Diagnostic Test Accuracy of Serum Procalcitonin.2

S YS T E M AT I C R E V I E W
Diagnostic test accuracy of serum procalcitonin

compared with C-reactive protein for bone and joint
Downloaded from http://journals.lww.com/jbisrir by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX
infection in children and adolescents: a systematic review

1AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8K2+Ya6H515kE= on 08/31/2023
and meta-analysis
Brett Ritchie 1 Kylie Porritt 1 Tania Marin 1 Nicole Williams 2,3
1
JBI, Faculty of Health and Medical Sciences, The University of Adelaide, Adelaide, SA, Australia, 2Centre for Orthopaedic and Trauma Research,
The University of Adelaide, Adelaide, SA, Australia, and 3Department of Orthopaedic Surgery, Women’s and Children’s Hospital, Adelaide, SA,
Australia
ABSTRACT
Objective: The objective of this review was to synthesize the best available evidence for the diagnostic test
accuracy of serum procalcitonin compared with serum C-reactive protein for suspected osteomyelitis and septic
arthritis in hospitalized children and adolescents.
Introduction: Measurement of serum C-reactive protein remains a routine investigation for the diagnosis of
osteoarticular infection in children and adolescents. Measurement of serum procalcitonin has been shown to
outperform C-reactive protein in adults with osteomyelitis and septic arthritis. Before procalcitonin can be
considered as a potential replacement or add-on test in children and adolescents, a systematic review and
meta-analysis targeting this population should be conducted.
Inclusion criteria: Original studies reporting the diagnostic accuracy of procalcitonin and/or C-reactive protein
in children and adolescents between one month and 18 years of age admitted to hospital with suspected
osteoarticular infection were included. Studies must have compared the index test to at least one reference test.
Reference test was defined as positive culture or polymerase chain reaction confirmation of a pathogen from
blood, bone biopsy, or joint fluid aspirate in combination with at least two of the following: i) purulent material
from sterile site, ii) positive radiological findings consistent with osteoarticular infection, and ii) symptoms and signs
consistent with osteomyelitis and/or septic arthritis.
Methods: The JBI methodology for systematic reviews of diagnostic test accuracy was followed. Information was
sourced from four databases (MEDLINE, Embase, Cochrane Central Register of Controlled Trials, and Web of Science)
and four gray literature sources (MedNar, OpenGrey, Google Scholar, and ProQuest Dissertations and Theses). Only
studies published in English were considered. The methodological quality of selected studies was formally evaluated,
sensitivity and specificity data were extracted, and 95% confidence intervals determined. Meta-analysis was
performed to estimate summary points using a bivariate model and to generate a hierarchical summary receiver
operating characteristic (HSROC) curve with global measures of test accuracy performance, such as likelihood ratio
and diagnostic odds ratio. A narrative was provided where meta-analysis was not appropriate.
Results: Eight studies were included in the review. Four of these studies used a common C-reactive protein test
threshold of 20 mg/L. At this threshold, the estimated pooled sensitivity of C-reactive protein was 0.86 (0.68–0.96)
and the pooled specificity was 0.9 (0.83–0.94). Using a hierarchical summary receiver operating characteristic model
from six studies, the diagnostic odds ratio for C-reactive protein was estimated to be 39.4 (14.8–104.9) with a positive
likelihood ratio 5.3 (2.3–11.9) and a negative likelihood ratio 0.1 (0.07–0.2). There were insufficient studies from this
review to statistically evaluate the diagnostic accuracy of procalcitonin.
Correspondence: Brett Ritchie, brett.ritchie@sa.gov.au

KP is an associate editor for JBI Evidence Synthesis, but had no influence on the editorial processing of this manuscript. The other authors declare
no conflicts of interest.
DOI: 10.11124/JBIES-20-00357
JBI Evidence Synthesis ß 2021 JBI 3209
© 2021 JBI. Unauthorized reproduction of this article is prohibited.

SYSTEMATIC REVIEW B. Ritchie et al.
Conclusion: Clinicians should continue to measure serum C-reactive protein as the preferred inflammatory marker
in hospitalized children and adolescents with suspected osteomyelitis or septic arthritis. More evidence is needed
before incorporating procalcitonin routinely into clinicians’ diagnostic test strategy. Improvements with the design,
quality, and reporting of procalcitonin diagnostic test assays in children and adolescents with osteoarticular infection
is needed.
Systematic review registration number: PROSPERO CRD42019140276
Keywords: C-reactive protein; osteoarticular; osteomyelitis; procalcitonin; septic arthritis
JBI Evid Synth 2021; 19(12):3209–3237.
Summary of Findings
Diagnostic test accuracy of serum procalcitonin compared with C-reactive protein for bone and joint infection in
children and adolescents
Bibliography: Ritchie B, Porritt K, Marin T, Williams N. Diagnostic test accuracy of serum procalcitonin compared with
C-reactive protein for bone and joint infection in children and adolescents: a systematic review and meta-analysis.
JBI Evid Synth.2021;19(12):3209-37.
Question: What is the test accuracy performance of measurement of serum C-reactive protein in children and adolescents
admitted to hospital with suspected joint infection?
New test: Serum measurement of C-reactive protein
Cut-off value: 20 mg/L
Reference test: Positive culture or polymerase chain reaction from blood, bone, or aspirate, or positive histopathology in
absence of positive culture.
Pooled sensitivity: 0.86 (95% CI: 0.76 to 0.92)
Pooled specificity: 0.90 (95% CI: 0.83 to 0.94)
Number of results per
1000 patients tested
(95% CI)
Number of Quality of the
Test result Prevalence Prevalence patients Evidence Comments
0.1% 1% (studies) (GRADE)
Typically Typically
seen in 1 seen in 9
patient patients
True
1 (1 to 1) 9 (8 to 9) Wide CIs were reported across some studies
positives
527 (4) with respect to estimates for both sensitivity and
False MODERATE specificity, and this was considered sufficient to
0 (0 to 0) 1 (1 to 2) downgrade by one level.
negatives
True 899 (829 to 891 (822 to Wide CIs were reported across some studies
negatives 939) 931)
527 (4) with respect to estimates for both sensitivity and
False 100 (60 to 99 (59 to MODERATE specificity, and this was considered sufficient to
positives 170) 168) downgrade by one level.
CI: Confidence interval

GRADE Working Group grades of evidence
High certainty: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate certainty: Further research is likely to have an important impact on our confidence in the estimate of effect and may
change the estimate.
Low certainty: Further research is very likely to have an important impact on our confidence in the estimate of effect and is
likely to change the estimate.
Very low certainty: Any estimate of effect is very uncertain.

Introduction or without septic arthritis16; however, no statistical

steomyelitis in children and adolescents is an analysis on the test characteristics of CRP was per-
O inflammation of the bone commonly caused by
an acute bacterial infection that spreads via the
formed from this systematic review.
Procalcitonin (PCT) is another serum inflamma-
blood.1 This infection can result in destruction of tory biomarker that has more recently emerged as
the bone and surrounding tissues. Septic arthritis more specific for diagnosing bacterial infections.17
occurs when the infection enters a joint space or This biomarker has diagnostic test accuracy charac-
the surrounding joint membrane. Infection of the teristics that appear to favor serum measurement of
bone and/or the joint is collectively termed osteo- PCT over CRP in suspected cases of acute osteo-
articular infection.2 articular infections.9,10 This evidence is based on one
The overall prevalence of osteoarticular infection systematic review and meta-analysis performed in
is estimated to be at least 2 per 10,000 children.3 2013 and a more recently conducted meta-analysis
Osteoarticular infections commonly affect children in 2017 using data extracted largely from adults.9,10
younger than five years of age and boys are 1.2 to 3.7 In both of these studies, however, concerns regarding
times more likely to experience infection than girls.4 inconsistencies and limitations may have potentially
In children, infection more often involves the bones biased the outcomes, raising some doubt over their
and joints of the lower extremities.5 The source of findings and recommendations.18 The diagnostic test
the bacteremia leading to infection is not clinically accuracy of PCT versus CRP in both pediatric and
evident in many cases, suggesting that colonization neonatal sepsis, however, does support PCT as the
of the mucosa or skin is likely to be an important preferred inflammatory marker.19,20
portal of entry.6 This is reflected by the spectrum of Although there are many additional factors that
common bacteria that both colonize and cause must be considered before any new test is intro-
osteoarticular infections in children, including duced, establishing the diagnostic test accuracy in
Staphylococcus aureus, Kingella kingae, Streptococ- the population of interest is an essential and prelimi-
cus pneumoniae, and Streptococcus pyogenes.1 The nary step. To date, there has been no meta-analysis
response to infection from these pathogens results in published for children and adolescents that examines
a cascade of local and systemic inflammatory medi- the test performance of serum CRP for suspected
ators that can be detected in the blood.7 osteoarticular infections, and no systematic review
Early and prompt diagnosis of osteoarticular infec- and meta-analysis conducted for serum PCT specifi-
tion as well as effective antibiotic treatment are essen- cally for children and adolescents with bone and
tial to avoid severe and permanent sequela.8 The joint infections. Estimates taken from individual
diagnosis of bone and joint infections can be difficult trials alone may be underpowered with the potential
in children because the presentation and clinical fea- for false-positive and false-negative errors.21 Ana-
tures are similar to other non-infective conditions.9,10 lyzing and summarizing the available data using a
Currently, there is no single, non-invasive, objective statistical meta-analysis approach may increase the
and error-free diagnostic test that can rule in or rule power and the precision of the diagnostic test effect
out osteoarticular infections in children.11 Clinicians estimates for both of these inflammatory markers.
must therefore continue to depend on history, exami- This information is most important when evaluating
nation, radiological imaging, and a panel of labora- the post-test probability of acute osteoarticular
tory investigations to assist them with the initial infection using these biomarkers.
diagnostic decision-making process.11 Before any formal clinical recommendations can
There is sufficient evidence to support the mea- be made regarding the applicability of measuring
surement of serum C-reactive protein (CRP) as a serum PCT in children and adolescents with sus-
non-invasive inflammatory mediator in suspected pected acute osteoarticular infections, further evi-
osteoarticular infections in children.12-14 This bio- dence from a systematic review and meta-analysis,
marker test is now performed routinely to help in the specifically for this population, would seem prudent.
diagnosis of acute bone and joint infections.15 A We therefore aimed to synthesize and analyze the
2012 systematic review of acute pediatric hematog- best available evidence to compare the diagnostic test
enous osteomyelitis concluded that CRP was the accuracy of PCT and CRP in children and adolescents
most useful marker to distinguish osteomyelitis with with suspected osteoarticular infection.

