or laboratory
markers that can accurately differentiate severe bacterial
infection from localized or viral infections in young children
with fever without source21-28; unfortunately, no single, ideal
marker has been identified.7,10,12,13,29-39
The WBC count is routinely recommended as an initial
screening marker in children with fever without source. Creactive protein has been thought to be a more sensitive and
specific biomarker than leukocyte count40,41; in addition,
procalcitonin, the prohormone of calcitonin, has been shown to
distinguish bacterial from viral infections and to correlate well
with clinical severity.4-13,39,42,43 In healthy individuals,
circulating levels of procalcitonin are generally very low but can
increase by hundreds- to thousands-fold within 4 to 6 hours in
response to systemic bacterial infection. During the last decade,
numerous studies have evaluated the accuracy of procalcitonin
as a marker of severe bacterial infection in children with fever
592 Annals of Emergency Medicine
Yo et al
without source, and most compared it with C-reactive protein
or leukocyte count. The purpose of this study was to
quantitatively summarize, by means of a meta-analysis, all
existing evidence in the literature from such reports.
MATERIALS AND METHODS
We adhered to the methods and procedures of the Cochrane
Collaboration44 and the Preferred Reporting Items for Systematic
Reviews and Meta-analyses guidelines45 for reporting systematic
reviews.
We performed a comprehensive search of the MEDLINE,
EMBASE, and Cochrane databases for pertinent studies published
since inception to April 2011. Procalcitonin has not yet been listed
as a Medical Subject Headings term, so for our initial search, we
used “procalcitonin” as the text word, and we did not set any
language restriction. We identified additional references by crosschecking bibliographies of retrieved full-text articles.
Study Design
We included studies that met all of the following criteria: (1) age
range between 7 days and 36 months; (2) evaluation of
procalcitonin alone or compared with other laboratory markers,
such as C-reactive protein or leukocyte count, to detect severe
bacterial infection in children with fever without source; and (3)
sufficient data to construct a 2⫻2 contingency table. We excluded
studies having significant overlap (more than 50%) of study
patients with the selected studies. Two authors (C.-H.Y. and P.S.H.) independently assessed all titles and abstracts to determine
that inclusion criteria were satisfied. Full-text articles were retrieved
if either of the reviewers considered the abstract potentially suitable.
The 2 reviewing authors then independently assessed the full text of
the retrieved studies for their suitability for inclusion. Discrepancies
were resolved by having an additional reviewer (C.-C.L.) assess the
full article, and then consensus was reached about inclusion in the
meta-analysis. The 2 original reviewers independently extracted
data from each study selected. Extracted data comprised the
following: overall study characteristics (including the first author,
country, language, and date of publication), patient characteristics
(including age range and percentage of male patients), quantitative
data required for construction of a 2⫻2 table (including number of
participants, sensitivity, specificity, and case number), information
about the procalcitonin test (including cutoff levels, quantitative or
semiquantitative nature of the test), study settings, and outcomes.
The quality of the selected studies was determined using Quality
Assessment of Diagnostic Accuracy Studies criteria (Table 1).46 We
consistently used data having the highest sensitivity and performed
a sensitivity analysis in which we used the data with the lowest
sensitivity instead of the data with the highest sensitivity. Two
studies did not report full data required for inclusion in our metaanalysis, and, unfortunately, we did not receive a response from the
corresponding authors to our requests.
Data Collection and Processing and Primary Data Analysis
We used the bivariate random-effects model for diagnostic
meta-analysis to obtain weighted overall estimates of the
Volume , .  : November 
Yo et al
Tests for Sources of Serious Bacterial Infections in Children
Table 1. Quality assessment of diagnostic accuracy studies criteria for included studies.
Study ID
Lacour, 2001
Galetto-Lacour, 2003
Thayyil, 2005
Andreola, 2007
Guen, 2007
Maniaci, 2008
Olaciregui, 2009
Manzano, 2010
Spect Select
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Ref
Yes
Yes
Yes
Yes
Unclear
Yes
Yes
Yes
Indep
Time
Index
Period Vfull Vbias Test Testdesc Refdesc Blintest
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Blinref
Clin Indeterm Withdraw
Unclear
Unclear
No
Unclear
Unclear
Unclear
Unclear
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
Unclear
No
No
No
No
No
Yes
Yes
Yes
Yes
Yes
Unclear
Yes
Unclear
Spect, Representative spectrum of patients; Select, selection criteria clearly described; Ref, adequate reference standard; Time Period, short time period between reference and index test; Vfull, all patient verified by reference standards; Vbias, same reference standard used; Indep Index Test, reference independent of index test;
Testdesc, adequate index test description; Refdesc, adequate reference test; Blintest, blinding for index test; Blinref, blinding for reference test; Clin, clinical data
available; Indeterm, uninterpretable test result reports; Withdraw, withdraw explained.
sensitivity and specificity of procalcitonin as a marker for severe
bacterial infection in children with fever without source. The
bivariate approach assumes a bivariate distribution for the logittransformed sensitivity and specificity. In addition to
accounting for study size, the bivariate model estimates and
adjusts for the negative correlation between sensitivity and
specificity of the index test that may arise from the use of
different thresholds in different studies. A further advantage of
the bivariate model is that it uses a random-effects approach for
sensitivity and specificity, which allows for any heterogeneity
beyond chance resulting from clinical and methodological
differences between and among studies. A summary receiver
operating characteristic curve was constructed as a way to
summarize the true- and false-positive rates of the different
diagnostic criteria. The overall sensitivity, specificity, positive
likelihood ratio, and negative likelihood ratio, as well as their
corresponding 95% confidence intervals (CIs), were calculated
on the basis of the binominal distributions of the true positives
and true negatives.
