can broad-range 16s ribosomal ribonucleic acid gene polymerase chain reactions improve the diagnosis...

INFECTIOUS DISEASE/ORIGINAL RESEARCH

Can Broad-Range 16S Ribosomal Ribonucleic Acid GenePolymerase Chain Reactions Improve the Diagnosis of Bacterial

Meningitis? A Systematic Review and Meta-analysisLakshmi Srinivasan, MBBS, Jared M. Pisapia, MD, MTR, Samir S. Shah, MD, MSCE, Casey H. Halpern, MD, Mary C. Harris, MD

From the Division of Neonatology, Department of Pediatrics, The Children’s Hospital of Philadelphia, Philadelphia, PA (Srinivasan, Harris); the PerelmanSchool of Medicine at the University of Pennsylvania, Philadelphia, PA (Srinivasan, Pisapia, Halpern, Harris); the Division of Infectious Diseases,

Department of Pediatrics, Cincinnati Children’s Hospital Medical Center and the University of Cincinnati College of Medicine, Cincinnati, OH (Shah); andthe Department of Neurosurgery, the Hospital of the University of Pennsylvania, Philadelphia, PA (Pisapia, Halpern).

Study objective: We aim to evaluate the accuracy of the broad-range 16S polymerase chain reaction test in thediagnosis of bacterial meningitis through a systematic review and meta-analysis.

Methods: We searched MEDLINE, EMBASE, and the Cochrane Controlled Trials Registry, using the MedicalSubject Headings terms “polymerase chain reaction,” “RNA, ribosomal, 16S,” and “bacterial meningitis.” Forour primary analysis, we examined the 16S polymerase chain reaction in culture-proven bacterial meningitis. Inancillary observations, we included studies of culture-negative presumed bacterial meningitis, in which there washigh clinical suspicion for bacterial meningitis despite negative cerebrospinal fluid culture results. We extractedinformation necessary to calculate sensitivity and specificity and used bivariate hierarchic modeling meta-analysis methods to obtain pooled statistics. We also estimated potential sources of error and bias such asbetween-study heterogeneity and publication bias.

Results: Fourteen of 299 studies met inclusion criteria for culture-proven bacterial meningitis; 448 (16.1%) of2,780 subjects had positive cerebrospinal fluid culture results. Pooled analysis demonstrated a sensitivity of92% (95% confidence interval [CI] 75% to 98%), specificity of 94% (95% CI 90% to 97%), positive likelihood ratioof 16.26 (95% CI 9.07 to 29.14), and negative likelihood ratio of 0.09 (95% CI 0.03 to 0.28) for culture-provenbacterial meningitis. The polymerase chain reaction test result was also positive in 30% of cases of culture-negative presumed bacterial meningitis. There was significant heterogeneity between studies.

Conclusion: This meta-analysis supports the role of 16S ribosomal ribonucleic acid polymerase chain reactionas a diagnostic tool in bacterial meningitis. With further refinements in technology, the polymerase chainreaction test has the potential to become a useful adjunct in the diagnosis of bacterial meningitis in theemergency department. [Ann Emerg Med. 2012;60:609-620.]

Please see page 610 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 matterCopyright © 2012 by the American College of Emergency Physicians.http://dx.doi.org/10.1016/j.annemergmed.2012.05.040

pae

iHprswf

SEE EDITORIAL, P. 621.

INTRODUCTIONBackground

Despite advances in intensive care and antibiotic therapy,bacterial meningitis remains associated with considerablemortality.1 Among survivors, morbidities include hearing lossand cognitive and motor deficits.1,2 Early diagnosis of bacterialmeningitis is crucial to optimize therapy and minimize the riskof complications.3 Challenges in verifying the diagnosis ofbacterial meningitis and identifying the causative organism have
led to the investigation of new modalities such as universal s
Volume , . : November

olymerase chain reaction tests, which have the potential for uses rapid diagnostic tests in acute care settings, such as themergency department (ED).

Culture of cerebrospinal fluid, obtained by lumbar puncture,s the criterion standard for the diagnosis of bacterial meningitis.

owever, the lumbar puncture is often deferred in critically illatients. Also, antibiotics administered presumptively mayeduce the yield of cerebrospinal fluid cultures. Studies havehown that sterilization of the cerebrospinal fluid can occurithin 2 to 6 hours of initiation of antibiotic therapy, leading to

alse-negative cerebrospinal fluid culture results.4 Even in
ituations in which lumbar punctures are performed before
Annals of Emergency Medicine 609

http://www.lexisnexis.com/dpartner/process.asp?qs_id=8156

http://www.annemergmed.com

http://annemergmed.com/content/podcast


http://dx.doi.org/10.1016/j.annemergmed.2012.05.040

isad

I

dDbtrpoat(maFttspspheIrsbttekcp

G

sri

MS

scEs

Improving the Diagnosis of Bacterial Meningitis Srinivasan et al

antibiotic administration, 15% to 30% of cerebrospinal fluidculture results may be negative in the face of true meningealdisease.5 Additionally, cerebrospinal fluid Gram stain does notreveal bacteria in up to 40% of cases of bacterial meningitis.5,6

Previous studies have also reported the underdiagnosis ofmeningitis in populations such as premature infants in thenosocomial setting.7 The absence of other laboratory features ofbacterial meningitis such as cerebrospinal fluid pleocytosis andelevated cerebrospinal fluid protein levels in high-risk infants,immunocompromised individuals, and patients withoverwhelming infections adds to difficulties in diagnosis.8-11

The above factors create uncertainty in determining theappropriate choice and duration of antibiotic therapy. Becausethe undertreatment of bacterial meningitis has undesirableconsequences, practitioners frequently continue broad spectrumantibiotics when faced with inconclusive test results. Therefore,there is need for an adjunctive diagnostic tool that is rapid,sensitive, accurate, and unaffected by antibiotic therapy. Such atest would not only reduce diagnostic uncertainty and promptrapid institution of appropriate therapy but also preventunnecessary prolongation of antibiotic use. Both interventionshave the potential to improve patient outcomes.12 Moleculartechniques for diagnosis of specific viral infections, such asenterovirus and herpesvirus polymerase chain reactions, are now

Editor’s Capsule Summary

What is already known about this topicIn suspected meningitis, diagnostic confusion oftenexists about the infectious cause. Polymerase chainreaction for bacterial ribonucleic acid has beeninvestigated as a potentially rapid test to diagnosebacterial meningitis.

What question this study addressedIs cerebrospinal fluid polymerase chain reaction anaccurate test to diagnose bacterial meningitis?

What this study adds to our knowledgeIn this meta-analysis of 14 studies, cerebrospinalfluid polymerase chain reaction appears to be bothsensitive and specific for diagnosis of bacterialmeningitis compared with standard cultures.Reports of positive polymerase chain reaction andnegative culture results suggest that this test may bemore sensitive than cultures in some circumstances.

How is this relevant to clinical practiceCerebrospinal fluid polymerase chain reactiontesting is a promising new technology indevelopment that may aid future emergencydepartment diagnosis of bacterial meningitis.

