 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
a Department of Pediatrics, University of Tennessee Health Science Center and Children's Foundation Research Center at Le Bonheur Children's Medical Center, Memphis, Tennessee
b Department of Infectious Diseases, St. Jude Children's Research Hospital, Memphis, Tennessee
c Department of Pediatrics, University of South Florida, Tampa, Florida
d Department of Pediatrics, Kosair Children's Hospital, Louisville, Kentucky
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
METHODS. Medical records of children with pneumococcal meningitis at a single institution from 1991–2001 were retrospectively reviewed. Vancomycin start time was defined as the number of hours from initiation of cefotaxime or ceftriaxone therapy until the administration of vancomycin therapy. Outcome variables were death, sensorineural hearing loss, and other neurologic deficits at discharge. Associations between independent variables and outcome variables were assessed in univariate and multiple logistic regression analyses.
RESULTS. Of 114 subjects, 109 received empiric vancomycin therapy in combination with cefotaxime or ceftriaxone. Ten subjects (9%) died, whereas 37 (55%) of 67 survivors who underwent audiometry had documented hearing loss, and 14 (13%) of 104 survivors were discharged with other neurologic deficits. Subjects with hearing loss had a significantly shorter median vancomycin start time than did those with normal hearing (<1 vs 4 hours). Vancomycin start time was not significantly associated with death or other neurologic deficits in univariate or multivariate analyses. Multiple logistic regression revealed that hearing loss was independently associated with vancomycin start time <2 hours, blood leukocyte count <15000/µL, and cerebrospinal fluid glucose concentration <30 mg/dL.
CONCLUSIONS. Early empiric vancomycin therapy was not clinically beneficial in children with pneumococcal meningitis but was associated with a substantially increased risk of hearing loss. It may be prudent to consider delaying the first dose of vancomycin therapy until 
2 hours after the first dose of parenteral cephalosporin in children beginning therapy for suspected or confirmed pneumococcal meningitis.
Key Words: Streptococcus pneumoniae • vancomycin • meningitis • dexamethasone • hearing loss
Abbreviations: CSF—cerebrospinal fluid
Children with meningitis caused by Streptococcus pneumoniae face a guarded prognosis: in developed nations, 
10% die and one third of survivors are discharged from the hospital with significant neurologic sequelae.1–3 Previous studies have identified a multitude of adverse prognostic indicators in children with pneumococcal meningitis; however, differences across studies in populations, therapeutic approaches, and statistical analyses make synthesis of their results problematic.2–4 Thus, it presently remains unclear which risk factors are independently associated with the outcomes of death, hearing loss, and other neurologic sequelae. Particular uncertainty continues to surround both the use of adjunctive dexamethasone therapy and the optimal antibiotic therapy for children with pneumococcal meningitis.5
During the 1990s, widespread pneumococcal resistance to penicillin and third-generation cephalosporins emerged, and some patients with cefotaxime-resistant pneumococcal meningitis experienced delayed cerebrospinal fluid (CSF) sterilization when treated initially with cefotaxime or ceftriaxone alone.6–10 In view of these developments, in 1997, the American Academy of Pediatrics Committee on Infectious Diseases recommended that patients with "definite or probable bacterial meningitis" should empirically receive combination therapy with vancomycin plus either cefotaxime or ceftriaxone until susceptibility testing results are available.11 The rationale for the inclusion of vancomycin in initial therapy was based on the known association between delayed CSF sterilization and neurologic sequelae in children with bacterial meningitis.12
Even before the Committee on Infectious Diseases issued its recommendation, combination empiric therapy including vancomycin had already become an accepted practice for children with suspected pneumococcal meningitis at our tertiary children's medical center. As we have noted previously, this strategy has effectively prevented bacteriologic therapeutic failures in children with pneumococcal meningitis in our institution: from 1991 to 1999, none of 80 children (30 of whom underwent repeat CSF examinations 24–96 hours into therapy) who received empiric vancomycin therapy suffered documented delayed CSF sterilization.13 Nevertheless, patients in this series suffered adverse outcomes at rates similar to or exceeding those documented elsewhere: 7% died, and 53% of survivors who underwent audiometric testing had moderate to profound sensorineural hearing loss.13 These observations raised 2 questions. First, is early empiric vancomycin therapy clinically beneficial in children with pneumococcal meningitis? Second, how promptly is vancomycin therapy initiated in these children, and would earlier administration improve outcomes?
