INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: Streptococcus pneumoniae; meningitis; outcome; corticosteroids

Streptococcus pneumoniae is the most frequent cause of community-acquired bacterial meningitis in adults. A population-based study performed in the United States found an annual incidence of 1.1 cases per 100,000 population (1). Despite advances in antibiotic therapy, the morbidity (20 to 30%) and case fatality (21 to 28%) (2) associated with pneumococcal meningitis have changed little in the past 30 yr. Several previous studies, based on the analysis of large cohorts, have identified baseline clinical features of disease severity predicting adverse clinical outcome (2, 4, 5). However, data on prognostic factors in adult patients admitted to the intensive care unit (ICU) for S. pneumoniae meningitis are scarce. In addition, the proportion of isolates of S. pneumoniae that are nonsusceptible to penicillin (PNSP) and cephalosporins has been increasing in frequency during the last decade in many countries (6). Because of the poor penetration of antibiotics into the cerebrospinal fluid (CSF), a significant increase of the minimal inhibitory concentration (MIC) may lead to clinical or microbiologic failure. Few studies have specifically evaluated the relationship between the susceptibility of S. pneumoniae to penicillin and the outcome (10, 11) in the most severe group of patients, namely those admitted to the ICU.

Randomized controlled trials (12, 13) support dexamethasone therapy in children with Haemophilus influenzae meningitis. Although some authorities recommend dexamethasone use for patients with severely impaired mental status or high intracranial pressure (14), the benefits of such adjunctive therapy are far less clear in adult patients suffering from pneumococcal meningitis.

The purpose of this retrospective study was: (1) to identify factors associated with in-hospital mortality and morbidity in adult patients admitted to the ICU for pneumococcal meningitis; (2) to determine whether prognosis is affected by nonsusceptibility to antibiotics; and (3) to assess the potential role of dexamethasone therapy for the most severely ill patients.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The charts of all consecutive patients older than 18 yr admitted to the Bichat-Claude Bernard ICU with pneumococcal meningitis between January 1988 and June 1999 were reviewed. A standardized case report form was completed for each episode using information abstracted from the ICU records. Pneumococcal meningitis was defined by CSF pleocytosis (> 10 cells/µl) and one or more of the following criteria: a positive culture of the CSF or blood; the presence in the CSF of gram-positive diplococci; the presence in the CSF of pneumococcal antigens detected by a latex agglutination method.

From medical records, we extracted information on demographic data, comorbidity, and immunocompetence. The severity of the underlying disease was assessed by the Knaus score (15), and the Charlson comorbidity scale (16). Conditions that were considered predisposing factors for the analysis included immunosuppression, proven or documented dural fistula, and severe chronic alcoholism. Altered immune state was defined by one of the following: human immunodeficiency virus (HIV) infection, malignancy not in remission, asplenia, diabetes mellitus, long-term use of corticosteroids or other immunosuppressive medications. Dural fistula was considered probable if there was a past history of head trauma or neurosurgery or a previous episode of pneumococcal meningitis or clinical CSF leak. In addition, the following parameters were determined within 48 h of admission to the ICU: Simplified Acute Severity Score (SAPS II) (17) and Glasgow Coma Score (GCS) (18), the need for mechanical ventilation, clinical laboratory parameters in serum, and CSF abnormalities, namely glucose level, protein level, and leukocyte count.

Determination of MIC for penicillin G, amoxicillin, and cefotaxime was performed either by standard microbroth dilution with Mueller-Hinton media supplemented with 3% lysed horse blood or since 1995 by the E test. Susceptibility categories were determined using the 1997 National Committee for Clinical Laboratory Standards guidelines for breakpoints: penicillin G:  0.06 µg/ml = susceptible, 0.12 to 1 µg/ml = intermediate,  2 µg/ml = resistant; amoxicillin and cefotaxime:  0.5 µg/ml = susceptible, 1 µg/ml = intermediate,  2 µg/ml = resistant (19). For the purpose of the analysis, any strain that was not fully susceptible to penicillin (MIC < 0.12 µg/ml) was considered as "nonsusceptible". Appropriate antibiotic therapy was defined as initial empiric therapy to which the isolated S. pneumoniae was fully susceptible. Antibiotic daily doses for the three main administered molecules were amoxicillin 200 mg/kg, cefotaxime 200 mg/kg, vancomycin 40 to 60 mg/kg. The recommended duration of antibiotic therapy was 14 d.