Food and Drug Administration (FDA)–approved and available d

610 Annals of Emergency Medicine

n some EDs after the demonstration of excellent sensitivity andpecificity.13,14 However, there has not yet been an equivalentdvance in the use of polymerase chain reaction for bacterialetection.

mportanceIn recent years, there has been great interest in the

evelopment of a universal bacterial detection strategy.ifferent approaches have been explored, including broad-based

acterial detection strategies using polymerase chain reactionechniques targeting gene sequences encoding the 16S and 23Segions of bacterial ribosomal ribonucleic acid (RNA), multiplexolymerase chain reaction methods targeting the identificationf a specific set of pathogenic bacteria, and other molecularpproaches such as matrix-assisted laser desorption/ionizationime-of-flight.15 Tests using the 16S ribosomal ribonucleic acidrRNA) gene polymerase chain reaction approach have beenost extensively studied and have the greatest potential for

dvancement toward a standardized protocol that could reachDA approval and clinical applicability. These tests are based onhe rationale that regions of the bacterial genome that encodehe 16S ribosomal unit are conserved across all bacterialpecies.16 Polymerase chain reaction tests developed withrimers and probes directed toward identification of geneequences encoding 16S ribosomal RNA can identify theresence of DNA from most bacterial species.17,18 These testsave the advantages of amplifying minute amounts of DNA,ven from nonviable bacteria, with results available in hours.19

n addition, species-specific sequence variability within the 16Segion could allow identification of pathogenic bacteria withequencing technology or species-specific primers and probes foracteria of interest.20-23 With continued refinements inechnology, the polymerase chain reaction test has the potentialo become an important adjunctive diagnostic tool in EDvaluation for bacterial meningitis. However, several gaps in ournowledge must be addressed before the bacterial polymerasehain reaction test can be successfully integrated into clinicalractice.

oals of This InvestigationWe performed a systematic review and meta-analysis of

tudies using broad-range 16S rRNA gene polymerase chaineaction on cerebrospinal fluid samples to determine its accuracyn diagnosing bacterial meningitis.

ATERIALS AND METHODStudy Design

We conducted a systematic review and meta-analysis oftudies using broad-range 16S polymerase chain reaction inerebrospinal fluid for diagnosis of bacterial meningitis (Table1, available online at http://www.annemergmed.com). We

elected studies according to quality and availability of reported
ata (as detailed below).


rrpps

O

feiaccsmspmcramncpsawsw

cpbTpcictc(

R

i(ioabi

Srinivasan et al Improving the Diagnosis of Bacterial Meningitis

SettingTwo independent reviewers (L.S. and J.M.P.) searched

MEDLINE, EMBASE, and the Cochrane Controlled TrialsRegistry, using the Medical Subject Headings terms“polymerase chain reaction,” “RNA, ribosomal, 16S,” and“bacterial meningitis.” The search was limited to studiesinvolving humans, in any language. Studies that evaluated the16S polymerase chain reaction as a diagnostic tool amongsubjects of all ages and subjects exposed to antibiotic therapywere included. The bibliographies of review articles were alsoexamined.

The 2 reviewers examined the articles retrieved from theinitial search and included only studies using the 16Spolymerase chain reaction to diagnose meningitis caused by anybacterial species. A quality assessment of diagnostic accuracystudies (QUADAS) checklist was created to ensure sufficientquality for inclusion in the meta-analysis (Table E2, availableonline at http://www.annemergmed.com). Studies using non-16S-based polymerase chain reaction tests and organism-specificpolymerase chain reaction tests were excluded. Case reports,editorials, letters, commentaries, and review articles were alsoexcluded.

For the purpose of this meta-analysis, “proven” bacterialmeningitis was defined by the growth of bacteria fromcerebrospinal fluid culture. Culture-negative presumed bacterialmeningitis was defined by individual authors according toclinical examination and ancillary laboratory tests in thepresence of negative cerebrospinal fluid culture results; theprecise definition varied between studies.

Methods of MeasurementThe 2 independent reviewers (L.S. and J.M.P.) assessed the

short-listed articles to verify that they met the inclusion criteria,were of sufficient methodological quality, and reported adequateinformation to be included in the analysis. We determined theadequacy of information according to the presence of keycomponents of the QUADAS checklist, such as sufficient detailabout the spectrum of participants, the index and referencestandard tests, and our ability to independently verifysensitivity, specificity, and likelihood ratios from the dataprovided (see Table E2, available online at http://www.annemergmed.com, for details of the QUADAS checklist). Incases of disagreement about the eligibility of a study forinclusion, a third reviewer (M.C.H.) arbitrated a consensusdecision. Interrater reliability was assessed with the � statistic.

Data Collection and ProcessingL.S. and J.M.P. extracted data from each study to perform

independent calculations of tests of diagnostic accuracy. Datacollected included sample size, demographic information, detailsof diagnosis, methods and results of polymerase chain reactionand bacterial identification, and turnaround times, when
available. In the case of studies using real-time polymerase chain p

eaction, we accepted the definitions of “positive polymerase chaineaction” provided by the individual studies according to theirredetermined cycle threshold limits. Although all included studiesrovided polymerase chain reaction results, not all identified thepecific bacteria detected by polymerase chain reaction.

utcome Measures and Primary Data AnalysisTwo-by-two tables (including true positives, true negatives,

alse positives, and false negatives) were created with raw dataxtracted from each study. If the sequencing or culture resultsdentified bacteria considered contaminants by individual studyuthors, those samples were excluded from the analysis. Wealculated sensitivity, specificity, and likelihood ratios,omparing polymerase chain reaction results with the criteriontandard of “proven” bacterial meningitis. Bivariate hierarchicodeling meta-analysis methods were used to compute pooled

tatistics (using the METANDI command in Stata). Thisrovides estimates based on both the bivariate random-effectsodel and the hierarchic summary receiver operating

haracteristic (sROC) model.24 The diagnostic odds ratio (theatio between the positive and negative likelihood ratios) was useds a measure of accuracy, providing a combined estimate of pairedeasures such as sensitivity and specificity, and positive and

egative likelihood ratios.25 The Cochran Q and I2 values werealculated to estimate heterogeneity between studies.26 Weerformed a linear regression of the log odds ratios on the inversequare root of the effective sample sizes to assess for funnel plotsymmetry and, hence, publication bias (Deeks test).27 Forest plotsere created to provide a graphic representation of the pooled

ensitivity and specificity of individual studies in relation to theeighted pooled overall measure.

We also provide data about the ability of the polymerasehain reaction to detect bacteria in cases of culture-negativeresumed bacterial meningitis in which the clinical suspicion foracterial meningitis was high despite negative culture results.his situation is analogous to the clinical scenario in whichatients receive antibiotics before the lumbar puncture, makingulture results less reliable. For this analysis, we included studiesn the main meta-analysis, as well as 4 additional studies of onlyulture-negative meningitis, if they provided adequate data onhe performance of the polymerase chain reaction in thisondition. All statistical analyses were performed with Stataversion 11.1; StataCorp, College Station, TX).26

ESULTSTwo hundred ninety-nine studies were identified by the

nitial search of MEDLINE, EMBASE, and Cochrane databasesFigure 1). On further assessment of these studies by the 2ndependent reviewers, 87 publications were excluded becausef the type of information provided (eg, case reports, reviewrticles). Of the remaining publications, 194 were excludedecause they did not conform to the definitions established fornclusion in the meta-analysis (eg, studies evaluating non-16S
olymerase chain reaction techniques or that targeted




caQ

C

ssfom

dies


identification of only a select few bacterial species). Two of thesestudies used a 16S component in their polymerase chainreaction testing but did not report adequate information tocalculate sensitivity and specificity for that portionseparately.28,29 Four studies were excluded from the mainanalysis because they focused on assessment of the polymerasechain reaction test in culture-negative presumed bacterialmeningitis but were included in the ancillary analysis of thelatter condition.30-33 The remaining 14 studies were found tomeet inclusion criteria. All 14 articles were assessed for qualityof reporting and included according to consensus between the

Pubmed, EMBASE and Cochrane databa"polymerase chain reac�

299 studies iden�

Screening of abstracts by

Studies excluded du

Case reports/seri

Review ar�cles:

