close

Вход

Забыли?

вход по аккаунту

?

Arisk score for unfavorable outcome in adults with bacterial meningitis.

код для вставкиСкачать
A Risk Score for Unfavorable Outcome in
Adults with Bacterial Meningitis
Martijn Weisfelt, PhD,1 Diederik van de Beek, PhD,1 Lodewijk Spanjaard, PhD,2 Johannes B. Reitsma, PhD,3
and Jan de Gans, PhD1
Objective: To derive and validate a bedside risk score for adverse outcome in adults with bacterial meningitis.
Methods: We derived a score for the risk for an unfavorable outcome (Glasgow Outcome Scale score 1– 4) by performing
logistic regression analyses of data from a prospective cohort study (Dutch Meningitis Cohort; N ⫽ 696). A key set of independent prognostic variables was selected from 22 potential predictors. A nomogram based on these key variables was constructed to facilitate use in clinical practice. To validate this nomogram, we used data from our randomized controlled trial on
adjunctive dexamethasone therapy in adults with bacterial meningitis (European Dexamethasone Study; N ⫽ 301).
Results: Unfavorable outcome occurred in 237 of 696 episodes (34%) in the Dutch Meningitis Cohort; 143 patients (21%)
died. In the analysis, 6 of 22 variables that are routinely available within 1 hour after admission were robust enough for inclusion
in the final risk score: age, heart rate, Glasgow Coma Scale score, cranial nerve palsies, a cerebrospinal fluid leukocyte count less
than 1,000 cells/mm3, and gram-positive cocci in cerebrospinal fluid Gram’s stain. The concordance index for the risk score was
0.84 (95% confidence interval, 0.80 – 0.87) in the original cohort and 0.81 (95% confidence interval, 0.74 – 0.87) in the external
validation cohort (European Dexamethasone Study).
Interpretation: This bedside risk score can be used to identify patients with a high risk for unfavorable outcome in adults with
bacterial meningitis within 1 hour after the initial presentation.
Ann Neurol 2008;63:90 –97
Bacterial meningitis is a life-threatening disease. In bacterial meningitis, clinical deterioration can occur rapidly and is often difficult to predict.1–3 A simple bedside risk score for unfavorable outcome would be
helpful in risk assessment and subsequent management
of individual patients. Recently, we described clinical
features and prognostic factors in 696 episodes of
community-acquired bacterial meningitis in adults.4,5
In this prospective cohort study, risk factors for an unfavorable outcome were those indicative of systemic
compromise, a low cerebrospinal fluid (CSF) leukocyte
count, impaired consciousness, and infection with
Streptococcus pneumoniae.4,5 Previous studies on prognostic factors in adults with bacterial meningitis were
hampered by retrospective data collection, heterogeneous patient populations (adults and children), small
sample size, high mathematical complexity, and predicted outcome with variables that will become available only within days after admission (eg, blood or
CSF culture) or with semioutcome measurements (eg,
mechanical ventilation).6 –22 The aim of this study was
to derive and validate a simple score to predict the risk
for unfavorable outcome in adults with communityacquired bacterial meningitis. To enhance the clinical
applicability of this model, we included only data available within 1 hour after presentation.
From the Departments of 1Neurology and 2Medical Microbiology,
Netherlands Reference Laboratory for Bacterial Meningitis; and
3
Department of Clinical Epidemiology and Biostatistics, Center of
Infection and Immunity Amsterdam (CINIMA), Academic Medical
Center, Amsterdam, the Netherlands.
Published online September 6, 2007 in Wiley InterScience
(www.interscience.wiley.com). DOI: 10.1002/ana.21216
Received May 26, 2007, and in revised form Jun 21. Accepted for
publication Jul 19, 2007.