Review question arthritis where the participant had one or both index
What is the diagnostic test accuracy of serum mea- tests and a reference test performed. This included
surement of PCT compared with CRP in patients studies where both sensitivity and specificity were
between one month and 18 years of age who are reported, and/or positive predictive value and/or
admitted to hospital with suspected osteoarticular negative predictive value, and/or receiver operating
infection? characteristic (ROC) curve. There were no limita-
tions on the laboratory test methodology for the
Inclusion criteria index tests employed in the studies.
Participants and setting
This review includes studies of children and adoles- Methods
cents from one month to 18 years of age who are This review was conducted according to a published
hospitalized with suspected acute osteoarticular a priori protocol as outlined in the JBI Manual for
infection. Studies involving immunocompromised Evidence Synthesis and following the PRISMA-DTA
children and adolescents were excluded. Studies check list.18,22,23 The protocol was registered in
involving both adults and children were included PROSPERO (CRD42019140276).
if the outcome data for children and adolescents
could be extracted separately. Search strategy
A comprehensive three-step search strategy was first
Index test performed to locate published and unpublished stud-
The current review considered original studies that ies written in English. The search was first run in July
evaluated, either alone or in combination, the diag- 2019 and updated in January 2021. The initial
nostic accuracy of PCT or CRP for identifying acute limited search (Step 1) was undertaken in MEDLINE
osteoarticular infection. (PubMed) and followed by an analysis of text words
contained in the title and abstracts, and of the index
Reference test terms used to describe the articles. This informed the
This review included studies that compared the development of a comprehensive search strategy that
index test to studies that diagnose acute osteomyeli- was adapted for each additional database searched
tis or septic arthritis as: (Step 2).
a. A positive bacterial culture or polymerase chain Databases and platforms searched in this system-
reaction confirmation of an accepted pathogen atic review included MEDLINE (PubMed), Embase
from blood, biopsy, or aspirate, and/or (Ovid), Cochrane Central Register of Controlled
b. Reported results for two or more of the follow- Trials (Ovid), and Web of Science. Sources of unpub-
ing criteria: lished studies and gray literature searched included
i. purulent material from biopsy or aspirate MedNar, OpenGrey, Google Scholar, and ProQuest
specimen; Dissertations and Theses.
ii. positive radiological findings of osteoartic- In the final step (Step 3), the reference lists of
ular infection; studies selected for critical appraisal were also
iii. signs and symptoms consistent with osteo- screened for any additional eligible studies. The full
myelitis/septic arthritis. search strategies used for each of the final searches,
together with total numbers of references retrieved
Diagnosis of interest are presented in Appendix I.
Studies where the diagnosis of interest was presumed
acute hematogenous osteomyelitis or septic arthritis Study selection
were included in this systematic review. Following the search, all identified citations were
collated and uploaded into EndNote v.X9 (Clarivate
Types of studies Analytics, PA, USA). All duplicates were removed.
This review considered any quantitative study that The titles and abstracts of studies were individually
examined the diagnostic accuracy of serum measure- screened for assessment against the inclusion and
ment of PCT or CRP for osteomyelitis or septic exclusion criteria by author BR. All potentially

relevant studies were retrieved in full and formally Data synthesis

assessed for eligibility against the inclusion and A statistician was consulted to assist with the statis-
exclusion criteria. The reason for exclusion of tical synthesis and analysis of this systematic review.
any study that did not meet the inclusion criteria All statistical data analyses were performed using
at full-text review was documented. The results of STATA v15.0 (Stata Corp LLC, Texas, USA). In
the study selection and inclusion process are pre- addition to sensitivity and specificity data, likelihood
sented in a Preferred Reporting Items for System- ratios, odds ratios, and predictive values with their
atic Reviews and Meta-Analyses (PRISMA) flow respective 95% confidence intervals (CIs) were cal-
diagram.24 culated and tabulated for all included studies. Where
meta-analysis was not possible, a synthesis of find-
Assessment of methodological quality ings was presented in a narrative format.
Citations of selected studies after the full-text review
were imported into the JBI System for the Unified Meta-analysis
Management, Assessment and Review of Infor- A random-effects meta-analysis requires more than
mation (JBI SUMARI; JBI, Adelaide, Australia). three studies for each index test reporting the same
Two independent reviewers (BR and SS) critically outcomes to provide sufficiently diverse data for
appraised each study for methodological quality. statistical convergence in STATA. Where meta-
This was based on 10 ‘‘signaling questions’’ from analysis was performed in this review, sensitivity
the revised Quality Assessment of Diagnostic Studies and specificity results were graphically displayed.
(QUADAS-2) critical appraisal tool for diagnostic For diagnostic tests sharing a common threshold level,
test accuracy.25 The four domains used to assess risk a forest plot using a random-effects model was gener-
of bias included patient selection (questions 1 to 3), ated, and where thresholds varied, a hierarchical
index tests (questions 4 and 5), reference standard summary receiver operating characteristic (HSROC)
(questions 6 and 7), and flow and timing (questions 8 was displayed. Results of all meta-analyses were
to 10). If all signaling questions are answered ‘‘yes,’’ presented with 95% CI.
then the risk of bias is judged as low. If any signaling The magnitude of heterogeneity was reported
question is answered as ‘‘no,’’ then potential for bias qualitatively as depicted graphically from the forest
exists.25 Disagreements between reviewers (BR and plots and by the degree of closeness of the scattered
SS) were resolved through discussion. The results observed study results from the HSROC.
from the critical appraisal process were exported
directly from JBI SUMARI. Assessing certainty in the findings
Studies using a common test threshold were included
Data extraction in a Summary of Findings using the Grading of Rec-
The data related to the diagnostic test accuracy of ommendations, Assessment, Development and Eval-
the index tests were extracted using a standardized uation (GRADE) approach for grading the body of
tool.26 Sensitivity and specificity were chosen as evidence on diagnostic tests. The quality of the evi-
principal discriminative measures of diagnostic dence was evaluated as either high, low, or very low.27
accuracy for both index tests in this review. Data
were extracted from all studies either directly where Results
2 x 2 tables were provided or indirectly using the Study inclusion
raw numbers to construct the contingency table. A total of 3013 articles were identified through the
Index test threshold levels were also extracted from systematic search strategy. After removing dupli-
each included study, and where common thresholds cates, 2152 records were screened by title and
were identified, the outcomes were evaluated abstract, and of these, 105 full-text studies were
statistically. Additional data extracted included assessed for eligibility. From these papers, 97 articles
the country where the study was performed, year were excluded, leaving eight studies in the final
of publication, basic patient demographic charac- analysis. A PRISMA flow diagram of the search
teristics (eg, age range), target and comparator and study selection process is presented in
conditions, clinical setting, study design, and fund- Figure 1. The list of excluded studies and the reasons
ing sources. for exclusion are reported in Appendix II.

Records idenfied through Addional records idenfied through

Idenficaon
database searching other sources

(n=3013) (n=0)
Records aer duplicates removed

(n=2152)
Screening
Records screened Records excluded

(n=2152) (n=2047)
Eligibility
Full-text arcles assessed Full-text arcles excluded,

for eligibility with reasons
(n=105) (n=97)
Review or guideline: 14
Insufficient data: 11
Language other than English: 11
Not diagnosc study: 11
CRP/PCT not included: 8
Included neonate data: 7
Included adult data: 8
Not limited to bone and joint
Included
infecon: 8
Studies appraised by Reference standard not meeng
reviewers to be included in criteria: 5
both quantave analysis Case report: 6
Other: 8
and qualitave synthesis
(n = 8)
CRP, C-reacve protein; PCT, procalcitonin

Figure 1: Search results and study selection and inclusion process24
Methodological quality interpretation of the index test and knowledge of
The methodological quality of the eight publications the results of the reference standard (question 4).30
evaluated is summarized in Table 1. The risk of Five studies specified a pre-test threshold for the
selection bias of patients existed across all included index test (question 5).12,29,31-33 One study included
studies (questions 1 to 3). Only two studies used a two threshold levels for the same index test.30 In all
case-control design (question 2).28,29 One study eight studies, the reference test as described was
commented on blinding of the reference test believed to correctly diagnose acute septic arthritis
result from interpretation of the index tests.30 This or osteomyelitis in children and adolescents (ques-
study provided sufficient information regarding tion 6). There was insufficient information provided

Table 1: Critical appraisal of included diagnostic test accuracy studies

Study Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10
33
Caird et al., 2006 Y Y Y U Y Y U U Y Y
28
Cui et al., 2017 Y N Y U N Y U U N Y
31
Eich et al., 1999 Y Y Y U Y Y U U N Y
32
Faesch et al., 2009 Y Y Y U Y Y U Y N N
30
Greeff, 2012 Y Y Y Y N Y U U N Y
34
Jung et al., 2003 U Y Y U N Y U U N Y
29
Singhal et al., 2011 Y N U U Y Y U U N Y
12
Sultan and Hughes, 2010 Y Y Y U Y Y U U N Y
Y, yes; U, unclear; N, no
Q1: Was a consecutive or random sample of patients enrolled?
Q2: Was a case-control design avoided?
Q3: Did the study avoid inappropriate exclusions?
Q4: Were the index test results interpreted without knowledge of the results of the reference standard?
Q5: If a threshold was used, was it pre-specified?
Q6: Is the reference standard likely to correctly classify the target condition?
Q7: Were reference standard results interpreted without knowledge of results of the index test?
Q8: Was there an appropriate interval between index test and reference standard?
Q9: Did all patients receive the same reference standard?
Q10: Were all patients included in the analysis?
to determine if the reference test results were per- accuracy studies.28,30,32,33 A consecutive cohort-
formed without knowledge of the performance of selected cross-sectional design was used in six of
the index test across all eight studies (question 7). the studies (ie, patient selection was based on suspi-
Only one of the eight included studies recorded the cion of having osteoarticular infection).12,30-33,34 In
time interval between index and reference tests the remaining two studies, a case-control design was
(question 8).32 This review identified only one study used.28,29 Five studies examined septic arthritis
where every patient enrolled received the identical exclusively as the target condition of investiga-
invasive procedure (hip aspiration) as their reference tion.12,29,31,33,34
test (question 9).33 All other studies included one or The remaining three studies included osteomye-
more options for the reference test as described in the litis.28,30,32 Only one study directly compared both
inclusion criteria above. One study excluded 21 CRP and PCT to the reference test.30 Two studies
patients due to lost or insufficient blood samples32; investigated PCT without CRP.28,32 Five included
all remaining patients were included and reported in CRP exclusively as the index test.12,29,31,33,34
the final analysis (question 10). Overall, none of the Index test thresholds reported for CRP included
eight studies included in this review were considered 20 mg/L12,29,31,33 and 10 mg/L,30,34 and for PCT
to be at low risk of bias as judged using the QUA- 0.2 ng/mL,30 0.5 ng/mL,32 and 3.56 ng/mL.28
DAS-2 tool.
Review findings
Characteristics of included studies The sensitivity and specificity of the index tests for
The key characteristics of the eight studies included each individual study are included in Table 2. Vari-
are summarized in Appendix III.12,28-33,34 The sam- ation in diagnostic accuracy was observed between
ple sizes of the eight studies ranged from 33 to 313, the two different index tests and within index tests
with a total of 1520 participants. Patient character- across different thresholds. For CRP, the highest
istics were relatively uniform overall and the clinical sensitivity and specificity values recorded from an
setting was predominantly specialized hospital- individual study were 100% and 85.9%, respec-
based medical services with pediatric expertise. This tively, at a threshold of 20 mg/L.31 In comparison,
review included four prospective diagnostic test for PCT, the highest sensitivity and specificity

JBI Evidence Synthesis
SYSTEMATIC REVIEW
Table 2: Summary of diagnostic test measures of procalcitonin and C-reactive protein for identifying osteoarticular infection in children
and adolescents
Sensitivity Specificity LRþ LR– DOR PPV NPV