Studies excluded due

Non PCR studies: 13

Species-directed diagn

Ot

(Study in non-

(Cannot extract results o

14 studies assessed, agreed uponan

Figure 1. Steps in selection of stu

reviewers, with 93% agreement (� statistic 0.818; 95% a


onfidence interval [CI] 0.787 to 0.866). (See Table E2,vailable online at http://www.annemergmed.com, forUADAS checklist.)

haracteristics of Study SubjectsTable 1 provides a chronologic listing of included studies,

howing sample size, number of positive samples, andensitivity, specificity, and likelihood ratios, calculated with datarom the individual 2�2 tables.20-23,34-43 Overall, 448 (16.1%)f 2,780 included subjects met criteria for proven bacterialeningitis. The study subjects were neonates, children, and

arched using MeSH terms "16S rRNA", "bacterial meningi�s"

on ini�al search

independent reviewers

study type (n=87):

3 Editorials: 2

Le�ers: 2

ethodology (n=194):

Non bacterial PCR: 11

(16S or non 16S): 166

4

cal se�ng:2)

component of PCR: 2)

reviewers for inclusion in meta-is

Studies examining 16S PCR in “culture-nega�ve meningi�s”: 4

to be included in meta-analysis.

se seon,"

fied

two

e to

es: 7

10

to m

osis

her:

clini

f 16S

byalys

dults (overall age range 0 to 91 years), but information about



odc

M

s(bpr(rapem1


the exact number of subjects in each age group was notextractable. Among the studies, there was a large variation insample size and number of patients with positive “criterionstandard” cerebrospinal fluid findings. A wide range ofsensitivities for the 16S polymerase chain reaction test wasreported (47% to 100%), whereas the specificity values weremore narrowly clustered (77% to 100%) (Table 1). Additionaldetails of the individual studies, including description of theextraction process, polymerase chain reaction technique, primersused, and details about bacterial species identification, areprovided in Table 2. There were several methodologicaldifferences reported. First, a variety of extraction processes andprimers specific to the 16S region of the bacterial genome wasused. In addition, polymerase chain reaction techniques variedbetween conventional polymerase chain reaction followed byagarose gel detection, and real-time technology. Although notuniversally used, bacterial identification was performed withsequencing technology in a number of studies.22,34,37,39-42

Table 1. Sensitivity, specificity, and likelihood ratios of 16S pol

StudySample

Size

PositiveCultureResult

ProbableContaminants TP FP FN TN

Radstrom,199420

304 140 15 112 6 28 15

Dicuonzo,199921

77 15 None listed 15 14 0 4

Lorino,200022

90 14 2 14 15 0 6

Lu, 200036 150 13 None listed 12 3 1 13Saravolatz,

20033874 17 None listed 17 1 0 5

Schuurman,200439

227 37 9 24 6 13 18

Poppert,200537

151 38 2 38 5 0 10

Xu, 200541 421 8 Numerous(exactnumber notprovided)

8 133 0 28

Deutch,200642

196 14 None listed 11 3 3 17

Deutch,200734

350 34 3 16 21 18 29

Welinder-Olsson,200740

345 39 7 25 26 14 28

Chakrabarti,200923

267 53 None listed 47 5 11 20

Duan,200935

20 4 None listed 4 3 0 1

Rothman,201043

108 22 4 16 0 6 8

Total 2,780 448 359 241 94 2,08

TP, True positive; FP, false positive; FN, false negative; TN, true negative; LR�, p*“Proven” bacterial meningitis was defined by the growth of bacteria from cerebro

Finally, most reported the detection limits (analytical sensitivity) m


f their assay by performing experiments of spiking and serialilution, with results varying from 3.25 to 1,000 or greaterolony forming units/mL (cfu/mL).

ain ResultsForest plots were created by pooling the sensitivity and

pecificity from individual studies, weighted by sample sizeFigure 2). The pooled sensitivity of the 14 studies was found toe 92% (95% CI 75% to 98%) (Figure 2, left), whereas theooled specificity was 94% (95% CI 90% to 97%) (Figure 2,ight). The analysis also revealed a high positive likelihood ratio16.26; 95% CI 9.07 to 29.14) and a low negative likelihoodatio (0.09; 95% CI 0.03 to 0.28). These results imply thatlthough a positive test result is highly predictive, a negativeolymerase chain reaction test result also performs well atxcluding a relatively uncommon disease such as bacterialeningitis. The diagnostic odds ratio was calculated to be

87.57 (95% CI 55.57 to 633.07), suggesting that the odds of

ase chain reaction in “proven” meningitis.*

nsitivity, %(95% CI)

Specificity,%(95% CI) LR� (95% CI) LR� (95% CI)

0 (72–86) 96 (92–99) 21.9 (9.9–48.2) 0.21 (0.15–0.29)

0 (78–100) 77 (65–87) 4.2 (2.7–6.7) 0.04 (0–0.62)

0 (77–100) 80 (70–89) 4.8 (3.1–7.6) 0.04 (0–0.64)

2 (64–100) 98 (94–100) 42.2 (13.6–130.5) 0.08 (0.01–0.52)0 (80–100) 98 (91–100) 37.6 (7.7–182.8) 0.03 (0–0.44)

5 (47–80) 97 (93–99) 20.5 (9–46.8) 0.36 (0.23–0.56)

0 (91–100) 96 (90–99) 20.5 (9–46.3) 0.01 (0–0.21)

0 (63–100) 68 (63–72) 2.9 (2.4–3.6) 0.08 (0.01–1.21)

9 (49–95) 98 (95–100) 47.7 (15–151.3) 0.22 (0.08–0.59)

7 (30–65) 93 (90–96) 7.1 (4.1–12.2) 0.57 (0.41–0.78)

4 (47–79) 92 (88–94) 7.5 (4.9–11.7) 0.39 (0.26–0.6)

1 (69–90) 98 (95–99) 33.9 (14.1–81.3) 0.19 (0.11–0.33)

0 (40–100) 81 (54–96) 4.4 (1.6–11.6) 0.13 (0.01–1.77)

3 (50–89) 100 (96–100) 124.8 (7.8–2,003.5) 0.28 (0.15–0.55)

likelihood ratio; LR–, negative likelihood ratio.l fluid culture.

ymer

Se

8 8

8 10

1 10

4 96 10

4 6

8 10

0 10

9 7

5 4

0 6

4 8

3 10

6 7

6

ositivespina

eningitis, given a positive test result, is close to 200 times that



Table 2. Methodological details of studies included in meta-analysis.