We undertook the present study to address these questions and to identify risk factors for adverse neurologic outcomes in children with pneumococcal meningitis in the era of early empiric vancomycin therapy. With these ends in mind, we expanded the data set of our previous study and subjected it to multivariate analysis to identify clinical and laboratory characteristics independently associated with the outcomes of death, hearing loss, and other neurologic deficits. In particular, we sought to investigate the relationship between timing of the first dose of vancomycin and the above outcomes after controlling for the effects of other predictive variables.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Data regarding demography, past medical history, clinical and laboratory findings, antibiotic susceptibility, hospital course, therapies administered, and outcome were abstracted from medical records. Audiometric testing results were abstracted from medical records and from the electrophysiology laboratory records of the hospital. The date and time of the initial lumbar puncture were abstracted from medical records, as were the date and time of the first dose of each parenterally administered antibiotic. These data were used to calculate the number of hours from the first dose of any parenteral antibiotic appropriate for pneumococcal meningitis (eg, penicillin G, cefotaxime, and ceftriaxone)14 until the first dose of vancomycin (henceforth called the vancomycin start time).
Data Analysis
Three outcome variables were investigated: death, moderate to profound sensorineural hearing loss, and other neurologic deficits at the time of hospital discharge. The associations of independent variables with each outcome variable were assessed in univariate analyses using Fisher's exact (categorical data) or Wilcoxon rank-sum (continuous data) tests. Hearing loss was defined by the absence of an observable response at 50 dB in 1 ear on audiometric testing. Other neurologic deficits were defined by the presence of functionally significant motor or cranial nerve deficits (other than hearing loss) or global encephalopathy at the time of hospital discharge. The analyses of hearing loss and other neurologic deficits only included subjects in whom these findings could not be attributed to underlying conditions. Penicillin susceptibilities of pneumococcal isolates were interpreted according to current Clinical Laboratory Standards Institute guidelines.15 Abnormal neuroimaging (ie, computed tomography or MRI) studies were defined as those interpreted by a staff radiologist as showing evidence of central nervous system pathology not referable to an underlying condition. Depressed consciousness at admission was defined by documentation of coma; a Glasgow coma score of <14; unresponsiveness; or opisthotonic, decorticate, or decerebrate posturing.
Separate multiple logistic regression models were developed for each outcome variable. Independent variables present in 10% of the study population and with univariate P < .20 were included in initial models. A backward elimination approach was used to yield the most parsimonious yet descriptive model possible. Independent variables with multivariate P < .05 were retained. After a tentative final model was created for each outcome, previously eliminated independent variables were individually forced back into the model and retained if P was then < 0.05. Vancomycin start time was initially included in all of the models regardless of univariate associations. Statistical calculations were performed using Statview (SAS Institute, Cary, NC).
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
Outcomes
Ten subjects died: 1 suffered a full arrest shortly after presentation to the emergency department, and 9 died after initial intensive care unit (ICU) admission and mechanical ventilation. Audiometric testing results were documented for 67 of 104 subjects who survived to hospital discharge. Of these, 37 subjects (55%) had moderate to profound sensorineural hearing loss in 1 ear. Subjects who underwent audiometric testing were less likely to have underlying disorders than were those who were not tested (8 of 67 vs 15 of 37; P = .001); otherwise, clinical characteristics of these groups did not significantly differ (data not shown). Fourteen surviving subjects were discharged from the hospital with functionally significant neurologic deficits other than hearing loss. These subjects' neurologic findings included (number of subjects): persistent vegetative state (2), spastic hemiplegia or diplegia (3), hemiparesis (4), and other motor and/or cranial nerve palsies (6).