Patients and Methods
Derivation Cohort
The Dutch Meningitis Cohort (DMC), a prospective, nationwide, observational cohort study in the Netherlands, included 696 episodes of community-acquired bacterial meningitis in adults. Inclusion and exclusion criteria, treatment,
and outcome measures are described elsewhere.4 In summary, patients were older than 16 years, had bacterial meningitis microbiologically confirmed by CSF culture, and were
listed in the database of the Netherlands Reference Laboratory for Bacterial Meningitis from October 1998 to April
2002. The study was performed in accordance with the
Dutch privacy legislation and was approved by our ethics
Address correspondence to Dr de Gans, Department of Neurology,
Academic Medical Center, P.O. Box 22660, 1100 DD Amsterdam,
the Netherlands. E-mail: j.degans@amc.uva.nl
This article includes supplementary materials available via the Internet at http://www.interscience.wiley.com/jpages/0364-5134/suppmat
90
© 2007 American Neurological Association
Published by Wiley-Liss, Inc., through Wiley Subscription Services
committee. Focal neurological abnormalities were divided
into focal cerebral deficits (aphasia, monoparesis, or hemiparesis) and cranial nerve palsies. Results of CSF Gram’s stain
were categorized as follows: (1) gram-positive cocci, (2)
gram-negative cocci, (3) other bacterial species (according to
judgment of treating physician), and (4) no bacteria on CSF
Gram’s stain. A total of 1,108 episodes of bacterial meningitis were identified by the reference laboratory. A total of
994 case-record forms were sent out, and the response rate
was 76% (754/994). The demographic characteristics of patients with meningitis identified by the laboratory and those
included in the analysis were similar for each causative organism. Fifty-eight patients were excluded (50 patients with
hospital-acquired meningitis, 3 patients with a recent history
of neurosurgery, and 5 patients with a neurosurgical device),
leaving a total of 696 episodes of community-acquired meningitis in 671 patients. A total of 25 of 671 patients (4%)
had more than 1 episode. Given this low frequency of multiple episodes, we analyzed all available data without specifically incorporating any potential correlation between multiple episodes from the same patient. At discharge, patients
underwent a neurological examination and outcome was
graded by means of the Glasgow Outcome Scale (GOS).
This is a well-validated measurement scale with scores varying from 1 (indicating death) to 5 (good recovery).23 A favorable outcome was defined as a score of 5, and an unfavorable outcome as a score of 1 to 4.4,5,24
Validation Cohort
The European Dexamethasone Study (EDS) was a doubleblind, placebo-controlled trial of adjunctive dexamethasone
therapy for adults with bacterial meningitis. Inclusion and
exclusion criteria are described more extensively elsewhere.24
In summary, eligible patients were older than 16 years and
had suspected meningitis in combination with cloudy CSF,
bacteria in CSF on Gram’s staining, or a CSF leukocyte
count of more than 1,000 cells/mm3.24 Patients were enrolled between June 1993 and December 2001. A total of
301 patients were randomly assigned to receive dexamethasone sodium phosphate, at a dose of 10mg given every 6
hours intravenously for 4 days, or placebo. The study medication was given 15 to 20 minutes before or with the first
dose of antibiotics. The primary outcome measure was the
GOS score 8 weeks after randomization, as assessed by the
patients’ physicians. A total of 36 of 301 patients (12%) in
the EDS were also included in the DMC (21 in the dexamethasone group and 15 in the placebo group). The results
of the validation after exclusion of these patients are shown
separately.
Statistical Analysis
Predictors for an unfavorable outcome in the DMC were examined by logistic regression analysis. From previous research and pathophysiological interest, a predefined set of 22
potentially relevant determinants of outcome that are rou-
tinely available within 1 hour after presentation were selected
to be entered in a multivariate logistic regression analysis.4 –22
This selection of variables is different from those in our previous description of the DMC in which we also included
potential predictors that become available later during clinical course.4 The number of events (n ⫽ 237) allowed us to
examine about 24 predictors. We choose to enter the complete set of predictors into a single multivariate model to
show the strength of each predictor. Odds ratios and 95%
confidence intervals (CIs) were used to show the strength of
the association between these potential prognostic factors and
the probability of an unfavorable outcome in a multivariate
model. Colinearity between potential predictors was assessed
by examining variance inflation factors and cross tabulations.
No pertinent problems were detected. Interactions were not
assessed because we had no specific interactions of interest
and considering interactions would add to the complexity of
the analysis and increase the risk of false-positive findings
because of overfitting. Although the median percentage of
missing values for individual variables in our study was low
(2%),4 data were complete on all potential predictors in only
355 of 696 episodes (51%). To maximize the benefits of our
multivariate model, we imputed missing values by using
multiple imputation techniques based on multivariate normal distributions.25 The final estimates of the multivariate
model were obtained by combining the coefficients of five
rounds of imputations. A comparison of results from the
same model using only cases with complete data is provided
in the supplemental materials available online (see Supplemental Table 1).
To obtain a nomogram for predicting individual risks that
can be used in daily practice, we applied backward selection
(significance level to stay in the model: p ⱕ 0.05) to reduce
the number of predictors. The number of points for each
predictor was based on the original coefficient from the logistic regression model (after backward elimination) by multiplying it by 10 and rounding it to the lowest whole number. The total number of points derived by specifying values
for all predictors was used to calculate the expected probability of a poor outcome based on the logistic regression
model. We used n ⫽ 600 bootstrap samples in which we
repeated the backward elimination procedure to increase the
likelihood of selecting variables that are genuinely related to
the outcome. Variables that remained in the model in more
than half of the bootstrap samples were included in the final
nomogram model.