We also performed the more conventional diagnostic odds
ratio meta-analysis. Unadjusted data were used exclusively in all
meta-analyses. Summary diagnostic odds ratios were estimated
by random- (DerSimonian-Laird) or fixed-effects models,
depending on whether I2 was greater than 50%. We used a
linear regression of log odds ratios on inverse roots of effective
sample sizes as a test for funnel plot asymmetry in diagnostic
meta-analyses. A nonzero slope coefficient is suggestive of
significant small-study bias (P⬍.10). To formally quantify the
extent of between-study variation (heterogeneity), we calculated
the I2 statistics. Statistically significant heterogeneity was
considered present at I2 greater than 50%. Sources of
heterogeneity then were examined by Galbraith plots and metaregression. We defined a priori the following clinical and design
characteristics of a study as potentially relevant covariates: cutoff
value, study setting, and age range of the study patients. We
performed an exploratory analysis by testing the covariates one
at a time in the meta-regression model. Statistical analyses were
conducted with Stata (version 10.0; StataCorp, College Station,
Volume , .  : November 
Figure 1. Flow chart of study identification and inclusion.
TX), notably, the midas and metandi commands. All statistical
tests were 2-sided, and statistical significance was defined as
P⬍.05.
RESULTS
Our search yielded 1,856 citations, 181 of which were
retrieved for full-text review. Of these, 173 articles were
excluded, mainly because related exposure or outcomes were not
studied or reported (Figure 1). A total of 8 citations were
selected for our meta-analysis, 6 citations of which included
analysis of C-reactive protein levels and 7 citations for leukocyte
Annals of Emergency Medicine 593
Tests for Sources of Serious Bacterial Infections in Children
Yo et al
Table 2. Summary of the characteristics of the included studies.
Age Range
Study Design
No. of
Participants
Lacour, 2001, Switzerland
7 days to 36 mo
Prospective, observational
124
22.5
PCT, CRP, WBC, IL-6, IL-1Ra
Galetto-Lacour, 2003, Switzerland5
7 days to 36 mo
Prospective, observational
99
29.2
PCT, CRP, WBC, IL-6
Thayyil, 2005, England6
7 days to 36 mo
Prospective, observational
72
11.1
PCT, CRP, WBC
Andreola, 2007, Italy7
7 days to 36 mo
Prospective, observational
408
23.1
PCT, CRP, WBC
Guen, 2007, France8
7 days to 36 mo
Prospective, observational
215
3.2
PCT, CRP, WBC
Maniaci, 2008, USA9
⬍3 mo
Prospective, observational
234
12.8
PCT, WBC
Olaciregui, 2009, Spain10
⬍3 mo
Prospective, observational
347
23.63
PCT-Q, CRP, WBC
Manzano, 2010, Canada11
1 to 36 mo
Randomized controlled trial
384
16.1
PCT, WBC
Author, Year, Country
4
Prevalence, %
Biomarker Tested
PCT, Procalcitonin; CRP, C-reactive protein; IL, interleukin; UTI, urinary tract infection.
Table 3. Summary of subgroup analysis of the included studies by different study characteristics.
Variables
Procalcitonin
Overall analysis4-11
Age ⱕ36 mo4-8,11
Cutoff⫽0.5 ng/mL5-7,10,11
ED setting4,5,7,9-11
CRP
Overall analysis4-8,10
Cutoff⫽40 mg/L4,5,7,8
Age ⱕ36 mo4-8
ED setting4,5,7,10
Leukocyte count
Overall analysis4-8,10,11
Cutoff⫽15,000/mm3(4-6,8,10,11)
Age ⱕ36 mo4-8,11
ED setting4,5,7,10,11
Number of
Studies
Sensitivity (95% CI)
Specificity (95% CI)
AUROC (95% CI)
8
6
5
6
0.83 (0.70–0.91)
0.82 (0.71–0.90)
0.78 (0.68–0.85)
0.84 (0.69–0.93)
0.69 (0.59–0.85)
0.72 (0.63–0.80)
0.72 (0.59–0.82)
0.69 (0.51–0.83)
0.84 (0.80–0.87)
0.84 (0.81–0.87)
0.82 (0.78–0.85)
0.85 (0.81–0.87)
6
4
5
4
0.74 (0.65–0.82)
0.74 (0.58–0.85)
0.74 (0.61–0.84)
0.75 (0.63–0.84)
0.76 (0.70–0.81)
0.76 (0.68–0.82)
0.75 (0.68–0.80)
0.80 (0.76–0.84)
0.81 (0.78–0.84)
0.81 (0.77–0.84)
0.81 (0.77–0.84)
0.84 (0.81–0.87)
7
6
6
5
0.58 (0.49–0.67)
0.61 (0.52–0.70)
0.58 (0.47–0.68)
0.59 (0.51–0.67)
0.73 (0.67–0.77)
0.72 (0.65–0.77)
0.71 (0.65–0.77)
0.76 (0.73–0.78)
0.70 (0.65–0.74)
0.68 (0.64–0.72)
0.70 (0.66–0.74)
0.77 (0.73–0.80)
OR, Odds ratio.
counts. In total, we included 1,883 patients tested for
procalcitonin, 1,265 patients tested for C-reactive protein, and
1,649 patients tested for leukocyte counts. The prevalence of
severe bacterial infection in each group was 340 of 1,883
(18.1%), 248 of 1,265 (19.6%), and 310 of 1,649 (18.8%),
respectively.