In our ICU, the use of corticosteroids as adjunctive treatment of pneumococcal meningitis is not based on protocol. In treated patients, dexamethasone was administered as 0.15 mg/kg doses every 6 h for 2 to 4 d (12). Patients were considered to have received corticosteroids only if administration (at least one dose) began on the date of admission. The use of hypertonic mannitol in case of severe cerebral edema led to physician's discretion. The amount of fluid administered was determined on the basis of serum sodium concentration with the goal of maintaining this concentration around 140 mmol/L. No standard protocol was used to achieve a specific Pa_CO2. All patients received anticonvulsant either to prevent or to treat seizures. Intracranial pressure monitoring devices are not used in our unit. Neuroimaging was performed when severe cerebral edema was suspected or in the presence of a focal neurologic deficit. Patients were considered to have neurologic impairment if at least one of the following conditions were present: motor deficit, hearing impairment clinically detected, behavioral or language disturbance, or hydrocephalus. As proposed by other investigators (4), adverse clinical outcome was defined as death during ICU stay or neurologic impairment at discharge from the ICU.

Statistical Methods

Association between single variables and the two end-points (death and adverse clinical outcome) was assessed by Wilcoxon rank sum test for quantitative variables and ² or Fisher exact test when more appropriate for qualitative variables. Multiple logistic regression was used to determine independent prognostic factors of death within hospital and adverse outcome. All variables achieving statistical significance at a 10% level in the univariate analysis were simultaneously considered in the multivariable model, provided their missing value rate was below 0.10. Missing values for quantitative variables were estimated by the minimum generalized variance method (using SAS PRINQUAL procedure), whereas binary variables missing values were withdrawn from the analysis. Biologic variables were dichotomized using cutoffs of clinical significance as follows: 100 G/L for platelet count, 150 µmol/L for serum creatinine, and 15 G/L for leukocytes. For the pH, in absence of such information, the median (7.48) was taken as threshold value. Knaus score was also dichotomized, separating null scores from positive ones. GCS and SAPS II were divided into three categories, namely less than 6, between 6 and 13, and more than 13 for the GCS and less than 30, between 30 and 50, and more than 50 for the SAPS II.

A backward stepwise variable elimination procedure was used to identify the set of independent prognostic variables of in-hospital mortality and adverse outcome. The validity of the logistic regression models was checked using Hosmer and Lemeshow goodness-of-fit test (20). We assessed the robustness of our prognostic set on the strategy used to handle missing data, by rerunning the multivariate model after recoding missing values of binary data either to 0 or 1. According to the relatively high number of potential predictors with respect to the number of events and to its dependence on medical decision, the effect of corticosteroids was assessed by forcing this predictor into the final identified prognostic model. All analyses were performed using SAS 6.12 (SAS Institute, Cary, NC) and S-Plus 4.5 (MathSoft Inc, Seattle, WA) software packages. All tests were two-sided, with significance levels fixed at 5%.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

