In children and young people with suspected meningitis, can CSF variables (white cell count, glucose, protein) distinguish between bacterial and viral meningitis?

Bibliographic information / Study type & evidence / Study details / Number of patients / Patient characteristics / Intervention & comparison / Follow up & outcome / Results & effect size / Reviewer comment
Chavanet
2007 79 / Retro-spective cohort study
EL=III / Aim: to establish a scoring tool to distinguish bacterial and viral meningitis (BM & VM) and compare it to other available decision making tools / N=275
100 adults and 175 children with meningitis
BM=19,
VM=105
Undetermined meningitis (UM)=51
Data collected 1995-2002 / Children aged 3 months to 15 years hospitalised for acute meningitis between 1995 and 2002 at a French hospital / Infectious meningitis defined as more than 7 leucocytes/ml of CSF with no alternative diagnosis
Diagnosis of BM by identified bacterial pathogen in CSF (visually, culture) or blood (culture) or a significant concentration of bacterial antigen by latex agglutination was obtained in blood, urine or CSF, or seroconversion for Leptospiral spp. was evident.
Aseptic or viral meningitis diagnoses were given if a viral pathogen was identified in CSF, blood or stool or if no antibacterial treatments were administered.
Meningitis entered into the medical record as “either” category was classified as “undetermined” / Demo-graphic data, signs and symptoms data and CSF test results recorded from notes for CSF white cell count, CSF protein. / CSF White cell count
mean (SD) cells/ml BM=2994 (3263)
VM=218 (280)
UM=431 (772)
%CSF-PMN
BM= 87 (12)
VM=43(31)
UM=63(30)
Suggested threshold as part of Meningitest scoring =1800 cells/ml
(AROC analysis =0.75)
CSF neutrophil percentage >=80
AROC analysis 0.94)
CSF protein
mean (SD) g/l
BM=2.3 (1.5)
VM=0.38(0.18)
UM=0.47 (0.24)
Suggested threshold as part of Meningitest scoring = 1.2 g/l (AROC analysis =0.92) / Country: France
Funding: Not reported
Note: this paper uses %CSF-PMN in tables and neutrophil percentage in the text and abstract – as far as I can tell they are using them to mean the same thing.
Corrall 1981 80 / Prospective cohort study
EL=II / Aim: to evaluate a CSF CRP latex agglutination system and compare its ability to discriminate between bacterial and non-bacterial meningitis with other CSF tests / N=56 children with meningitis
BM=24 (12/24 Hib)
VM=8
UM=24
VM + UM = nonbacterial meningitis
Data collection 1978-1980 / Children aged 1 month to 16 years with signs and symptoms of meningitis and more than 10WBC/mm3 in CSF were prospectively enrolled in the study between 1978 and 1980 in US / Diagnosis of bacterial meningitis by CSF culture and by bacterial antigen detection in CSF, serum and urine by counterimmunoelectrophoresis and immunoassay
Diagnosis of viral meningitis by CSF, serum, throat and rectal sample culture / CSF test results prospectively recorded:
CSF white cell count
CSF protein
CSF glucose / CSF White cell count
mean (SD) cells/ mm3
BM=2417 (1380)
VM=149 (116)
UM=264 (204)
CSF PMN
mean (SD) cells/ mm3
BM=2171 (1382)
VM=60 (93)
UM=196 (184)
To differentiate between BM vs. VM+UM (@CSF WCC >500cells/mm3) sensitivity 74%, specificity 94%, PPV 89%, NPV 83%
To differentiate between BM vs. VM+UM (@CSF PMN concentration >200cells/mm3)
sensitivity 91%, specificity 84%, PPV 81%, NPV 93%
CSF protein
mean (SD) mg/dl
BM=174 (36)
VM=74 (35)
UM= 46 (11)
To differentiate between BM vs. VM+UM (at CSF [Protein] >100mg/dl)
sens 74%, spec 94%, PPV 89%, NPV 83%
CSF Glucose
mean (SD) mg/dl
BM=28 (8)
VM=56 (8)
UM=63 (6).