Study Age Range PCR Details Bacterial Identification Detection Limit, cfu/mL

Radstrom,199420

N/A* Extraction: boiling2-step seminested PCRStep 1: universal 16S primers u3, ru8Amplified segment: 990 bpStep 2: 4 species-specific primersAgarose gel electrophoresis

Yes; 4 species identified byspecies-specific primers (step2)†(‡§�¶#**††‡‡)

N meningitidis: 3�102

Dicuonzo,199921

N/A* Extraction: boilingUniversal 16S primers PM27, PM8Amplified segment: 600 bpAgarose gel electrophoresis

Yes; 4 species identified byspecies-specificoligoprobes†(§�#**‡‡§§��)

N meningitidisS pneumoniaeH influenzae:2–5�102

Lorino,200022

N/A* Extraction: boilingUniversal 16S primers PM27, PM8Amplified segment: 600 bpAgarose gel electrophoresis

Yes; sequencing (Perkin Elmer,BLAST); also, 4 speciesidentified by species-specificoligoprobes†(‡§�¶#††‡‡§§¶¶)

N meningitidisS pneumoniaeH influenzae:2–5�102

Lu, 200036 1–83 y Extraction: QIAamp tissue kit (QIAGEN)Universal 16S primers U1, U2Amplified segment: 996 bpAgarose gel electrophoresisTurnaround time: 1 day

Yes; restriction endonucleasedigestion†(§�¶#**††§§¶)

E coli: 1�101

S aureus: 2.5�102

Saravolatz,200338

0–79 y Extraction: boilingUniversal 16S primers: 5=-GGAGGAAGGTGGG

GATGACG-3=,5=-ATGGTGTGACGGGCGGTGTG-3=;Amplified segment: 241 bpAgarose gel electrophoresis;Internal amplification control for inhibition

No S pneumoniae: 3.25

Schuurman,200439

0–88 y Extraction: bead beating, silica guanidiniumthiocyanate

Universal 16S primers 27F,1492RAmplified segment: 1,500 bpAgarose gel electrophoresisUracil-N-glycosylase system to prevent amplicon

contaminationInternal amplification control for inhibition

Yes; sequencing (AppliedBiosystems, BLAST, RDP-II)†(‡§�**)

E coli: 1�102

S aureus: 2�102

Poppert,200537

N/A* Extraction: QIAamp mini kit (QIAGEN)Universal 16S primers RW01, DG74Amplified segment: 360–380 bpReal-time PCR (Light Cycler; SYBR green)

Yes; specific hybridization probeLight Cycler assays;sequencing (AppliedBiosystems, BLAST)†(‡�¶‡‡§§)

N meningitidis:�2.5�101

S epidermidis:�2.5�101

Xu, 200541 N/A* Extraction: boiling; alkali lysis; High Pure PCRTemplate Preparation Kit (Roche); QIAamp blood kit(QIAGEN); OmniPrep kit (G Biosciences)

Universal 16S primers P11P, P13PAmplified segment: 216 bpAgarose gel electrophoresis

Yes; sequencing (AppliedBiosystems,BLAST)†(‡§�¶**††‡‡¶¶)

Boiling:S aureus: 1.5�104

E coli: 2.8�103

C jejuni: 3.4�102

QIAamp kit:S aureus: 7.6�102

E coli: 1.1�103

C jejuni: 1.4�102

Deutch,200642

6 days–86 y Extraction: MagNA Pure LC Microbiology Kit M-GRADE(Roche)

Universal 16S primers: forward: 5=-TCCTACGGGAGGCAGCAGT-3=, reverse 5=-GGACTACCAGGGTATCTAATCCTGTT-3=Amplified segment: 466 bpAgarose gel electrophoresis and real-time PCR (Light

Cycler, SYBR green);Internal amplification control for inhibition

Yes; sequencing (AppliedBiosystems, BLAST)†(‡§�**)

N/A*

Turnaround time: PCR: 2–4 h; sequencing: 8–10 h

614 Annals of Emergency Medicine Volume , . : November

phutOip

L

t


of a negative test result.26 The Cochran Q value was highlysignificant for both sensitivity and specificity (57.9 and 651.8,respectively; P�.001), reflecting considerable heterogeneitybetween studies. I2 values, which denote the proportion ofvariability accounted for by between-study heterogeneity, werehigh for both parameters (I2 for sensitivity 79.2%, 95% CI68.4% to 90%; I2 for specificity 98%, 95% CI 97.5% to98.5%). The results of Deeks test for funnel plot asymmetryindicate that there is no significant asymmetry, suggesting a lowlikelihood of publication bias (P�.62).

We also provide data on the performance of the 16S

Table 2. Continued.

Study Age Range PCR Details

Deutch,200734

N/A* Extraction: MagNA Pure LC Microbiolo(Roche)

Universal 16S primers: forward: 5=-TCGAGGCAGCAGT-3=, reverse 5=-GGACTCCAGGGTATCTAATCCTGTT-3=Amplified segment: 466 bpReal-time PCR (Light Cycler, TaqMan

6-FAM probe)Turnaround time: PCR: 4 h; sequenci

Welinder-Olsson,200740

1 day–91 y Extraction: DNA preparation kit (RochUniversal 16S primers SSU1, 806RAmplified segment: 766 bpAgarose gel electrophoresisTurnaround time: 30 h

Chakrabarti,200923

0–14 y Extraction: boilingSeminested PCRUniversal 16S primers u3, ru8Amplified segment: 1,000 bpThree species-specific primersAgarose gel electrophoresis

Duan,200935

3 days–11 y Extraction: lysis, boilingUniversal 16S primers P690F, P829RReal-time PCR (universal fluorescent

800–778 bp)Turnaround time: 2–3 h

Rothman,201043

N/A* Extraction: lysis and ultrafiltrationUniversal 16S primers p891F, p1033Amplified segment: 161 bpReal-time PCR (TaqMan universal fluoTurnaround time: 3 h

BLAST, Basic Local Alignment Search Tool; bp, base pairs; PCR, polymerase chai*N/A: Data not available.†Bacteria identified by PCR.‡Neisseria meningitidis.§Hemophilus influenzae.�Streptococcus.¶Staphylococcus.#Mycobacterium tuberculosis.**Escherichia coli.††Other Gram-negative bacteria (Klebsiella, Pasteurella, Serratia, Proteus, Citroba‡‡Other bacteria (Borrelia, Acinetobacter, Bacillus, Streptomyces, Corynebacterium§§Enterococcus.��Listeria.¶¶Pseudomonas.

polymerase chain reaction test in cases of culture-negative c


resumed bacterial meningitis (Table 3). There waseterogeneity in the definition of culture-negative meningitissed by the studies.The ability of the polymerase chain reactionest to detect bacterial pathogens varied from 3% to 100%.verall, the 16S polymerase chain reaction provided bacterial

dentification in 120 (30%) of 360 cases of culture-negativeresumed bacterial meningitis.

IMITATIONSThis study has several important limitations. We compared

he 16S rRNA gene polymerase chain reaction test to the

Bacterial Identification Detection Limit, cfu/mL

it M-GRADE

GG

orescent

day

Yes; sequencing(BLAST)†(�¶**††‡‡)

N/A*

Yes; sequencing(Applied Biosystems,BLAST)†(‡§�¶**††‡‡§§�)

S pneumoniae: 1�101

E coli, S aureus: 1�103

Yes; 3 species ofbacteria identified byspecies-specificprimers†(�)

E coli: 1�103

S pneumoniae: 4�103

:

Yes; sequencing(Invitrogen)†(**††)

N/A*

nt probe)

Yes; 7 species ofbacteria identified byspecies-specificprobes and Gram-typingprobes†(‡§�¶**��)

S pneumonia: 70S aureus: 50L monocytogenes: 110N meningitides: 20H influenza: 10E coli: 30S epidermidis: 10

tion; RDP II, Ribosomal Database Project.