Study Index test cut-off TP FN FP TN (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) (95% CI) (95% CI)
Caird et al., CRP > 20 mg/L 29 5 4 10 85.3 71.4 3.0 0.2 14.5 87.9 66.7
200633 (68.9–95.0) (41.9–91.6) (1.3–6.9) (0.0–0.5) (3.4–62.5) (71.8–96.6) (38.4–88.2)
Cui et al., PCT 3.56 ng/mL 71 21 53 122 77.2 69.7 2.5 0.3 7.8 57.3 85.3
201728 (67.2–85.3) (62.3–76.4) (1.9–3.2) (0.2–0.5) (4.3–13.9) (48.1–66.1) (78.4–90.7)
Eich et al., CRP 20 mg/L 8 0 9 55 100 85.9 6.5 0.0 99.3 47.1 100
199931 (63.1–100.0) (75.0–93.4) (3.5–11.9) (0.0–0.9) (5.3–1867) (23.0–72.2) (93.5–100.0)
Faesch et al., PCT > 0.5 ng/mL 2 6 9 282 25.0 96.9 8.08 0.77 10.5 18.2 97.7
200932 (3.19–65.1) (94.2–98.6) (2.0–31.5) (0.5–1.7) (0–53.1) (2.3–51.8) (95.5–99.2)
Greeff, PCT > 0.2 ng/mL 11 1 4 17 91.7 81.0 4.8 0.1 46.8 73.3 94.4
201230 (61.5–99.8) (58.1–94.6) (1.9–11.8) (0.01–0.7) (5.51–475) (44.9–92.2) (72.7–99.9)
CRP > 10 mg/L 12 0 14 5 100 26.3 1.3 0.1 9.5 46.2 100
(73.5–100) (9.2–51.2) (0.9–1.8) (0.0–2.3) (0.5–189) (26.6–66.6) (47.8–100)
Jung et al., CRP > 10 mg/L 24 3 7 90 88.5 92.8 12.3 0.1 98.6 76.7 96.8
200334 (69.8–97.6) (85.7–97.0) (5.9–25.4) (0.0–0.3) (24.6–387) (57.7–90.1) (90.9–99.3)
Singhal et al., CRP > 20 mg/L 25 4 20 262 86.2 92.9 12.2 0.1 81.9 55.6 98.5-
201129 (68.3–96.1) (89.3–95.6) (7.8–19.0) (0.0–0.4) (26.9–247) (40.0–70.4) (96.2–99.6)
Sultan and CRP 20 mg/L 3 2 9 82 60.0 90.1 6.0 0.4 13.7 25.0 97.6
Hughes, 201012 (14.7–94.7) (82.1–95.4) (2.3–15.6) (0.2–1.3) (2.4–78.2) (5.5–57.2) (91.7–99.7)
CI, confidence interval; CRP, C-reactive protein; DOR, diagnostic odds ratio; FN, false negative; FP, false positive; LRþ, positive likelihood ratio; LR–,
negative likelihood ratio; NPV, negative predictive value; PCT, procalcitonin; PPV, positive predictive value; TN, true negative; TP, true positive
ß 2021 JBI
B. Ritchie et al.
3216

StudyId SENSITIVITY (95% CI) StudyId SPECIFICITY (95% CI)

Eich 1999 1.00 [0.63 - 1.00] Eich 1999 0.86 [0.75 - 0.93]
Caird 2006 0.85 [0.69 - 0.95] Caird 2006 0.71 [0.42 - 0.92]
Sultan 2010 0.60 [0.15 - 0.95] Sultan 2010 0.90 [0.82 - 0.95]
Singhal 2011 0.86 [0.68 - 0.96] Singhal 2011 0.93 [0.89 - 0.96]
COMBINED 0.86[0.76 - 0.92] COMBINED 0.90[0.83 - 0.94]
Q = 3.99, df = 3.00, p = 0.26 Q = 9.31, df = 3.00, p = 0.03
I2 = 24.73 [0.00 - 100.00] I2 = 67.79 [33.40 - 100.00]
0.1 1.0 0.4 1.0

SENSITIVITY SPECIFICITY
Figure 2: Forest plots of sensitivity and specificity from studies using a C-reactive protein threshold of 20
mg/L in adolescents and children with septic arthritis
recorded from an individual study were 91.7% and The HSROC curve plot best describes how sensitiv-
81.0%, respectively, at a threshold of 0.2 ng/mL.30 ity and specificity trade off at different positivity
The other calculated measures of diagnostic accu- thresholds values. This was achieved by using only
racy included in Table 2 for comparison between one pair of sensitivity and specificity from each of the
index tests are positive and negative likelihood four studies12,26,28,29 using a CRP threshold of
ratios, odds ratio, and positive and negative predic- 20 mg/L together with one pair each from the two
tive values. STATA uses a continuity correction by additional studies27,32 using a CRP threshold of
adding 0.5 when there is a zero in one or more cells. 10 mg/L (Figure 3).
Using the bivariate model, the first meta-analysis Summary data measures for CRP included a diag-
was performed for CRP studies with a common nostic odds ratio (DOR) of 39.48 (14.85–104.96),
threshold of 20 mg/L. The summary point for pooled positive likelihood ratio (LRþ) of 5.31 (2.35–11.98),
sensitivity with 95% CI was 0.86 (0.76–0.92) and for and negative likelihood ratio (LR–) of 0.13 (0.07–
specificity 0.90 (0.83–0.94; see Figure 2). In these 0.24). A pre-test versus post-test probability curve
four studies, the target condition included septic for osteoarticular infection using CRP is shown in
arthritis only.12,26,28,29 Figure 4.
For the second meta-analysis, hierarchical logistic Because only three studies were included for eval-
regression models were used to display estimates of uating the diagnostic test accuracy of PCT for osteo-
HSROC parameters and simple summary measures. articular infection in children and adolescents,25,27,30

we were unable to perform meta-analysis in STATA

1
to construct either a forest plot or HSROC curve for

PCT. Hence, it was also not possible to statistically
compare the diagnostic accuracy of CRP with PCT.
0.8
In the only pediatric osteoarticular infection

study directly comparing the sensitivity and specific-
ity of PCT at a threshold of 0.2 ng/mL with CRP at
0.6
a threshold of 10 mg/mL, PCT was shown to be

Sensivity
less sensitive (91.7% [61.5–99.8] versus 100%

[73.5–100]) than CRP, but more specific (81%
0.4
[58.1–94.6] versus 26.3% [9.2–51.2]) than CRP.30

From the calculated DOR as a global measure
of diagnostic accuracy in this study, PCT outper-
0.2
formed CRP at 46.8 (5.51–475) to 9.5 (0.5–189),

respectively.30
In the second PCT study that also included both
0
septic arthritis and osteomyelitis cases, a test thresh-

1 0.8 0.6 0.4 0.2 0
Specificity old of 0.5 ng/mL resulted in a sensitivity of 25%
(3.19–65.1) and a specificity of 96.9% (94.2–
Study esmate Summary point
98.6).32 The diagnostic DOR from this study, how-
95% confidence
HSROC curve region ever, was lower at 10.5 (0.0–53.1).
95% predicon In the final study using the highest reported
region threshold for PCT at 3.56 ng/mL in children with
osteomyelitis only, the sensitivity was 77.2 (67.2–
Figure 3: Hierarchical summary receiver operating 85.3) and specificity 69.7% (62.3–76.4) with a
characteristic (HSROC) curve for C-reactive protein DOR of 7.8 (4.3–13.9).28
Heterogeneity
Visual inspection of the forest plots for CRP
(Figure 2) demonstrates overlapping CIs for both
1.0 sensitivity and specificity; while this does not
Posive Test Result
LR+ =5.31 [2.35 - 11.98] exclude heterogeneity, it does infer this may not
Negave Test Result
0.8 LR– =0.13 [0.07 - 0.24] be excessive. The distribution of individual studies
from the HSROC curve, together with the number of
Posterior Probability
0.6 studies lying outside the 95% prediction area plus a

prediction region that is larger than the 95% CI
0.4 region, confirm a degree of heterogeneity (Figure 3).
0.2 Publication bias

Deeks’ funnel plot for examining publication bias for
0.0
0.0 0.2 0.4 0.6 0.8 1.0
meta-analysis of diagnostic test accuracy studies was
Prior Probability
Prevalence Heterogeneity
performed and did not demonstrate statistically sig-
Uniform Prior Distribuon = [0.00 - 1.00]
Uncondional NPV =0.80 [0.72 - 0.87] nificant (P ¼ 0.11) asymmetry indicative of publica-
Uncondional PPV =0.75 [0.68 - 0.83]
LR+, posive likelihood rao; LR–, negave likelihood rao; NPV, negave predicve
tion bias (Figure 5).
value; PPV, posive predicve value
Discussion
Figure 4: Line graph of post-test probability versus
prior probability of osteoarticular infection based This systematic review and meta-analysis is the first
on positive or negative C-reactive protein test published review to formally evaluate the diagnostic
result test accuracy of CRP and PCT in children and

disease among studies, and is exacerbated by the

smaller studies included.35 When compared with the
Deeks' Funnel Plot Asymmetry Test

pvalue = 0.11
5
.1
4
Study two previous published studies using combined adult
Regression
Line and pediatric participants,9,10 the test accuracy esti-
mates using DOR were both lower at 3.56 and

.15
22.25. Furthermore, the likelihood ratio for CRP
1
from all six pediatric studies combined in meta-
1/root(ESS)
3 2
analysis demonstrated a higher LRþ value of 5.3
(2.3–11.9) versus 1.3 (1.1–1.6) and a lower LR–
.2
value of 0.1 (0.07–0.2) versus 0.4 (0.1–1.3) when
compared with adult and pediatric data combined.9
6
Although these results suggest that CRP may per-
.25
form better in the pediatric population if osteoar-
1 10
Diagnosc Odds Rao

100 1000
ticular infection is suspected, further supporting
evidence is required.
Figure 5: Deeks funnel plot for publication bias for By applying precise inclusion criteria with this
C-reactive protein studies systematic review, we aimed to minimize differences
between the studies at the time of statistical analysis.
This approach resulted in a limited number of eligi-
adolescents presenting to hospital with suspected ble studies for evaluating PCT, and we were there-
acute osteoarticular infection. This review evaluated fore unable to perform a meta-analysis. From among
the test performance of PCT in this population for the limited PCT studies included, considerable vari-
whom CRP is an established inflammatory marker in ability was observed across the test performance
clinical practice.13 outcome measures. This makes interpretation of
For children and adolescents presenting with sus- the overall accuracy of this novel test more difficult.
pected septic arthritis only, a common diagnostic test From our comprehensive search strategy, we
threshold of 20 mg/L for CRP was reported in a total included only one study that directly compared
of four of the included studies. At this threshold, the CRP with PCT in children and adolescents.30
bivariate pooled sensitivity (with 95% CI) was 0.86 Although the number of participants was small,
(0.68–0.96) and the corresponding pooled specific- PCT demonstrated superior overall test accuracy
ity was 0.90 (0.83–0.94). This corresponds to a LRþ to CRP. In two other published studies that were
of 8.60 (4.75–15.5) and a LR– of 0.15 (0.09–0.25). excluded from this review but that directly compared
Results from this review compare with the lower PCT with CRP, the diagnostic accuracy of PCT in
reported performance of CRP when using combined bone and joint infection in children was inconsis-
adult and pediatric septic arthritis data, with a tent.36,37 One study reported better performance
sensitivity of 0.45 (0.35–0.55) and specificity of from PCT whereas the other reported lower accu-
0.07 (0.02–0.25).10 The summary point findings racy compared with CRP.
observed in this study provide a more precise esti- From this review, it was surprising to discover
mate of the diagnostic accuracy of CRP as an index that a study design comparing alternative index tests
test for children and adolescents. These results also directly with the reference standard was not rou-
add validity to the use of CRP as an investigational tinely employed; therefore, it was not possible to
test in this setting beyond individual published evaluate the performance of each index test by
studies. comparing their individual HSROC models. To
By combining six studies that included both sus- overcome this limitation, we attempted to compare
pected osteomyelitis and septic arthritis cases, and the results of PCT and CRP studies indirectly with
incorporating CRP at the different reported thresh- the reference standard using network meta-analysis
olds (20 mg/L and 10 mg/L) using an HSROC model, techniques applied to diagnostic test accuracy.38
the overall test accuracy (using DOR) of CRP was However, due to the small number of PCT studies,
39.4 (14.8–104.9). This wide CI is partly reflected this novel approach also proved to be statistically
by the differences of subjects with and without unworkable.