To differentiate between BM vs. VM+UM (@CSF [glucose] <40mg/dl)
sens 78%, spec 100%, PPV 100%, NPV 86% / Country: US
Funding: Not reported
Dubos 2006 63 / Retrospective cohort study
EL=III / Aim: to identify the blood and CSF tests that best distinguish between bacterial and aseptic meningitis in an emergency department / N=167 children
BM=21 (1/21 Hib)
AM=146 / Children aged 28 days to 16 years admitted to a French university hospital between 2000 and 2004 for acute meningitis.
Period of inclusion was extended to 1995-2004 to include more BM cases only
Exclusion criteria: known neurosurgical disease, immunosuppression, traumatic Lumbar Puncture (LP), pre-treated meningitis or referral from another hospital after diagnosis. Those with incomplete files were also excluded secondarily / Acute meningitis defined as CSF WBC count of ≥7WBCs/mm3
BM defined as the acute onset of meningitis with documented CSF bacterial infection (direct examination, culture or latex agglutination)
Aseptic meningitis was defined as the acute onset of meningitis and the absence of any bacterial meningitis criteria / Data were extracted from medical notes including:
CSF white cell count; CSF protein, CSF glucose concentration. / CSF white cell count mean/median/[range] cells/ mm3
BM=3072/1120/[7-10,600]
AM=179/85.5/[7-2520] p<10-5.
To differentiate between BM vs. AM (@WBC count threshold 200/ mm3)
OR=9 (95% CI 3 to 32); p<10-5
CSF neutrophils/mm3 mean (range)
BM=2789 (7-10,600)
AM=68 (0-952)
p<10-6
To differentiate between BM vs. AM (@ CSF neutrophil count threshold 100/ mm3)
OR=18 (95% CI 5.1 to 71.2); p<10-8
CSF proteinmean/median/[range] g/l BM=1.8/1.4/[0.4-5.3]
AM=0.4/0.4/[0.1-1.8] p<10-6
To differentiate between BM vs. AM (@ [Protein] threshold 0.5g/l)
OR=22 (95% CI 6 to 101)
CSF glucose mean/median/[range] mmol/l
BM=1.8/1.4/[0.0-4.4]
AM=3.0/3.0/[1.3-4.6] p=0.01.
To differentiate between BM vs. AM (@ [Glucose] threshold 2.5mmol/l i.e. 45mg/dL)
OR=6 (95% CI 2 to 17)
Procalcitonin diagnostic accuracy at >= 0.5ng/ml:
Sens=89%
Spec=89%
C reactive protein diagnostic statistics at a threshold of >=20mg/l:
Sens= 91%
Spec= 71%
Blood Procalcitonin mean/median/[range] ng/ml
BM=20.5/9.1/[0.2-107] AM=0.3/0.2/[0.1-4.4] p<10-6
To differentiate between BM vs. AM (PCT threshold 0.5ng/ml)
OR = 64 (95% CI 12 to 452)
Blood CRP mean/median/[range] mg/l
BM=190/178/[8.5-426] AM=18.6/8/[3-213]
p<10-6
To differentiate between BM vs. AM (CRP threshold 20 mg/L)
OR=24 (95% CI 5 to 155)
Blood white blood cell count mean/median/[range] cells/ mm3
BM=18,495 /18,400/ [2400-43,200]
AM=12,031 /10,600 / [2,00-67,200]
p=0.01.
To differentiate between BM vs. AM (WBC count threshold 15000/ mm3)
OR=7 (95% CI 3 to 22)
Neutrophil count mean/median/[range] cells/ mm3
BM=13,748/14,245/[740-36,290]
AM=8,403/7,300/[1,180-51,740]
p=0.06
To differentiate between BM vs. AM (neutrophil count threshold 10000/ mm3)
Sensitivity=60%
Specificity=71%
OR=4 (95% CI 1 to 11)
C reactive protein diagnostic statistics at a threshold of >=20mg/l:
Sens= 91%
Spec= 71% / Country: France
Funding: Not reported
BenGershom 1986 81 / Prospective cohort study
EL=II / Aim: to prospectively study the diagnostic value of CSF [CRP] in neonatal and childhood meningitis / N=45
BM=20 (inc 3 neonates)
VM=25 / 42 children (aged 1 month to 13 years) and 3 neonates referred to a US hospital with suspected viral or bacterial meningitis and for whom a sufficient residual CSF sample was available for CRP testing.