Table 2 lists the univariate associations of selected independent variables with outcome variables, with continuous variables dichotomized to simplify the presentation of results. Table 3 presents univariate associations of vancomycin start time (also dichotomized) with outcome variables. As shown, early vancomycin administration was associated with an increased risk of hearing loss, whereas vancomycin start time was not significantly associated with mortality or other neurologic deficits. Among children with hearing loss, the median vancomycin start time was <1 hour (interquartile range: 0–1.5 hours), whereas that of children without hearing loss was 4 hours (interquartile range: 1–12 hours; P = .0005). With increasing vancomycin start time, the proportion of tested children with hearing loss decreased in stepwise fashion: <1 hour, 18 (78%) of 23; 1 to 2 hours, 6 (67%) of 9; 2 to 5 hours, 3 (33%) of 9; >5 hours, 5 (28%) of 18 (P = .006).
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Until now, only limited data have been available regarding the safety and efficacy of empiric vancomycin therapy, administered in combination with a third-generation cephalosporin, in children with pneumococcal meningitis. In 2002, investigators in Australia published their analysis of secular trends and concluded that empiric vancomycin use was not related to the outcome of pneumococcal meningitis.17 Also in 2002, Kellner et al18 reported that, among children with penicillin-nonsusceptible pneumococcal meningitis, those who received empiric vancomycin therapy received fewer mean days of intravenous antibiotics (12 vs 18.5 days; P = .04) but did not demonstrate reductions in fever duration, ICU admission, hearing loss, or any neurologic sequelae. Data from adult patients with pneumococcal meningitis are similarly sparse; however, a retrospective cohort study in West Virginia (1978–1997) found that mortality was increased among adults treated with the combination of vancomycin and cephalosporin (3 of 5) compared with those treated with chloramphenicol (1 of 9) or cephalosporin alone (0 of 9).19 Our study differs from these previous studies in that we used multivariate analysis of patient-level data from a sizeable pediatric population in which empiric vancomycin therapy is routine. Thus, we believe that the present report represents the most complete body of evidence to date concerning the safety and efficacy of early empiric vancomycin therapy in children with pneumococcal meningitis.
The mechanisms by which early vancomycin therapy might increase the risk of hearing loss are uncertain, but likely involve effects on the host inflammatory response. Experimental data indicate that the combination of vancomycin and ceftriaxone induces more rapid killing of pneumococci than is achieved by either agent alone.20–22 This rapid bacterial lysis may then result in increased liberation of bacterial proinflammatory components (eg, lipoteichoic acid), resulting in an enhanced host inflammatory response that putatively mediates neurologic injury.23 Counterbalancing the potential risks related to rapid bacterial lysis, however, is the increased risk of neurologic sequelae associated with delayed (ie, >24 hours) CSF sterilization.12 Therefore, optimal therapy for pneumococcal meningitis must achieve the dual goals of sterilizing the CSF while minimizing inflammatory damage to host tissues. The most commonly advocated strategy for achieving these dual ends is the administration of adjunctive corticosteroid therapy. Although this approach remains controversial, a growing body of evidence suggests that dexamethasone, initiated concomitantly with or before the first dose of parenteral antibiotics, decreases rates of mortality in adults and of neurologic sequelae (including hearing loss) in children.24–26 Another potentially attractive strategy for sterilizing CSF while minimizing inflammation is to empirically administer a nonbacteriolytic antibiotic, such as rifampin, in combination with ceftriaxone or cefotaxime. Currently, evidence favoring this strategy is primarily experimental: mortality rates are lower in rifampin-treated mice with pneumococcal meningitis than in those treated with ceftriaxone,27 and compared with those treated with ceftriaxone alone, animals who receive rifampin plus ceftriaxone demonstrate reductions in CSF pneumolysin concentrations,28 CSF lipoteichoic acid concentrations, and density of apoptotic neurons in the hippocampal dentate gyrus.29 The inclusion of nonbacteriolytic antibiotics in initial therapy for bacterial meningitis merits further evaluation in human clinical trials.