Calibration or goodness of fit of the nomogram model
was assessed by examining the differences in observed versus
predicted number of patients with unfavorable outcome
across the range of predicted risk. Calibration was formally
tested by performing the Hosmer–Lemeshow goodness of fit
test for logistic regression models. The discriminative ability
of the nomogram model in the derivation and validation
dataset was assessed by visualizing the distribution and overlap in risk scores of patients with and without an unfavorable
outcome. The concordance or c statistic was used to quantify
the discriminative ability in a single number. Bootstrap sampling was used to obtain CIs around the c statistic. All anal-
Weisfelt et al: Meningitis Risk Score
91
yses were undertaken with SAS software (version 9.11; SAS
Institute, Cary, NC).
Role of the Funding Source
The funding source had no role in the study design; in the
collection, analysis, and interpretation of data; in the writing
of the report; and in the decision to submit the manuscript
for publication.
Results
Characteristics of the DMC (derivation cohort) and
the EDS (validation cohort) are noted in Table 1. In
the DMC, a total of 696 episodes of bacterial meningitis occurred in 671 patients; 25 patients had a second
episode during the study period. CSF culture yielded S.
pneumoniae in 352 episodes (51%), Neisseria meningi-
Table 1. Characteristics of Patients Included in the Derivation and Validation Cohorts
Characteristics
Derivation Cohort
(N ⴝ 696)a
Validation Cohort
(N ⴝ 301)b
50 ⫾ 20
45 ⫾ 19c
317/661 (48)
121/299 (40)
32/666 (5)
22 (7)
64/692 (9)
5 (2)c
312 (45)
Not recorded
Neck stiffness, n (%)
569/685 (83)
267/285 (94)c
Heart rate ⬎ 120 beats/min, n (%)
77/652 (12)
20/299 (7)
Diastolic blood pressure ⬍ 60mm Hg, n (%)
61/670 (9)
22/298 (7)
Body temperature ⱖ 38°C, n (%)
522/678 (77)
245/300 (82)
Rash, n (%)
176/683 (26)
Not recorded
11 ⫾ 3
12 ⫾ 3c
305 (44)
144/291 (49)
Focal cerebral deficits, n (%)
157 (23)
59 (20)
Cranial nerve palsies, n (%)
89 (13)
32 (11)
184/645 (29)
21/296 (7)c
Mean protein ⫾ SD, gm/L
4.9 ⫾ 4.5
4.5 ⫾ 3.1
Mean CSF/blood glucose ratio ⫾ SD
0.2 ⫾ 0.2
0.2 ⫾ 0.2
46 ⫾ 37
Not recorded
198,000 ⫾ 100,000
222,000 ⫾ 95,000c
Gram-positive cocci, n (%)
320/654 (49)
108/290 (37)c
Gram-negative cocci, n (%)
225/654 (34)
104/290 (36)
Other bacteria, n (%)
22/654 (3)
5/290 (2)
Negative, n (%)
87/654 (13)
73/290 (25)c
Patient characteristics before admission
Mean age ⫾ SD, yr
Duration of symptoms ⬍ 24 hr, n (%)
Seizures, n (%)
Antibiotics before admission, n (%)
d
Predisposing factors, n (%)
Clinical characteristics on presentation
Mean GCS score ⫾ SD
e
Triad of fever, neck stiffness, and change in mental status, n (%)
f
Laboratory features
Indexes of inflammation in CSF
White-cell count ⬍ 1,000/mm3, n (%)
Mean ESR ⫾ SD, mm/hr
Mean thrombocyte count ⫾ SD, platelets/mm3
Gram stain
a
The Glasgow Coma Scale score was obtained in 694 episodes, cerebrospinal fluid (CSF) protein level was determined in 634 episodes,
CSF/blood glucose ratio in 617 episodes, erythrocyte sedimentation rate (ESR) in 549 episodes, and thrombocyte count in 653 episodes.
b
CSF protein level was determined in 292 patients, CSF/blood glucose ratio in 288 patients, and thrombocyte count in 294 patients.
c
Significant difference between derivation cohort and validation cohort.
d
Defined as the presence of otitis/sinusitis, pneumonia, or immunocompromise.
e
Scores on the Glasgow Coma Scale (GCS) range from 3 to 14, with 14 indicating a normal level of consciousness (abnormal flexion
was omitted from the scale).
f
Defined as aphasia, monoparesis, or hemiparesis.
SD ⫽ standard deviation.
92
Annals of Neurology
Vol 63
No 1
January 2008
tidis in 257 episodes (37%), and other bacteria in 87
episodes (13%). A detailed description of presenting
features, clinical course, and outcome has been published previously.4 Outcome was graded as favorable
(GOS score, 5) in 459 episodes (66%) and as unfavorable in 237 of 696 episodes (34%); GOS score was 4
in 67 episodes, 3 in 24 episodes (3%), and 2 in 3 episodes (⬍1%), and 143 patients (21%) died.
In the EDS, a total of 301 patients were included.