Exclusion criteria: Pre-existent hydrocephalus and traumatic LP. / 17/20 patients had CSF culture proven bacterial meningitis
21/25 had CSF culture proven viral meningitis. / Blood & CSF test results and other study data prospectively recorded for study including:
CSF white cell count; CSF protein, CSF glucose concentration. / CSF white cell count
to differentiate BM vs. VM: WBC count threshold >500 cells/ul, sens 15/17 88%, spec 18/25 72%, PPV 15/22 68%, NPV 18/20 90%
CSF protein
to differentiate BM vs. VM: [Protein] threshold >100mg/dl,
sens 16/17 94%, spec 23/25 92%, PPV 16/18 89%, NPV 23/24 96%
CSF glucose
to differentiate BM vs. VM: [Glucose] threshold <2.2mmol/l,
sens 8/17 47%, spec 22/23 96%, PPV 8/9 89%, NPV 22/31 71% / Country: Netherlands
Funding: Not reported
Sormunen 1999 66 / Retrospective cohort study
EL=III / Aim: to clarify the extent to which Gram negative bacterial meningitis can be distinguished from viral meningitis by assessment of CSF and blood indices in children over 3m of age / N=237
BM=55 (23/55 Hib, all gram negative)
VM=182 / Children aged 3 months to 15 years with either bacterial meningitis (recruited 1984 to 1991 into 2 large Finnish studies) or with a diagnosis of viral meningitis (given by one of three Finnish hospitals during 1977 to 1992)
Exclusion criteria for BM patients: immunocompromise, prosthetic device (e.g. a shunt) or receipt of more than one parenteral dose of antimicrobials prior to diagnosis / BM diagnosis by positive CSF culture of Gram stain negative bacteria
VM or “presumed VM” diagnosis by VM diagnosis @ discharge, no parenteral antibiotics being given during hospitalisation, encephalitis not diagnosed and CSF WBC count >5x 106/l / Results were retrospectively extracted from patient notes including:
CSF white cell count; CSF protein, CSF glucose concentration. / Blood CRP
mean (SD) mg/l
BM=115 (63)
VM=<20
p<0.0001.
To differentiate between BM vs. VM
@ CRP threshold >20.0 mg/l
sens 96%, spec 93%, PPV 83%, NPV 99%
@ CRP threshold >40.0 mg/l
sens 86%, spec 100%, PPV 100%, NPV 95%
Blood white blood cell count
mean (SD) cells x 109/l
BM=18.0 (8.1)
VM=10.6 (4.3)
p<0.0001
To differentiate between BM vs. VM
@ WBC threshold 15.0 (x109/l)
sens 62%, spec 85%, PPV 58%, NPV 87%
@ WBC threshold 20.0 (x109/l)
sens 32%, spec 97%, PPV 79%, NPV 82%
@ WBC threshold >25.0 (x109/l)
sens 20%, spec 100%, PPV 100%, NPV 80%
CRP level (mg/l, mean):
BM= 16.3 (median= 11.1, SD= 21.8, range= 1.4-85.3)
VM= 3.8 (median= 3.3, SD= 1.8, range= 1.3-9.4)
P<10-4
CSF white cell count
mean (SD) cells x 106
BM=4540 (4040)
VM=240 (310)
p<0.0001.
