Comparison of cerebrospinal fluid lactate with physical, cytological, and other biochemical characteristics as prognostic factors in acute bacterial meningitis.

OBJECTIVE
Bacterial meningitis (BM) is associated with a high morbidity and mortality. Cerebrospinal fluid (CSF) lactate may be used as a prognostic marker of this condition. We hypothesized that CSF lactate levels would remain elevated in participants who died of acute BM compared with those who recovered from this disease. To evaluate the potential use of lactate and other CSF biomarkers as prognostic markers of acute BM outcome.


METHODS
This retrospective, longitudinal study evaluated dynamic CSF biomarkers in 223 CSF samples from 49 patients who fulfilled the inclusion criteria of acute BM, with bacteria identified by CSF culturing. The participants were grouped according to outcome: death (n = 9; 18.37%) and survival (n = 40; 81.63%). All participants received appropriate antibiotic treatment.


RESULTS
In the logistic regression model, lactate concentration in the final CSF sample, xanthochromia, and CSF glucose variation between the first and last CSF samples were predictors of a poor outcome (death). In contrast, decrease in CSF white blood cell count and CSF percentage of neutrophils, increase in the percentage of lymphocytes, and normalization of the CSF lactate concentration in the last CSF sample were predictors of a good prognosis.


CONCLUSION
The study confirmed the initial hypothesis. The longitudinal analysis of CSF lactate is an important predictor of prognosis in acute BM.

Acute bacterial meningitis (BM) is a life-threatening syndrome, with mortality between 3% and 21%, and may have severe neurological sequelae 1 . Early diagnosis and antibacterial therapy are essential for a positive outcome 2 . The differential diagnosis between BM and viral meningitis is crucial for treatment and prognosis 3,4,5 . The concentration of CSF lactate, in addition to providing information about the differential diagnosis between BM and viral meningitis 1,6,7 , holds potential as a prognostic marker in acute BM. Experimental animal studies have shown that elevated CSF lactate levels persisting after appropriate antibiotic treatment are associated with increased morbidity and mortality, and may facilitate prognostic prediction 8 . Several studies showed that CSF lactate levels were significantly different in clinical samples obtained from patients with a positive and a negative prognosis 9,10,11,12,13,14,15 .
We hypothesized that CSF lactate levels remained high among participants with acute BM who died, in contrast to the group who survived. To test this hypothesis, we calculated CSF lactate levels longitudinally, and evaluated whether these values could be used as a prognostic factor in BM with that of other physical, biochemical, and cytological characteristics of the CSF.

Participants
A retrospective study was conducted with CSF reports from the Clinical Pathology Laboratory of the Hospital de Clínicas, Universidade Federal do Paraná (HC-UFPR) during the period between 2006 and 2017. The Institutional Research Review Board of the HC-UFPR, Brazil, approved this study.
Patients clinically suspected of having acute BM were selected randomly, based on microbiological identification of the etiological agent. Data regarding the CSF biochemical and cytological characteristics, as well as demographic data, were obtained from the hospital computer records. The inclusion criteria were as follows: 1) CSF samples from patients attending HC-UFPR with clinical suspicion of acute meningitis; 2) CSF cultures positive for bacteria; and 3) quantification of lactate concentration in the CSF. The exclusion criteria of the study were as follows: 1) patients with CSF cultures negative for bacteria; 2) virus identified in the CSF; 3) samples obtained from other services. Lactate was measured in all CSF samples included throughout the study period.

Sample collection
The CSF samples were collected for clinical purposes. To study the dynamics of CSF biomarkers, all CSF samples collected during each patient's hospitalization period were included. Two hundred and twenty-three CSF samples from 49 patients fulfilled the inclusion criteria of the study.
Patients were allocated to two groups according to the outcome: negative outcome (death; n = 9; 18.37%) and positive outcome (survival; n = 40; 81.63%). All participants received appropriate antibiotic treatment according to the sensitivity of the CSF bacteria isolated.
The sites of CSF puncture were comparable between groups. In the first CSF sample, 19 samples (38.78%) were collected by lumbar puncture ( The time interval between the last CSF puncture and hospital discharge in the whole group was median (interquartile range [IQR]:) 5 days (2; 13).
Overall, a median of four CSF punctures (IQR: 1; 6) were performed during hospitalization. In the positive-outcome group, the median number of punctures performed was four (1.5; 6); in the negative-outcome group, it was two (1; 7) (p = 0.518). The median interval between the first and last CSF sample in all patients who underwent more than one CSF puncture (n = 36) was 26 days (13; 44.5). In the positive-outcome group, this median interval was 26 days (13.25; 40.25); in the negative-outcome group it was 34.5 days (8.75; 47.5).