Heterogeneity between studies in this meta-analysis The reference test criteria included options that
was confirmed by the prediction eclipse in the HSROC were not uniformly applied across the studies we
space. Previous published meta-analyses evaluating analyzed. Although more consistency was observed
CRP and PCT have relied on measures such as I2 to with the reference standard for joint infections, few
evaluate heterogeneity. However, because sensitivity patients with an improbable diagnosis were sub-
and specificity are correlated, univariate measures jected to the more invasive reference tests for ethical
such as I2 may not be appropriate for diagnostic and safety reasons. Theoretically, this approach may
test accuracy studies.39 A previously published system- misclassify some patients, but in practice, the clinical
atic review also found considerable variation in progression of the acute condition in patients with
approaches to investigating and reporting on hetero- actual infection would prevent this occurrence. The
geneity in systematic reviews of diagnostic studies.40 risk from this bias is difficult to eliminate.
Clearly this remains an area where more formal Potential bias from patients lost to follow-up was
statistical guidance is needed as well as an improved not considered to be an important factor in this
understanding by both authors and readers. There systematic review.
was an insufficient number of studies to formally
investigate the sources of heterogeneity in this meta- Strengths and limitations
analysis. The strengths of this review included following a
The combined approach of including any child or published protocol without deviation, and reporting
adolescent with bone and/or joint infection was a systematic review and meta-analysis according to
taken to optimize the number of eligible studies, the PRISMA-DTA checklist.23
remain consistent with the literature, and reflect Our search strategy included gray literature sour-
the clinical dilemma that differentiation may not ces and avoided methodological filters shown to miss
always be possible at the time the index test is relevant diagnostic test accuracy studies.41 The uni-
performed. The potential disadvantage of combining form inclusion criteria stated in this review helped
studies with two different sites of infection is that facilitate consistency across eligible studies. Impor-
any difference in the diagnostic test performance tantly, most of these studies did not use a case-
between these sites could not be identified without control design, which is known to overestimate both
formal subgroup analysis. sensitivity and specificity.42
From the critical appraisal of the methodological We identified several limitations with this system-
quality of the included studies, we considered the atic review and meta-analysis. Only studies in
risk of selection bias to be low. We were unable to English were included. The small number and size
establish whether the index tests were interpreted of eligible studies available for analysis precluded
without knowledge of the results of the reference our capacity to address the research question and
standard and vice versa. Fortunately, the index test decreased the precision of our final diagnostic test
results are objective measures performed by scien- parameters using bivariate and HSROC models. The
tific staff using laboratory equipment and therefore studies included were of moderate quality, exposing
are less exposed to risk of bias without blinding. Bias several sources of risk of bias that may impact on the
may occur, however, if the index test threshold was outcomes observed. We did not report on the extent
not pre-specified and the investigator had knowledge of heterogeneity using statistical tests, and relied on
of the reference test outcome. Three of the eight graphical visualization from CIs with forest plots
studies included in our analysis did not pre-specify a and prediction regions from the HSROC curve. The
test threshold.28,30,34 narrative commentary on PCT as an index test in this
The reference standard inclusion criteria applied setting was a subjective assessment. The risk of
in this review appeared to correctly and accurately publication bias in this meta-analysis of diagnostic
diagnose pediatric patients with bone and/or joint test accuracy was evaluated as recommended using
infection. Knowledge of the index test result in this Deeks’ funnel plot for asymmetry.43 Although this
setting would be unlikely to alter the interpretation test was not significant, it is underpowered when
of the reference test classification, as this included using a small number of studies and may have been
objective laboratory microbiological criteria. falsely reassuring.44 The results from our systematic

review and meta-analysis are applicable only to a formally evaluating these index tests. The assessment,
comparable population and clinical setting. measurement, interpretation, and presentation of

heterogeneity in meta-analysis of diagnostic test accu-
Conclusion racy remains complex, and more formal guidance is
This systematic review and meta-analysis provide required to assist researchers in performing these
clinicians and researchers with the best available tasks. This review will permit researchers to update
evidence related to the test accuracy of CRP in the meta-analysis for CRP in this setting and, as
children and adolescents with suspected osteoartic- new publications appear, will help facilitate the first
ular infection. There was an insufficient number of meta-analysis comparison for PCT in children and
scientific studies to facilitate a meta-analysis of the adolescents with suspected osteoarticular infection.
test accuracy of PCT and to compare this statistically
Acknowledgments
with CRP. Clinicians should remain vigilant to the
indiscriminate use of PCT in routine practice with- Soumya Soumya for her contribution as the inde-
out evidence of its accuracy in this setting. pendent second reviewer analyzing the methodolog-
ical quality of the data, and Suzanne Edwards for her
Recommendations for clinical practice statistical advice and input. Natalie Dempster,
The results from this study provide clinicians with research librarian, for her guidance with the search
increased statistical precision regarding the diagnos- strategies used in this systematic review.
tic test accuracy of CRP in children and adolescents
with suspected acute bone and joint infection. This References
1. Gutierrez K. Bone and joint infections in children. Pediatr
test should remain the preferred adjunctive investi-
Clin North Am 2005;52(3):779–94; vi.
gation to help define the post-test probability of
2. Gigante A, Coppa V, Marinelli M, Giampaolini N, Falcioni D,
suspected acute osteoarticular infection (Grade A, Specchia N. Acute osteomyelitis and septic arthritis in
JBI Grades of Recommendation).45 We are unable to children: a systematic review of systematic reviews. Eur
recommend the use of PCT in this specific setting as Rev Med Pharmacol Sci 2019;23(2 Suppl):145–58.
an add-on or replacement test based on a compre- 3. Agarwal A, Aggarwal AN. Bone and joint infections in
hensive review of the literature (Grade A, JBI Grades children: acute hematogenous osteomyelitis. Indian J
of Recommendation).45 Clinicians should be cau- Pediatr 2016;83(8):817–24.
tious when requesting or recommending PCT rou- 4. Faust SN, Clark J, Pallett A, Clarke NMP. Managing bone
tinely in isolation for children and adolescents with and joint infection in children. Arch Dis Child
suspected acute bone and joint infection until we can 2012;97(6):545–53.
5. Saavedra-Lozano J, Falup-Pecurariu O, Faust SN, Girschick H,
make a better informed decision regarding the accu-
Hartwig N, Kaplan S, et al. Bone and joint infections. Pediatr
racy of this test.
Infect Dis J 2017;36(8):788–99.
6. Le Saux N. Diagnosis and management of acute osteo-
Recommendations for research articular infections in children. Paediatr Child Health
This study has identified several important implica- 2018;23(5):336–43.
tions for research, beginning with improving the 7. Lew DP, Waldvogel FA. Osteomyelitis. Lancet 2004;364
design of studies investigating the diagnostic accu- (9431): 369–79.
racy of biomarkers CRP and PCT in children and 8. Wang C, Zhong D, Liao Q, Kong L, Liu A, Xiao H. Procalcitonin
adolescents. More studies using head-to-head or levels in fresh serum and fresh synovial fluid for the
direct comparisons of these index tests with a single differential diagnosis of knee septic arthritis from rheuma-
reference standard are required. Where this toid arthritis, osteoarthritis and gouty arthritis. Exp Ther
Med 2014;8(4):1075–80.
approach is not feasible, the use of network meta-
9. Shen CJ, Wu MS, Lin KH, Lin WL, Chen HC, Wu JY, et al. The
analysis may provide an alternative approach for
use of procalcitonin in the diagnosis of bone and joint
clinicians but requires researchers to become famil- infection: a systemic review and meta-analysis. Eur J Clin
iar with this test methodology. Diagnostic test accu- Microbiol Infect Dis 2013;32(6):807–14.
racy studies should also include a pre-specified 10. Zhao J, Zhang S, Zhang L, Dong X, Li J, Wang Y, et al. Serum
hypothesis that can be statistically tested against a procalcitonin levels as a diagnostic marker for septic arthri-
null hypothesis using sample size calculations when tis: a meta-analysis. Am J Emerg Med 2017;35(8):1166–71.

11. Dodwell ER. Osteomyelitis and septic arthritis in children: meta-analyses: the PRISMA statement. Ann Intern Med
current concepts. Curr Opin Pediatr 2013;25(1):58–63. 2009;151(4):264–9; W64.

12. Sultan J, Hughes PJ. Septic arthritis or transient synovitis of 25. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ,
the hip in children: the value of clinical prediction algo- Reitsma JB, et al. QUADAS-2: a revised tool for the quality
rithms. J Bone Joint Surg Br 2010;92(9):1289–93. assessment of diagnostic accuracy studies. Ann Intern Med
13. Paakkonen M, Kallio MJ, Kallio PE, Peltola H. Sensitivity of 2011;155(8):529–36.