Enterobacter).obacterium).

gy K

CTACA

5= flu

ng: 1e)

probe

R

resce

n reac

cter,, Fus

urrent criterion standard for universal bacterial detection in


D

oaiaatdacpbApde

tbsirbkpiSm(t

rwiltTroecdcp

bodstns


meningitis, the cerebrospinal fluid culture. However, the latteris itself an imperfect test, affected by antibiotic pretreatment,bacterial density, and culture techniques, and thus the results ofour comparison may be subject to intrinsic bias. As noted byprevious authors, meta-analyses of diagnostic tests face a specificset of issues, and our study is no exception.25 Although weensured the quality of included studies by using the QUADASchecklist, we recognize that there was a wide range of patientand sample accrual methods. The variability in the design andconduct of studies contributes to difficulty in estimatingtheir quality and in ensuring that the data are comparable.This is reflected in the significant heterogeneity uncoveredin the meta-analysis and is likely due to the fact that thistechnique is currently not subject to regulatory oversight oragreement on a single methodology. Despite theheterogeneity, the estimates of sensitivity and specificityappeared to show a similar direction of effect, supporting thepooling of data into a meta-analysis. There is also always theconcern for publication bias because studies with positiveresults are more likely to be published. Though this wasdifficult to assess in the face of between-study heterogeneity,results of the Deeks test (for funnel plot asymmetry) showeda low likelihood of publication bias. We have also providedinformation about the performance of this test in culture-negative presumed bacterial meningitis. Despite the potentialbias associated with the variability in the definition of thiscondition and the wide range of sensitivities reported in thisancillary analysis, our data suggest that the 16S polymerasechain reaction may have clinical utility in this situation. Anadditional limitation includes our inability to extract data

Radstrom, 1994

Dicuonzo, 1999

Lorino, 2000

Lu, 2000

Saravolatz, 2003

Schuurman, 2004

Poppert, 2005

STUDY, YEAR

Xu, 2005

Deutch, 2006

Deutch, 2007

Welinder, 2007

Chakrabarti, 2009

Duan, 2009

Rothman, 2010

20 40 60 80 100

95% CI Point estimate

FOREST PLOT FOR SENSITIVITYSTUDY, YEAR

Radstrom, 1994

Dicuonzo, 1999

Lorino, 2000

Lu, 2000

Saravolatz, 2003

Schuurman, 2004

Poppert, 2005

Xu, 2005

Deutch, 2006

Deutch, 2007

Welinder, 2007

Chakrabarti, 2009

Duan, 2009

Rothman, 2010

60 80 1004020

95% CI Point estimate

FOREST PLOT FOR SPECIFICITY

Figure 2. The graph on the left shows the plot of thesensitivity of individual studies (filled squares), with CIs(horizontal lines). The diamond at the bottom shows thepooled sensitivity. The graph on the right shows the plot ofthe specificity of individual studies (filled squares), with CIs(horizontal lines). The diamond at the bottom shows thepooled specificity.

from a few studies.29,44 p


ISCUSSIONBacterial meningitis remains a cause of significant adverse

utcomes.1,2 Early diagnosis has the potential to guide specificntibiotic therapy, reduce unnecessary antibiotic use, andmprove patient outcomes. Currently available diagnostic testsre often unreliable, particularly when antibiotics aredministered before the lumbar puncture.4 Our meta-analysis ofhe 16S ribosomal RNA gene polymerase chain reactionemonstrates that this diagnostic tool has excellent sensitivitynd specificity. We suggest that these tests, when used inombination with methods to identify bacterial species, mayrove useful for confirmation of bacterial meningitis and foracterial identification in acute care settings such as the ED.lso, a negative polymerase chain reaction result has a highrobability of ruling out disease and may be helpful in makingecisions about the discontinuation of antibiotic therapy andarly discharge of noninfected patients.

Our review reveals that there was considerable variability inhe sensitivity of the 16S polymerase chain reaction, which maye partially attributable to methodological differences acrosstudies. The preliminary step of DNA extraction from sampless a key determinant of the success of the polymerase chaineaction. In our meta-analysis, extraction processes ranged fromoiling techniques20-23,35,38,41 to use of commercial extractionits that rely on lysis35,41,43 or bead beating.39 Studies did notrovide a reliable estimate of the amount of DNA extracted, sot is difficult to compare the success of each of these methods.ix studies reported a sensitivity of 100%, however, and aajority of these studies used a crude DNA extraction process

boiling method).21,22,35,37,38,41 This leads us to speculate thathis extraction method may improve the yield of bacterial DNA.

Other factors affecting the sensitivity of the polymerase chaineaction test include the primer and probe regions selected,hich should be of adequate length to identify all bacteria of

nterest while reliably excluding nonbacterial DNA. The studiesisted in this meta-analysis used multiple primer/probe designsargeting various lengths and regions of the 16S rRNA gene.he methodology ranged from conventional polymerase chain

eaction followed by agarose gel detection20-23,36,38-42 to the usef real-time technology34,35,37,42,43; the latter may allow us tostimate bacterial load when results are positive, whereasonventional polymerase chain reactions followed by agarose-geletection can provide only a positive or negative result. Weould not identify primers, probes, or polymerase chain reactionrotocols consistently linked to high or low sensitivities.

The polymerase chain reaction test requires a complementaryacterial identification technique to provide the specific identityf the detected bacteria. Studies included in this review adoptedifferent approaches to bacterial identification. Eight of 14tudies performed sequencing to confirm the specific identity ofhe pathogen.22,34,35,37,39-42 Seven studies used a limitedumber of species-specific primers and probes to identify a fewpecific organisms,20-23,36,37,43 whereas 1 study reported positive
olymerase chain reaction results without further identification

cdsnvptcp

tr

negati


of bacteria.38 Although sequencing allows identification of anybacterial species detected by polymerase chain reaction, possiblyleading to overidentification of contaminants, it is limited in itsability to detect multiple pathogens. Species-specific primers, onthe other hand, may miss unusual pathogens.

We believe that our ancillary observation about thesensitivity of the polymerase chain reaction test in culture-negative presumed bacterial meningitis also has clinicalrelevance. Practitioners often base decisions about diagnosis andduration of antibiotic therapy on the overall clinical scenario,not just the cerebrospinal fluid culture result. Though limitedby the variation in definition of culture-negative presumed

Table 3. Sensitivity of 16S polymerase chain reaction in culture

StudyDefinition of Culture-Negative Presumed

Bacterial Meningitis

Radstrom, 199420 Positive by bacterial antigen testingDicuonzo, 199921 At least 1 abnormal CSF parameter:

glucose �40 mg/dL, protein �45mg/dL, WBC �40 cell/mm3

Lorino, 200022 Clinical signs � at least 1 abnormal CSFparameter: glucose �40 mg/dL, totalprotein �40 mg/dL, WBC �40 cells/mm3

Lu, 200036 Clinical symptoms or positive bybacterial antigen testing

Saravolatz, 200338 CSF WBC count �100 cells/mm3

Schuurman, 200439 Clinical symptomsPoppert, 200537 Clinical symptoms, elevated CSF WBC

counts, or same bacteria inbloodstream

Deutch, 200642 Clinical symptoms or earlier positive CSFculture result with same bacteria, orpositive CSF antigen testing result

Welinder-Olsson, 200740 Clinical signs, CSF WBC�10�106 cells/L, �/– same bacteria in bloodstream

Chakrabarti, 200923 Clinical signs, elevated CSF WBC count,�/– positive Gram stain result,abnormal CSF protein and glucoselevels

Duan, 200935 Clinical meningitis as defined bylymphocytes and neutrophils in theCSF, low concentration CSF glucose,sterile blood and CSF cultures

Margall Coscojuela,200232†

Clinical signs

Pandit, 200533†Clinical signs and cytochemical criteria

(not detailed)Arosio, 200830†

Clinical symptoms and elevated CSFWBC (range 322–16,000 cells/mm3)�/– positive Gram stain result �/–elevated CSF protein level �/– lowCSF glucose level

Chen, 200931†Clinical symptoms

CSF, cerebrospinal fluid.*Culture-negative presumed bacterial meningitis defined by individual authors acctive CSF culture results.†Studies not included in main meta-analysis because they examined only culture-

bacterial meningitis, our data suggest that the 16S polymerase c


hain reaction test may provide supportive evidence for theiagnosis of bacterial meningitis when there is a clinicaluspicion of disease and cerebrospinal fluid culture results areegative. A prospective study of patients withentriculoperitoneal shunt infections suggested that the 16Solymerase chain reaction test result remained positive later inhe course of illness than that of conventional bacterialultures.34 This test may thus provide useful information aboutatients who receive antibiotics before the lumbar puncture.

