To Tap or Not to Tap: High Likelihood of Meningitis Without Sepsis Among Very Low Birth Weight Infants
Context. Neonatal meningitis is associated with significant morbidity and mortality. We speculated that meningitis may be underdiagnosed among very low birth weight (VLBW) infants because of the failure to perform lumbar punctures (LPs) in infants with suspected sepsis.
Objective. This study was undertaken to review the epidemiology of late-onset meningitis in VLBW (401–1500 g) infants and to evaluate the concordance of cerebrospinal fluid (CSF) and blood culture (BC) results.
Methods. VLBW infants (excluding those with intraventricular shunts) born at centers of the National Institute of Child Health and Human Development Neonatal Research Network from September 1, 1998, through December 31, 2001, were studied. Late-onset meningitis was defined by culture-based criteria and classified as meningitis with or without associated sepsis. Unadjusted comparisons were made using χ2 tests and adjusted comparisons using regression models.
Results. Of 9641 VLBW infants who survived >3 days, 2877 (30%) had ≥1 LPs, and 6056 (63%) had ≥1 BC performed after day 3. One hundred thirty-four infants had late-onset meningitis (1.4% of all patients; 5% of those with an LP). Pathogens associated with meningitis were similar to those associated with sepsis. One third (45 of 134) of the infants with meningitis had negative BCs. Lower gestational age and prior sepsis increased risk for meningitis. Compared with uninfected infants, those with meningitis had a longer time on mechanical ventilation (28 vs 18 days), had longer hospitalizations (91 vs 79 days), were more likely to have seizures (25% vs 2%), and were more likely to die (23% vs 2%).
Conclusions. Meningitis is a serious complication among VLBW infants, associated with increased severity of illness and risk of death. Of note, one third of the infants with meningitis had meningitis in the absence of sepsis. Because CSF cultures were performed only half as often as BCs, this discordance in blood and CSF culture results suggests that meningitis may be underdiagnosed among VLBW infants.
Neonatal meningitis is an uncommon but serious disorder with high rates of morbidity and mortality.1–5 The importance of a lumbar puncture (LP) as part of the diagnostic evaluation of the neonate with suspected sepsis has been the subject of debate, and clinical practice varies.5–10 Neonatologists often defer the LP when evaluating the very low birth weight (VLBW) infant. Reasons stated include the perceived low risk of meningitis versus the risk of the procedure in the often unstable VLBW patient. This study was undertaken to review the epidemiology of late-onset meningitis (LOM) in VLBW infants (401–1500 g) cared for at the centers of the National Institute of Child Health and Human Development (NICHD) Neonatal Research Network and to evaluate the concordance of blood cultures (BCs) and cerebrospinal fluid (CSF) cultures.
We suspected that there would be center differences in the frequency with which an LP was performed and speculated that meningitis may be underdiagnosed in VLBW infants because of the failure to routinely perform LPs in VLBW infants with suspected sepsis.
The NICHD Neonatal Research Network is a consortium of tertiary neonatal centers. The network maintains a registry of all VLBW infants born and/or admitted to participating centers within 14 days of birth. Trained research nurses collect maternal data soon after birth and infant data until 120 days, discharge, or death.11 The registry includes data on late-onset sepsis and meningitis, infecting organisms, maternal demographics, pregnancy, labor and delivery complications, infant characteristics, morbidities, and final status. In September 1998, infection surveillance was expanded to include results of all BCs and CSF cultures, antibiotic therapy, and selected risk factors for infection. For this study, we excluded infants with intraventricular shunts and CSF cultures obtained from daily LPs or indwelling reservoirs that were performed only to treat posthemorrhagic hydrocephalus.
