Abstract
OBJECTIVE. We evaluated the diagnostic utility of the presence and number of cerebrospinal fluid (CSF) bands in distinguishing bacterial from aseptic meningitis among children with CSF pleocytosis.
METHODS. We identified retrospectively a cohort of children 29 days to 19 years of age with CSF pleocytosis (≥10 × 106 leukocytes per L) who were treated in the emergency departments of 8 pediatric centers between January 2001 and June 2004 and whose CSF was evaluated for the presence of bands. We performed bivariate and multivariate analyses to determine the ability of CSF bands to distinguish bacterial from aseptic meningitis.
RESULTS. Among 1116 children whose CSF was evaluated for the presence of bands, 48 children (4% of study patients) had bacterial meningitis. Bacterial meningitis, compared with aseptic meningitis, was associated with a greater CSF band proportion (0.03 vs 0.01; difference: 0.02; 95% confidence interval: 0.00–0.04) and CSF absolute band count (392 × 106 cells per L vs 3 × 106 cells per L; difference: 389 × 106 cells per L; 95% confidence interval: −77 × 106 cells per L to 855 × 106 cells per L). In addition, 29% of patients with bacterial meningitis, compared with 18% of patients with aseptic meningitis, had any bands detected in the CSF. After adjustment for other factors associated with bacterial meningitis, however, CSF band presence, CSF absolute band count, and CSF band proportion were not independently associated with bacterial meningitis.
CONCLUSION. In this multicenter study, neither the presence nor quantity of CSF bands independently predicted bacterial meningitis among children with CSF pleocytosis.
The widespread introduction of highly effective conjugate vaccines against Haemophilus influenzae type b1,2 and Streptococcus pneumoniae3–7 has reduced the probability that a child with cerebrospinal fluid (CSF) pleocytosis will have bacterial meningitis. However, CSF pleocytosis in a pediatric patient still raises concern regarding the possibility of bacterial meningitis. Although published evidence is lacking, some clinicians think that the presence of immature neutrophils (bands) in the CSF increases the risk of bacterial meningitis and, as a consequence, they may treat patients with CSF bands differently from those without CSF bands. One published reference states that CSF characterized by the “presence of immature polymorphonuclear leukocytes is suggestive of bacterial infection.”8 Previous studies identified increased risk of bacterial meningitis with peripheral blood absolute band counts (ABCs) of >0.5 × 109 cells per L,9 peripheral blood band/neutrophil ratios of >0.11,10 or peripheral blood immature neutrophil/total neutrophil ratios of >0.12.11
We sought to determine whether the presence or quantity of bands in the CSF would predict bacterial meningitis reliably. To this end, we identified, from our larger, multicenter meningitis cohort,12 the subset of children whose CSF evaluations included determination of CSF bands. The purpose of this study was to evaluate the diagnostic utility of the presence and number of bands in the CSF in distinguishing bacterial from aseptic meningitis among children with CSF pleocytosis who were evaluated in pediatric emergency departments (EDs).
METHODS
Details of the study design were described previously.12 The Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics endorsed this study protocol, and the institutional review boards at the participating institutions approved this study and granted permission for data sharing. For this study, we included the 8 EDs (of the 20 participating in the larger cohort study) whose laboratories reported the presence and quantity of bands in CSF differential counts.
Briefly, we reviewed the medical records of all patients 29 days to 19 years of age who were diagnosed as having CSF pleocytosis in the EDs of the study institutions between January 2001 and June 2004. We identified study patients by using International Classification of Diseases, Ninth Revision codes. Because the original study of 3295 patients with CSF pleocytosis was performed to create and to validate a decision rule to identify children with bacterial meningitis presenting to a participating ED, we excluded patients who would require hospital admission regardless of the risk of bacterial meningitis (presence of critical illness, purpura, an intracranial shunt device, a neurosurgical procedure in the previous 1 month, immunosuppression, other bacterial infections necessitating inpatient antibiotic therapy, or known active Lyme disease) or who had received any antibiotics within 72 hours before their lumbar punctures. We defined bacterial meningitis on the basis of a CSF culture positive for a bacterial pathogen or CSF pleocytosis (CSF white blood cell [WBC] count of ≥10 × 106 cells per L) in association with a blood culture or CSF latex agglutination test positive for a bacterial pathogen. We defined aseptic meningitis on the basis of the presence of CSF pleocytosis with negative bacterial culture results for blood and CSF.
For this study, we included only the 1116 (34%) of 3295 patients who were evaluated at centers whose hematology technicians examined all CSF specimens for the presence and number of bands, through manual counting. We defined CSF bands as immature neutrophils with nonsegmented nuclei, and we expressed their quantities as ABCs and as proportions of bands in the differential WBC counts. We also collected information on clinical and laboratory findings for all study patients.
