Neonatal Meningitis: What Is the Correlation Among Cerebrospinal Fluid Cultures, Blood Cultures, and Cerebrospinal Fluid Parameters?
BACKGROUND. Meningitis is a substantial cause of morbidity and mortality in neonates. Clinicians frequently use the presence of positive blood cultures to determine whether neonates should undergo lumbar puncture. Abnormal cerebrospinal fluid (CSF) parameters are often used to predict neonatal meningitis and determine length and type of antibiotic therapy in neonates with a positive blood culture and negative CSF culture.
METHODS. We evaluated the first lumbar puncture of 9111 neonates at ≥34 weeks' estimated gestational age from 150 NICUs, managed by the Pediatrix Medical Group, Inc. CSF culture results were compared with results of blood cultures and CSF parameters (white blood cells [WBCs], glucose, and protein) to establish the concordance of these values in culture-proven meningitis. CSF cultures positive for coagulase-negative staphylococci and other probable contaminants, as well as fungal and viral pathogens, were excluded from analyses.
RESULTS. Meningitis was confirmed by culture in 95 (1.0%) neonates. Of the 95 patients with meningitis, 92 had a documented blood culture. Only 57 (62%) of 92 patients had a concomitant-positive blood culture; 35 (38%) of 92 had a negative blood culture. In neonates with both positive blood and CSF cultures, the organisms isolated were discordant in 2 (3.5%) of 57 cases. In each case, the CSF pathogen required different antimicrobial therapy than the blood pathogen. For culture-proven meningitis, CSF WBC counts of >0 cells per mm3 had sensitivity at 97% and specificity at 11%. CSF WBC counts of >21 cells per mm3 had sensitivity at 79% and specificity at 81%. Culture-proven meningitis was not diagnosed accurately by CSF glucose or by protein.
CONCLUSIONS. Neonatal meningitis frequently occurs in the absence of bacteremia and in the presence of normal CSF parameters. No single CSF value can reliably exclude the presence of meningitis in neonates. The CSF culture is critical to establishing the diagnosis of neonatal meningitis.
- CSF—cerebrospinal fluid
- LP—lumbar puncture
- WBC—white blood cell
- EGA—estimated gestational age
- CBC—complete blood cell count
- IQR—interquartile range
- RBC—red blood cell
Neonatal meningitis is a devastating infection that is often difficult to diagnose.1–5 Signs of meningitis are often subtle in the neonate; thus, the diagnosis of meningitis must be made by cerebrospinal fluid (CSF) examination.6 However, lumbar puncture (LP) is often not performed as part of the initial neonatal sepsis evaluation in the first days of life.7,8 Instead, the decision to perform an LP is often based on the isolation of a potential pathogen from blood cultures. Although basing the decision to perform LP on blood culture results occurs in clinical practice, the diagnostic accuracy of this strategy is unknown.
Maternal antibiotic prophylaxis or delayed LP in antibiotic-treated neonates make interpreting CSF cultures difficult, because CSF cultures may be negative within hours of antibiotic administration.9,10 In these cases of antibiotic pretreatment, clinicians must rely on the CSF parameters to determine the presence of meningitis. The cellular and biochemical abnormalities in the CSF of older patients with bacterial meningitis are present for 44 to 68 hours.11 Similar values for neonates are unknown. Several studies have examined the CSF parameters of the noninfected neonate to define the normal values. These studies, although small and often from a single center, have resulted in the widespread teaching that 21 WBCs per mm3 in the CSF is a normal value.12–14 However, the clinical use of CSF parameters in the identification of subsequent culture-confirmed meningitis has not been evaluated previously.
The objective of this investigation was to evaluate the accuracy of current clinical diagnostic strategies in the identification of meningitis among term or near-term neonates. To accomplish this goal, we examined CSF parameters from >9000 term or near-term neonates in an attempt to develop an algorithm for predicting neonatal meningitis. We compared blood and CSF culture positivity rates and looked for the presence of concordance between these results.