erythrocyte sedimentation rate and C-reactive protein in 26. Campbell JM, Klugar M, Ding S, Carmody DP, Hakonsen SJ,
childhood bone and joint infections. Clin Orthop Relat Res Jadotte YT, et al. Diagnostic test accuracy: methods for
2010;468(3):861–6. systematic review and meta-analysis. Int J Evid Based
14. Kallio MJ, Unkila-Kallio L, Aalto K, Peltola H. Serum C-reac- Healthc 2015;13(3):154–62.
tive protein, erythrocyte sedimentation rate and white 27. Schunemann H, Brozek J, Guyatt G, Oxman A, editors.
blood cell count in septic arthritis of children. Pediatr Infect Handbook for grading the quality of evidence and the
Dis J 1997;16(4):411–3. strength of recommendations using the GRADE approach.
15. Reitzenstein JE, Yamamoto LG, Mavoori H. Similar erythro- Updated October 2013. The Grade Working Group [inter-
cyte sedimentation rate and C-reactive protein sensitivities net]. 2013 [cited 2021 Jan 27]. Available from: https://
at the onset of septic arthritis, osteomyelitis, acute rheu- gdt.gradepro.org/app/handbook/handbook.html.
matic fever. Pediatr Rep 2010;2(1):e10. 28. Cui C, Fu M, Gao B. Procalcitonin and pancreatic stone
16. Dartnell J, Ramachandran M, Katchburian M. Haematoge- protein function as biomarkers in early diagnosis of pediat-
nous acute and subacute paediatric osteomyelitis: a sys- ric acute osteomyelitis. Med Sci Monit 2017;23:5211–7.
tematic review of the literature. J Bone Joint Surg Br 29. Singhal R, Perry DC, Khan FN, Cohen D, Stevenson HL,
2012;94(5):584–95. James LA, et al. The use of CRP within a clinical prediction
17. Samsudin I, Vasikaran SD. Clinical utility and measurement algorithm for the differentiation of septic arthritis and
of procalcitonin. Clin Biochem Rev 2017;38(2):59–68. transient synovitis in children. J Bone Joint Surg Br
18. Ritchie B, Porritt K, Marin T, Williams N. Diagnostic test 2011;93(11):1556–61.
accuracy of serum measurement of procalcitonin and C- 30. Greeff E. Is procalcitonin useful in diagnosing septic arthri-
reactive protein for bone and joint infection in children and tis and osteomyelitis in children? SA Orthop J 2012;11(1):
adolescents: a systematic review protocol. JBI Evid Synth 52–6.
2020;18(3):564–70. 31. Eich GF, Superti-Furga A, Umbricht FS, Willi UV. The painful
19. Yo CH, Hsieh PS, Lee SH, Wu JY, Chang SS, Tasi KC, et al. hip: evaluation of criteria for clinical decision-making. Eur J
Comparison of the test characteristics of procalcitonin to C- Pediatr 1999;158(11):923–8.
reactive protein and leukocytosis for the detection of seri- 32. Faesch S, Cojocaru B, Hennequin C, Pannier S, Glorion C,
ous bacterial infections in children presenting with fever Lacour B, et al. Can procalcitonin measurement help the
without source: a systematic review and meta-analysis. Ann diagnosis of osteomyelitis and septic arthritis? A prospec-
Emerg Med 2012;60(5):591–600. tive trial. Ital J Pediatr 2009;35(1):33.
20. Vouloumanou EK, Plessa E, Karageorgopoulos DE, Manta- 33. Caird MS, Flynn JM, Leung YL, Millman JE, D’Italia JG, Dormans
dakis E, Falagas ME. Serum procalcitonin as a diagnostic JP. Factors distinguishing septic arthritis from transient syno-
marker for neonatal sepsis: a systematic review and meta- vitis of the hip in children. A prospective study. J Bone Joint
analysis. Intensive Care Med 2011;37(5):747–62. Surg Am 2006;88(6):1251–7.
21. Ridgeon EE, Bellomo R, Aberegg SK, Sweeney RM, Varugh- 34. Jung ST, Rowe SM, Moon ES, Song EK, Yoon TR, Seo HY.
ese RS, Landoni G, et al. Effect sizes in ongoing randomized Significance of laboratory and radiologic findings for dif-
controlled critical care trials. Crit Care 2017;21(1):132. ferentiating between septic arthritis and transient synovitis
22. Campbell JM, Klugar M, Ding S, Carmody DP, Hakonsen SJ, of the hip. J Pediatr Orthop 2003;23(3):368–72.
Jadotte YT, et al. Diagnostic test accuracy systematic 35. Fischer JE, Bachmann LM, Jaeschke R. A readers’ guide
reviews. In: Aromatis E, Munn Z, editors. JBI Manual for to the interpretation of diagnostic test properties: clini-
Evidence Synthesis [internet]. Adelaide: JBI, 2000. [cited cal example of sepsis. Intensive Care Med 2003;29(7):
2021 Jan 27]. Available from: https://synthesismanual.jbi.- 1043–51.
global. 36. McMichael BS, Nickel AJ, Christensen EW, Frenn KA, Truong
23. McInnes MDF, Moher D, Thombs BD, McGrath TA, Bossuyt WH, Laine JC, et al. Discriminative accuracy of procalcitonin
PM, et al. the PRISMA DTA Group. Preferred reporting items and traditional biomarkers in pediatric acute musculoskel-
for a systematic review and meta-analysis of diagnostic test etal infection. Pediatr Emerg Care 2020. [Epub ahead of
accuracy studies: the PRISMA-DTA Statement. JAMA print]
2018;319(4):388–96. 37. Butbul-Aviel Y, Koren A, Halevy R, Sakran W. Procalcitonin as
24. Moher D, Liberati A, Tetzlaff J, Altman DG; The PRISMA a diagnostic aid in osteomyelitis and septic arthritis. Pediatr
Group. Preferred reporting items for systematic reviews and Emerg Care 2005;21(12):828–32.

38. White IR. Network meta-analysis. Stata J 2015;15(4): 42. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Group Q-S.
951–85. A systematic review classifies sources of bias and variation

39. McGrath TA, Alabousi M, Skidmore B, Korevaar DA, Bossuyt in diagnostic test accuracy studies. J Clin Epidemiol
PMM, Moher D, et al. Recommendations for reporting of 2013;66(10):1093–104.
systematic reviews and meta-analyses of diagnostic test 43. Deeks JJ, Macaskill P, Irwig L. The performance of tests of
accuracy: a systematic review. Syst Rev 2017;6(1):194. publication bias and other sample size effects in systematic
40. Naaktgeboren CA, Ochodo EA, Van Enst WA, de Groot JAH, reviews of diagnostic test accuracy was assessed. J Clin
Hooft L, Leeflang MMG, et al. Assessing variability in results Epidemiol 2005;58(9):882–93.
in systematic reviews of diagnostic studies. BMC Med Res 44. Sotiriadis A, Papatheodorou SI, Martins WP. Synthesizing
Methodol 2016;16:6. evidence from diagnostic accuracy tests: the SEDATE guide-
41. Whiting P, Westwood M, Beynon R, Burke M, Sterne JA, line. Ultrasound Obstet Gynecol 2016;47(3):386–95.
Glanville J. Inclusion of methodological filters in searches for 45. JBI Grades of Recommendation [internet]. 2014 [cited 2021
diagnostic test accuracy studies misses relevant studies. J Jan 27]. Available from: https://jbi.global/sites/default/files/
Clin Epidemiol 2011;64(6):602–7. 2019-05/JBI-grades-of-recommendation_2014.pdf.

Appendix I: Search strategy

All searches were performed on January 27, 2021.

MEDLINE (PubMed)
Search performed in English.
Total references retrieved: 483
(((((osteomyelitis[MeSH Terms]) OR (osteomyelitis AND infection[MeSH Terms])) OR arthritis, infectious
[MeSH Terms]) OR osteomyelitis[Title/Abstract]) OR (osteoarticular infection[Title/Abstract] OR bone
infection[Title/Abstract] OR joint infection[Title/Abstract])) OR (septic arthritis[Title/Abstract] OR infectious
arthritis[Title/Abstract] OR bacterial arthritis[Title/Abstract]) AND ((((procalcitonin[MeSH Terms]) OR (pro-
calcitonin[Title/Abstract] OR PCT[Title/Abstract])) OR C-reactive protein[MeSH Terms]) OR (C-reactive
protein[Title/Abstract] OR CRP[Title/Abstract])) AND (((Infant or child or adolescent or pediatrics[MeSH
Terms])) OR (infant[Title/Abstract] OR child[Title/Abstract] OR adolescent[Title/Abstract] OR pediatric
[Title/Abstract] OR paediatric[Title/Abstract])) NOT ((animals[MeSH Terms]) NOT humans[MeSH Terms])))
Embase (Ovid)
Search performed in English.
#1 exp Osteomyelitis
#2 exp Osteomyelitis/ and exp Infection
#3 Arthritis, Infectious/
#4 Osteomyelitis.ti,ab,kw.
#5 ((osteoarticular or bone or joint) adj4 infect).ti,ab,kw.
#6 (arthritis adj2 (septic or sepsis or infectious or bacterial)).ti,ab,kw.
#7 or/1–6
#8 Procalcitonin/
#9 (pro?calcitonin or PCT).ti,ab,kw.
#10 C-reactive protein/
#11 (c-reactive protein or CRP).ti,ab,kw.
#12 or/8–11
#13 Infant/ or exp Child/ or Adolescent/ or Pediatrics/
#14 (infant or child or adolescent or juvenileor p?ediatric or girl or boy).af.
#15 13 or 14
#16 7 and 12 and 15
#17 16 not (Animals/ not (Animals/ and Human/)
#18 limit 17 to English language

Cochrane Central Register of Controlled Trials (Ovid)

Limited to English.
#1 exp Osteomyelitis
#2 exp Osteomyelitis/ and exp Infection
#3 Arthritis, Infectious/
#4 Osteomyelitis.ti,ab,kw.
#5 ((osteoarticular or bone or joint) adj4 infect).ti,ab,kw.
#6 (arthritis adj2 (septic or sepsis or infectious or bacterial)).ti,ab,kw.
#7 or/1–6
#8 Procalcitonin/
#9 (pro?calcitonin or PCT).ti,ab,kw.
#10 C-reactive protein/
#11 (c-reactive protein or CRP).ti,ab,kw.
#12 or/8–11
#13 Infant/ or exp Child/ or Adolescent/ or Pediatrics/
#14 (infant or child or adolescent or juvenileor p?ediatric or girl or boy).af.
#15 13 or 14
#16 7 and 12 and 15
#17 16 not (Animals/ not (Animals/ and Human/)
#18 limit 17 to English language
Web of Science
Limited to English.
#1 TS¼(Osteomyelitis OR septic arthritis OR osteoarticular infection OR bone infection OR joint

infection OR infectious arthritis) AND
#2 TS¼(C-reactive protein OR CRP) OR
#3 TS¼(Procalcitonin OR PCT)AND
#4 TS¼(infant OR child OR adolescent OR pediatric)
Search refined by Web of Science category, PEDIATRICS

MedNar

Osteomyelitis OR ‘‘septic arthritis’’ OR ‘‘osteoarticular infection’’ OR ‘‘bone infection’’ OR ‘‘joint
infection’’ OR ‘‘infectious arthritis’’ AND C-reactive protein OR CRP OR Procalcitonin OR PCT AND
Infant OR child OR adolescent OR p?diatric

OpenGrey
Search performed on the first 100 articles.
osteomyelitis OR ‘‘septic arthritis’’ OR ‘‘osteoarticular infection’’ OR ‘‘bone infection’’ OR ‘‘joint infection’’
OR ‘‘infectious arthritis’’ AND ‘‘C-reactive protein’’ OR CRP OR procalcitonin OR PCT AND Infant OR
child OR adolescent OR /p.ediatric/
The search was refined by selecting discipline medicine AND English language.
Google Scholar
Advanced search
#1 ‘‘procalcitonin’’ OR ‘‘C-reactive protein’’ AND ‘‘osteomyelitis’’ OR ‘‘septic arthritis’’
#2 without ‘‘adult’’
#3 selected in the title only.
ProQuest Dissertations and Theses Global

#1 osteomyelitis OR
#2 septic arthritis OR
#3 osteoarticular infection OR
#4 bone infection OR
#5 OR joint infection OR
#6 infectious arthritis AND
#7 C-reactive protein OR
#8 CRP OR
#9 Procalcitonin OR
#10 PCT AND
#11 infant OR
#12 child OR
#13 adolescent OR
#14 pediatric OR
#15 Paediatric