During the study period a total of 5,243 patients were admitted to our ICU. Of these, 80 (1.5%) had S. pneumoniae meningitis. The number of admissions for meningitis ranged from 6 to 12 per year over the 10-yr study period. The main characteristics were: mean age: 59 ± 18 yr (range: 18 to 87); sex ratio (M/F): 1.25; mean SAPS II: 39 ± 14.3; mean GCS: 8.5 ± 3.5. Forty-four patients were admitted directly from the emergency department, whereas the other 36 patients were transferred from other hospitals for intensive care (26 from medical units and 10 from other ICUs). A total of 56 patients (70%) required mechanical ventilation, 32 of them being intubated before the admission to the ICU and 24 within 48 h of admission. On presentation, 17 patients (21%) had generalized or focal seizures. Only two patients had hypotension (defined as systolic arterial pressure  90 mm Hg). Predisposing factors for invasive pneumococcal infection or for pneumococcal meningitis were found in 64 patients, 32 of them having more than one factor. Half of the patients had a concomitant ear or sinus infection which was considered as the portal of entry (Table 1). Blood cultures were performed at the time of admission in 75 of the 80 patients and grew S. pneumoniae in 50 cases (62%). Six patients had concomitant S. pneumoniae endocarditis documented in all cases by the presence on echocardiography of valvular vegetations: two on mitral valve, two on aortic valve, and two on both mitral and aortic valves.

S. pneumoniae was isolated into both the CSF and blood cultures in 45 patients, only into CSF in 21, and only in blood cultures in 5. In three of these latter patients, Gram stain on CSF showed gram-positive diplococci and the other two patients had more than 1,000 cells/µl, a CSF glucose level < 2 mmol/L, and a CSF protein level > 2 g/L. In nine patients the diagnosis was made either on CSF Gram staining or by detection of pneumococcal antigens into the CSF. The susceptibility to penicillin and to cefotaxime of pneumococcal isolates causing meningitis varied according to time; the overall prevalence of nonsusceptible was 24% (17/71) but the number of nonsusceptible strains increased from 0/30 during the 1988-1993 period to 17/41 (41%) during the 1994-1999 period. Among the 17 nonsusceptible strains, 14 were intermediate to penicillin G and three were resistant. Sixteen PNSP strains were tested against amoxicillin: two strains were intermediate and one was resistant. Among the 12 PNSP strains tested against cefotaxime, two were intermediate but none was resistant. On admission no significant differences were found between patients with susceptible (n = 54) or nonsusceptible (n = 17) strains for the following clinical characteristics: age: 51 ± 19 yr versus 58 ± 15 yr (p = 0.16); SAPS II: 39 ± 14 versus 41 ± 11 (p = 0.68); and GCS: 8 ± 3 versus 9.5 ± 3 (p = 0.21), respectively.

The following antibiotic regimens were administered: amoxicillin (n = 32), cefotaxime alone (n = 6) or associated with vancomycin (n = 9), amoxicillin followed by cefotaxime (n = 10), cefotaxime followed by amoxicillin (n = 15), others (n = 8) (penicillin: two patients; cefotaxime and fosfomycin: two patients; cefotaxime and rifampin: three patients; meropenem: one patient). In all cases, antibiotic therapy was started before ICU admission, either in the emergency department or in the other hospital. All patients received appropriate initial empiric therapy. Systemic corticosteroids were administered on the day of admission to 27% of the cohort (22 patients). In 10 patients, the first dose was given concomitantly to the first antibiotic dose. In the other 12 patients, the intervals between the initiation of antibiotic therapy were as follows: < 6 h, (n = 4), 6 to < 12 h (n = 7), and 12 to 24 h (n = 1). The 22 steroid-treated patients received the following antibiotic regimens: penicillin (n = 1), amoxicillin (n = 8), cefotaxime (n = 5), ceftriaxone (n = 1), cefotaxime and vancomycin (n = 6), cefotaxime and rifampin (n = 1). With respect to all characteristics tested, there were no significant differences between patients receiving or not receiving corticosteroids (see Table E1 in online data supplement).