Our results are subject to limitations inherent in retrospective studies. First, because of our retrospective study design, we could only analyze data that was available in medical and laboratory records. In particular, audiometry test results were documented in only 64% of surviving subjects. It is possible that subjects who underwent audiometry differed systematically from those who did not; however, the only observed difference between these groups was a higher rate of underlying conditions (which, in many cases, would have obviated the attribution of hearing loss to the meningitis episode in any event) in the latter. Second, because this is a nonrandomized, observational study, our results could be compromised by the effects of confounding variables; therefore, we controlled for such confounders using multiple logistic regression modeling. It is conceivable, for example, that subjects with more severe disease at the time of presentation might have an inherently higher risk of hearing loss and might receive vancomycin therapy sooner than those with less severe disease. In our multivariate analysis, however, clinical measures of disease severity (eg, ICU admission and respiratory failure) were not significantly associated with hearing loss, and, importantly, their inclusion in models did not attenuate the observed association between vancomycin start time and hearing loss. Another limitation is that we could not assess the effect, if any, of adjunctive corticosteroid therapy on outcomes, because corticosteroids are infrequently prescribed for children with pneumococcal meningitis at this center. Although corticosteroid therapy was not significantly associated with any outcome variable in our multivariate analyses, this lack of association may result from insufficient power rather than absence of efficacy.
Several lines of evidence support the validity of the observed association between vancomycin start time and hearing loss. First, the univariate strength of this association is substantial. Second, a dose-response gradient is apparent: the risk of hearing loss progressively increases with decreasing vancomycin start time. Finally, the association strength increases in multivariate analysis and is independent of other variables, including those associated with disease severity. It further merits mentioning that the magnitude of this association is clinically relevant, as well as statistically significant: for every 2 patients who received vancomycin therapy <2 hours after the first dose of cefotaxime or ceftriaxone, 1 suffered otherwise unexpected hearing loss.
We recognize that only limited conclusions can be drawn from a retrospective study and would welcome a randomized, controlled trial to confirm the safety and efficacy of early empiric vancomycin therapy in children with pneumococcal meningitis. Until such a study is performed, we propose that the approach to empiric therapy of pneumococcal meningitis in children should be reconsidered. For the present, empiric vancomycin use remains justifiable because of the risks of delayed CSF sterilization in cases of meningitis caused by antimicrobial-resistant pneumococci, although the incidence of such cases has been reduced in the United States in the era of conjugate pneumococcal vaccination.30 Our results suggest that administration of vancomycin <2 hours after the first dose of cefotaxime or ceftriaxone confers no clinical benefit and, in fact, is associated with a substantially increased risk of hearing loss. Given these data, it may be prudent to consider delaying the first dose of vancomycin therapy until 2 hours after the first dose of parenteral cephalosporin in children beginning therapy for suspected or confirmed pneumococcal meningitis. Further investigation is required to determine whether such a delay is beneficial or harmful in children who receive early adjunctive corticosteroid therapy and to determine whether nonbacteriolytic antibiotics, such as rifampin, might prove to be safer than vancomycin in this clinical scenario.
|
FOOTNOTES |
|---|
Address correspondence to Steven C. Buckingham, MD, Le Bonheur Children's Medical Center, Room 301, 50 N Dunlap St, Memphis, TN 38103. E-mail: sbuckingham{at}utmem.edu
The authors have indicated they have no financial relationships relevant to this article to disclose.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S. C. Buckingham, B. K. English, J. A. McCullers, K. M. Knapp, J. Lujan-Zilbermann, and K. L. Orman Could Hearing Loss Be Related to Delay in Administration of Other Antibiotics Rather Than Early Use of Vancomycin?: In Reply Pediatrics, February 1, 2007; 119(2): 416 - 417. [Full Text] [PDF]
|
|
|
|
|
|
B. E. Lee, S. E. Forgie, A. R. Joffe, J. L. Robinson, and W. L. Vaudry Could Hearing Loss Be Related to Delay in Administration of Other Antibiotics Rather Than Early Use of Vancomycin? Pediatrics, February 1, 2007; 119(2): 415 - 416. [Full Text] [PDF]
|
|
Read all P3Rs
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved. |
||
|
||