LOM was defined by a CSF culture positive for bacteria or fungi performed after 72 hours of age. Data on other CSF parameters (cells, glucose, protein, and Gram stain) were not collected. CSF cultures positive for organisms generally considered to be contaminants were considered negative for this analysis. BCs taken within 7 days of each positive CSF culture were reviewed to classify meningitis further as: 1) meningitis without sepsis (2 positive CSF cultures for the same organism [within a week of each other] or a single positive CSF culture for an organism other than coagulase-negative staphylococcus [CONS] and BC negative for that organism); 2) meningitis with sepsis (CSF and BC positive for the same organism [included some cases in which the CSF was positive for 2 organisms and BC positive for at least 1 of the organisms]); or 3) probable contaminant (CONS in a single CSF culture and BC negative for CONS; a single CSF culture positive for multiple organisms and associated BC negative; or a single positive CSF culture thought to be a contaminant by the study site [ie, not treated]). All cases of meningitis (excluding probable contaminants) were used to evaluate risk factors, infecting pathogens, and outcome.
Gestational age (GA), intraventricular hemorrhage (IVH), periventricular leukomalacia (PVL), and necrotizing enterocolitis (NEC) were assessed as described in previous network publications.11,12 The participating centers and the numbers of patients that each contributed to the study are listed in the Appendix 1.
BCs and CSF cultures were processed by the clinical microbiology laboratories at each academic center using standard methods. BCs were processed using either the Bactec (Becton Dickinson, Sparks, MD) or BacT/Alert (Organon Teknika, Durham, NC) systems; 0.5 to 1 mL of blood was inoculated per BC.
A binary variable was created to indicate infants who were and were not evaluated by CSF culture after day 3 of life, and statistical significance for comparisons of this variable across groups was determined by the χ2 test. Most analyses focused on the subset of patients who were evaluated by CSF culture. LOM was treated as a binary variable. Statistical significance for unadjusted comparisons (eg, the variation in incidence of meningitis across study centers) was determined by using the χ2 test. Possible risk factors for meningitis were evaluated in logistic regression models fit to the binary outcome, and statistical significance of covariates was determined by Wald χ2 tests. Characteristics included in the model were study center, GA, birth weight, race, sex, prior IVH grade 3 or 4, PVL, and prior sepsis. Infants with meningitis were considered to have prior IVH or prior sepsis if the diagnosis date was on or before the date of the positive CSF culture. Logistic regression models were used also to evaluate the relationship between meningitis (and other infection groups) and risk of seizures and death after adjusting for study center and GA. Mean days on mechanical ventilation, time to attain full feeds, and length of hospital stay were compared among infection groups using linear regression models, which included study center and GA in addition to the infection group. P values < .05 were considered significant. Analyses were completed by using SAS software (SAS Institute, Cary, NC).13
Populations Studied and Incidence of Meningitis
Between September 1, 1998, and December 31, 2001, 11 028 VLBW infants were born and/or cared for at network centers, and 9828 infants survived at least 72 hours. After excluding 187 infants with shunts (2%), 9641 were included in the analyses. Although 6056 of the 9641 infants (63%) had at least 1 BC performed after day 3 of life, only 2877 (30%) infants had an LP performed. Almost all infants who had an LP (98%) had an associated BC. The average age at the time of the first LP was 22 days (median: 16 days; range: 4–120 days).
Among infants with a BC, the percentage that also had at least 1 CSF culture after day 3 varied significantly by clinical center, ranging from 22% to 85% (P < .001). Infants with a positive BC were significantly more likely than those with negative BCs to have an LP. Among those who had a positive BC after day 3, infants ≥25 weeks old and >750 g were most likely to have an LP performed (Table 1). Infants with seizures were more likely to have an LP performed; 54% of those with seizures had an LP, compared with 29% of those without seizures (P < .001).
Two hundred twenty-eight infants had a positive CSF culture after day 3; 94 infants whose results were considered probable contaminants were not counted as having meningitis. These 94 infants had 96 CSF cultures: 85 positive for CONS/negative BC; 2 positive for bacillus; 1 positive for aspergillus; and 8 positive for multiple organisms. Thus, using strict criteria, 134 infants had meningitis. This represents 1.4% of all infants (134 of 9641) and 5% of patients evaluated with a CSF culture after day 3 (134 of 2877).