The Duke University Institutional Review Board approved this investigation. We assembled a cohort of neonates from an administrative database. Clinical data for these neonates were recorded prospectively for the database and analyzed retrospectively for this article. Neonates eligible for inclusion in the study were discharged from 150 NICUs managed by the Pediatrix Medical Group, Inc, from 1997 through 2004, were ≥34 weeks in estimated gestational age (EGA), and had an LP performed. The Pediatrix Group cared for 16395 neonates during this time who underwent LP; 9171 neonates were ≥34 weeks' EGA and had a report of a CSF culture. We excluded 60 patients in whom the culture was reported from a ventricular tap or shunt. The total sample size in this analysis was 9111 patients.
The Pediatrix Group database is an administrative database and the method of data collection has been described previously.15–18 We collected data on birth weight, EGA (based on the examination of the neonatologist), gender, race, Apgar scores, admission type, day of life, maternal age, peripheral complete blood count (CBC), blood and CSF culture results, CSF parameters and mortality. CSF data were codified and triple checked for accuracy.
Meningitis was defined by a positive CSF culture for bacteria. CSF cultures positive for organisms generally considered contaminants (coagulase-negative staphylococci, other skin flora [viridans streptococci and diptheroids], or mixed organisms) were not included in the positive CSF results. Blood cultures taken within 3 days (average time for bacterial cultures to be read and organisms identified) of the CSF culture were reviewed for concordance. Blood and CSF cultures were considered concordant if both were positive with the same organism. CSF and blood cultures were considered discordant if the CSF was positive and the blood culture was negative or if the organism isolated from the CSF was a different organisms than the blood isolate.
Only the first CSF result for any patient was included in this investigation. Only blood cultures obtained within 3 days of the LP were considered in the analyses. In the primary analysis, positive blood and CSF cultures and CSF parameters were tabulated. We calculated sensitivity, specificity, and likelihood ratios for CSF parameters to diagnose meningitis. We evaluated different CSF WBC counts as threshold values for a negative test. We conducted the data analysis using Stata 8.1 (Stata Corp, College Station, TX). Secondary analyses using blood cultures obtained within 10 days of the LP, CSF and RBCs <100 cells per mm3, were also conducted and yielded similar results to the primary analysis. Therefore, only the primary analysis is reported.
A total of 9111 infants ≥34 weeks' EGA with no ventricular shunts underwent LP during the study period (Table 1). The mean EGA was 38 weeks (range: 34–44 weeks), and the mean birth weight was 3.16 kg (range: 0.55–5.7 kg); 56.4% of the cohort were male infants, and 43.5% were female. Admission to the NICU was variable, with 41.5% of admissions from the newborn nursery, 35.8% of admissions from the hospital delivery, 3% of admissions from home delivery, and 11.7% of admissions from hospital transfers. Median Apgar scores at 1 and 5 minutes were >7 (78% and 92%, respectively), and the majority of the LPs, 6988 (76.6%) of 9111, were performed in the first 3 days of life.
Of the 9111 LPs performed, 184 CSF cultures were positive for bacterial growth. CSF cultures positive with a mixture of bacterial organisms, coagulase-negative staphylococci, and other skin flora were excluded. Ninety-five (1.0%) of 9111 CSF cultures grew a potential bacterial pathogen (Table 2) and included Gram-positive organisms other than skin flora (62 of 95 [65.3%]) and Gram-negative rods (31 of 95 [33.6%]). Mortality data are available on 8550 patients, and those with culture-confirmed meningitis were significantly more likely to die than neonates with CSF cultures yielding no growth (4 of 73 vs 40 of 8477; relative risk: 11.61; 95% confidence interval: 4.27–31.61).
Blood Culture Discordance
CSF cultures were matched to a blood culture within 3 days of the LP in 8596 (94.1%) infants, and 398 (4.5%) of 8596 blood cultures were positive (Table 2). Of the 95 patients with a CSF culture positive for a bacterial pathogen, 92 (96.8%) had a documented blood culture within 3 days of LP. Of those 92 neonates, the majority 65 (71%) of 92 had blood and LP cultures performed on the same day. The remaining neonates had a time difference between the blood culture and CSF culture of 1 day (21 of 92 [22.8%]), 2 days (4 of 92 [4.3%]), or 3 days (2 of 92 [2.2%]).