CSF culture was performed in 299 patients (99%) and
yielded S. pneumoniae in 108 patients (36%), N. meningitidis in 97 patients (32%), and other bacterial species in 29 patients (10%); in 65 patients (22%), CSF
cultures were negative. In 26 of 65 patients (40%) with
negative CSF culture, bacteria were detected by CSF
Gram’s stain or blood culture. Characteristics of the
study population are described more extensively elsewhere.24 Outcome was graded as favorable in 242 patients (80%) and as unfavorable in 59 patients (20%);
GOS score was 4 in 20 (7%) and 3 in 7 (2%), and 32
patients died (GOS score, 1; 11%).
Baseline characteristics of the DMC (derivation) and
the EDS (validation) on admission were highly similar.
Due to inclusion and exclusion criteria of the EDS (see
earlier), the proportion of patients who were pretreated
with antibiotics, had a CSF leukocyte count of less
than 1,000 cells/mm3, or had a positive CSF Gram’s
stain was smaller in the EDS than in DMC. In addition, patients included in the EDS were younger than
patients in the DMC, had a greater rate of neck stiffness, a lower score on the Glasgow Coma Scale (GCS),
and a greater thrombocyte count. Naturally, patients in
the EDS were more likely to receive adjunctive treatment with corticosteroids than patients in the DMC
(157/301 patients [52%] vs 118/696 episodes [17%];
p ⬍ 0.0001). In the DMC, corticosteroids were administered before or with the first dose of antibiotics in
24 of these 118 episodes (20%); the remaining 94 episodes received steroids later during clinical course.
Several potential prognostic factors in the DMC
were significantly associated with an unfavorable outcome in the univariate analysis; however, most factors
were no longer significant in the multivariate analysis
(see Supplemental Table 2). In the multivariate model,
advanced age, duration of symptoms less than 24
hours, tachycardia (heart rate ⬎ 120 beats/min), a low
GCS score on presentation, presence of focal cerebral
deficits, presence of cranial nerve palsies, a CSF leukocyte count of less than 1,000 cells/mm3, a high erythrocyte sedimentation rate, a low thrombocyte count,
and the presence of gram-positive cocci in CSF Gram’s
stain were significantly associated with an unfavorable
outcome. In the backward variable elimination procedure (including bootstrapping), advanced age, tachycardia, a low GCS score, presence of cranial nerve palsies, a CSF leukocyte count of less than 1,000 cells/
mm3, and gram-positive cocci in CSF Gram’s stain
were identified as consistent independent prognostic
factors (Table 2). Calibration of the model was good
because differences between observed versus predicted
number of events across the range of predicted risk
were small, resulting in a nonsignificant and high p
value for the Hosmer–Lemeshow goodness of fit test
(␹2 ⫽ 6.36; degrees of freedom ⫽ 9; p ⫽ 0.70). Results of the analysis with and without patients with a
second episode of meningitis within the studied period
were similar (696 episodes vs 671 patients).
The resulting nomogram of these variables is shown
in the Figure. The scores for individual patients ranged
from 0 to 65 points, and the associated risk estimates
for an unfavorable outcome varied between 3.2 and
96%. In the DMC, data were complete for all 6 predictors included in the risk score in 574 of 696 episodes (82%); the concordance index for the nomogram
model was 0.84 (95% CI, 0.80 – 0.87).
Use of the nomogram can be illustrated as follows: a
patient aged 60 years (8 points), with a score on the
GCS of 7 (10 points), a heart rate of 130 beats/min
(10 points), without cranial nerve palsies (0 points), a
CSF leukocyte count of 400 cells/mm3 (13 points),
and gram-positive cocci in CSF Gram’s stain (12
points) has a total of 53 points, which is associated
with an estimated risk for an unfavorable outcome of
approximately 85%. In contrast, a 20-year-old patient
(0 points), with a score on the GCS of 14 (1 point),
without tachycardia (0 points), a CSF leukocyte count
of 2,700 cells/mm3 (0 points), and gram-negative diplococci in CSF Gram’s stain (0 points) has a total of 1
Table 2. Independent Risk Factors for an
Unfavorable Outcome in the Derivation Cohort
Characteristics
OR (95% CI)
Age, yra
1.24 (1.1–1.38)
Heart rate ⬎ 120 beats/min
2.97 (1.67–5.28)
b
GCS score
0.87 (0.82–0.93)
Cranial nerve palsies
2.48 (1.45–4.24)
White-cell count ⬍ 1,000/mm
3
3.91 (2.50–6.11)
Gram stain
Gram-positive cocci
3.63 (2.18–6.04)
Gram-negative cocci
1.00 (reference)
Other bacteria
1.27 (0.40–4.09)
Negative
1.21 (0.59–2.49)
a
Odds ratios (ORs) are calculated in 10-year increments for
age.
b
Glasgow Coma Scale (GCS) scores range from 3 to 14, with
14 indicating a normal level of consciousness (abnormal
flexion was omitted from the scale). The GCS score was
obtained in 694 episodes.