No significant differences were found in the percentage of infants with meningitis by clinical center in the subgroup of infants evaluated with a CSF culture (Fig 1). Infants with a positive BC were significantly more likely to have meningitis than those with a negative BC (7.2% vs 1.5%; P < .001).
The majority of infants had 1 episode of meningitis (only 7 of 134 [5%] had 2 episodes). The average age for the first episode was 26 days (median: 19 days; range: 4–102 days). Sixty-seven percent of the infants (90 of 134) had a repeat LP performed. Of the 44 patients without a follow-up LP, 11 (25%) died within 6 days of their initial culture. For most patients (70 of 90 [78%]), the follow-up LP was performed within 10 days of the original LP. Of the repeat LPs, 11 of 90 (12%) were positive. The same organism grew on the original and repeat cultures for 10 of these patients despite treatment with presumably appropriate antimicrobial agents in 9 cases. One patient with Candida albicans meningitis on both the original and repeat culture was not begun on antifungal therapy until after the second positive culture.
One third of those with meningitis (45 of 134) had a positive CSF culture without an associated positive BC, and two thirds (89 of 134) had meningitis with sepsis (Table 2). Of note, none of the 45 patients with meningitis/no sepsis had an earlier episode of sepsis with the organism causing meningitis. Because CONS may be a contaminant, if we exclude cases with CONS in the CSF, the percentage of infants who had meningitis without sepsis is even greater (47% vs 34%). For the vast majority of infants, the BC and LP were closely linked. For 87 of the 134 (65%) infants with meningitis, the positive CSF culture was taken on the same day as the associated BC; for 37 patients (28%), the LP was performed 1 to 3 days after the BC. Among these 124 infants with meningitis and closely linked BCs and CSF cultures, the percentages with and without sepsis were similar to those overall: 83 with sepsis (67%) vs 41 without sepsis (33%).
Gram-positive agents were the most frequent organisms (63% of first episodes), with CONS accounting for 29% of first episodes. Gram-negative pathogens accounted for 19% and fungal agents for 18% of cases (Table 3). Among 89 patients with meningitis/sepsis, CONS (44%), Staphylococcus aureus (10%), and group B streptococci (9%) were the most common Gram-positive agents. Escherichia coli (6%) and Klebsiella (6%) were the most frequent Gram-negative organisms, and C albicans (9%) was the most frequent fungal agent. Among the 45 infants with meningitis alone, Enterococcus species (27%), C albicans (22%), and Streptococcus viridans (16%) were the most frequent organisms.
Risk Factors for Meningitis
Selected neonatal characteristics and the risk of meningitis among infants evaluated by LP are presented in Table 4. No association was found between meningitis and an infant’s race, sex, birth weight, prior IVH, or PVL. However, lower GA was associated with greater risk of meningitis, as was prior sepsis (7% vs 2%; odds ratio [OR]: 2.9; 95% confidence interval [CI]: 1.9–4.5; P < .001).
For the purpose of evaluating hospital outcomes, infants who had an LP performed were classified into the following groups: 1) LOM (n = 134); 2) culture-confirmed sepsis but no meningitis and no NEC (n = 1263); 3) NEC with or without culture-positive sepsis and no meningitis (n = 329); 4) culture-negative clinical infection alone (n = 637); and 5) uninfected (no meningitis, no culture-positive or clinical sepsis, and no NEC [n = 514]). Seizures were significantly more common in infants with LOM (25%) than in infants with sepsis alone (9%) or who were uninfected (2%) (P < .001). Adjusting for GA and study center, infants with meningitis were nearly 15 times more likely to have seizures (OR: 14.8; 95% CI: 6.5–33.8) than infants who were uninfected. Compared with those who were uninfected, infants with meningitis spent more time on mechanical ventilation (28 ± 1.9 vs 18 ± 1.0 days), took longer to attain full feeds (31 ± 1.5 vs 21 ± 1.8 days), and had longer hospital stays (91 ± 3.9 vs 79 ± 2.1 days) (each: P < .001).