CSF biochemistry and cytology
After CSF puncture, the CSF samples were kept at room temperature and analyzed up to 30 minutes after reception in the clinical pathology laboratory.
Lactate was quantified in the CSF by amperometry (RAPID Point 500, Siemens, New York, NY, USA). Total protein in the CSF was quantified using benzethonium chloride as the precipitation reagent. Glucose in the CSF and the plasma were determined using the hexokinase/G-6-PDH method (Architect, Abbott Diagnostics, Abbot Park, IL). Total white blood cell count (WBC) per mm 3 in the CSF was quantified in fresh, non-centrifuged CSF, in a Fuchs-Rosenthal chamber. Cerebrospinal fluid pleocytosis was defined as a WBC count of > 5 cells/mm 3 . For differential WBC counts, CSF samples were concentrated using a Shandon cytospin (Thermo-Shandon; Pittsburgh, PA) and the slides were stained using May-Grünwald-Giemsa staining. Xanthochromia in the CFS 16 was assessed by visual inspection and quantified using the color index by comparing the supernatant after centrifugation with a solution of potassium bichromate 17 .

Microbiological diagnostic tests for bacteriology
All CSF samples were subjected to direct microscopic examination (Gram-stain smear) and culturing. The CSF specimens were inoculated onto agar plates (5% sheep's blood agar and supplemented chocolate agar plates) and incubated at 37°C for 24-48 hours. For positive cultures, bacteria were identified using a VITEK ® 2 compact system (BioMérieux Vitek, Inc., Hazelwood, MO).

Statistical analyses
The variables studied in the CSF were the color index, WBC, neutrophils (%), lymphocytes (%), glucose (mg/dL), CSF/blood glucose ratio, total protein (mg/dL), and lactate concentration (mmol/L). All variables were studied in the first and last CSF samples. In addition, the mean value and variation (value of first CSF sample subtracted from that of last CSF sample) of all CSF samples from the same participant were assessed. For the variation in the CSF variables, the median of the variable in serial CSF samples was used as a cut-off point.
Results are presented as median (IQR) or number (n) and percentage (%), as appropriate. Categorical variables were compared between groups using Fisher's exact test and continuous variables were compared using the Mann-Whitney or Kruskal-Wallis test for non-parametric data, as appropriate. Dependent continuous variables were compared using the Wilcoxon test.
To estimate survival from lifetime data, Kaplan-Meier curves were constructed and compared using the log-rank test. For the cut-off point of variables, the median of overall samples or the reference values of the laboratory parameters were used.
Logistic regression analysis was used to estimate or predict the value of potential predictors of prognosis, such as sex, race, age, and CSF characteristics (color index, leukocyte, neutrophil percentage lymphocyte percentage, glucose, protein, and lactate levels, and CSF/blood glucose ratio). The odds ratio and 95% confidence intervals (CI) were used to quantify the strength of the associations. Results were considered significant at the 5% alpha level.
The bacteria isolated by CSF culture in the first CSF sample for each of the groups are shown in Table 1. Overall, 31 samples (63.26%) were positive for Gram-positive bacteria and 18 samples (36.73%) for Gram-negative bacteria. There was no difference in CSF lactate levels between the groups with Gram-positive or Gram-negative bacteria in the CSF (p = 0.736).
Overall, the mortality rate in this series was 18.37%. In the positive-outcome group, 26 patients (65%) had Gram-positive and 14 patients (35%) had Gram-negative bacteria. In the negative-outcome group, five patients (55.55%) had Grampositive and four patients (44.44%) had Gram-negative bacteria. Thus, mortality was not dependent on the type of bacteria involved (p = 0.708).
The physical, cytological, and biochemical characteristics of the CSF in the first and last CSF samples based on the   [5] ; analyzed with Fisher's exact test.
outcome are shown in Table 2 and Figure 1. The dynamics of CSF biochemical and cell characteristics, described by the variation and mean of the characteristics of the sequential CSF samples, are shown in Table 3 and Figure 1. The percentage of lymphocytes and glucose levels were higher in the first CSF sample in the negative-outcome group than in the positive-outcome group (p = 0.025 and 0.032, respectively). In the last CSF sample, lactate (Figure 1), color index, neutrophil percentage, and total protein levels were higher in the negative-outcome than in the positive-outcome group (p = 0.002; p = 0.034; p = 0.014; and p = 0.048, respectively). The lymphocyte percentage was lower in the negative-outcome group (p = 0.041).
The variation between the first and last CSF sample showed increased WBCs and lactate levels in the last CSF sample, and decreased lymphocyte percentage and glucose levels (p = 0.017; p < 0.001; p = 0.010; p = 0.006, respectively; Table 3). The mean value of all CSF samples collected per patient was higher in the negative-outcome than in the positive-outcome group for color index (p = 0. 047). The lactate concentration was numerically higher in the negative-outcome group, but this difference was not statistically significant (Table 3).
In the negative-outcome group, CSF cultures and Gramstain smears were positive in the last CSF sample in two of the nine cases (22.22%). Both cases showed the same bacteria that were identified in the first CSF samples. In the positive-outcome group, the CSF culture was negative in the last sample for all participants with more than one CSF sample (n = 40). All findings in the last CSF sample of the negative-outcome group suggested a lack of response to the antibiotic therapy.
There was a significant difference in survival time between groups, i.e., more deaths occurred among patients who were older than 29 years, which was the overall median age of patients included in the study (p = 0.027). Death was more frequent where the mean lactate concentration in serial CSF samples was higher than 3.5 mmol/L; CSF lactate in the last CSF sample was higher than 3.5 mmol/L; and CSF  was xanthochromic in the last CSF sample, independent of the value of the CSF color index (p = 0.050; p < 0.001; p = 0.027, respectively; Figure 2). There was no difference in survival time in relation to the other analyzed variables.
In multivariable logistic regression analysis, the following variables differed significantly between the negative-outcome and positive-outcome groups: the lactate concentration in the last CSF sample (odds ratio [OR]: 1.513; 95% CI: 1.178-2.527;