The mean and median lengths of stay in the ICU were 18 d and 11 d, respectively. The in-hospital mortality rate for the total cohort was 25% (20/80). The mean delay between admission to the ICU and death was 13.5 d (range: 1 to 35). Six patients died within the first 2 d of admission. The causes of death were: brain death (n = 15), cardiogenic shock in the setting of endocarditis (n = 2), ventilator-associated pneumonia (n = 2), and gastrointestinal bleeding (n = 1). There was only one death among the 17 patients infected with a nonsusceptible strain (6%). Of the 80 patients, 32 (53% of survivors; 40% of the total cohort) had a neurologic deficit that persisted at discharge. Neurologic impairments noted at hospital discharge were as follows: hearing impairment (15 patients), behavior disturbances (9 patients), language disturbances (7 patients), motor deficit (6 patients), hydrocephalus (3 patients). Thus, an adverse clinical outcome was observed in 52 patients.

Univariate analysis relating baseline characteristics to death is shown in Table 2. Nonsurvivors had a higher SAPS II (p = 0.0006), a lower GCS (p = 0.005), and were more often on mechanical ventilation than survivors (p = 0.004). In addition, nonsurvivors had a lower platelet count (p = 0.0021), a higher arterial pH (p = 0.0002), a lower Pa_CO2 (p = 0.019), and a higher serum creatinine level (p = 0.035), but were more often infected with a susceptible strain (p = 0.03). By multivariate analysis only three variables remained independently associated with in-hospital mortality, namely a platelet count < 100 G/L, mechanical ventilation, and an arterial pH > 7.47. 

When adjusting for the identified prognostic factors, the use of corticosteroids was found to significantly reduce the risk of death (Table 3). Univariate analysis relating baseline characteristics to an adverse clinical outcome is shown in Table 4. Patients with adverse outcome had a higher SAPS II (p = 0.005), a lower GCS (p = 0.009), a higher arterial pH (p = 0.03), a lower Pa_CO2 (p = 0.04), and were more often on mechanical ventilation (p = 0.022) than survivors without adverse outcome. By multivariate analysis, only SAPS II remained independently associated with an adverse clinical outcome (Table 5).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Pneumococcal meningitis remains a potentially devastating disease that carries high mortality and morbidity rates. Until the last decade, antibiotic therapy regimens were simple because all isolates of S. pneumoniae were susceptible to penicillin. Thus, this molecule or amoxicillin were the antibiotics of choice for the treatment of pneumococcal meningitis. The dramatic increase in the prevalence of strains with reduced susceptibility or even resistant to penicillin, amoxicillin, or to third-generation cephalosporins led to the publication of new guidelines for the treatment of S. pneumoniae meningitis. According to these guidelines, patients with suspected pneumococcal meningitis should receive a combination of cefotaxime or ceftriaxone plus vancomycin at least until the results of susceptibility testing are available (14). The purpose of the present study was to identify factors associated with early outcome in patients admitted for pneumococcal meningitis. Among the strengths of this study were: the cohort was relatively homogenous with regard to the severity of the disease because all patients were hospitalized in our medical ICU; all patients were adults; and the study was limited to a 10-yr period, during which there were no major modifications in the supporting care of bacterial meningitis in our unit.

In the present study, three characteristics, namely the need for mechanical ventilation, thrombopenia, and alkalosis, were independently associated with death. The first two variables, especially requirement for mechanical ventilation, represent feature of severe disease at the time of initial antibiotic therapy. Thrombocytopenia is a marker of severe sepsis. Thus, our finding that a platelet count inferior to 100 G/L is predictive of death is not unexpected, keeping in mind that 62% of the patients had positive blood cultures. Mental status determined by the GCS was associated with poor outcome only in the univariate analysis. Unlike another recent study involving 269 patients with bacterial meningitis, 130 of them infected with S. pneumoniae (4), seizures and hypotension were not associated with a poorer outcome. However, in our study very few patients had hypotension. Surprisingly, we found that arterial alkalosis, mainly explained by hyperventilation, was predictive of death. To our knowledge, this association has not been previously reported. At least two factors may be involved to explain the relationship between respiratory alkalosis and outcome. First, in nonventilated patients severe hyperventilation may identify a subgroup of patients with a more severe disease. Alternatively, it should be emphasized that hyperventilation is commonly used in acute neurologic disorders to decrease intracranial pressure by causing cerebral vasoconstriction and decreasing cerebral blood flow. However, such treatment could potentially result in severe cerebral ischemia. Indeed, in severe head injury, it has been recently shown that hyperventilation may cause a significant reduction in brain tissue oxygen pressure (21). Moreover, in a series of 20 seriously ill children with acute bacterial meningitis, it has been shown that 28% of the children had baseline cerebral blood flow values already significantly reduced. Hyperventilation in these patients could further reduce cerebral blood flow below ischemic thresholds (22).