Previous investigators have observed some of the challengeso developing a clinically useful bacterial polymerase chaineaction, including problems with bacterial and human DNA

ative presumed bacterial meningitis.*

Number of PositivePCR Test Results

Cases of Culture-NegativePresumed Bacterial

MeningitisSensitivity, %

(95% CI)

5 5 100 (46–100)14 37 38 (23–55)

18 47 38 (25–54)

3 3 100 (31–100)

1 37 3 (0.1–16)6 6 100 (52–100)3 3 100 (31–100)

3 3 100 (31–100)

19 19 100 (79–100)

5 88 6 (2–13)

3 3 100 (31–100)

9 51 18 (9–31)

19 41 46 (31–62)

9 12 100 (46–100)

3 5 38 (23–55)

to clinical examination and ancillary laboratory tests in the presence of nega-

ve presumed bacterial meningitis.

-neg

ording

ontamination,45 the presence of inhibitors in the substrate,20


wacpcrrgsaba

baiTwdpaHmdWtd

S

Apdaar

Ftrpar(Cf(i

PRAA

PRA


and difficulty in cell wall lysis.46 The highly sensitive nature ofthe polymerase chain reaction technique often leads to detectionof contamination with nonpathogenic bacterial DNA.Investigators have addressed this by using rigorous steriletechniques, amplifying larger amplicons of the 16S rRNA gene,and setting strict upper-cycle threshold limits for regarding real-time polymerase chain reaction results as truly positive.39,45 Inrecent times, the development of ultrapure DNA-free Taqpolymerases and reagents has also aided in the effort toovercome this problem.45 Other investigators have observed therole of human DNA in causing a nonspecific background signal,thus decreasing the sensitivity of the polymerase chain reaction,especially in situations with a low bacterial load.47 Furthermore,high protein content or other molecules can nonspecificallyinhibit the polymerase chain reaction and decrease its analyticalsensitivity.20 Deutch et al,34,42 and Saravolatz et al38 usedinternal positive controls to test for inhibitors. Investigatorshave also observed the decreased analytical sensitivity of the 16Spolymerase chain reaction test with regard to Gram-positiveorganisms. The addition of steps to facilitate cell wall lysis hasimproved this problem.39

The development of a new diagnostic test spans severallayers, from analytic testing to identify limits of detection andscope of bacterial identification, to clinical validation studiesthat contribute to refinement of the technology. Studies at thisstage often use locally developed versions of technology, whichmust then be incorporated into a single best technique approvedby regulatory bodies such as the FDA. Although our studyhighlights the variability in current 16S polymerase chainreaction tests, significant progress has been made in refining andperfecting this technology, bringing it closer to the bedside.Real-time polymerase chain reaction instruments, which provideresults with a turnaround time of 2 to 6 hours, are increasinglyavailable at hospital-based central laboratories for bothenteroviral and herpes virus testing. As the field of bacterialpolymerase chain reaction testing moves toward bedsideperformance, we believe that the turnaround times in the contextof the studies in this meta-analysis do not yet reflect the truepotential of this test. At least 2 commercial 16S polymerase chainreaction–based kits (Sepsitest, Molzym, Bremen, Germany;Septifast, Roche, Basel, Switzerland) have become available forresearch purposes in studies of sepsis.48,49 As additional studiesemerge that demonstrate reliable detection of bacteria with 16Spolymerase chain reaction technology, clinical application of thistest is closer. However, broad-range assays require validation fordetection of a wide range of pathogens, which is much morecomplicated than validation against a single pathogen. Thus, theFDA approval process is particularly complex and time consumingin these situations.

To our knowledge, this is the first meta-analysis to evaluatethe accuracy of the 16S polymerase chain reaction test for thediagnosis of bacterial meningitis. Our study has severalstrengths. We have used a consistent a priori definition of
bacterial meningitis as the criterion standard for comparison 2

ith the polymerase chain reaction. There was a high level ofgreement between the independent reviewers about dataollection and analysis. Furthermore, our study includesublications spanning the last decade of research with a largeomposite sample size and a large number of positive cultureesults. A wide spectrum of the population with differing ageanges was included in this meta-analysis, supporting theeneralizability of these results. Furthermore, our ancillary datauggest the potential utility of the polymerase chain reaction asn adjunctive test in subjects with culture-negative presumedacterial meningitis, many of whom may have receivedntibiotics before the lumbar puncture.

In conclusion, our meta-analysis supports the utility of aroad-range 16S rRNA polymerase chain reaction as andjunctive tool for the diagnosis of bacterial meningitis, givents high sensitivity, specificity, and promising likelihood ratios.his test could provide early, accurate diagnosis of infectionhile decreasing antibiotic exposure and facilitating earlierischarge of noninfected patients. We also suggest its utility foratients with culture-negative presumed bacterial meningitisnd those receiving antibiotic therapy before lumbar puncture.owever, additional studies are needed to define the optimalethodology for this test so that greater consensus can be

eveloped to achieve meaningful comparisons across studies.ith further refinements in polymerase chain reaction

echnology, this test has the potential to improve clinicalecisionmaking within the time frame of ED care.

upervising editor: David A. Talan, MD

uthor contributions: LS conceived the study idea. LS and JMPerformed the literature search, data collection, and primaryata analysis. All authors contributed to the study design. Alluthors contributed to the analysis and interpretation of datand article revision. LS drafted the article and takesesponsibility for the paper as a whole.

unding and support: By Annals policy, all authors are requiredo disclose any and all commercial, financial, and otherelationships in any way related to the subject of this article aser ICMJE conflict of interest guidelines (see www.icmje.org). Theuthors have stated that no such relationships exist. Dr. Harriseceived support from the University Research FoundationUniversity of Pennsylvania) and the Foerderer Murray Award (Thehildren’s Hospital of Philadelphia). Dr. Shah received support

rom the National Institute of Allergy and Infectious DiseasesK01 AI73729) and the Robert Wood Johnson Foundation underts Physician Faculty Scholar Program.

ublication dates: Received for publication August 1, 2011.evisions received January 23, 2012, and May 7, 2012.ccepted for publication May 31, 2012. Available onlineugust 9, 2012.

resented as an abstract at the Eastern Society for Pediatricesearch, Philadelphia, PA, March 2010; and at the Pediatriccademic Societies, Vancouver, British Columbia, Canada, May
010.

http://www.icmje.org

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3


The content of the article is solely the responsibility of theauthors and does not necessarily represent the official viewsof the National Institutes of Health.

Address for correspondence: Lakshmi Srinivasan, MBBS,E-mail [email protected].

REFERENCES1. Heath PT, Nik Yusoff NK, Baker CJ. Neonatal meningitis. Arch Dis

Child Fetal Neonatal Ed. 2003;88:F173-178.2. Grimwood K, Anderson P, Anderson V, et al. Twelve year

outcomes following bacterial meningitis: further evidence forpersisting effects. Arch Dis Child. 2000;83:111-116.

3. Aronin SI, Peduzzi P, Quagliarello VJ. Community-acquired bacterialmeningitis: risk stratification for adverse clinical outcome and effectof antibiotic timing. Ann Intern Med. 1998;129:862-869.