CI ⫽ confidence interval.
Weisfelt et al: Meningitis Risk Score
93
degrees of freedom ⫽ 4; p ⫽ 0.89. These results were
similar if patients with a negative CSF culture in the
EDS (n ⫽ 65; 22%) were excluded or after exclusion
of the 36 patients who were included in both the EDS
and the DMC. The distribution of risk scores in the
derivation and validation cohorts is shown in the supplemental materials (see Supplemental Fig 2). Although
c statistics, indicative of discriminatory ability, are
more than 0.80 for both cohorts, overlap in risk scores
in patients with a favorable and unfavorable outcome
does occur.
Fig. Prediction rule for risk for unfavorable outcome in adults
with bacterial meningitis. Tachycardia was defined as a heart
rate greater than 120 beats/min. low cerebrospinal fluid
(CSF) leukocyte count was defined as ⬍1,000 cells/mm3. Result of CSF Gram’s stain: G⫺ ⫽ gram-negative cocci; No ⫽
no bacteria; Other ⫽ other bacterial species; G⫹ ⫽ grampositive cocci. Instruction: Locate the age of the patient on the
top axis and determine how many points the patient receives.
Repeat this for the remaining five axes. Sum the points for all
six predictors and locate the total sum on the total point axis.
Draw a line straight down to the axis labeled “% unfavorable
outcome” to find the estimated probability of an unfavorable
outcome for this patient.
point, which is associated with an approximately 3%
risk for an unfavorable outcome.
The receiver operating characteristic curves of the
full multivariate model and the final nomogram model
are shown in the electronic supplement (see Supplemental Fig 1). The area under the receiver operating
characteristic curves remained largely similar after reduction of the number of prognostic variables.
The distribution of the risk scores and associated
risks for unfavorable outcome in the EDS is shown in
Table 3. Data in this cohort were complete for all 6
predictors included in the risk score in 284 of 301 patients (94%). The expected proportion of patients with
an unfavorable outcome based on nomogram was 18%
(interquartile range, 8 –24%). In 187 of 284 patients
(66%) the risk score was between 11 and 30 points,
associated with an estimated risk for an unfavorable
outcome between 8 and 40%. The concordance index
for the final risk score in the EDS was 0.81 (95% CI,
0.74 – 0.87); 0.82 (95% CI, 0.72– 0.91) in the placebo
group versus 0.79 (95% CI, 0.69 – 0.88) in the dexamethasone group. The results of the Hosmer–Lemeshow goodness of fit test were as follows: ␹2 ⫽ 1.11;
94
Annals of Neurology
Vol 63
No 1
January 2008
Discussion
This bedside risk score helps physicians to reliably
stratify bacterial meningitis patients at initial presentation with respect to the risk for an unfavorable outcome. Risk assessment can be important for physicians
because of decisions about the level of care (ward or
high-care facility) but may be even more important for
informing the patient and his or her relatives. Our risk
score is not a prediction of clinical course but of unfavorable outcome. Outcome must be regarded as the
result of clinical course. We deliberately did not include surrogate outcome measurements, such as mechanical ventilation or admission to an intensive care
unit. Our risk score is helpful only in decision making
in the early phase of admission. The use of our model
may also facilitate the interpretation of future clinical
studies in adults with bacterial meningitis because the
risk for adverse outcome or the effect of therapeutic
interventions is likely to vary in different patient
groups.26
There are many potential pitfalls in the development
and validation of prognostic models, such as overfitting
and unspecified coding or selection of variables in the
model.27,28 To reduce the risk for overfitting, we used
separate cohorts for derivation and validation of our
risk score. The large sample size in the derivation cohort enabled us to include all 22 potential prognostic
factors in a multivariate analysis, thereby preventing inappropriate selection of variables. The use of multiple
imputation is encouraged in prognostic research because it will reduce the bias that may arise from missing values that are not missing at random (ie, neck
stiffness may be absent in patient with meningitis in
deep coma).1 The analytic strategy of bootstrapping
was used to prevent the inclusion of spurious variables
in the final model. The nomogram model was then
validated in a new series of patients.