Of those neonates with accompanying blood cultures, 57 (62.0%) of 92 patients had a concomitant positive blood culture; 35 (38.0%) of 92 had a negative blood culture. Of those 35 neonates with discordant blood and CSF cultures, 21 (60%) of 35 had a blood culture and LP on the same day, and 11 (31%) of 35 had 1 day between the time of blood and CSF culture. Of the remaining neonates, 3 (9%) of 35 had 2 or 3 days between blood and CSF culture. In those neonates with both positive blood and CSF cultures, the organisms isolated were discordant in 2 (3.5%) of 57, as shown in Table 3. In both discordant pairs, the CSF pathogen required different antimicrobial therapy than the blood pathogen.
We evaluated CSF parameters to determine their predictive value in diagnosing meningitis, and these results are shown in Table 4. Neonates with negative CSF cultures had a range of 0 to 90000 CSF WBCs per mm3, with a median of 6/mm3 (interquartile range [IQR]: 2–15). In neonates with bacterial meningitis, the range of CSF WBCs was from 0 to 15900/mm3, with a median of 477/mm3 (IQR: 38–1950). However, 5% of neonates with bacterial meningitis had either 0 or 1 CSF WBCs per mm3, and 10% of neonates with bacterial meningitis had ≤3 CSF WBCs per mm3.
CSF glucose and protein values were highly variable in infants both with and without meningitis. Neonates without meningitis had CSF glucose values ranging from 0 to 1089 mg/dL, (median: 49 mg/dL; IQR: 43–58). Neonates with bacterial meningitis had CSF glucoses ranging from 0 to 199 mg/dL (median: 20 mg/dL; IQR: 3–55). Infants without meningitis had CSF protein levels ranging from 3 to 4122 mg/dL (median: 103 mg/dL; IQR: 77–142). Infants with bacterial meningitis had CSF protein ranging from 41 to 1964 mg/dL (median: 273 mg/dL; IQR: 125–550). CSF RBC values are likewise highly variable. CSF RBCs ranged from 0 to 4070000 cells per mm3, with a median of 190 RBCs per mm3 (IQR: 12–2250). Infants with bacterial meningitis had a median of 257 RBCs per mm3 (IQR: 26–1400).
We determined the sensitivity, specificity, and likelihood ratios of CSF WBC counts, glucose, and protein to predict the presence of meningitis using different thresholds for these values (Table 4). Highest sensitivity was obtained when the threshold was any presence of WBCs in the CSF (97%), but this also led to the lowest specificity (11%). Using 21 WBCs as the upper limit of the threshold led to a sensitivity of 79% and a specificity of 81%. The CSF glucose values were less sensitive predictors of culture positivity but had higher specificities than CSF WBCs, as shown in Table 4. We attempted to construct an algorithm that could be used to predict meningitis in the absence of a CSF culture. However, given the variability in the CSF parameters and the lack of sensitivity and specificity of traditional threshold values, we were unable to develop an algorithm that would accurately and precisely predict meningitis based on CSF parameters alone. Detailed information regarding the 12 neonates with culture-proven meningitis and CSF WBC counts of ≤21 cells per mm3 is shown in Table 5.
Peripheral WBC data obtained within 3 days of the LP was also analyzed, because the CBC is often used in conjunction with blood culture data to determine whether an LP should be completed. CBC data were available on 8312 (91.2%) of 9111 patients. Of the 8312 neonates with CBC data, 66 (0.8%) had WBC counts of <3000/mm3, 1438 (17.3%) had WBC counts between 3000 and 10000/mm3, 2416 (29.1%) had WBC counts between 10001 and 15000/mm3, and 2156 (25.9%) had WBC counts between 15001 and 20000/mm3. There were 2236 neonates (26.9%) with WBC counts >20000/mm3. The peripheral WBC count was neither sensitive nor specific for bacterial meningitis. There were 85 neonates with a positive CSF culture with a pathogen and with a peripheral CBC within 3 days. Of these neonates, 17 (20%) of 85 had WBC counts of <3000/mm3, 36 (43.4%) of 85 had WBC counts of ≥3000 and <10000/mm3, 13 (15.3%) of 85 had WBC counts of ≥10000 and <15000/mm3, and 9 (10.6%) of 85 had WBC counts of ≥15000 and <20000/mm3. In all of the cases, using peripheral WBCs as a predictor for meningitis had a positive likelihood ratio <1.0.