To differentiate between BM vs. VM
@ WBC threshold >500 (x106/l)
sens 78%, spec 89%, PPV 69%, NPV 93%
@ WBC threshold >1000 (x106/l)
sens 75%, spec 97%, PPV 89%, NPV 93%
@WBC threshold >2000 (x106/l)
sens 64%, spec 99%, PPV 97%, NPV 90%
CSF protein
mean (SD) mmol/l
BM=1.88 (1.5)
VM=0.52 (0.24)
p<0.0001
To differentiate between BM vs. VM
@ CSF [protein] threshold >1.0 g/l
sens 64%, spec 96%, PPV 84%, NPV 88%
@ CSF [protein] threshold >1.5 g/l
sens 50%, spec 99%, PPV 96%, NPV 85%
CSF glucose
mean (SD) mmol/l
BM=2.9 (1.6)
VM=3.3 (0.6)
p<0.1
To differentiate between BM vs. VM
@ CSF [glucose] threshold <2.0 mmol/l)
sens 31%, spec 100%, PPV 100%, NPV 79%
@ CSF [glucose] threshold <2.5 mmol/l)
sens 35%, spec 96%, PPV 79%, NPV 79%
CRP level (mg/l, mean):
BM= 16.3 (median= 11.1, SD= 21.8, range= 1.4-85.3)
VM= 3.8 (median= 3.3, SD= 1.8, range= 1.3-9.4)
P<10-4 / Country: Finland
Funding: Not reported
Baker 1989 82 / Retrospective cohort study
EL=III / Aim: to test the hypothesis that atypical lymphocyte presence in the CSF correlates with viral meningitis and that band forms correlate with bacterial meningitis / N=95 children BM = 54 (mean age 24 months, range 2 days to 17 years) (36/54 Hib)
VM=41 (mean age 15 months, range 3 days to 5 years) / Children aged 2 days to 17 years admitted to a US hospital with CSF confirmed viral or bacterial meningitis during 1985 to 1986
Exclusion criteria: CSF culture of a contaminant organism with no clinical evidence of meningitis on follow up or traumatic LP. / Viral or bacterial meningitis was confirmed by CSF culture / Review of hospital records. Outcomes include:
CSF white cell count. / CSF white cell count median (range) (cells/ mm3) BM=2500 (2 - 48,180)
VM=167 (2 - 1990)
p<0.001
Absolute PMNs/mm3
BM=2289 (1-40,950)
VM=32 (0-1791)
P<0.001
Absolute lymphocytes/mm3
BM=72 (0-6232)
VM=24 (0-546)
NS
Absolute monocytes/mm3
BM=48 (0-2992)
VM=28 (0-1170)
NS
To differentiate between BM vs. VM (@ WBC count threshold 500/ mm3)
sens 83%, spec 78%, PPV 83%, NPV not recorded / Country: US
Funding: Not reported
Jaeger 2000 83 / Retrospective cohort study
EL=2- / Aim: to validate a diagnosis model that provides a figure for the probability of meningitis based on 4 CSF and blood parameters / N=84 children BM=48
VM=36 / Children between 1 month - 3.5 years admitted to a French hospital during 1984-1996 and whose final diagnosis was acute meningitis.
Exclusion criteria: tuberculous meningitis, neoplastic meningitis, meningitis secondary to neurosurgery / BM defined by bacterial CSF culture or bacterial antigen in CSF, blood or urine by latex agglutination.
VM defined by viral culture from a blood, stool or CSF sample or discharge diagnosis of viral meningitis, and no aetiology other than viral infection was found and no antibiotics were given for meningitis treatment. / Review of medical records. Outcomes:
CSF white cell count; CSF protein, CSF glucose concentration. / CSF white cell count
mean (SD)(106/l)
BM=3445 (4131)
VM=305 (329)
p<0.0001
CSF PNM count (106/l) mean (SD)
BM=3152 (3976)
VM=49 (75)
P<0.0001
CSF protein
mean (SD) g/l
BM=1.5 (1.0)
VM=0.4 (0.2)
p<0.0001
CSF glucose
mean (SD) mmol/l
BM=1.6 (1.3)
VM=3.2 (0.7)
p<0.0001 / Country: France
Funding: Not reported
Bonsu 2003 60 / Prospective cohort study
EL=II / Aim: to validate the decision making that relies upon the peripheral blood WBC count alone to perform LP and compare this test to CSF WBC among young infants evaluated for infection / N=5375 infants
BM=22
Control (Children with fever evaluated for serious bacterial infection) =5353 / Infants aged 3 to 89 days with fever evaluated for serious bacterial infection in a US emergency department between 1992 and 1999
Exclusion criteria: CSF contaminated with Red Blood Cells (RBCs) >10000 cells/mm3 or children with leukaemia / BM defined by bacterial CSF culture or CSF culture of gram negative rods in neonates. / Prospective data entry into a database at time of evaluation in the emergency department.