Appendix II: Studies ineligible following full-text review

1. Aigner RM, Fueger GF, Vejda M. Follow-up of osteomyelitis of infants with systemic serum parameters
and bone scintigraphy. Nuklearmedizin 1996;35(4):116–21.
Reason for exclusion: No C-reactive protein (CRP) cut-off in article and no raw data included in analysis.
2. Akinkugbe O, Stewart C, McKenna C. Presentation and investigation of pediatric bone and joint
infections in the pediatric emergency department. Pediatr Emerg Care 2019;35(10):700–4.
Reason for exclusion: No criteria included for the diagnosis of osteoarticular disease. No defined
case definition.
3. Arnold JC, Cannavino CR, Ross MK, Westley B, Miller TC, Riffenburgh RH, et al. Acute bacterial
osteoarticular infections: eight-year analysis of C-reactive protein for oral step-down therapy. Pediatrics
2012;130(4):e821–828.
Reason for exclusion: CRP for oral stepdown and not a diagnostic paper.
4. Aupiais C, Basmaci R, Ilharreborde B, Blachier A, Desmarest M, Job-Deslandre C, et al. Arthritis in
children: comparison of clinical and biological characteristics of septic arthritis and juvenile idiopathic
arthritis. Arch Dis Child 2017;102(4):316–22.
Reason for exclusion: This study does not allow for extraction or calculation of any of the parameters to
determine diagnostic test accuracy.
5. Bacon MA, Bailie HC, Stohr KK. Kocher’s criteria: is it always useful? Arch Dis Child 2016;101 (Suppl 1):
A36-A37.
Reason for exclusion: No CRP raw data included and no definition of acute osteoaricular infection included.
6. Baker ADL, Macnicol MF. Haematogenous osteomyelitis in children: epidemiology, classification,
aetiology and treatment. Paediatr Child Health 2008;18(2):75–84.
Reason for exclusion: Review article only; no raw data included in this paper.
7. Barriocanal MB, Jimenez MR, Amador JTR, Insuga VS, Sanchez AB, Jareno MLL. [Acute osteomyelitis:
epidemiology, clinical manifestations, diagnosis and treatment.] An Pediatr (Barc) 2013;78(6):367–73.
[Spanish]
Reason for exclusion: Language other than English.
8. Basmaci R, Ilharreborde B, Bonacorsi S, Kahil M, Mallet C, Aupiais C, et al. [Septic arthritis in children
with normal initial C-reactive protein: clinical and biological features.] Arch Pediatr 2014;21(11):1195–9.
[French]
9. Benvenuti MA, An TJ, Mignemi ME, Martus JE, Mencio GA, Lovejoy SA, et al. A clinical prediction
algorithm to stratify pediatric musculoskeletal infection by severity. J Pediatr Orthop 2019;39(3):153–7.
Reason for exclusion: This paper includes soft tissue infections and does not discriminate between bone and
joint and soft tissue infections.
10. Bonhoeffer J, Haeberle B, Schaad UB, Heininger U. Diagnosis of acute haematogenous osteomyelitis and
septic arthritis: 20 years experience at the University Children’s Hospital Basel. Swiss Med Wkly
2001;131(39–40):575–81.
Reason for exclusion: Criteria for reference test are insufficient.

11. Brown MD. Test characteristics of C-reactive protein (CRP) for pediatric septic arthritis. J Pediatr
Orthop 2004;24(3):344.
Reason for exclusion: Letter to editor.
12. Bueno Barriocanal M, Ruiz Jimenez M, Ramos Amador JT, Soto Insuga V, Bueno Sanchez A, Lorente
Jareno ML. Acute osteomyelitis: epidemiology, clinical manifestations, diagnosis and treatment. An Pediatr
(Barc) 2013;78(6):367–73. [Spanish]
13. Butbul-Aviel Y, Koren A, Halevy R, Sakran W. Procalcitonin as a diagnostic aid in osteomyelitis and
septic arthritis. Pediatr Emerg Care 2005;21(12):828–32.
Reason for exclusion: Two weeks to 19 years age group and this does not meet inclusion criteria.
14. Calvo C, Nunez E, Camacho M, Clemente D, Fernandez-Cooke E, Alcobendas R, et al. Epidemiology
and management of acute, uncomplicated septic arthritis and osteomyelitis: Spanish Multicenter Study.
Pediatr Infect Dis J 2016;35(12):1288–93.
Reason for exclusion: Sensitvity outcomes only provided without corresponding specificity or other out-
comes of diagnositic test accuracy.
15. Ceroni D, Regusci M, Pazos J, Dayer R, Kaelin A. Acute bone and joint infections in children: how much
attention should be paid to persistent fever during intravenous antibiotic therapy? Rev Chir Orthop
Reparatrice Appar Mot 2003;89(3):250–6.
Reason for exclusion: French article, with only abstract in English.
16. Dartnell J, Ramachandran M, Katchburian M. Haematogenous acute and subacute paediatric osteo-
myelitis: a systematic review of the literature. J Bone Joint Surg Br 2012;94(5):584–95.
Reason for exclusion: Systematic review without raw data for diagnosis test accuracy evaluation.
17. De Boeck H. Osteomyelitis and septic arthritis in children. Acta Orthop Belg 2005;71(5):505.
Reason for exclusion: Summary paper only without raw data for CRP diagnostic accuracy testing.
18. Delaney RA, Lenehan B, O’Sullivan L, McGuinness AJ, Street JT. The limping child: an algorithm to
outrule musculoskeletal sepsis. Ir J Med Sci 2007;176(3):181–7.
Reason for exclusion: No raw data provided and no case definitions of osteoarthritis/septic arthritis
included.
19. Digman JM, Roddy R, Kaczenski A, Godbold S, Abramo T. Evaluation of procalcitonin as a negative
predictor for serious bacterial infection in pediatrics. Ann Emerg Med 2016;68(4 Suppl 1):S93.
Reason for exclusion: This article included any serious infections and it was not possible to extract
osteoarticular infections. Abstract-only format.
20. Dubois-Ferrière V, Belaieff W, Lascombes P, de Coulon G, Ceroni D. Transient synovitis of the hip:
which investigations are truly useful? Swiss Med Wkly 2015;145.
Reason for exclusion: This paper does not include septic arthritis or osteomyelitis.
21. Ernst AA, Weiss SJ, Tracy LA, Weiss NR. Usefulness of CRP and ESR in predicting septic joints. South
Med J 2010;103(6):522–6.
Reason for exclusion: Adult data only provided.

22. Gage MJ, Twomey KD, Sala DA, Maguire KJ, Hanstein R, Hennrikus WL, et al. Identifying
predictive factors of pediatric septic arthritis of the knee in a lyme endemic area. Bull Hosp Jt Dis
2018;76(3):161–4.
Reason for exclusion: Study type does not allow extraction of either sensitivity or specificity data for CRP.
23. Gandini D. Acute septic arthritis of the hip in children in northern Australia. ANZ J Surg
2003;73(3):136–9.
Reason for exclusion: Sensitivity data only is provided.
24. Garcı́a-Arias M, Balsa A, Mola EM. Septic arthritis. Best Pract Res Clin Rheumatol 2011;25(3):407–21.
Reason for exclusion: Summary paper; no diagnostic test accuracy data available.
25. Gibian JT, Daryoush JR, Wollenman CC, Johnson SR, Henry A, Koehler RJ, et al. The heterogeneity of
pediatric knee infections: a retrospective analysis. J Pediatr Orthop 2020;40(6):314–21.
Reason for exclusion: Calculation of sensitivity only, therefore unable to determine diagnostic accuracy.
26. Giordano M, Aulisa AG, Guzzanti V, Careri S, Krzysztofiak A, Toniolo RM. Managing of musculo-
skeletal infections in children. Eur Rev Med Pharmacol Sci 2019;23(2 Suppl):179–86.
Reason for exclusion: No raw data provided in this article.
27. Grummert E, Michael K. Pediatric hip pain: transient synovitis versus septic arthritis. J Diagn Med Sonog
2006;22(3):185–8.
Reason for exclusion: Case presentation only.
28. Hariharan P, Kabrhel C. Sensitivity of erythrocyte sedimentation rate and C-reactive protein for the
exclusion of septic arthritis in emergency department patients. J Emerg Med 2011;40(4):428–31.
Reason for exclusion: Adult patients only presented.
29. Harris JC, Caesar DH, Davison C, Phibbs R, Than MP. How useful are laboratory investigations in the
emergency department evaluation of possible osteomyelitis? Emerg Med Australas 2011;23(3):317–30.
Reason for exclusion: Summary information; no raw data provided.
30. Highton E, Perez MG, Cedillo Villamagua C, Sormani MI, Mussini MS, Isasmendi A, et al. Osteo-
articular infections in a tertiary care children’s hospital: epidemiology and clinical characteristics in
association with bacteremia. Arch Argent Pediatr 2018;116(2):e204-e209. [Spanish]
31. Holloway E, Ruffles T, Godden C. Diagnosis and management in the child with suspected septic arthritis
a 14 year retrospective study and review of the evidence. Arch Dis Child 2011;96:A62.
Reason for exclusion: Abstract without raw data, and case definitions not provided.
32. Howard A, Wilson M. Easily missed? Septic arthritis in children. BMJ (Online) 2010;341(7776):
776–7.
Reason for exclusion: Letter to the editor.
33. Htun S. Phogenic osteomyelitis of long bones in children. Liverpool, England: The University of
Liverpool; 1975. p. 1.
Reason for exclusion: No test accuracy data.

34. Hugle T, Schuetz P, Mueller B, Laifer G, Tyndall A, Regenass S, et al. Serum procalcitonin for
discrimination between septic and non-septic arthritis. Clin Exp Rheumatol 2008;26(3):453–6.
Reason for exclusion: Unable to separate patients 16 to 18 years from general adult cohort.
35. Inusa BPD, Oyewo A, Brokke F, Santhikumaran G, Jogeesvaran KH. Dilemma in differentiating between
acute osteomyelitis and bone infarction in children with sickle cell disease: the role of ultrasound. PLoS One
2013;8(6).
Reason for exclusion: Variable case definitions included without being clearly defined.
36. Jain S, Tittal P, Rohilla N, Sud A, Yadav C, Kanojia RK, et al. Acute septic arthritis revisited: a
prospective study in 93 patients correlating C-reactive protein levels with duration of intravenous antibiotic
therapy, clinical and radiological outcomes. Eur J Orthop Surg Tr 2009;19(7):447–55.
Reason for exclusion: Case definitions of patients included were vague.
37. Jana FN, Ortega CS, Goiano EO. Epidemiological study of osteoarticular infections in children. Acta
Ortop Bras 2018;26(3):201–5.
Reason for exclusion: Includes neonates potentially, and no clarity with definitins of oateoarticular
infections in cases.
38. Journeau P, Wein F, Popkov D, Philippe R, Haumont T, Lascombes P. Hip septic arthritis in children:
assessment of treatment using needle aspiration/irrigation. Orthop Traumatol Surg Res 2011;97(3):308–13.
Reason for exclusion: Not diagnostic testing assay article.
39. Kallio MJ, Unkila-Kallio L, Aalto K, Peltola H. Serum C-reactive protein, erythrocyte sedimentation rate
and white blood cell count in septic arthritis of children. Pediatr Infect Dis J 1997;16(4):411–13.
Reason for exclusion: Sensitivity only provided, and unable to calculate diagnostic test accuracy.
40. Kao HC, Huang YC, Chiu C-H, Chang LY, Lee ZL, Chung PW, et al. Acute hematogenous osteomyelitis
and septic arthritis in children. J Microbiol Immunol Infect 2003;36(4):260–5.
Reason for exclusion: Summary data only included in this review.
41. Karambin MM, Hashemian H. Childhood arthritis: rate of different types. Acta Medica Iranica
2009;47(1):31–4.
Reason for exclusion: Examined only arthritis not a diagnostic test paper to evaluate CRP.
42. Kocher MS, Lee B, Dolan M, Weinberg J, Shulman ST. Pediatric orthopedic infections: early detection
and treatment. Pediatr Ann 2006;35(2):112–22.
Reason for exclusion: Review article with only summary information.
43. Kocher MS, Mandiga R, Zurakowski D, Barnewolt C, Kasser JR. Validation of a clinical prediction rule
for the differentiation between septic arthritis and transient synovitis of the hip in children. J Bone Joint Surg
Am 2004;86(8) 1629–35.
Reason for exclusion: CRP not included in the analysis.
44. Kopchak O, Kostik M. Differential diagnosis of non-bacterial and acute hemotogenous osteomyelitis.
Pediatr Rheumatol 2017;15 (Suppl 2):102.
Reason for exclusion: Not diagnostic test evaluation; summary data only included.