Although “too sick to tap” and “risk of complications” are cited often as reasons to defer the LP, we found no difference in the risk of death between infants who did and did not have an LP (284 of 2877 [10%] vs 661 of 6764 [10%]). However, meningitis increased the risk of death substantially. Overall, 23% (31 of 134) of infants with meningitis died, compared with 9% (253 of 2743) of those who had an LP performed but did not have meningitis (P < .001). This relationship remained after adjustment for GA and study center (OR for death/meningitis versus no meningitis: 2.4; 95% CI: 1.6-3.8; P < .001). The likelihood of death was increased in all 4 groups of infected patients (Table 5). The proportion of infants with meningitis who died was significantly higher for infants infected with Gram-negative (41%) and fungal (32%) agents than for those with gram-positive (15%) (P = .01) agents. Three of the 31 infants with meningitis who died (10%) did so on the day of the CSF culture, 6 (19%) between 1 and 3 days, 5 (16%) between 4 and 7 days, and 17 (55%) >7 days after the diagnosis of meningitis.
Among VLBW infants, meningitis is associated with acute complications, increased risk of death, and late neurodevelopmental sequelae among survivors.14,15 This study reviews the epidemiology of LOM in VLBW infants cared for at the centers of the NICHD Neonatal Research Network, with emphasis on the concordance of BCs and CSF cultures, rates of infection, infecting pathogens, risk factors for disease, in-hospital morbidity, and mortality. Five percent of the infants evaluated with an LP had meningitis. Lower GA and prior sepsis increased risk of meningitis. One in 4 infants with meningitis had seizures, a clinical sign that could lead to earlier LP and diagnosis.
Perhaps the most important finding of this study is that one third of the patients with meningitis had negative BCs within a week of the CSF culture. The importance of the LP in the evaluation of VLBW infants with suspected late-onset infection remains controversial. Because VLBW infants often suffer respiratory distress and cardiovascular instability, it has been argued that many infants are too ill to undergo an LP. In this study, only one half of all infants who had a BC performed also had an LP. Moreover, there were major practice differences: the percentage of infants who had a BC and an LP performed ranged from 22% to 85% across centers. Although LPs were performed more frequently in infants with positive BCs, only 66% of the infants with culture-confirmed sepsis had an LP performed. We speculate that the failure to perform an LP routinely in infants with suspected late-onset sepsis might result in an underdiagnosis of meningitis.
A possible limitation of this analysis is that there is physician variation in the decision to perform an LP. Because meningitis can be diagnosed only by LP, there may be selection bias in the patients diagnosed with meningitis. Infants with a positive BC were significantly more likely to have an LP performed at all centers. We have no information on why an LP was not performed if a sepsis evaluation was done (ie, if a BC was drawn to rule out sepsis). Better understanding of why physicians do or do not perform an LP and the diagnostic accuracy of factors used to make this decision might lead to improved patient care. However, among those patients who had an LP performed, there was no significant difference across centers in the rate of positive CSF cultures. This finding suggests that, although there are LP practice differences across centers, they probably are not explained by better clinical acumen (ie, some clinicians are not necessarily better at predicting which VLBW infant has meningitis and therefore requires an LP).
A wide range of bacterial and fungal agents was associated with LOM in this cohort. The pathogens were similar to those reported in network patients with late-onset sepsis.12 Although Gram-positive agents were most frequent overall, Gram-negative and fungal agents were more common among patients with meningitis alone than in those with both meningitis and sepsis (24% vs 17% and 27% vs 14%, respectively). One third of infants with meningitis did not have a follow-up LP; 11 of these 44 patients (25%) died within 6 days of the initial culture and were likely very sick. Of concern is the fact that 10 of 90 repeat LPs grew the same organism as the original CSF culture, underscoring the importance of a repeat LP to determine that meningitis is being appropriately treated.