Outcomes:
CSF white cell count; CSF protein, CSF glucose concentration. / CSF white cell count
median (IQ range) (1000 cells/ mm3)
BM=79 (8 -1315)
Control=3 (2 - 6)
p<0.0001.
Area Under Curve (AUC) for CSF WBC count =0.82.
To differentiate between BM vs. Control
at CSF WBC count threshold ≥8 cells/ mm3
sens 77%, spec 79%, PPV 1.5%, NPV 99.9%.
at CSF WBC count threshold ≥ 10 cells/ mm3
sens 73%, spec 83.5%, PPV 1.8%, NPV 99.9%.
at CSF WBC count threshold ≥100 cells/ mm3
sens 41%, spec 96%, PPV 4.2%, NPV 99.8%
at CSF WBC count threshold ≥ 1000 cells/ mm3
sens 23%, spec 99.3%, PPV 13.9%, NPV 99.7%.
Median WBC count:
BM= 10,000 (4000 to 15,200)
Control= 11,200 (8,500 to 14,600)
P=0.26
WBC count <5000 cells/μl:
BM= 3.2 (2.3 to 4)
Control= 4.2 (3.7 to 4.6)
P=0.005
Sens= 31.9%
Spec= 96%
PPV= 1.0%
NPV= 99.7%
WBC count ≥ 5000 cells/μl:
BM=13.3 (9.9 to 17.1)
Control= 11.4 (8.8 to 14.8)
P=0.13
WBC count ≥10,000 cells/mm3:
Sens= 50%
Spec= 38.4%
PPV= 0.3%
NPV= 99.5%
WBC count ≥15,000 cells/mm3:
Sens= 27.2%
Spec= 77.1%
PPV= 0.5%
NPV= 99.6%
WBC count ≥20,000 cells/mm3:
Sens= 4.5%
Spec= 93.0%
PPV= 0.0%
NPV= 99.6% / Country: US
Funding: Not reported
Bonadio
1989 84 / Retrospective cohort study
EL=III / Aim: to determine whether the CBC differential ratio has utility in differentiating neonates with bacterial meningitis from those with non-bacterial meningitis / N=72 neonates
BM=18
VM=13
AM=41 / Neonates (under 4wks old) born at term who received a LP at a US hospital (1985 – 1988) and whose CSF contained pleocytosis or was culture positive for pathogenic organisms.
Exclusion criteria: immunodeficiency, antibiotic therapy within 72h of evaluation*, traumatic LP*
*Unless the CSF culture was positive for a bacterial pathogen / BM defined by bacterial CSF culture or positive CSF antigen latex agglutination test
VM confirmed by CSF, stool or throat sample culture
AM defined by negative CSF viral culture for 6/41 neonates. No further information provided / Retrospective view of admission records / CSF white cell count
WBC >22mm3
BM=15/18 (83%)
VM=13/13 (100%)
AM=41/41 (100%)
CSF protein
[Protein] >170mg/dl
BM=10/18 (55%)
VM=0/13 (0%)
AM=0/41 (0%)
To differentiate BM vs. Non-BM
sens 55.6%, spec 100%, PPV 100%, NPV 87.1%
CSF glucose
[Glucose] <34mg/dl
BM=11/18 (61%)
VM=7/13 (54%)
AM=7/41 (17%)
To differentiate BM vs. Non-BM
sens 61.1%, spec 74.1%, PPV 44%, NPV 85.1%
Blood CBC Differential ratio Blood CBC DR <1
BM=18/18 (100%)
VM=3/13 (23%)
AM=5/41 (12%)
To differentiate BM vs. Non-BM
sens 100%, spec 81.8%, PPV 69.2%, NPV 100% / Country: US
Funding: Not reported