Multibacterial meningitis
Multibacterial meningitis, i.e. different bacteria isolated from sequential CSF samples, was identified in eight patients (16.33%). Of these, two patients (22.22%) were in the negative-outcome group and six patients (15.63%) were in the positive-outcome group (p = 0.628). In the group with multibacterial meningitis, two patients (25%) died, whereas in the group with a single bacterial strain, six patients (75%) died (p = 0.601).
The relationship between the CSF lactate concentration in patients with multibacterial meningitis and the identified bacteria according to the outcome is shown in Figure 3. A peak of CSF lactate levels (> 3.5 mmol/L) was associated with a positive bacterial culture in six of the 19 (31.58%) CSF samples.

DISCUSSION
In the present series, the mortality rate in patients with acute BM was 18% and this was not dependent on the type of causal bacteria. In developed countries, the mortality rate of acute BM is 10-30%, while in low-income countries it can reach 50% 2 .
In this study, multivariable logistic regression analysis showed that the persistence of a high lactate concentration in the last CSF sample; xanthochromia, indicated by a high mean color index in longitudinal CSF samples; and lower variation in glucose levels between the first and last CSF sample differed significantly between the negative-outcome group and the positive-outcome group.
The value of CSF lactate levels for predicting the outcome of BM is still debatable. In this study, in a univariate analysis, the CSF lactate concentration in the first CSF sample did not discriminate the negative-outcome from the positive-outcome group and was not useful for predicting outcome, in agreement with previous reports 14,18,19 but in contrast to results from several other studies 9,10,11,12,13 . In the present study, the CSF lactate levels of the last sample were higher in the negative-outcome than in the positive-outcome group, corroborating the findings of a previous study 14 . In addition, we demonstrated that variation in lactate levels, calculated by the difference between the first and last CSF, differed between the two groups.
On the other hand, the present study showed that a rapid decrease in the CSF lactate concentration during treatment was indicative of therapeutic efficacy and good prognosis, confirming the results of previous studies 14,20,21 . The CSF samples with lactate concentrations higher than 3.5 mmol/L were less frequent in the positive-outcome group and more common in the negative-outcome group. Thus, lactate may be used as a marker of positive as well as negative outcome. In the present study, xanthochromia, quantified by color index, differed significantly between the negative-outcome and positive-outcome groups; specifically, xanthochromia was associated with a negative prognosis. Color index is a visual comparative method for quantifying CSF xanthochromia; it is easy and is not expensive to perform, although it has a low sensitivity (47%) 17 . Xanthochromia has not been investigated in the past as a prognostic marker in acute BM; however, this finding must be interpreted with caution, as the present study included CSF samples collected from different puncture sites. Further studies are necessary to confirm our findings. The causes of CSF xanthochromia in acute BM are increased CSF WBC (> 200 cells/mm³), bacterial concentration > 10 5 CFU/mL, and high protein concentration (> 150 mg/dL). Additionally, other causes of xanthochromia can be considered as the presence of red blood cells (> 400 red blood cells/mm³), high intake of carotenoids, hyperbilirubinemia, and meningeal melanomatosis 16 .