45. Krzysztofiak A, Bozzola E, Lancella L, Marchesi A, Villani A. Osteomyelitis in children. Italian Journal
of Pediatrics Conference: 72th Congress of the Italian Society of Pediatrics Italy 2017;43(1 Suppl).
Reason for exclusion: Summary abstract only; no CRP or procacitonin included.
46. Kunnamo I, Kallio P, Pelkonen P, Hovi T. Clinical signs and laboratory tests in the differential diagnosis
of arthritis in children. Am J Dis Child 1987;141(1):34–0.
Reason for exclusion: Could not exclude neonates in this review.
47. Kuong EE, To M, Chow W, Yuen MH, Choi AKY, Fong CM. Pitfalls in diagnosing septic arthritis in
hong kong children: ten years’ experience. Hong Kong Med J 2012;18(6):482–7.
Reason for exclusion: Only included sensitivity data, therefore unable to calculate diagnostic accuracy
parameters.
48. Lalanda M, Alonso JA. Improving the management of the child with an unexplained limp. Clinical
Governance 2006;11(4):308–15.
Reason for exclusion: Testing algorithim only provided.
49. Lavy CBD. The clinical features and surgical treatment of acute septic arthritis in Malawian children.
University College London (United Kingdom); 2006. p. 154.
Reason for exclusion: No diagnositic accuracy test accuracy data for CRP or procalcitonin.
50. Lazzarini L, Mader JT, Calhoun JH. Osteomyelitis in long bones. J Bone Joint Surg Am 2004;86(10):
2305–18.
Reason for exclusion: Treatment article in summary format, not diagnostic test accuracy paper.
51. Le Saux N. The diagnosis and treatment of acute osteoarticular infections in children. Paediatr Child
Health 2018;23(5):344–52. [French]
52. Lee S, Kim HW, Cho HK, Yun YH, Ryu KH, Kim KH. Clinical presentations and causative organisms in
children and adolescents with osteoarticular infections: a retrospective study. Pediatric Infection and Vaccine
2015;22(3):154–63. [Korean]
53. Levine MJ, McGuire KJ, McGowan KL, Flynn JM. Assessment of the test characteristics of C-reactive
protein for septic arthritis in children. J Pediatr Orthop 2003;23(3):373–7.
Reason for exclusion: Unable to determine upper or lower age limits of patients included.
54. Lew DP, Waldvogel FA. Osteomyelitis. Lancet 2004;364(9431):369–79.
Reason for exclusion: Summary review article, not diagnostic test accuracy data.
55. Li SF, Cassidy C, Chang C, Gharib S, Torres J. Diagnostic utility of laboratory tests in septic arthritis.
Emerg Med J 2007;24(2):75–7.
Reason for exclusion: No CRP or procalcitonin included in this article.
56. Lin Z, Vasudevan A, Tambyah PA. Use of erythrocyte sedimentation rate and C-reactive protein to
predict osteomyelitis recurrence. J Orthop Surg (Hong Kong) 2016;24(1):77–3.
Reason for exclusion: No cases of acute osteomyelitis.

57. Lorrot M, Fitoussi F, Faye A, Mariani P, Job-Deslandre C, Penneçot G, et al. Laboratory studies in
pediatric bone and joint infections. Arch Pediatr 2007;14:S86–90. [French]

58. Macnicol MF. Patterns of musculoskeletal infection in childhood. J Bone Joint Surg Br. 2001;83(1):1–2.
Reason for exclusion: Editorial only.
59. Macnicol MF, Watts AC. Haematogenous osteomyelitis. Surgery (Oxford) 2005;23(1):25–30.
Reason for exclusion: Summary article.
60. Madruga Dias J, Costa MM, Pereira da Silva JA, Viana de Queiroz M. Septic arthritis: patients with or
without isolated infectious agents have similar characteristics. Infection 2014;42(2):385–91.
Reason for exclusion: Adult cohort of patients.
61. Maharajan K, Patro DK, Menon J, Hariharan AP, Parija SC, Poduval M, et al. Serum procalcitonin is a
sensitive and specific marker in the diagnosis of septic arthritis and acute osteomyelitis. J Orthop Surg Res
2013;8:19.
Reason for exclusion: Included aldults and unable to differentiate children.
62. Mahmoudi S, Pourakbari B, Borhani K, Khodabandeh M, Valian SK, Aziz-Ahari A, et al. Acute
osteomyelitis and septic arthritis in children: a referral hospital-based study in Iran. Wiener Medizinische
Wochenschrift 2017;167(11–12):259–63.
Reason for exclusion: Sensitivity only provided and unable to extract or calculate diagnostic accuracy.
63. Mamishi S, Kalantari N, Pourakbari B. Clinical feature and etiology of septic arthritis and osteomyelitis
in children in a 10 year period. Acta Medica Iranica 2007;45(1):58–62.
Reason for exclusion: Review article.
64. Matan AJ, Smith JT. Pediatric septic arthritis. Orthopedics 1997;20(7):630–5.
Reason for exclusion: No CRP or procalcitonin included.
65. McMahon A-M. Bone and joint infections. Paediatrics and Child Health 2011;21(12):546–51.
Reason for exclusion: Summary article only.
66. Mistry RM, Lennon D, Boyle MJ, Chivers K, Frampton C, Nicholson R, et al. Septic arthritis and acute
rheumatic fever in children: the diagnostic value of serological inflammatory markers. J Pediatr Orthop
2015;35(3):318–22.
Reason for exclusion: Not a diagnostic test accuracy study; recommendations only.
67. Mitchell PD, Viswanath A, Obi N, Littlewood A, Latimer M. A prospective study of screening for
musculoskeletal pathology in the child with a limp or pseudoparalysis using erythrocyte sedimentation rate,
C-reactive protein and MRI. J Child Orthop 2018;12(4):398–405.
Reason for exclusion: Included all infections and unable to isolate osteoarticular infections.
68. Monachello Araujo D, Cordero Garcia JM, Garcia Zogby L, Guzman Cruz AJ, Rodado Marina S,
Coronado Poggio M, et al. Osteomyelitis in children: association between analytical parameters and three-
phase bone scan findings. Eur J Nucl Med Mol I 2018;45 (Suppl 1):S350.
Reason for exclusion: Abstract without raw data for analysis.

69. Moro-Lago I, Talavera G, Moraleda L, Gonzalez-Moran G. Clinical presentation and treatment of septic
arthritis in children. Rev Esp Cir Ortop Traumatol 2017;61(3):170–5. [Spanish]

70. Nannery R, Heinz P. Approach to joint pain in children. Paediatr Child Health (United Kingdom)
2018;28(2):43–9.
Reason for exclusion: Summary article.
71. Paakkonen M, Kallio MJ, Kallio PE, Peltola H. Sensitivity of erythrocyte sedimentation rate and C-
reactive protein in childhood bone and joint infections. Clin Orthop Relat Res 2010;468(3):861–6.
Reason for exclusion: Sensitivity data only provided and unable to calculate diagnostic accuracy
test performance.
72. Paakkonen M, Kallio MJ, Kallio PE, Peltola H. C-reactive protein versus erythrocyte sedimentation rate,
white blood cell count and alkaline phosphatase in diagnosing bacteraemia in bone and joint infections. J
Paediatr Child Health 2013;49(3):E189–192.
Reason for exclusion: Focus on children with bacteraemia and testing accuracy, and not primary
osteoarticular infection.
73. Patel L, Michael J, Schroeder L, Sherman AK, Berglund L, Newland JG. Can a septic hip decision rule aid
in the evaluation of suspected pediatric musculoskeletal infections? J Emerg Med 2019;56(3):241–7.
Reason for exclusion: Included all musculskeletal infections not restricted to bone and joint.
74. Peltola H, Vahvanen V, Aalto K. Fever, C-reactive protein, and erythrocyte sedimentation rate in
monitoring recovery from septic arthritis: a preliminary study. J Pediatr Orthop 1984;4(2):170–4.
Reason for exclusion: Markers used to monitor the effectiveness of treatment only.
75. Perry M. Erythrocyte sedimentation rate and C reactive protein in the assessment of suspected bone
infection–are they reliable indices? J R Coll Surg Edinb 1996;41(2):116–18.
Reason for exclusion: Single case report.
76. Reed L, Baskett A, Watkins N. Managing children with acute non-traumatic limp: the utility of clinical
findings, laboratory inflammatory markers and X-rays: paediatric emergency medicine. Emerg Med
Australas 2009;21(2):136–42.
Reason for exclusion: Included neonates; no defined reference test description and no case definitions.
77. Reitzenstein JE, Yamamoto LG, Mavoori H. Similar erythrocyte sedimentation rate and C-reactive
protein sensitivities at the onset of septic arthritis, osteomyelitis, acute rheumatic fever. Pediatr Rep 2010;
2(1):e10.
Reason for exclusion: No criteria provided to determine the reference test standard.
78. Riise ØR, Kirkhus E, Handeland KS, Flatø B, Reiseter T, Cvancarova M, et al. Childhood osteomyelitis-
incidence and differentiation from other acute onset musculoskeletal features in a population-based study.
BMC Pediatr 2008;8:45.
Reason for exclusion: Included all infections and not examining diagnostic test performance.
79. Robinson S, Leonard P. C reactive protein, erythrocyte sedimentation rate, or both, in the diagnosis of
atraumatic paediatric limb pain? Emerg Med J 2012;29(12):969–71.
Reason for exclusion: Gold standard reference test opinion by orthopedic team. Not objective.

80. Rutz E, Spoerri M. Septic arthritis of the paediatric hip - a review of current diagnostic approaches and
therapeutic concepts. Acta Orthop Belg 2013;79(2):123–34.