Prior studies have described the difficulties of diagnosing neonatal meningitis based on clinical examination and laboratory data (such as CBC and blood culture).1,2,6,19–22 These studies have been limited by sample size20–22 and lack of CSF parameters2 or focused on premature neonates.1 Our study focuses on the near-term and term infants, has a large sample size, and includes data on CSF parameters.
The most common pathogens (group B streptococcus and Escherichia coli) we report are similar to those reported by Wiswell et al.2 Similar to previous reports,22–24 we also found that the finding of a positive CSF culture in neonates who have an LP is rare: 10 in 1000. However, these results underestimate the true incidence of disease, because many of the CSF cultures were obtained after antibiotics were started or from neonates born to mothers who had received antibiotics.
We had hoped to identify factors that would allow clinicians to rapidly assess the likelihood of meningitis in neonates based on CSF findings. However, we were unable to do this, because our analysis of CSF parameters confirmed that meningitis can occur in the presence of normal CSF WBC, glucose, and protein levels. Our study results reinforce the finding of others studies that report that a substantial proportion (33% [ref 2] to 53% [ref 25]) of neonates with culture-proven meningitis have negative blood cultures. We suspect that our rate (38%) may be related to an increased use of antibiotics that has been advocated to prevent group B streptococcal disease.25 Even more disturbing are the 2 neonates with Gram-negative rod meningitis who had Gram-positive organisms in blood culture (Table 3). If LP had not been performed in these neonates and presumptive therapy based on the blood culture, these cases would have been missed, or the diagnosis delayed, with likely serious consequences.
Even more problematic, we were unable to find any CSF parameter that could be used to exclude meningitis. The commonly used threshold value of 21 cells as the upper limit of normal for the term neonate12–14 would have lead to 12.6% of meningitis cases being missed. Our analysis of CSF parameters demonstrates that meningitis can occur in the presence of normal CSF WBC, glucose, and protein levels. All of the cases outlined in Table 5 would have been missed without LP results except for the 2 that also had a positive blood culture with the same organisms. Because meningitis occurs in the face of normal CSF parameters, it is impossible to construct an algorithm to predict meningitis based on abnormal CSF values. This finding reinforces the need to perform the LP at the onset of the sepsis evaluation.
The strengths of this study include the large size, the focus on term or near-term infants, and the incorporation of CSF parameters with blood and CSF culture results. The study is limited in that it is a retrospective analysis of an administrative data set and has missing data. Although 95% of patients with meningitis had blood culture data available, 42% of patients with meningitis and 50% overall were missing CSF parameter information, which decreased the sample size and may affect the results. Although data are missing, the conclusions are based on ∼4500 complete observations and provide evidence that neonatal meningitis can be missed if the LP is not performed as a routine part of the sepsis evaluation.
In addition, the study cohort is based on those infants who had a CSF culture obtained rather than those infants who had bacteremia. Therefore, we do not capture those neonates who had a blood culture obtained but did not have an LP performed or those neonates with bacteremia who underwent LP but did not have a CSF culture sent. Despite eliminating the obvious bacterial contaminants, some of the pathogenic organisms isolated from CSF cultures may have been contaminants. However, given the immunocompromised status of neonates and the substantial impact that meningitis has on neurodevelopmental outcomes, failure to treat pathogens isolated from the CSF is ill advised, and these pathogens, therefore, were included in our analysis.