We evaluated the quality of included studies using Quality
Assessment of Diagnostic Accuracy Studies criteria. The 2
reviewers (C.-H.Y., P.-S.H.) agreed 86% of the time (14 items);
the 2 items on which the 2 reviewers disagreed were resolved by
a consensus meeting with the 3 coauthors (C.-H.Y., P.-S.H.,
and C.-C.L.). All studies were prospective, enrolled consecutive
outpatients presenting with fever without source, and had an
independent reference examination for severe bacterial infection
594 Annals of Emergency Medicine
outcome (Table 2). Among the possible sources of bias, about
80% of the studies did not indicate whether physicians were
blinded to the index tests when making a final diagnosis of
severe bacterial infection. Several studies did not explicitly
explain withdrawals, define an acceptable delay between tests, or
report uninterpretable results. A small proportion of studies did
not adequately describe reference standards or reference tests.
No evidence of publication bias was found (Table 3).
Characteristics of Study Subjects
Table 2 lists study and population characteristics of all 8
patient populations. Most enrolled patients were aged between
7 days and 36 months (Table 2). Patients were treated
exclusively in the ED setting in 6 of the 8 studies, and 2
Volume , .  : November 
Yo et al
Tests for Sources of Serious Bacterial Infections in Children
Table 2. Continued.
Cutoff, PCT, ng/mL; CRP, mg/L;
WBC, /mm3
PCT⫽0.9, CRP⫽40, WBC⫽15,000
PCT⫽0.5, CRP⫽40, WBC⫽15,000
PCT⫽0.5, CRP⫽50, WBC⫽15,000
PCT⫽0.5; CRP⫽40, 80;
WBC⫽10,000
PCT⫽2; CRP⫽40, 80;
WBC⫽15,000
PCT⫽0.12, WBC: NA
PCT⫽0.5, CRP⫽30, WBC⫽15,000
PCT⫽0.5, WBC⫽15,000
PCT
Sensitivity,
Specificity, %
Patient
Setting
Outcomes
Bacteremia, pyelonephritis lobar pneumonia,
meningitis osteoarthritis
Bacteremia, pyelonephritis lobar pneumonia,
meningitis osteoarthritis, deep abscess
Bacterial pneumonia, bacterial meningitis,
septicemia/occult bacteremia,
pyelonephritis
Bacteremia, pyelonephritis, lobar pneumonia,
bacterial meningitis, bone or joint
infections sepsis
Occult bacteremia
CRP
WBC
Sensitivity,
Sensitivity,
Specificity, % Specificity, %
ED
92.9, 78.1
89.3, 75.0
67.9, 77.1
ED
93.1, 74.3
79.3, 78.6
51.7, 74.3
Pediatric unit
87.5, 50
75.0, 68.8
50.0, 53.1
ED
73.4, 76.4
71.3, 81.2
50.0, 75.4
Pediatric unit
57.1, 84.6
42.9, 64.9
62.5, 65.9
96.7, 25.5
NA
NA
63.4, 87.2
63.4, 84.2
58.5, 78.6
77.4, 64.0
NA
71.0, 75.1
UTI, bacteremia bacterial meningitis bacterial ED
gastroenteritis bacterial pneumonia
UTI, bacteremia, cellulitis, sepsis, bacterial
ED
gastroenteritis, pneumonia
UTI, pneumonia occult bacteremia, bacterial ED
meningitis, neutropenia
ED, emergency department; UTI, urinary tract infection.
Table 3. Continued.
Positive
Likelihood Ratio
Negative
Likelihood Ratio
Diagnostic OR
(95% CI)
I2 (95% CI)
Publication Bias
(Egger’s Test P Value)
2.69 (1.87–3.87)
2.98 (2.20–4.05)
2.75 (1.98–3.81)
2.70 (1.75–4.18)
0.25 (0.15–0.40)
0.25 (0.15–0.41)
0.31 (0.23–0.42)
0.23 (0.13–0.41)
10.6 (6.9–16.0)
9.25 (6.4–13.3)
9.38 (6.81–12.9)
10.1 (7.40–13.8)
30.2 (0–68.9)
47.6 (0–79.25)
30.9 (0–73.5)
48.8 (0–79.7)
.326
.276
.545
.169
3.10 (2.48–3.87)
3.02 (2.06–4.25)
2.90 (2.14–3.94)
3.82 (3.19–4.58)
0.34 (0.25–0.46)
0.35 (0.20–0.61)
0.35 (0.22–0.56)
0.31 (0.21–0.46)
9.83 (7.05–13.7)
9.49 (3.90–23.1)
9.17 (4.31–19.5)
11.1 (7.82–15.7)
45.5 (0–78.4)
66.0 (0.4–88.4)
56.0 (0–83.7)
0 (0–84.69)
.631
.738
.612
.095
2.11 (1.63–2.74)
2.15 (1.59–2.90)
2.09 (1.49–2.95)
2.48 (2.08–2.95)
0.58 (0.46–0.73)
0.54 (0.41–0.72)
0.58 (0.43–0.79)
0.53 (0.44–0.65)
4.26 (3.22–5.63)
5.05 (3.58–7.12)
4.04 (2.34–6.97)
4.53 (3.39–6.05)
41.5 (0–75.39)
33.5 (0–73.26)
58.8 (0–84.7)
41.4 (0–78.39)
.612
.079
.744
.556
included pediatric inpatients. The prevalence of severe bacterial
infection in these 8 studies ranged from 3.2% to 29.2%, with a
median of 18.1%.