Reason for exclusion: Diagnostic algorithm.
81. Saavedra-Lozano J, Falup-Pecurariu O, Faust SN, Girschick H, Hartwig N, Kaplan S, et al. Bone and
joint infections. Pediatr Infect Dis J 2017;36(8):788–99.
Reason for exclusion: Guideline.
82. Sanabria-Baez G, Delgado V, Arbo A. Clinical and microbiological features of septic arthritis in children:
experience of one Latin American country. J Infect Dev Ctries 2017;11(12):971–3.
Reason for exclusion: CRP evaluation not included in this paper.
83. Sanchez-Moreno P, Ardanuy-Pizarro AV, Navarro L, Melon M, Falcon-Neyra MD, Camacho-Lovillo
M. Acute osteoarticular infections in children in a tertiary hospital: our experience across 5 years. Ann
Rheum Dis 2015;74:1233.
Reason for exclusion: Not diagnostic test study.
84. Simon L, Gauvin F, Amre DK, Saint-Louis P, Lacroix J. Serum procalcitonin and C-reactive protein levels
as markers of bacterial infection: a systematic review and meta-analysis. Clin Infect Dis 2004;39(2):206–17.
Reason for exclusion: Meta-analysis for all bacterial infections in general and not osteoarticular infections.
85. Smith NSM, Wilson N, Janet GM. Differentiating between jia and septic arthritis: Identification of
important features in history, examination and laboratory investigation. Rheumatology (United Kingdom)
2012;51:viii11.
Reason for exclusion: Abstract; no CRP raw data that can be extracted provided.
86. Talebi-Taher M, Shirani F, Nikanjam N, Shekarabi M. Septic versus inflammatory arthritis: discrimi-
nating the ability of serum inflammatory markers. Rheumatol Int 2013;33(2):319–24.
Reason for exclusion: Adult population only.
87. Unkila-Kallio L, Kallio M, Peltola H. Acute haematogenous osteomyelitis in children in Finland. Finnish
Study Group. Ann Med 1993;25(6):545–9.
Reason for exclusion: Only sensitivity data provided without corresponding specificity. Unable to calculate
diagnostic accuracy.
88. Unkila-Kallio L, Kallio M, Peltola H. The usefulness of C-reactive protein levels in the identification of
concurrent septic arthritis in children who have acute hematogenous osteomyelitis. A comparison with the
usefulness of the erythrocyte sedimentation rate and the white blood-cell count. J Bone Joint Surg Am
1994;76(6):848–53.
Reason for exclusion: Comparing osteomyeltitis and septic arthritis in neonates only.
89. Unkila-Kallio L, Kallio MJ, Eskola J, Peltola H. Serum C-reactive protein, erythrocyte sedimentation
rate, and white blood cell count in acute hematogenous osteomyelitis of children. Pediatrics 1994;93(1):
59–62.
Reason for exclusion: Only sensitivity data provided and unable to calculate the diagnostic accuracy.
90. Wasz-Hockert O, Stenman UH, Kostia J. Monarthritis in children. An etiological, clinical and differential
diagnostical study on 264 children. Ann Paediatr Fenn 1965;11(3):119–33.
Reason for exclusion: No CRP included

91. Welling BD, Haruno LS, Rosenfeld SB. Validating an algorithm to predict adjacent musculoskeletal
infections in pediatric patients with septic arthritis. Clin Orthop Relat Res 2018;476(1):153–9.
Reason for exclusion: CRP raw data or diagnostic test accuracy parameters not provided.
92. Yeo A, Ramachandran M. Acute haematogenous osteomyelitis in children. BMJ 2014;348:g66.

Reason for exclusion: Review article.
93. Zamani A, Raeeskarami SR, Akbrai Asbagh P, Oloomi Yazdi Z, Matloob R, Zamani N, et al. Pediatric
septic arthritis: A 10- year epidemiologic study in Imam Khomeini Hospital Complex. Tehran U Med J
2010;67(10):736–42. [Arabic]
94. Boccuzzi E, Bounseno D, Ferro V, Raucci U, Real A, Piga S, et al. The osteoarticular infection in pediatric
emergency setting: a challenging diagnosis. Pediatr Emerg Care 2020;36(2):e108-e114.
Reason for exclusion: Not diagnostic test accuracy study.
95. McMichael B, Nickel A, Christensen E, Frenn K, Truong W, Laine J, et al. Discriminative accuracy of
procalcitonin and traditional biomarkers in pediatric acute musculoskeletal infection. Pediatr Emerg Care
2020. Online ahead of print.
Reason for exclusion: Included soft tissue infections.
96. Bayram S, Bigili F, Kiral D, Yagci T, Yildrim A, Demirel M. Which inflammatory marker is more reliable
in diagnosing acute septic arthritis in pediatric population. Pediatr Int 2020. Online ahead of print.
Reason for exclusion: Included neonates.
97. Danilov C, Ihle C, Frenandez F, Blummerstock, G, Wirth T, Oliver E. Pseudo paralysis of the shoulder
and increased C-reactive protein are predictive factors for septic arthritis in children superior to other clinical
symptoms a case series of 25 patients. J Child Orthop 2020;14(1):85–90.
Reason for exclusion: Included neonates.

SYSTEMATIC REVIEW
Appendix III: Characteristics of included studies
Participants
% female Clinical Study
Study Country Age range Context setting design Target condition Index test Reference test Funding sources
Caird UK N ¼ 48 Presenting to clinic or Tertiary chil- Prospective, Septic arthritis hip Serum CRP WBC > 50.0 x 109/L No grants or outside
et al., 52.0% emergency department dren’s hospital consecutive, > 20 mg/L from hip aspirate funding received.
200633 7 months to with history suspicious for cohort-selected, culture positive from hip
16.0 years simple septic arthritis: cross-sectional study or blood, or bacteria
having an ultrasound and seen on Gram stain
hip aspiration. from hip aspirate.
All 48 patients under-
went hip aspiration.
Cui et al., China N ¼ 267 Presenting to hospital with Hospital Prospective, Acute Serum PCT Bone biopsy culture. Nil funding
201728 44.4% suspected cases of osteo- case-control-selected, osteomyelitis 3.56 ng/L recorded.
Mean age myelitis (187) and health cross-sectional study No conflict of
12.4 years volunteers (80). interest.
In cases of suspected acute
osteomyelitis.
Based on mean weight and
standard deviation of
groups, >99.7% probability
of no neonates included in
this study.
Eich et al., Switzerland N ¼ 89 Presenting to children’s hos- University Retrospective, Septic arthritis hip Serum CRP Pus aspirated from the Nil recorded.
199931 41.5% pital with acute hip pain. children’s consecutive, 20 mg/L hip joint and/or growth
1 month to hospital cohort-selected, of pathogenic bacteria
12.3 years cross-sectional study from the aspirate.
(18 patients had joint
aspiration).
Faesch France N ¼ 339 Children presenting with Academic Prospective, consecu- Osteomyelitis and Serum PCT Positive bacteriological No competing
et al., 1 month to non-traumatic decreased tertiary care tive, cohort-selected, septic arthritis > 0.5 ng/mL culture from blood, interests.
200932 14.0 years motion of a skeletal hospital cross-sectional study bone aspiration, or joint
segment. fluid.
Greeff, South Africa N ¼ 33 Children younger than 14 Hospital Prospective, consecu- Osteomyelitis and Serum PCT Pus at arthrotomy or No competing
201230 1 month to years of age presenting to tive, cohort-selected, septic arthritis > 0.2 ng/mL from bone and interests.
14.0 years hospital with signs and cross-sectional study Data extracted microscopy and culture
symptoms of osteomyelitis only and serum of tissue.
or septic arthritis. No child CRP > 10 mg/L
ß 2021 JBI
younger than 30 days threshold

(personal correspondence).
Jung Korea N ¼ 124 Children with acute hip University Retrospective, consecu- Septic arthritis Serum CRP Positive culture of hip Nil recorded
B. Ritchie et al.
et al., 1 month to pain. hospital tive, cohort-selected, > 10 mg/L fluid from arthrotomy.
200334 15.0 years cross-sectional cohort
study
3236

SYSTEMATIC REVIEW
(Continued)
Participants
% female Clinical Study
Study Country Age range Context setting design Target condition Index test Reference test Funding sources
Singhal UK N ¼ 311 All children and adolescents Children’s Retrospective, Septic arthritis Serum CRP Positive culture hip Nil recorded.
et al., 30.2% with acute, new-onset hospital consecutive, > 20 mg/L aspirate or abundance
201129 2.4 months to atraumatic limp or hip pain. cohort-selected, WBC on microscopy
15.1 years cross-sectional study (total 42 patients
underwent arthrotomy
hip).
Sultan and UK N ¼ 96 All children and adolescents Teaching Retrospective, Septic arthritis Raised serum Positive culture from hip Nil recorded.
Hughes, 35.4% admitted to hospital with hospital consecutive, CRP 20 mg/L joint/ positive blood
201012 1.0 to 12.0 irritable hip. cohort-selected, culture and numerous
years 91 (94.8%) transient synovi- cross-sectional study WBC on hip microscopy.
tis; 5 (5.2%) septic arthritis.
CRP, C-reactive protein; PCT, procalcitonin; WBC, white blood cell.
ß 2021 JBI
B. Ritchie et al.
3237

Diagnostic Test Accuracy of Serum Procalcitonin.2

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Diagnostic Test Accuracy of Serum Procalcitonin.2

Uploaded by

Copyright:

Available Formats

S YS T E M AT I C R E V I E W

Diagnostic test accuracy of serum procalcitonin

infection in children and adolescents: a systematic review

Correspondence: Brett Ritchie, brett.ritchie@sa.gov.au

JBI Evidence Synthesis ß 2021 JBI 3209

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

CI: Confidence interval

JBI Evidence Synthesis ß 2021 JBI 3210

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

Introduction or without septic arthritis16; however, no statistical

JBI Evidence Synthesis ß 2021 JBI 3211

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

JBI Evidence Synthesis ß 2021 JBI 3212

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

relevant studies were retrieved in full and formally Data synthesis

JBI Evidence Synthesis ß 2021 JBI 3213

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

Records idenﬁed through Addional records idenﬁed through

database searching other sources

Records aer duplicates removed

Records screened Records excluded

Full-text arcles assessed Full-text arcles excluded,

CRP, C-reacve protein; PCT, procalcitonin

JBI Evidence Synthesis ß 2021 JBI 3214

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

Table 1: Critical appraisal of included diagnostic test accuracy studies

JBI Evidence Synthesis ß 2021 JBI 3215

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

JBI Evidence Synthesis

Sensitivity Specificity LRþ LR– DOR PPV NPV

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

StudyId SENSITIVITY (95% CI) StudyId SPECIFICITY (95% CI)

COMBINED 0.86[0.76 - 0.92] COMBINED 0.90[0.83 - 0.94]

Q = 3.99, df = 3.00, p = 0.26 Q = 9.31, df = 3.00, p = 0.03

I2 = 24.73 [0.00 - 100.00] I2 = 67.79 [33.40 - 100.00]

0.1 1.0 0.4 1.0

JBI Evidence Synthesis ß 2021 JBI 3217

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

we were unable to perform meta-analysis in STATA

to construct either a forest plot or HSROC curve for

In the only pediatric osteoarticular infection

a threshold of 10 mg/mL, PCT was shown to be

less sensitive (91.7% [61.5–99.8] versus 100%

[58.1–94.6] versus 26.3% [9.2–51.2]) than CRP.30

formed CRP at 46.8 (5.51–475) to 9.5 (0.5–189),

septic arthritis and osteomyelitis cases, a test thresh-

0.6 studies lying outside the 95% prediction area plus a

0.2 Publication bias

JBI Evidence Synthesis ß 2021 JBI 3218

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

disease among studies, and is exacerbated by the

Deeks' Funnel Plot Asymmetry Test

mates using DOR were both lower at 3.56 and

Diagnosc Odds Rao

JBI Evidence Synthesis ß 2021 JBI 3219

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

JBI Evidence Synthesis ß 2021 JBI 3220

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

comparable population and clinical setting. measurement, interpretation, and presentation of

JBI Evidence Synthesis ß 2021 JBI 3221

© 2021 JBI. Unauthorized reproduction of this article is prohibited.

current concepts. Curr Opin Pediatr 2013;25(1):58–63. 2009;151(4):264–9; W64.