4. Kanegaye JT, Soliemanzadeh P, Bradley JS. Lumbar puncture inpediatric bacterial meningitis: defining the time interval forrecovery of cerebrospinal fluid pathogens after parenteralantibiotic pretreatment. Pediatrics. 2001;108:1169-1174.

5. Tunkel AR, Scheld WM. Acute bacterial meningitis. Lancet. 1995;346:1675-1680.

6. Neuman MI, Tolford S, Harper MB. Test characteristics andinterpretation of cerebrospinal fluid gram stain in children. PediatrInfect Dis J. 2008;27:309-313.

7. Stoll BJ, Hansen N, Fanaroff AA, et al. To tap or not to tap: highlikelihood of meningitis without sepsis among very low birthweight infants. Pediatrics. 2004;113:1181-1186.

8. Garges HP, Moody MA, Cotten CM, et al. Neonatal meningitis:what is the correlation among cerebrospinal fluid cultures, bloodcultures, and cerebrospinal fluid parameters? Pediatrics. 2006;117:1094-1100.

9. La Scolea LJ Jr, Dryja D. Quantitation of bacteria in cerebrospinalfluid and blood of children with meningitis and its diagnosticsignificance. J Clin Microbiol. 1984;19:187-190.

10. Fitch MT, Abrahamian FM, Moran GJ, et al. Emergencydepartment management of meningitis and encephalitis. InfectDis Clin North Am. 2008;22:33-52, v-vi.

11. Fitch MT, van de Beek D. Emergency diagnosis and treatment ofadult meningitis. Lancet Infect Dis. 2007;7:191-200.

12. Kumar A, Roberts D, Wood KE, et al. Duration of hypotensionbefore initiation of effective antimicrobial therapy is the criticaldeterminant of survival in human septic shock. Crit Care Med.2006;34:1589-1596.

13. Dewan M, Zorc JJ, Hodinka RL, et al. Cerebrospinal fluidenterovirus testing in infants 56 days or younger. Arch PediatrAdolesc Med. 2010;164:824-830.

14. Seme K, Mocilnik T, Komlos KF, et al. GeneXpert enterovirusassay: one-year experience in a routine laboratory setting andevaluation on three proficiency panels. J Clin Microbiol. 2008;46:1510-1513.

15. Dark PM, Dean P, Warhurst G. Bench-to-bedside review: thepromise of rapid infection diagnosis during sepsis usingpolymerase chain reaction-based pathogen detection. Crit Care.2009;13:217.

16. Janda JM, Abbott SL. 16S rRNA gene sequencing for bacterialidentification in the diagnostic laboratory: pluses, perils, andpitfalls. J Clin Microbiol. 2007;45:2761-2764.

17. Greisen K, Loeffelholz M, Purohit A, et al. PCR primers andprobes for the 16S rRNA gene of most species of pathogenicbacteria, including bacteria found in cerebrospinal fluid. J ClinMicrobiol. 1994;32:335-351.

18. Harris KA, Hartley JC. Development of broad-range 16S rDNA PCRfor use in the routine diagnostic clinical microbiology service.
J Med Microbiol. 2003;52(pt 8):685-691.

9. Lehmann LE, Hunfeld KP, Emrich T, et al. A multiplex real-timePCR assay for rapid detection and differentiation of 25 bacterialand fungal pathogens from whole blood samples. Med MicrobiolImmunol. 2008;197:313-324.

0. Radstrom P, Backman A, Qian N, et al. Detection of bacterialDNA in cerebrospinal fluid by an assay for simultaneous detectionof Neisseria meningitidis, Haemophilus influenzae, andstreptococci using a seminested PCR strategy. J Clin Microbiol.1994;32:2738-2744.

1. Dicuonzo G, Lorino G, Lilli D, et al. Use of oligoprobes onamplified DNA in the diagnosis of bacterial meningitis. Eur J ClinMicrobiol Infect Dis. 1999;18:352-357.

2. Lorino G, Angeletti S, Gherardi G, et al. Diagnostic value ofcytokine assays in cerebrospinal fluid in culture-negative,polymerase chain reaction–positive bacterial meningitis. Eur J ClinMicrobiol Infect Dis. 2000;19:388-392.

3. Chakrabarti P, Das BK, Kapil A. Application of 16S rDNA basedseminested PCR for diagnosis of acute bacterial meningitis.Indian J Med Res. 2009;129:182-188.

4. Sterne JAC, ed. Meta-analysis in Stata: An Updated CollectionFrom the Stata Journal. College Station, TX: Stata Press; 2009.

5. Reitsma JB, Glas AS, Rutjes AW, et al. Bivariate analysis ofsensitivity and specificity produces informative summarymeasures in diagnostic reviews. J Clin Epidemiol. 2005;58:982-990.

6. Deeks JJ. Systematic reviews in health care: systematic reviewsof evaluations of diagnostic and screening tests. BMJ. 2001;323:157-162.

7. Deeks JJ, Macaskill P, Irwig L. The performance of tests ofpublication bias and other sample size effects in systematicreviews of diagnostic test accuracy was assessed. J ClinEpidemiol. 2005;58:882-893.

8. Hall LM, Duke B, Urwin G. An approach to the identification of thepathogens of bacterial meningitis by the polymerase chainreaction. Eur J Clin Microbiol Infect Dis. 1995;14:1090-1094.

9. Kotilainen P, Jalava J, Meurman O, et al. Diagnosis ofmeningococcal meningitis by broad-range bacterial PCR withcerebrospinal fluid. J Clin Microbiol. 1998;36:2205-2209.

0. Arosio M, Nozza F, Rizzi M, et al. Evaluation of the MicroSeq 50016S rDNA-based gene sequencing for the diagnosis of culture-negative bacterial meningitis. New Microbiol. 2008;31:343-349.

1. Chen LH, Duan QJ, Cai MT, et al. Rapid diagnosis of sepsis andbacterial meningitis in children with real-time fluorescentquantitative polymerase chain reaction amplification in thebacterial 16S rRNA gene. Clin Pediatr (Phila). 2009;48:641-647.

2. Margall Coscojuela N, Majo Moreno M, Latorre Otin C, et al. Useof universal PCR on cerebrospinal fluid to diagnose bacterialmeningitis in culture-negative patients. Eur J Clin Microbiol InfectDis. 2002;21:67-69.

3. Pandit L, Kumar S, Karunasagar I. Diagnosis of partially treatedculture-negative bacterial meningitis using 16S rRNA universalprimers and restriction endonuclease digestion. J Med Microbiol.2005;54(pt 6):539-542.

4. Deutch S, Dahlberg D, Hedegaard J, et al. Diagnosis ofventricular drainage-related bacterial meningitis by broad-rangereal-time polymerase chain reaction. Neurosurgery. 2007;61:306-311; discussion 311–312.

5. Duan QJ, Shang SQ, Wu YD. Rapid diagnosis of bacterialmeningitis in children with fluorescence quantitative polymerasechain reaction amplification in the bacterial 16S rRNA gene. EurJ Pediatr. 2009;168:211-216.

6. Lu JJ, Perng CL, Lee SY, et al. Use of PCR with universal primers
and restriction endonuclease digestions for detection and

mailto:[email protected]

4

4

4

4

4

4


identification of common bacterial pathogens in cerebrospinalfluid. J Clin Microbiol. 2000;38:2076-2080.

37. Poppert S, Essig A, Stoehr B, et al. Rapid diagnosis of bacterialmeningitis by real-time PCR and fluorescence in situ hybridization.J Clin Microbiol. 2005;43:3390-3397.

38. Saravolatz LD, Manzor O, VanderVelde N, et al. Broad-rangebacterial polymerase chain reaction for early detection ofbacterial meningitis. Clin Infect Dis. 2003;36:40-45.