Risk estimates from the nomogram model were well
calibrated with small differences in observed and predicted number of events, and the discriminatory ability
was good but not perfect; for example, overlap in risk
scores between patients with a favorable and unfavorable outcome did occur. This means that the model
cannot fully discriminate between patients with and
Table 3. Predicted and Observed Number of Patients with an Unfavorable Outcome across the Range of Predicted
Risk Scores in the Validation Cohort
Risk Score
ⱕ10
Observed Number
of Patients in
Category
Observed Number of
Patients with an
Unfavorable Outcome
Expected Proportion of Patients
with an Unfavorable Outcome
Based on Nomogram
74 (26%)
8/74 (11%)
6%
11–20
127 (45%)
17/127 (13%)
13%
21–30
62 (22%)
20/62 (32%)
28%
31–40
16 (6%)
9/16 (56%)
50%
3/5 (60%)
79%
57/284 (20%)
18%
⬎40
5 (2%)
Total
284 (100%)
Data included 284 patients in whom data were complete for all 6 prognostic factors included in the risk score.
without an event. Discrimination becomes important if
there is a need to establish a cutoff value above and
below which patients will be treated differently. Currently, such a cutoff value is not known; therefore, this
risk score should not be used as a decision rule in the
management of individual patients. Although physicians should be careful not to interpret the estimated
risks from the model as absolute certainties, we believe
our score will predominantly be used to inform patients and families. Good calibration (eg, closeness between predicted and observed probabilities) is then a
key virtue to stratify patients in different risk categories. Furthermore, the use of our score must not lead to
a bias for the treating medical physician. For example,
a favorable score value on admission could influence
the alertness of the physician; thus, complications arising during the clinical course might be missed or diagnosed too late. Clinical deterioration of a patient
with bacterial meningitis might occur rapidly and is
still difficult to predict for individual patients.
To enhance the clinical applicability of our risk
score, we included only data that will be routinely
available within hours after presentation. Therefore,
characteristics such as length of stay on the intensive
care unit, microbial coverage of the antibiotic therapy,
and results of blood and CSF cultures were not included in our prognostic model. Knowledge of the
causative organism of meningitis is important in predicting the risk for an unfavorable outcome. Gram’s
staining of CSF, when positive, permits rapid identification of bacteria and can accurately discriminate between gram-positive and -negative bacteria as the cause
of meningitis.1,2,4
Our study has several limitations. First, most patients in the DMC did not receive treatment with adjunctive dexamethasone. Therefore, steroid therapy
could not be included in our analyses as a potential
prognostic factor. The EDS showed that treatment
with adjunctive dexamethasone, started before or with
the first dose of antibiotics, reduced the risk for unfa-
vorable outcome, including mortality, from 25 to
15%.24 A subsequent meta-analysis also showed a beneficial effect of dexamethasone on neurological sequelae.29 Recent studies showed that use of dexamethasone
was not associated with increase of cognitive impairment in surviving patients.30,31 After publication of
these results, dexamethasone is recommended in most
adults with suspected bacterial meningitis.1,32,33 In a
previous post hoc analysis of the EDS, we showed that
treatment of dexamethasone affected risk stratification.
In this study, risk factors for an unfavorable outcome
in patients treated with placebo failed to achieve statistical significance in patients who received dexamethasone therapy.34 However, logistic regression analyses in
this previous study were less refined compared with
this study (eg, imputation of missing values, bootstrapping, and backward elimination procedures were not
performed). Although adjunctive dexamethasone therapy was not included in our current prognostic model,
the concordance index remained robust after validation
in the dexamethasone-treated patients in the validation
cohort. The consistency of discriminative power in patients treated with dexamethasone and in patients
treated with placebo in our validation cohort indicates
that our risk score is likely to have good external validity. So, this is also true in patients with bacterial
meningitis treated with adjunctive dexamethasone therapy.
Second, different criteria were used for inclusion in
the derivation and validation cohorts. The derivation
cohort consisted of only CSF culture–positive patients.
Negative CSF cultures are estimated to occur in 11 to
30% of patients with bacterial meningitis.4,6 However,
the clinical presentation in patients with positive and
negative CSF cultures was similar in several studies, so
this is unlikely to have confound our results.6,7 A substantial proportion of patients with identified bacterial
meningitis (37%) were not included in the DMC,
which might have resulted in a selection bias. As described previously in the Netherlands, rates of antibi-
Weisfelt et al: Meningitis Risk Score
95
otic resistance among pneumococcal and meningococcal isolates in the DMC were low (⬍1%).4,5,35 In
other parts of the world, the rate of antimicrobial resistance of S. pneumoniae to penicillin and cephalosporins has emerged as a problem in the treatment of patients with pneumococcal meningitis.2 Nevertheless,
the DMC is the first prospective and most comprehensive cohort study on bacterial meningitis to date.
The validation cohort consisted of patients included
in a clinical trial. Patients in a clinical trial have to
fulfil explicit inclusion and exclusion criteria. Perhaps
one of the most important inclusion criteria of the
EDS was a CSF white-cell count of more than 1,000/
mm3. Patients could be included with less than 1,000/
mm3, but only with cloudy CSF or bacteria seen in
CSF Gram’s stain. In the EDS, 7% of patients had a
CSF white-cell count less than 1,000/mm3, versus 29%
in the DMC. Because low CSF white-cell count is one
of the risk factors in our model, this might have affected the validation of our model. Nevertheless, the
placebo-controlled design of the EDS presented a
unique opportunity to evaluate the validity of our risk
score in patients who did and did not receive steroid
therapy. Currently, prospective data on patients with
bacterial meningitis after implementation of routine
dexamethasone therapy are not available. We started a
new nationwide prospective study on bacterial meningitis and the first results are expected in 2007. In addition, we encourage the validation of our risk score
(“Dutch Meningitis Risk Score”) in other future datasets with dexamethasone-treated patients.