Procalcitonin has the best diagnostic accuracy to detect
severe bacterial infection among children with fever without
source, followed by leukocyte count and C-reactive protein.
According to the bivariate model, procalcitonin has greater
sensitivity than C-reactive protein or leukocyte count (overall
0.83 [95% CI 0.70 to 0.91], 0.74 [95% CI 0.65 to 0.82], and
0.58 [95% CI 0.49 to 0.67], respectively), and the 3 markers
have roughly comparable specificity (overall 0.69 [0.59 to 0.85],
0.76 [0.70 to 0.81], and 0.73 [0.67 to 0.77], respectively)
(Table 3). The summary receiver operating characteristic curves
of the 3 markers are presented in Figure 2. For studies
Volume , .  : November 
examining procalcitonin, the study heterogeneity appears to be
more dichotomous than a gradation. Three studies4,5,9 showed
strong sensitivity and the rest showed unacceptably poor
sensitivity (Figure 2A).
Most studies were performed exclusively in the ED
environment. The diagnostic odds ratios for procalcitonin, Creactive protein, and leukocyte count were 10.6 (6.9 to 16.0),
9.83 (7.05 to 13.7), and 4.26 (3.22 to 5.63), respectively
(Figure 3). The positive likelihood ratios for procalcitonin, Creactive protein, and leukocyte count were 2.69 (1.87 to 3.87),
3.10 (2.48 to 3.87), and 2.11 (1.63 to 2.74), respectively. The
negative likelihood ratios for procalcitonin, C-reactive protein,
and leukocyte count were 0.25 (0.15 to 0.40), 0.34 (0.25 to
0.46), and 0.58 (0.46 to 0.73), respectively. The degree of
Annals of Emergency Medicine 595
Tests for Sources of Serious Bacterial Infections in Children
Yo et al
Figure 2. Plot of sensitivity and specificity for studies using procalcitonin (PCT) (A), C-reactive protein (B), or leukocyte
count (C) for the detection of serious bacterial infection (SBI) among children with fever without source (FWS), together
with the summary receiver operating characteristic curve (solid line) and the bivariate summary estimate (solid square),
together with the corresponding 95% confidence ellipse (inner dashed line) and 95% prediction ellipse (outer dotted line).
The symbol size for each study is proportional to the study size. The 95% CIs were determined on the basis of the
assumption that the 2 variables follow a bivariate normal distribution. The 95% prediction region is based on independent
variables, which gives a range of values around which an additional observation of the dependent variable can be expected
to be located.
consistency was calculated by I2. The I2 values for procalcitonin,
C-reactive protein, and leukocyte count were 30.2 (95% CI 0 to
68.9), 45.5 (95% CI 0 to 78.4), and 41.5 (95% CI 0 to 75.39),
respectively (Table 3). We did not observe a substantial degree
596 Annals of Emergency Medicine
of inconsistency in studies included for the meta-analyses for
procalcitonin (I2⫽30.2%), C-reactive protein (I2⫽45.5 %), or
leukocyte count (I2⫽41.5%). We performed subgroup analysis
by restricting studies with similar cutoff value, study settings, or
Volume , .  : November 
Yo et al
Tests for Sources of Serious Bacterial Infections in Children
aged 36 months or younger revealed mildly decreased sensitivity
(from 0.83 to 0.82) and moderate improvement in specificity
(from 0.69 to 0.72). No significant evidence of potential
publication bias was observed by Egger’s test for asymmetry of
the funnel plot (Table 3). Exploratory meta-regression analysis
did not find that any prespecified covariate significantly changed
the effect estimate.
LIMITATIONS
Several limitations must be considered when interpreting the
findings of this meta-analysis. By pooling studies dealing with a
variety of sample types, clinical settings, and study populations,
we may have introduced heterogeneity. Hence, the results of
this meta-analysis are applicable mainly to febrile children with
fever without source who present to the ED. We conducted
multiple comparisons for meta-analysis or subgroup analysis,
but the number of studies was too small to protect from type I
errors. Not all included studies used the same commercial kit to
test for procalcitonin and C-reactive protein, but all were of the
immunometric assay type, so the values obtained were assumed
to be comparable. Given the imperfect sensitivity and specificity
of the procalcitonin test, we do not recommend prescribing
empirical antibiotics simply on the basis of the biomarker test
results. Instead, we recommend developing an algorithm similar
to that developed by Philipp et al for adult pneumonia
patients, which interprets gray-zone procalcitonin results (eg,
0.25 to 0.5 ng/mL) in the context of clinical findings.51 At
present, there is no formal cost-effectiveness analysis based
on US data. However, an analysis from the United Kingdom
shows that procalcitonin incurs an additional cost of £45 to
£125 (US $70 to $200) in the treatment course of pneumonia,
which is small relative to the overall costs of pneumonia
treatment. Moreover, the cost should be weighed against the
potential adverse effects of antibiotics and emerging antibiotics
resistance.52
DISCUSSION
Figure 3. Forest plot of diagnostic odds ratio for studies
using PCT (A), C-reactive protein (B), or leukocyte count (C)
to detect SBI among children with fever without source.