We compared proportions by using the χ2 test and means by using the Student's t test, and we determined differences in proportions and means with 95% confidence intervals (CIs). We performed receiver operating characteristic (ROC) analyses and compared the areas under curves (AUCs) for CSF total WBC count, CSF absolute neutrophil count (ANC), CSF band proportion, and CSF ABC with respect to bacterial meningitis.13 We then assessed CSF band presence, CSF band percentage, CSF ABC, duration of fever and other characteristics known to be associated with bacterial meningitis (patient age, peripheral blood WBC count, CSF WBC count, CSF protein level, and CSF glucose level) while adjusting for study center by performing multivariate analyses using a general estimating equation. Patients with missing values for any of the covariates were excluded from the multivariate analyses. All statistical analyses were performed by using SPSS 16.1 (SPSS Inc, Chicago, IL).
RESULTS
There were 1476 children who met the inclusion criteria. We excluded 173 patients (12%) who would have required admission regardless of their risk of bacterial meningitis and 187 (13%) who had received antibiotic pretreatment. Of the 1116 remaining patients with CSF pleocytosis, 48 (4.3%) had bacterial meningitis and 1068 (96%) had aseptic meningitis (similar to the proportions of disease in the original study population, in which 121 of 3295 patients [3.7%] had bacterial meningitis12).
The patients with bacterial and aseptic meningitis differed with respect to age and laboratory characteristics (Table 1). There was a wide range in the reported presence of bands among the 8 centers that evaluated bands in peripheral blood and CSF samples. Bands were noted in 16% to 82% of peripheral blood samples at each center (14%–80% of aseptic meningitis cases and 75%–100% of bacterial meningitis cases) and in 6% to 32% of CSF specimens at each center (4%–32% of aseptic meningitis cases and 0%–100% of bacterial meningitis cases). Overall, patients with bacterial meningitis had a higher mean CSF band proportion and CSF ABC. However, CSF band presence was common for patients with both bacterial and aseptic meningitis, and the overlap was substantial.
Characteristics of the 1116 Study Patients
In the ROC analysis, CSF band proportion and CSF ABC were inferior to CSF total WBC count and CSF ANC as predictors of bacterial meningitis (Fig 1). The AUCs for these cell counts for bacterial meningitis were as follows: CSF ANC, 0.87 (95% CI: 0.81–0.93); CSF total WBC count, 0.82 (95% CI: 0.75–0.90); CSF band proportion, 0.57 (95% CI: 0.48–0.67); CSF ABC, 0.58 (95% CI: 0.48–0.67). The AUCs for CSF total WBC count, CSF ABC, and CSF band proportion were significantly lower than that for CSF ANC. In addition, the AUCs for CSF band proportion and CSF ABC were significantly lower than that for CSF total WBC count. The lower bounds for the 95% CIs of the AUCs for CSF ABC and band proportion included 0.5, as expected for tests without predictive value. The ROC curves provided estimates of optimal cutoff points but also highlighted the limitations of CSF band counts. CSF band proportions of ≥0.04 and CSF ABCs of ≥25 × 106 cells per L had high specificities (93% and 98%, respectively). However, the corresponding sensitivities of CSF band proportion and ABC were only 29%, and sensitivities decreased precipitously for values above the optimal cutoff points.
Bacterial meningitis ROC curves for CSF ABC and band proportion, compared with CSF ANC and total WBC count. AUC values were as follows: CSF ANC, 0.87 (95% CI: 0.81–0.93); CSF total WBC count, 0.82 (95% CI: 0.75–0.90); CSF band proportion, 0.57 (95% CI: 0.48–0.67); CSF ABC, 0.58 (95% CI: 0.48–0.67).
We then determined the independent discriminative ability of the presence of CSF bands, CSF band proportion, and CSF ABC. After adjustment for patient age, peripheral blood WBC, CSF WBC, CSF protein level, CSF glucose level, and center, multivariate analyses revealed that none of the following was independently associated with bacterial meningitis: CSF band presence, CSF band proportion, or CSF ABC (Table 2).
Bivariate and Multivariate Analyses of the Risk of Bacterial Meningitis on the Basis of the Presence and Quantity of CSF Bands
DISCUSSION
In this study, we report the diagnostic performance of the presence of bands in the CSF for identifying bacterial meningitis in children found to have CSF pleocytosis during their ED evaluations. Although bands were present with greater relative frequency and number in the CSF of patients with bacterial meningitis, compared with those with aseptic meningitis, the difference was not useful for clinical decision-making. Substantial proportions of patients with aseptic and bacterial meningitis had CSF bands. Moreover, CSF bands were inferior to CSF ANC in the ROC analysis and failed to emerge as an independent risk factor for bacterial meningitis in multivariate analyses. Previous work with blood smears revealed a similar lack of diagnostic utility of peripheral blood band counts in distinguishing bacterial from viral infections outside the central nervous system.14
Previous investigators demonstrated the diagnostic utility of CSF neutrophil levels in differentiating bacterial from aseptic meningitis in children.9,12,15–19 Several studies reported on the early predominance of CSF neutrophils in children with aseptic meningitis and the changes in CSF neutrophil proportions and ANCs with illness duration among these patients.20,21 However, of the previous investigations that analyzed total ANCs or neutrophil proportions, only 1 reported a combined total of mature and immature cells,20 and none reported on CSF bands separately.