Third, several patient characteristics were recorded
according to the judgment of the treating physician.
We did not specify definitions of alcoholism, diabetes
mellitus, and immunosuppressive drugs. Although this
may be considered as a limitation, we aimed on maximal adherence of physicians. This has resulted in an
exceptionally high response rate of 76% in the DMC.
Finally, all patients in our derivation and validation
cohorts underwent lumbar puncture. In patients with
septic shock or space-occupying lesions on computed
tomography, lumbar puncture is generally not performed or postponed.1 Therefore, these patient groups
were probably only partly represented in both cohorts,
which may have resulted in an underestimation of the
rate of adverse outcome. The low rate of patients with
sepsis in the DMC may also explain why hypotension
was not identified as an independent prognostic factor
in our multivariate model, although tachycardia and a
low CSF leukocyte count can be regarded as predictors
for sepsis.4,36
In conclusion, we derived and validated the first
bedside risk score based on routinely collected data to
predict risk for adverse outcome in individual adults
with community-acquired bacterial meningitis. Several
prognostic models in bacterial meningitis have been
96
Annals of Neurology
Vol 63
No 1
January 2008
published previously3–22; however, few or none reached
clinical practice. Our model is based on the largest prospective cohort study to date, is easy to use, and includes data that will be available within 1 hour after
presentation. The score helps to identify high-risk individuals and provides important information for patients and their relatives.
The DMC study was supported by Roche Pharmaceuticals, The
Netherlands (J.d.G.). The European Dexamethasone in Adulthood
Bacterial Meningitis was supported NV Organon (J.d.G.), which
also supplied the study medication. This study was supported by
Baxter BV (J.d.G.), the Dr. Jan Meerwaldt Foundation (D.v.d.B.),
the Meningitis Research Foundation, UK (grant number 03/03,
D.v.d.B.), and the Netherlands Organization for Health Research
and Development (ZonMw; NWO-Rubicon grant 2006, grant
number 019.2006.310.001 and NWO-Veni grant 2007, grant
number 916.76.023; both to D.v.d.B.).
We are indebted to many physicians in the Netherlands and other
European countries who participated in the EDS and DMC.
References
1. van de Beek D, de Gans J, Tunkel AR, Wijdicks EF.
Community-acquired bacterial meningitis in adults. N Engl
J Med 2006;354:44 –53.
2. Weisfelt M, de Gans J, van der Poll T, van de Beek D. Pneumococcal meningitis in adults: new approaches to management
and prevention. Lancet Neurol 2006;5:332–342.
3. Aronin SI, Peduzzi P, Quagliarello VJ. Community-acquired
bacterial meningitis: risk stratification for adverse clinical outcome and effect of antibiotic timing. Ann Intern Med 1998;
129:862– 869.
4. van de Beek D, de Gans J, Spanjaard L, et al. Clinical features
and prognostic factors in adults with bacterial meningitis.
N Engl J Med 2004;351:1849 –1859.
5. Weisfelt M, van de Beek D, Spanjaard L, et al. Clinical features, complications, and outcome in adults with pneumococcal
meningitis: a prospective case series. Lancet Neurol 2006;5:
123–129.
6. Durand ML, Calderwood SB, Weber DJ, et al. Acute bacterial
meningitis in adults. A review of 493 episodes. N Engl J Med
1993;328:21–28.
7. Sigurdardottir B, Bjornsson OM, Jonsdottir KE, et al. Acute
bacterial meningitis in adults. A 20-year overview. Arch Intern
Med 1997;157:425– 430.
8. Hussein AS, Shafran SD. Acute bacterial meningitis in adults.
A 12-year review. Medicine (Baltimore) 2000;79:360 –368.
9. Olsson RA, Kirby JC, Romansky MJ. Pneumococcal meningitis
in the adult. Clinical, therapeutic, and prognostic aspects in
forty-three patients. Ann Intern Med 1961;55:545–549.
10. Weiss W, Figueroa W, Shapiro WH, Flippin HF. Prognostic
factors in pneumococcal meningitis. Arch Intern Med 1967;
120:517–524.
11. Magnussen CR. Meningitis in adults; ten-year retrospective
analysis at community hospital. N Y State J Med 1980;80:
901–906.
12. Bohr V, Rasmussen N, Hansen B, et al. Pneumococcal
meningitis: an evaluation of prognostic factors in 164 cases
based on mortality and on a study of lasting sequelae. J Infect
1985;10:143–157.