age distribution but did not find significant heterogeneity
improvement (Table 3). We used Galbraith plots to determine
sources of heterogeneity. The Galbraith plot did not reveal any
significant outlying study. The subgroup analysis on patients
Volume , .  : November 
The prevalence of severe bacterial infection in children with
fever without source who are younger than 3 years is
approximately 20%.4-19 Differentiating the majority of patients
who will have a benign course from those who have serious
infections poses a great challenge to front-line clinicians. Missed
diagnoses may cause delayed administration of antibiotics,
potentially having long-term effects on morbidity and mortality.
Procalcitonin is a biomarker that has been shown to
differentiate bacterial from nonbacterial infection and which
correlates well with clinical severity. Several studies have shown
that procalcitonin can be used to detect severe bacterial
infection in children with fever without source.4-13 Our study
was designed to assess the diagnostic accuracy of the
procalcitonin test for detecting severe bacterial infection among
pediatric patients presenting with fever without source and to
compare it with the more conventional C-reactive protein test
and leukocyte count. Our meta-analysis, which included 8
Annals of Emergency Medicine 597
Tests for Sources of Serious Bacterial Infections in Children
studies comprising a total of 1,883 patients, demonstrated the
superior discriminative capability of procalcitonin over
conventional laboratory markers, as revealed by area under the
summary receiver operating characteristic curve data of 0.84
(95% CI 0.80 to 0.87) for procalcitonin, 0.81 (95% CI 0.78 to
0.84) for C-reactive protein, and 0.70 (95% CI 0.65 to 0.74)
for leukocyte count. The diagnostic odds ratio for procalcitonin
(10.6; 95% CI 6.9 to 16.0) was also superior to that of Creactive protein (9.83; 95% CI 7.05 to 13.7) and leukocyte
count (4.26; 95% CI 3.22 to 5.63). The positive likelihood
ratios for procalcitonin, C-reactive protein, and leukocyte count
were 2.69 (95% CI 1.87 to 3.87), 3.10 (95% CI 2.48 to 3.87),
and 2.11 (95% CI 1.63 to 2.74), respectively. The negative
likelihood ratios for procalcitonin, C-reactive protein, and
leukocyte count were 0.25 (95% CI 0.15 to 0.40), 0.34 (95%
CI 0.25 to 0.46), and 0.58 (95% CI 0.46 to 0.73), respectively.
Procalcitonin outperforms C-reactive protein and leukocyte
count in sensitivity rather than specificity, making it a better
rule-out diagnostic tool. In a typical ED setting, where
prevalence of severe bacterial infection is approximately 20% for
febrile children younger than 3 years, the posttest to test
probabilities after a positive test result are therefore 40%, 44%,
and 35% for procalcitonin, C-reactive protein, and leukocyte
count, respectively; those after a negative test are 6%, 8%, and
13%, respectively.
The overall positive likelihood ratio (2.69; 95% CI 1.87 to
3.87) for the procalcitonin test was not sufficiently high to be
used as a reliable rule-in tool for the diagnosis of severe bacterial
infection. For example, in a population with a 20% prevalence
(pretest probability) of severe bacterial infection, a positive
likelihood ratio of 2.69 translates into a positive predictive value
(posttest probability) of 40%. In other words, approximately 2
in 5 patients with positive procalcitonin test results can be
expected to have either clinically or microbiologically confirmed
severe bacterial infection. The diagnostic value of procalcitonin
to rule out severe bacterial infection in children with fever
without source performed as well as its rule-in value. In the
same population with a 20% prevalence of severe bacterial
infection, a negative likelihood ratio of 0.25 translates into a
negative predictive value of 94%. In other words, only 1 in 20
patients with negative procalcitonin results will have either
clinically or microbiologically confirmed severe bacterial
infection. However, given the huge social and medical costs
associated with missed severe bacterial infection diagnoses, we
recommend the procalcitonin test not be used as a stand-alone
test. Several clinical trials show that an algorithm integrating
clinical information and procalcitonin results or repeated
procalcitonin measurements in clinically suspected cases may
further reduce the false-negative rate. Results of studies
examining procalcitonin appears to be dichotomous. Three
studies4,5,9 showed strong sensitivity, whereas the rest showed
unacceptably poor sensitivity (Figure 2A). However, the
Galbraith plot analysis did not show any of the 3 studies to be a
significant outlier. The high sensitivity of the study by Maniaci
598 Annals of Emergency Medicine
Yo et al
et al9 may be due to the use of a low cutoff value (0.12 ng/mL),
whereas the high sensitivity for the other 2 studies4,5 may well
be ascribed to the high prevalence of case patients (22% and
29%, respectively) in the study population.