In the present study, we found that S. pneumoniae strains nonsusceptible to penicillin were not associated with a poorer outcome. Several previous studies (10, 11) have shown that the severity of disease and the outcome of pneumococcal meningitis caused by nonsusceptible strains to penicillin have been similar to those caused by organisms that are susceptible to penicillin. The lack of association between -lactam susceptibility and death or adverse clinical outcome may have at least three explanations. First, all studies, including ours, have limited power to detect a difference in prognosis on the basis of susceptibility to cefotaxime. Second, as in the present study, most, if not all patients, received appropriate antibiotic therapy. In our study, the prevalence of strains intermediate or resistant to amoxicillin or to cefotaxime was low. Thus -lactam concentrations achieved in the CSF were probably far above the MIC for most strains, and some patients received a combination of cefotaxime and vancomycin as recently recommended (14). Third, it can be speculated that nonsusceptible strains are intrinsically less virulent (23).

Systemic corticosteroids were administered on the day of admission to 27% of our patients. Although there is considerable evidence from clinical trials in children that adjunctive anti-inflammatory therapy with corticosteroids is effective in improving long-term outcome in Haemophilus influenzae meningitis in children (24), similar data involving adults with S. pneumoniae meningitis are lacking. Only one trial has suggested an overall reduction in mortality among adult patients suffering from pneumococcal meningitis and who received dexamethasone (25). In contrast, a recent trial failed to show any significant benefit of dexamethasone, but the study was prematurely stopped after the inclusion of only 31 patients with S. pneumoniae meningitis (26). According to our model, a protective effect of corticosteroids was observed in the most severely ill patients. The rather wide 95% confidence interval of the odds ratio may be explained by the small sample size. The beneficial effect of dexamethasone observed in our study was perhaps reduced by the long interval between the onset of antibiotics and the administration of steroid therapy in nearly half of the treated patients. A meta-analysis has suggested that corticosteroids should be administered before or with the first dose of antibiotics (24). The protection conferred by high-dose corticosteroids could be related to a favorable effect either on meningitis or on concomitant shock. The latter hypothesis seems unlikely, because only two patients presented with hypotension.

The present study has several limitations. First, the data were collected retrospectively. As a consequence, patients did not receive supportive care according to a standardized protocol. Second, the outcome was evaluated only at hospital discharge. Thus, we have no information on long-term mortality and morbidity. Despite these limitations, our data, derived from a relatively large number of patients admitted to the ICU for S. pneumoniae meningitis, may lead to four conclusions: (1) Pneumococcal meningitis continues to be a major cause of mortality and morbidity, especially in patients requiring admission to the ICU; (2) In patients under mechanical ventilation for pneumococcal meningitis, hyperventilation should be used with caution; (3) The prevalence of nonsusceptible strains is increasing, as reported in all surveillance studies performed worldwide. However, provided that empirical antibiotic therapy is appropriate, nonsusceptibility is not associated with a poorer outcome; and (4) corticosteroids may be beneficial, but only a well-designed, placebo-controlled trial of adjunctive dexamethasone therapy for pneumococcal meningitis will be able to determine the efficacy, if any, of this drug (27).