13. Bruyn GA, Kremer HP, de Marie S, Padberg GW, et al. Clinical evaluation of pneumococcal meningitis in adults over a
twelve-year period. Eur J Clin Microbiol Infect Dis 1989;8:
695–700.
14. Hoen B, Viel JF, Gerard A, et al. Mortality in pneumococcal
meningitis: a multivariate analysis of prognostic factors. Eur
J Med 1993;2:28 –32.
15. Kirkpatrick B, Reeves DS, MacGowan AP. A review of the clinical presentation, laboratory features, antimicrobial therapy and
outcome of 77 episodes of pneumococcal meningitis occurring
in children and adults. J Infect 1994;29:171–182.
16. Attia J, Hatala R, Cook DJ, Wong JG. The rational clinical
examination. Does this adult patient have acute meningitis?
JAMA 1999;282:175–181.
17. Kragsbjerg P, Kallman J, Olcen P. Pneumococcal meningitis in
adults. Scand J Infect Dis 1994;26:659 – 666.
18. Auburtin M, Porcher R, Bruneel F, et al. Pneumococcal meningitis in the intensive care unit: prognostic factors of clinical
outcome in a series of 80 cases. Am J Respir Crit Care Med
2002;165:713–717.
19. Lu CH, Huang CR, Chang WN, et al. Community-acquired
bacterial meningitis in adults: the epidemiology, timing of appropriate antimicrobial therapy, and prognostic factors. Clin
Neurol Neurosurg 2002;104:352–358.
20. Kastenbauer S, Pfister HW. Pneumococcal meningitis in adults:
spectrum of complications and prognostic factors in a series of
87 cases. Brain 2003;126:1015–1025.
21. Flores-Cordero JM, Maya-Villar R, Rincon-Ferrari MD, et al.
Acute community-acquired bacterial meningitis in adults admitted to the intensive care unit: clinical manifestations, management and prognostic factors. Intensive Care Med 2003;29:
1967–1973.
22. Ostergaard C, Konradsen HB, Samuelsson S. Clinical presentation and prognostic factors of Streptococcus pneumoniae
meningitis according to the focus of infection. BMC Infect Dis
2005;5:93.
23. Jennet B, Bond M. Assessment of outcome after severe brain
damage. Lancet 1975;1:480 – 484.
24. de Gans J, van de Beek D. Dexamethasone in adults with bacterial meningitis. N Engl J Med 2002;347:1549 –1556.
25. Little R, Rubin D. Statistical analysis with missing data. New
York: Wiley, 1987.
26. Rothwell PM, Warlow CP. Prediction of benefit from carotid
endarterectomy in individual patients: a risk-modelling study.
European Carotid Surgery Trialists’ Collaborative Group. Lancet 1999;353:2105–2110.
27. Altman DG, Royston P. What do we mean by validating a
prognostic model? Stat Med 2000;19:453– 473.
28. Concato J, Feinstein AR, Holford TR. The risk of determining
risk with multivariable models. Ann Intern Med 1993;118:
201–210.
29. van de Beek D, de Gans J, McIntyre P, Prasad K. Steroids in
adults with acute bacterial meningitis: a systematic review. Lancet Infect Dis 2004;4:139 –143.
30. Weisfelt M, Hoogman M, van de Beek D, et al. Dexamethasone and long-term outcome in adults with bacterial meningitis. Ann Neurol 2006;60:456 – 468.
31. Hoogman M, van de Beek D, Weisfelt M, et al. Cognitive outcome in adults after bacterial meningitis. J Neurol Neurosurg
Psychiatry (in press).
32. Fitch MT, van de Beek D. Emergency diagnosis and treatment
of adult meningitis. Lancet Infect Dis 2007;7:191–200.
33. van de Beek D, de Gans J. Dexamethasone in adults with
community-acquired bacterial meningitis. Drugs 2006;66:
415– 427.
34. Weisfelt M, van de Beek D, de Haan RJ, de Gans J. Dexamethasone and prognostic factors in adults with bacterial meningitis. J Neurol 2006;5:73–78.
35. van de Beek D, de Gans J, Spanjaard L, et al. Antibiotic guidelines and antibiotic use in adult bacterial meningitis in The
Netherlands. J Antimicrob Chemother 2002;49:661– 666.
36. Weisfelt M, van de Beek D, Spanjaard L, et al. Communityacquired bacterial meningitis in older people. J Am Geriatr Soc
2006;54:1500 –1507.
Weisfelt et al: Meningitis Risk Score
97
Документ
Категория
Без категории
Просмотров
0
Размер файла
122 Кб
Теги
adults, outcomes, scorm, meningitis, aris, bacterial, unfavorable
1/--страниц
Пожаловаться на содержимое документа