C-reactive protein is an acute-phase protein released by the
liver in response to systemic inflammation of infectious or
noninfectious cause.47 Likewise, leukocytosis has long been
recognized as a nonspecific marker of systemic infection, tissue
damage, or stress events. In contrast, procalcitonin responds
specifically to systemic infection, particularly bacterial
infection.48 However, we did not find greater specificity for
procalcitonin compared with C-reactive protein. A probable
explanation may be the confounding presence in this special
group of patients of the relatively few having noninfectious
causes for fever without source. Our analysis revealed that
procalcitonin outperforms C-reactive protein and leukocyte
count mainly in sensitivity. We thought this may be explained
by the kinetics of serum levels of procalcitonin and C-reactive
protein. In response to systemic infection, procalcitonin is
rapidly released from all tissues of the body and peaks as early as
12 to 24 hours after onset,49 whereas C-reactive protein level
increases slowly during the first 12 hours, peaking 48 to 72
hours after infection onset.47 Unlike adult patients, febrile
children, especially febrile infants, are usually brought to the ED
within the first few hours of fever onset, which may further
reduce the sensitivity of the C-reactive protein test for this
group of patients. Although it seems the relatively low
sensitivity of serum C-reactive protein for predicting severe
bacterial infection may severely limit its value in clinical
practice, recent studies have shown that C-reactive protein has
an interdependent diagnostic value with the procalcitonin test
and that the combination of the procalcitonin test and urine
dipstick test provides better accuracy than any single test used
alone.12,50
One study included in our meta-analysis used bacteremia as
the only definition of severe bacterial infection,8 but the others
used a broader definition, including bacterial isolation from
sterile body fluids or clinical and radiologic criteria highly
suggestive of invasive bacterial infection.4-7,9-13,42 Therefore,
our study confirms the usefulness of biomarkers in recognizing
all potentially invasive bacterial processes, even in the absence of
bacteremia. Variability in the ages of evaluated patients was
noted in the included studies. A child’s age is important because
neonates are more vulnerable to bacterial pathogens and
different age groups are susceptible to different pathogenic
spectra. This can affect the diagnostic accuracy of the reference
standard test. Most of the studies we analyzed included children
aged 3 to 36 months as one group, although this age grouping is
thought to be arbitrary. Two studies included children younger
than 3 months,9,10 probably because the widely recognized
Rochester and Philadelphia criteria emphasize that febrile
infants younger than 90 days belong to a specific age group for
diagnostic and treatment purposes. Sensitivity analysis and
meta-regression did not reveal significant effect modification by
Volume , .  : November 
Yo et al
different age ranges. Therefore, it was assumed that the
populations of all studies were homogeneous.
In summary, our study found that, compared with
conventional leukocyte counts and C-reactive protein level,
procalcitonin performs better for detecting serious bacterial
infection among children with fever without source.
Considering the poor pooled positive likelihood ratio and
acceptable pooled negative likelihood ratio, procalcitonin is
better for ruling out serious bacterial infection than for ruling it
in. Existing studies do not define how best to combine the
diagnostic value of procalcitonin with other clinical information
to improve overall diagnostic accuracy.
Supervising editor: Gregory J. Moran, MD
Author contributions: C-HY, J-YW, and C-CL were responsible
for study concept and design. C-HY, P-SH, S-HL, S-SC, K-CT,
and C-CL were responsible for acquisition of data. C-HY, P-SH,
S-HL, J-YW, and C-CL were responsible for analysis and
interpretation of data. C-HY and C-CL were responsible for
drafting the article. K-CT and C-CL were responsible for critical
revision of the article for important intellectual content. C-HY,
P-SH, and C-CL were responsible for statistical analysis. C-CL
takes responsibility for the paper as a whole.
Funding and support: By Annals policy, all authors are required
to disclose any and all commercial, financial, and other
relationships in any way related to the subject of this article
as per ICMJE conflict of interest guidelines (see
www.icmje.org). The authors have stated that no such
relationships exist.
Publication dates: Received for publication November 2,
2011. Revisions received March 9, 2012, and May 14, 2012.
Accepted for publication May 17, 2012. Available online
August 22, 2012.
Address for correspondence: Chien-Chang Lee, MD, MSc,
E-mail [email protected].
REFERENCES
1. Baraff LJ. Management of infants and young children with fever
without source. Pediatr Ann. 2008;37:673-679.
2. Baraff LJ. Management of fever without source in infants and
children. Ann Emerg Med. 2000;36:602-614.
3. Kuzmanović S, Roncević N, Stojadinović A. [Fever without a focus
in children 0-36 months of age]. Med Pregl. 2006;59:187-191.
4. Lacour AG, Gervaix A, Zamora SA, et al. Procalcitonin, IL-6, IL-8,
IL-1 receptor antagonist and C-reactive protein as identificators of
serious bacterial infections in children with fever without
localising signs. Eur J Pediatr. 2001;160:95-100.
5. Galetto-Lacour A, Zamora SA, Gervaix A. Bedside procalcitonin
and C-reactive protein tests in children with fever without
localizing signs of infection seen in a referral center. Pediatrics.
2003;112:1054-1060.