39. Schuurman T, de Boer RF, Kooistra-Smid AM, et al. Prospectivestudy of use of PCR amplification and sequencing of 16Sribosomal DNA from cerebrospinal fluid for diagnosis of bacterialmeningitis in a clinical setting. J Clin Microbiol. 2004;42:734-740.

40. Welinder-Olsson C, Dotevall L, Hogevik H, et al. Comparison ofbroad-range bacterial PCR and culture of cerebrospinal fluid fordiagnosis of community-acquired bacterial meningitis. ClinMicrobiol Infect. 2007;13:879-886.

41. Xu J, Moore JE, Millar BC, et al. Employment of broad range 16SrDNA PCR and sequencing in the detection of aetiological agentsof meningitis. New Microbiol. 2005;28:135-143.

42. Deutch S, Pedersen LN, Podenphant L, et al. Broad-range realtime PCR and DNA sequencing for the diagnosis of bacterialmeningitis. Scand J Infect Dis. 2006;38:27-35.

43. Rothman R, Ramachandran P, Yang S, et al. Use of quantitative
broad-based polymerase chain reaction for detection and

identification of common bacterial pathogens in cerebrospinalfluid. Acad Emerg Med. 2010;17:741-747.

4. Hall LM, Duke B, Urwin G. An approach to the identification of thepathogens of bacterial meningitis by the polymerase chainreaction. Eur J Clin Microbiol Infect Dis. 1995;14:1090-1094.

5. Corless CE, Guiver M, Borrow R, et al. Contamination andsensitivity issues with a real-time universal 16S rRNA PCR. J ClinMicrobiol. 2000;38:1747-1752.

6. Jordan JA, Durso MB, Butchko AR, et al. Evaluating the near-terminfant for early onset sepsis: progress and challenges to considerwith 16S rDNA polymerase chain reaction testing. J Mol Diagn.2006;8:357-363.

7. Handschur M, Karlic H, Hertel C, et al. Preanalytic removal ofhuman DNA eliminates false signals in general 16S rDNA PCRmonitoring of bacterial pathogens in blood. Comp ImmunolMicrobiol Infect Dis. 2009;32:207-219.

8. Wellinghausen N, Kochem AJ, Disque C, et al. Diagnosis ofbacteremia in whole-blood samples by use of a commercialuniversal 16S rRNA gene-based PCR and sequence analysis.J Clin Microbiol. 2009;47:2759-2765.

9. Yanagihara K, Kitagawa Y, Tomonaga M, et al. Evaluation ofpathogen detection from clinical samples by real-time polymerasechain reaction using a sepsis pathogen DNA detection kit. Crit
Care. 2010;14:R159.

res of consistency (eg, I2) for each meta-analysis.


Table E1. PRISMA checklist.

Section/Topic No.

TitleTitle 1 Identify the report as a sAbstractStructured summary 2 Provide a structured sum

data sources; study eliappraisal and synthesiimplications of key find

IntroductionRationale 3 Describe the rationale foObjectives 4 Provide an explicit statem

participants, interventio(PICOS).

MethodsProtocol and registration 5 Indicate whether a review

Web address), and, if aregistration number.

Eligibility criteria 6 Specify study characterischaracteristics (eg, yeacriteria for eligibility, gi

Information sources 7 Describe all information scontact with study authdate last searched.

Search 8 Present full electronic selimits used, such that

Study selection 9 State the process for selsystematic review, and

Data collection process 10 Describe method of dataindependently, in duplidata from investigators

Data items 11 List and define all variablsources) and any assu

Risk of bias in individualstudies

12 Describe methods used f(including specificationlevel) and how this info

Summary measures 13 State the principal summSynthesis of results 14 Describe the methods of

done, including measu

Checklist Item

Reportedon Page

No.

ystematic review, meta-analysis, or both. Title page

mary, including, as applicable, background; objectives;gibility criteria, participants, and interventions; studys methods; results; limitations; conclusions andings; systematic review registration number.

1

r the review in the context of what is already known. 2–3ent of questions being addressed with reference tons, comparisons, outcomes, and study design

3

protocol exists, if and where it can be accessed (eg,vailable, provide registration information, including

tics (eg, PICOS, length of follow-up) and reportrs considered, language, publication status) used asving rationale.

4

ources (eg, databases with dates of coverage,ors to identify additional studies) in the search and

4

arch strategy for at least 1 database, including anyit could be repeated.

4

ecting studies (ie, screening, eligibility, included in, if applicable, included in the meta-analysis).

4, QUADAS

extraction from reports (eg, piloted forms,cate) and any processes for obtaining and confirming.

4–5

es for which data were sought (eg, PICOS, fundingmptions and simplifications made.

4–5

or assessing risk of bias of individual studiesof whether this was done at the study or outcomermation is to be used in any data synthesis.

5

ary measures (eg, risk ratio, difference in means). 5handling data and combining results of studies, if 5

Volume , . : November Annals of Emergency Medicine 620.e1


Table E2. QUADAS checklist.

Study Number*

Characteristics 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Type of study R, B P, NB P, NB U, UB P, UB P, NB U, UB U, UB P, B P, B R, B P, NB P, NB U, B R, B R, B R, B U, UBSpectrum of participants

representative?Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y

Selection criteriadescribed?

Y Y Y U U Y U Y Y Y Y Y Y Y Y Y Y Y

Choice of referencestandard appropriate?

Y Y Y Y Y Y Y Y Y Y Y Y Y Y U U U U

Whole/random samplereceive verificationwith referencestandard?

Y Y Y Y Y Y Y Y Y Y Y Y Y Y N Y Y Y

Did participants receivesame referencestandard regardless ofindex test result?

Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y

Was reference standardindependent of theindex test?


Was index testdescribed in sufficientdetail to permitreplication?

Y Y Y Y Y Y Y Y Y Y Y Y Y Y N Y Y Y

Was execution ofreference standarddescribed in sufficientdetail to permitreplication?

N N N N Y Y Y N Y N N Y Y Y N N N N

Were index test resultsinterpreted withoutknowledge of resultsof referencestandard?

U U U U U U U U Y Y U U U Y U U N N

Were reference standardresults interpretedwithout knowledge ofresults of index test?


Were same clinical dataavailable when testresults wereinterpreted as wouldbe available when testis used in practice?

Y Y Y Y Y Y Y Y Y Y Y Y Y U U Y Y Y

Were indeterminate orintermediate testresults reported?

Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y U U Y

R, Retrospective; B, clinicians blinded to PCR results; P, prospective; NB, clinicians not blinded to PCR results; U, unclear; UB, blinding status unclear; Y, yes; N, no.*Names of studies listed by study number: 1: Radstrom, 199420; 2: Dicuonzo, 199921; 3: Lorino, 200022; 4: Lu, 200036; 5: Saravolatz, 200338; 6: Schuurman,200439; 7: Poppert, 200537; 8: Xu, 200541; 9: Deutch, 200642; 10: Deutch, 200734; 11: Welinder-Olsson, 200740; 12: Chakrabarti, 200923; 13: Duan, 200935; 14:Rothman, 201043; 15: Margall Coscojuela, 200232; 16: Pandit, 200533; 17: Arosio, 200830; 18: Chen, 2009.31 We did not list information on the following checklistdetails: Was time interval between reference standard and index test short? All samples for the index test were obtained during the same lumbar puncture as the ref-
erence standard. Withdrawals: None of the studies listed any withdrawals.
620.e2 Annals of Emergency Medicine Volume , . : November

can broad-range 16s ribosomal ribonucleic acid gene polymerase chain reactions improve the diagnosis...

Documents