In the present study, the CSF WBC count was numerically lower in the negative-outcome than in the positiveoutcome group, although the difference was not statistically significant. Previous studies showed that a low CSF WBC count is related to a poor prognosis 22 . A low CSF WBC count was also described in sepsis in adults with pneumococcal meningitis 23 . In our study, the last CSF sample showed a persistent elevation of WBC in both groups. In the positive-outcome group, the differential WBC showed predominance of lymphocytes, while in the negative-outcome group, neutrophils predominated. All findings in the last CSF sample in the negative-outcome group suggested a lack of response to antibiotic therapy, independent of the presence of bacteria in the last sample. The CSF findings in patients who have received previous antibiotics may differ 24,25 ; the positivity of conventional bacteriological diagnostic methods (CSF culture and Gram stain) in partiallytreated BM is reduced from 50% to 20%; the relationship between polymorphonuclear cells and lymphocytes in CSF may be reversed [26][27][28] .
In this study, multibacterial meningitis was not related to an increased rate of mortality, although the number of participants was small. Multibacterial infection is reported as a negative prognostic factor by some authors 6,29,30 . In the present study, the dynamics of the CSF lactate concentration in patients with multibacterial meningitis showed a peak of CSF lactate in those samples with positive bacterial cultures.
In the present study, older age was also considered as a factor of a negative prognosis, confirming the results of previous studies. Nevertheless, the age of participants considered in previous studies was higher than in our study 29,30 .
In this study, in a univariate analysis, decreased lactate levels, CSF WBC count, CSF percentage of neutrophils, increased lymphocyte percentage, and inversion of the CSF neutrophil/lymphocyte ratio were identified as factors predicting a good prognosis in a longitudinal CSF analysis, corroborating the findings of an earlier study 20 . In agreement with this previous study 20 , we did not find the CSF total protein levels to have a prognostic value.
Several studies addressing prognostic factors in BM have been performed; all were retrospective and relatively small. The present study differed from other studies in that it examined the dynamics of several CSF biomarkers to evaluate mortality-predicting factors from serial CSF samples. Other strengths of this study were the high number of overall CSF samples. Furthermore, in contrast to previous studies, our analysis included communitarian as well as nosocomial bacterial meningitis cases.
The main limitations of this study are its retrospective design; the inclusion of children and adults; the limited number of patients in the negative-outcome group; and the inclusion of only patients with definite acute BM, although the latter point could also be considered a positive point. The aim of this study was to investigate the impact of CSF biomarkers on prognoses, and thus no other measures of disease severity were analyzed. Meningitis prognosis is associated with several factors of diverse nature. Another important limitation was that this study included CSF samples collected from different sites of puncture; therefore, the results of total protein and xanthochromia should be interpreted with caution, as there is a gradient of CSF total protein from the ventricular to the spinal CSF. To address this, we performed a qualitative analysis, using the reference values for the location of the CSF puncture 5 . Concerning lactate, the site of CSF puncture did not interfere with the lactate results, as there was no difference in the reference values of lactate between ventricular and lumbar CSF 31 .
In conclusion, this longitudinal analysis adds important information about lactate and glucose levels as risk factors in the prognosis of acute BM. The current study suggests that persistently elevated CSF lactate and lower variation in glucose levels in patients with acute BM receiving adequate antibiotic therapy may indicate a negative prognosis, with increased mortality. The present study enforces the importance of longitudinal studies of CSF biomarkers and the fact that dynamic studies may provide more useful information for predicting disease outcome.