6. Thayyil S, Shenoy M, Hamaluba M, et al. Is procalcitonin useful in
early diagnosis of serious bacterial infections in children? Acta
Paediatr. 2005;94:155-158.
7. Andreola B, Bressan S, Callegaro S, et al. Procalcitonin and Creactive protein as diagnostic markers of severe bacterial
infections in febrile infants and children in the emergency
department. Pediatr Infect Dis J. 2007;26:672-677.
Volume , .  : November 
Tests for Sources of Serious Bacterial Infections in Children
8. Guen CG, Delmas C, Launay E, et al. Contribution of procalcitonin
to occult bacteraemia detection in children. Scand J Infect Dis.
2007;39:157-159.
9. Maniaci V, Dauber A, Weiss S, et al. Procalcitonin in young febrile
infants for the detection of serious bacterial infections.
Pediatrics. 2008;122:701-710.
10. Olaciregui I, Hernández U, Muñoz JA, et al. Markers that predict
serious bacterial infection in infants under 3 months of age
presenting with fever of unknown origin. Arch Dis Child. 2009;94:
501-505.
11. Manzano S, Bailey B, Girodias JB, et al. Impact of procalcitonin
on the management of children aged 1 to 36 months presenting
with fever without source: a randomized controlled trial. Am J
Emerg Med. 2010;28:647-653.
12. Galetto-Lacour A, Zamora SA, Andreola B, et al. Validation of a
laboratory risk index score for the identification of severe
bacterial infection in children with fever without source. Arch Dis
Child. 2010;95:968-973.
13. Lacour AG, Zamora SA, Gervaix A. A score identifying serious
bacterial infections in children with fever without source. Pediatr
Infect Dis J. 2008;27:654-656.
14. Kadish HA, Loveridge B, Tobey J, et al. Applying outpatient
protocols in febrile infants 1-28 days of age: can the threshold be
lowered? Clin Pediatr (Phila). 2000;39:81-88.
15. Bleeker SE, Derksen-Lubsen G, Grobbee DE, et al. Validating and
updating a prediction rule for serious bacterial infection in
patients with fever without source. Acta Paediatr. 2007;96:100104.
16. Garra G, Cunningham SJ, Crain EF. Reappraisal of criteria used to
predict serious bacterial illness in febrile infants less than 8
weeks of age. Acad Emerg Med. 2005;12:921-925.
17. Hsiao AL, Chen L, Baker MD. Incidence and predictors of serious
bacterial infections among 57- to 180-day-old infants. Pediatrics.
2006;117:1695-1701.
18. Nademi Z, Clark J, Richards CG, et al. The causes of fever in
children attending hospital in the north of England. J Infect.
2001;43:221-225.
19. Trautner BW, Caviness AC, Gerlacher GR, et al. Prospective
evaluation of the risk of serious bacterial infection in children
who present to the emergency department with hyperpyrexia
(temperature of 106 degrees F or higher). Pediatrics. 2006;118:
34-40.
20. Chancey RJ, Jhaveri R. Fever without localizing signs in children: a
review in the post-Hib and postpneumococcal era. Minerva
Pediatr. 2009;61:489-501.
21. Dubos F, Korczowski B, Aygun DA, et al. Distinguishing between
bacterial and aseptic meningitis in children: European comparison
of two clinical decision rules. Arch Dis Child. 2010;95:963-967.
22. Dubos F, Korczowski B, Aygun DA, et al. Serum procalcitonin level
and other biological markers to distinguish between bacterial and
aseptic meningitis in children: a European multicenter case
cohort study. Arch Pediatr Adolesc Med. 2008;162:1157-1163.
23. Korppi M, Don M, Valent F, et al. The value of clinical features in
differentiating between viral, pneumococcal and atypical bacterial
pneumonia in children. Acta Paediatr. 2008;97:943-947.
24. Dubos F, Moulin F, Raymond J, et al. Distinction between
bacterial and aseptic meningitis in children: refinement of a
clinical decision rule. Arch Pediatr. 2007;14:434-438.
25. Dubos F, Moulin F, Gajdos V, et al. Serum procalcitonin and other
biologic markers to distinguish between bacterial and aseptic
meningitis. J Pediatr. 2006;149:72-76.
26. Liu CF, Cai XX, Xu W. Serum procalcitonin levels in children with
bacterial or viral meningitis. Zhongguo Dang Dai Er Ke Za Zhi.
2006;8:17-20.
Annals of Emergency Medicine 599
Tests for Sources of Serious Bacterial Infections in Children
27. Verboon-Maciolek MA, Thijsen SF, Hemels MA, et al. Inflammatory
mediators for the diagnosis and treatment of sepsis in early
infancy. Pediatr Res. 2006;59:457-461.
28. Korppi M. Non-specific host response markers in the
differentiation between pneumococcal and viral pneumonia: what
is the most accurate combination? Pediatr Int. 2004;46:545-550.
29. Groselj-Grenc M, Ihan A, Pavcnik-Arnol M, et al. Neutrophil and
monocyte CD64 indexes, lipopolysaccharide-binding protein,
procalcitonin and C-reactive protein in sepsis of critically ill
neonates and children. Intensive Care Med. 2009;35:1950-1958.