In our study, we noted high interinstitutional variability in the identification and reporting of CSF and peripheral blood band counts. Comparable interinstitutional variability occurs in the reporting of peripheral blood band counts for febrile infants.22 Laboratory technicians tested with slides prepared from single samples23 or with identical photomicrographs24 varied widely in the identification of peripheral blood bands. The maturation of the myeloid series from the band form to the mature neutrophil is marked by segmentation of the nucleus into ≥2 lobes. However, the degree of nuclear constriction required to distinguish a segmented neutrophil from a band ranges from partial (one half to one third of the nuclear width) to complete (leaving only a filamentous strand).25,26 The variability in band reporting among the participating centers in our study represents an important weakness of the use of CSF bands as a discriminator between viral and bacterial meningitis. Of note, 12 of our original participating centers were excluded from this substudy because of laboratory practices of combining band and mature forms in a single CSF neutrophil count.
We recognize several limitations to our study. Our retrospective study depended on the accuracy and completeness of the medical records, with respect to clinical data and laboratory results. Erroneous documentation of times of ED antibiotic administration and lumbar punctures might have resulted in the inclusion of patients with early CSF sterilization, which might be particularly problematic with meningococcal disease.27 Because of our enrollment criteria, we were unable to describe the diagnostic utility of CSF bands when the total CSF WBC count was <10 × 106 cells per L. However, bacterial meningitis is rare in the absence of pleocytosis, even when immature neutrophils are present,28 and usually occurs only in patients with overwhelming sepsis,29 without presenting a diagnostic quandary. Therefore, when neutrophils are present in CSF without pleocytosis, the clinician will likely weigh other clinical and laboratory features more heavily when making medical decisions.
CONCLUSIONS
In this large multicenter study investigating the statistical significance and clinical relevance of bands in the CSF of children with pleocytosis, we found that neither the proportion nor the quantity of CSF bands predicted bacterial meningitis independently. Therefore, the detection of CSF bands does not seem to be useful for decision-making regarding the cause of CSF pleocytosis.
Acknowledgments
This work was supported in part by an Ambulatory Pediatric Association Young Investigator Grant.
We thank our coinvestigators in the multicenter meningitis study group who assisted with data abstraction and those at centers that did not evaluate patients for the presence of CSF bands, as follows: Dewesh Agrawal, MD (Children's National Medical Center and School of Medicine, George Washington University), Subhankar Bandyopadhyay, MD (Children's Hospital of Wisconsin and Medical College of Wisconsin, Children's Healthcare of Atlanta at Scottish Rite, and Mayo Clinic Jacksonville, Mayo School of Graduate Medical Studies), Bema Bonsu, MD (Columbus Children's Hospital and Ohio State University), Blake Bulloch, MD (Phoenix Children's Hospital and University of Arizona College of Medicine), Jennifer Chapman, MD (Columbus Children's Hospital and Ohio State University), Peter Dayan, MD, MSc (Morgan Stanley Children's Hospital of New York-Presbyterian Hospital and College of Physicians and Surgeons, Columbia University), Paul Ishimine, MD (Rady Children's Hospital San Diego and School of Medicine, University of California, San Diego), Patrick Johnston, ScD (Children's Hospital Boston), Ron Kaplan, MD (Children's Hospital and Regional Medical Center and School of Medicine, University of Washington), John Leake, MD (Rady Children's Hospital San Diego and School of Medicine, University of California, San Diego), Charles G. Macias, MD, MPH (Texas Children's Hospital and College of Medicine, Baylor University), Karim Mansour, MD (Children's Hospital and Research Center Oakland and School of Medicine, University of California, San Francisco), R. Ian McCaslin, MD, MPH (Rady Children's Hospital San Diego and School of Medicine, University of California, San Diego), Donna Moro-Sutherland, MD (WakeMed Health and Hospitals and University of North Carolina at Chapel Hill), Robert H. Riffenburgh, PhD (Naval Medical Center San Diego), Robert Schremmer, MD (Children's Mercy Hospital and Clinics and School of Medicine, University of Missouri-Kansas City), Dale Steele, MD (Hasbro Children's Hospital and School of Medicine, Brown University), and Uyen Truong, MD (University of California, Davis, Medical Center and School of Medicine, University of California, Davis, and Children's National Medical Center and School of Medicine, George Washington University).
We thank the members of the Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics for their critical review and endorsement of the study protocol.
Footnotes
- Accepted February 12, 2009.
- Address correspondence to John T. Kanegaye, MD, Division of Emergency Medicine, Rady Children's Hospital San Diego, 3020 Children's Way, MC 5075, San Diego, CA 92123. E-mail: jkanegaye{at}rchsd.org
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject
Clinicians treating pediatric patients with CSF pleocytosis may rely on band forms in the CSF to make assessments regarding the risk of bacterial meningitis. However, evidence to guide the interpretation of CSF band counts is scant.
What This Study Adds
In a large cohort of children with CSF pleocytosis evaluated in EDs, bands were frequently present in the CSF, regardless of viral or bacterial cause. CSF band presence, proportions, and absolute counts lacked independent association with bacterial etiology.
REFERENCES
- Copyright © 2009 by the American Academy of Pediatrics
In this issue