We are left with the dilemma of how to best treat the neonate who was pretreated with antibiotics or born to a mother who was pretreated with antibiotics. Although some authors have suggested that the “asymptomatic” term neonates are at low risk of having meningitis, defining symptomatology to accurately identify the newborn with serious sepsis or meningitis has not been accomplished.24 Symptoms of partially treated meningitis begin to appear 72 hours after the last dose of antibiotics.20 Thus, a short course of antibiotics (per local standard of care) and close follow-up of these infants should be considered. For example, in a pretreated, asymptomatic neonate with a negative blood culture, negative CSF culture, and elevated CSF WBC count (21 cells per mm3), 48 hours of antibiotics21 followed by either in-hospital observation or outpatient scheduled clinic visit 48 to 72 hours after discharge may be warranted.20
Neonatal meningitis remains a substantial cause of sepsis-related morbidity and mortality in the term and near-term infant. Our data demonstrate that there is no set of clinical parameters that excludes the diagnosis of meningitis in a neonate other than CSF cultures. The diagnosis of meningitis is, therefore, dependent on obtaining a timely and adequate culture of the spinal fluid.
Neonatal meningitis occurs in the absence of bacteremia and in the presence of normal CSF values. No single CSF value can be used to exclude meningitis, and peripheral WBC counts are also poor predictors of neonatal meningitis. The CSF culture is critical to the diagnosis, regardless of other laboratory results. These data suggest that an LP should be incorporated in a sepsis evaluation of an infant.
Dr Benjamin received support from National Institute of Child Health and Human Development grant HD044799, and Dr Garges received support from Ruth L. Kirschstein National Research Service Award Training grant T32-HD43029. We thank the Pediatrix Medical Group, Inc, and the Duke Clinical Research Institute for supporting this study. It was completed under a collaborative agreement between Pediatrix Medical Group, Inc, and Duke University.
- Accepted October 3, 2005.
- Address correspondence to Daniel K. Benjamin, Jr, MD, PhD, MPH, Department of Pediatrics, PO Box 17969, Duke Clinical Research Institute, Durham, NC 27715. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Stoll BJ, Hansen N, Fanaroff AA, et al. To tap or not to tap: high likelihood of meningitis without sepsis among very low birth weight infants. Pediatrics.2004;113 :1181– 1186
- ↵Wiswell TE, Baumgart S, Gannon CM, Spitzer AR. No lumbar puncture in the evaluation for early neonatal sepsis: will meningitis be missed? Pediatrics.1995;119 :803– 806
- Klein JO. Bacterial sepsis and meningitis. In: Remington JS, Klein JO, eds. Infectious Diseases of the Fetus and Newborn Infant. 5th ed. Philadelphia, PA: WB Saunders; 2001:943– 998
- Davies PA, Rudd PT. Neonatal Meningitis. London, United Kingdom: MacKeith Press; 1994:1– 177
- ↵Haworth JC. The diagnosis of acute meningitis in infancy. Lancet.1953;1(19):911– 914
- ↵Riordan FAI, Cant AJ. When to do a lumbar puncture. Arch Dis Child.2002;87 :235– 237
- ↵Kanegaye JT, Soliemanzadeh P, Bradley JS. Lumbar puncture in pediatric bacterial meningitis: determining the time interval for recovery of cerebrospinal fluid pathogens after parenteral antibiotic pretreatment. Pediatrics.2001;108 :1169– 1174
- ↵Feigin RD, Pearlman E. Bacterial meningitis beyond the neonatal period. In: Feigin RD, Cherry JD, eds. Textbook of Pediatric Infectious Diseases. 4th ed. Philadelphia, PA: WB Saunders; 1998
- ↵Berman RE, Kliegman RM, Arvin AM. Nelson Textbook of Pediatrics. 17th ed. Philadelphia, PA: WB Saunders; 2004
- Oski FA. Principles and Practice of Pediatrics. 2nd ed. Philadelphia, PA: JB Lippincott; 1994
- ↵Benjamin DK Jr, DeLong ER, Steinbach WJ, Cotton CM, Walsh TJ, Clark RH. Empirical therapy for neonatal candidemia in very low birth weight infants. Pediatrics.2003;112 :543– 547
- ↵Johnson CE, Whitwell JK, Pethe K, Saxena K, Super DM. Term newborns who are at risk for sepsis: are lumbar punctures necessary? Pediatrics.1997;99 (4). Available at: www.pediatrics.org/cgi/content/full/99/4/e10
- ↵Baker CJ, Kanto WP Jr. Implementing new GBS guidelines requires coordinated care. AAP News.2003;22 :79– 86
- Copyright © 2006 by the American Academy of Pediatrics