30. Manzano S, Bailey B, Girodias JB, et al. Comparison of
procalcitonin measurement by a semi-quantitative method and an
ultra-sensitive quantitative method in a pediatric emergency
department. Clin Biochem. 2009;42:1557-1560.
31. Rudensky B, Sirota G, Erlichman M, et al. Neutrophil CD64
expression as a diagnostic marker of bacterial infection in febrile
children presenting to a hospital emergency department. Pediatr
Emerg Care. 2008;24:745-748.
32. Fioretto JR, Martin JG, Kurokawa CS, et al. Interleukin-6 and
procalcitonin in children with sepsis and septic shock. Cytokine.
2008;43:160-164.
33. Lorrot M, Fitoussi F, Faye A, et al. Laboratory studies in pediatric
bone and joint infections. Arch Pediatr. 2007;14(suppl 2):S86-90.
34. Pavcnik-Arnol M, Hojker S, Derganc M. Lipopolysaccharide-binding
protein, lipopolysaccharide, and soluble CD14 in sepsis of
critically ill neonates and children. Intensive Care Med. 2007;33:
1025-1032.
35. Herd D. In children under age three does procalcitonin help
exclude serious bacterial infection in fever without focus? Arch
Dis Child. 2007;92:362-364.
36. Makhoul IR, Yacoub A, Smolkin T, et al. Values of C-reactive
protein, procalcitonin, and Staphylococcus-specific PCR in
neonatal late-onset sepsis. Acta Paediatr. 2006;95:1218-1223.
37. Pavcnik-Arnol M, Hojker S, Derganc M. Lipopolysaccharide-binding
protein in critically ill neonates and children with suspected
infection: comparison with procalcitonin, interleukin-6, and Creactive protein. Intensive Care Med. 2004;30:1454-1460.
38. Laskowska-Klita T, Czerwińska B. Concentration of C-reactive
protein, procalcitonin and alpha-1-antitrypsin in blood of neonates
and infants with signs of inflammation. Med Wieku Rozwoj. 2002;
6:5-11.
39. Hsiao AL, Baker MD. Fever in the new millennium: a review of
recent studies of markers of serious bacterial infection in febrile
children. Curr Opin Pediatr. 2005;17:56-61.
Yo et al
40. Pulliam PN, Attia MW, Cronan KM. C-reactive protein in febrile
children 1 to 36 months of age with clinically undetectable
serious bacterial infection. Pediatrics. 2001;108:1275-1279.
41. Isaacman DJ, Burke BL. Utility of the serum C-reactive protein for
detection of occult bacterial infection in children. Arch Pediatr
Adolesc Med. 2002;156:905-909.
42. Fernández Lopez A, Luaces Cubells C, García García JJ, et al.
Procalcitonin in pediatric emergency departments for the early
diagnosis of invasive bacterial infections in febrile infants: results
of a multicenter study and utility of a rapid qualitative test for this
marker. Pediatr Infect Dis J. 2003;22:895-903.
43. Kourtis AP, Sullivan DT, Sathian U. Practice guidelines for the
management of febrile infants less than 90 days of age at the
ambulatory network of a large pediatric health care system in the
United States: summary of new evidence. Clin Pediatr (Phila).
2004;43:11-16.
44. Higgins J, Green S, eds. Cochrane Handbook for Systematic
Reviews of Interventions. Version 5.0.1. London, England:
Cochrane Collaboration; 2008.
45. Moher D, Liberati A, Tetzlaff J, et al. PRISMA Group. Preferred
Reporting Items for Systematic Reviews and Meta-analyses: the
PRISMA statement. PLoS Med. 2009;6:e1000097.
46. Whiting P, Rutjes AW, Reitsma JB, et al. The development of
QUADAS: a tool for the quality assessment of studies of
diagnostic accuracy included in systematic reviews. BMC Med
Res Methodol. 2003;3:25.
47. Clyne B, Olshaker JS. The C-reactive protein. J Emerg Med. 1999;
17:1019-1025.
48. Simon L, Saint-Louis P, Amre DK, et al. Procalcitonin and Creactive protein as markers of bacterial infection in critically ill
children at onset of systemic inflammatory response syndrome.
Pediatr Crit Care Med. 2008;9:407-413.
49. Van Rossum AM, Wulkan RW, Oudesluys-Murphy AM.
Procalcitonin as an early marker of infection in neonates and
children. Lancet Infect Dis. 2004;4:620-630.
50. Galetto-Lacour A, Gervaix A. Identifying severe bacterial infection
in children with fever without source. Expert Rev Anti Infect Ther.
2010;8:1231-1237.
51. Schuetz P, Batschwaroff M, Dusemund F, et al. Effectiveness of a
procalcitonin algorithm to guide antibiotic therapy in respiratory
tract infections outside of study conditions: a post-study survey.
Eur J Clin Microbiol Infect Dis. 2010;29:269-277.
52. Cleves A, Williams J, Carolan-Rees G. Economic Report:
Procalcitonin to Differentiate Bacterial Lower Respiratory
Infections From Non-bacterial Causes. Centre for Evidence-based
Purchasing, London: NHS; 2010.
Did you know?
Podcasts are available for almost every article in Annals.
Visit http://www.annemergmed.com/content/podcast to find out more.
600 Annals of Emergency Medicine
Volume , .  : November 