VLBW infants with meningitis were significantly more likely to die (23%), compared with infants with sepsis alone (9%) or those who were uninfected (2%). Infants with Gram-negative or fungal meningitis were at particularly high risk for death (41% and 32%, respectively). Moreover, infants with meningitis were significantly more likely to have seizures, longer time on mechanical ventilation, longer time to reach full feeds, and longer hospital stays than infants who were uninfected.
Neonatal meningitis remains a serious complication among VLBW infants that is associated with increased severity of illness and risk of death. Because at least one third of infants with meningitis had an associated negative BC, these data suggest that LOM may be underdiagnosed in VLBW neonatal intensive care unit patients, many of whom do not have an LP performed because they are considered too sick to tap. These data support the need to perform an LP as part of the diagnostic evaluation of the VLBW infant with suspected late-onset infection. An LP is warranted, because meningitis may occur without a positive BC. Furthermore, the identification of Gram-negative or fungal agents in the CSF may influence the choice of drug and length of therapy. Finally, novel strategies to decrease the risk of meningitis among VLBW infants would reduce short- and long-term neonatal morbidity, the high cost of VLBW infant care, and improve survival.
Financial support was provided by National Institutes of Health grants U10 HD27851, U01 HD36790, U10 HD21364, U10 HD34216, U10 HD27871, M01 RR 06022, U10 HD27856, M01 RR 00750, U10 HD27853, M01 RR 08084, U10 HD34167, M01 RR 02635, M01 RR 02172, M01 RR 01032, U10 HD21373, U10 HD27904, U10 HD21397, U10 HD21415, U10 HD21385, U10 HD40689, U10 HD27880, M01 RR 00070, U10 HD27881, and M01 RR 00997.
We thank Mazie Tinsley for manuscript preparation.
- Received July 24, 2003.
- Accepted October 1, 2003.
- Address correspondence to Barbara J. Stoll, MD, Department of Pediatrics, Emory University School of Medicine, 2040 Ridgewood Dr, NE, Atlanta, GA 30322
- ↵Davies PA, Rudd PT. Neonatal Meningitis. London, United Kingdom: MacKeith Press; 1994:1–177
- Holt DE, Halket S, de Louvois J, Harvey D. Neonatal meningitis in England and Wales: 10 years on. Arch Dis Child Fetal Neonatal Ed. 2001;84 :F85– F89
- de Louvois J. Acute bacterial meningitis in the newborn. J Antimicrob Chemother. 1994;34(suppl A) :61– 73
- ↵Hristeva L, Booy R, Bowler I, Wilkinson AR. Prospective surveillance of neonatal meningitis. Arch Dis Child. 1993;69 :14– 18
- ↵Hack M, Horbar JD, Malloy MH, Tyson JE, Wright E, Wright L. Very low birth weight outcomes of the National Institute of Child Health and Human Development Neonatal Network. Pediatrics. 1991;87 :587– 597
- ↵Stoll BJ, Hansen N, Fanaroff AA, et al. Late-onset sepsis in very low birth weight neonates: the experience of the NICHD Neonatal Research Network. Pediatrics. 2002;110 :285– 291
- ↵SAS Institute. SAS/STAT User’s Guide, Version 8. Cary, NC: SAS Institute Inc; 1999
- ↵Doctor BA, Newman N, Minich NM, Taylor HG, Fanaroff AA, Hack M. Clinical outcomes of neonatal meningitis in very-low birth-weight infants. Clin Pediatr (Phila). 2001;40 :473– 480
- ↵Klinger G, Chin CN, Beyene J, Perlman M. Predicting the outcome of neonatal bacterial meningitis. Pediatrics. 2000;106 :477– 482
- Copyright © 2004 by the American Academy of Pediatrics