Eur J Clin Microbiol Infect Dis 2016 Mar; 35(3):397-403.

Neutrophil-to-lymphocyte ratio in the differential diagnosis of acute bacterial meningitis

Mentis AFA, Kyprianou MA, Xirogianni A, Kesanopoulos K, Tzanakaki G.

National Meningitis Reference Laboratory, Department of Public Health, National School of Public Health, Athens, Greece



The differential diagnosis of acute community-acquired meningitis is of paramount importance in both therapeutic and healthcare-related economic terms. Despite the routinely used markers, novel, easily calculated, and rapidly available biomarkers are needed particularly in resource-poor settings. A promising, exponentially studied inflammatory marker is the neutrophil-to-lymphocyte ratio (NLR), albeit not assessed in meningitis. The aim of this study was to investigate the utility of the NLR in the differential diagnosis of acute meningitis. Data on cerebrospinal fluid (CSF) and blood leukocyte parameters from more than 4,000 patients diagnosed with either bacterial or viral meningitis in Greece during the period 2006-2013 were retrospectively examined. The diagnostic accuracy of the NLR and neutrophil counts in CSF and blood were evaluated by receiver operating characteristic curves. The discrimination ability of both the NLR and neutrophil counts was significantly higher in CSF than in blood. The optimal cutoff values of the NLR and neutrophil counts were 2 in CSF vs 8 in blood, and 287 cells in CSF vs 12,100 cells in blood, respectively. For these values, sensitivity, negative predictive value, and odds ratio were statistically significantly higher in CSF than blood for both markers. Logistic regression analysis showed that the CSF NLR carries independent and additive information to neutrophil counts in the differential diagnosis of acute meningitis. This study is the first one to assess NLR in acute meningitis, providing promising results for its differential diagnosis.

PMID: 26792137



Acute microbial meningitis is a major global health problem with an annual incidence exceeding 15 million people worldwide and a global mortality of more than 300,000 (1). The vast majority of these deaths are a consequence of the bacterial cause of the infection, whereas the viral form has low comorbidity and rarely leads to adverse clinical outcomes. Therefore, the early identification of the source of the infection is crucial (2). As a matter of fact, the increased mortality in bacterial meningitis is usually associated with delay in receiving antimicrobial treatment. Also, in the case of viral meningitis, unnecessary treatment with antibiotics may lead to drug resistance and adverse effects, as well as increased expenses in health care (3).

Because bacteria cause an upsurge of the organism’s neutrophil production, the predominant biomarker used for the differential diagnosis (DD) of bacterial from viral meningitis is the neutrophil count (NC) in the patient’s cerebrospinal fluid (CSF) or, when this is not available, in the patient’s blood. Unfortunately, the exclusive use of this biomarker leads to a high proportion of both false positive and false negative misidentifications (2).

During the last few years, a novel biomarker has been widely studied: the “neutrophil-to-lymphocyte ratio” (NLR) which is the ratio of NC divided by lymphocyte count. The blood NLR seems to be a good indicator of the severity and predictor of clinical outcome in a range of diseases from cancer and cardiovascular events to community-acquired infections, i.e., pneumonia and bacteremia.

However, the NLR and especially, the CSF’s NLR has not been studied on community-acquired meningitis. In our study, we ask two questions regarding the DD of acute bacterial meningitis based on the NC and the NLR.

The first question is which fluid – CSF or blood –  provides measurements of the above parameters that help discriminate more efficiently the bacterial from viral meningitis. It may seem rather superfluous to ask this question since meningitis affects the central nervous system causing inflammation of the meninges. Although the first fluid where the organism’s response to that kind of infection is expectedly the CSF, this observation does not respond conclusively whether or not the CSF parameters are significantly better in differentiating bacterial from viral meningitis. Also, it does not provide an answer to whether or not blood parameters can be used as a viable alternative in the DD, given the simpler and less invasive and character of blood collection.

The second question refers to whether or not the NLR can provide independent and additional information to NC and thus, improve the above DD, which will by then be based on parameters that are accessible upon the patient’s initial samples collections, in contrast to laborious techniques.

The sample of the study is comprehensive and epidemiologically robust, as it practically entails all cases of meningitis reported in Greece from the beginning of 2006 up to the end of 2013. The adult and children’s hospitals, in which all 4,339 patients were hospitalized, followed the standard protocol of sending CSF or/ and blood samples to the Greek National Meningitis Reference Laboratory. All samples were analyzed for the presence of bacterial meningitis both with conventional bacteriological procedures, i.e., culture for blood samples, gram stain and latex agglutination tests for CSF samples, as well as multiple polymerase chain reaction (mPCR) assays that have been developed as in-house methods by our laboratory (4, 5).

To determine the ability of the NC or the NLR, in blood or CSF, to discriminate against the binary variable, i.e., “bacterial” vs. “viral” meningitis, we used the so-called Receiver Operating Characteristics (ROC) analysis. When an area under the curve (AUC), while charting “Sensitivity” vs. “1-Specificity” of each examined parameter, is significantly greater than 0.5, then the parameter in question can predict the outcome better than pure chance. Once the significance of the discrimination ability is established, we then calculate the “cut-off” value of the parameter that yields the best tradeoff between sensitivity and specificity. On this basis, the parameter itself is, also, converted to a binary variable with the corresponding values being the presence or absence of the criterion. After that, the analysis boils down to that of a split 2×2 contingency table (the joint presence or absence of the two criteria versus the source of infection), where the important point is the relative distribution of true positive, true negative, false positive and false negative predictions.


Figure 1Figure 1. ROC analysis for the discrimination ability of blood NC and blood NLR on bacterial vs. viral meningitis. The analysis was based on 1,979 blood samples.


The ROC analysis of the NC and NLR in the blood (Figure 1) proves that they both have significant discrimination ability to differentiate bacterial from viral meningitis. However, the values of the AUCs are not anticipative of satisfactory DD. As shown in Figure 2, more than half of blood samples meet none of the criteria, as both the NC and NLR values are below the cutoff values of 12,100 cells and 8, respectively. However, almost 35% of these samples are from patients diagnosed with bacterial meningitis; this means that the DD produces a high proportion of false negatives.



Figure 2

Figure 2. Distribution of meningitis cases based on the combination of the two biomarkers in blood. Inside the columns, the percentage of cases of bacterial meningitis depending on the presence of positive or negative predictors in the blood (odds ratios [OR] with 95% confidence intervals [CI]).


The AUCs of the same parameters in the CSF are significantly higher than the respective ones in blood, and thus make a more promising model for DD (Figure 3). Once again, as figure 4 shows, more than half of the CSF samples do not meet any of the two criteria, as both the NC and NLR values are below the cutoff values of 287 cells and 2, respectively. However, in this fluid, only 7.8% of these samples were from patients subsequently diagnosed with bacterial meningitis. At the other side, the presence of both CSF criteria is significantly associated (88.9%) with the presence of bacterial meningitis. The presence of only one of the two criteria leads to inconclusive conclusions but luckily, the number of patients with only one positive criterion is relatively small.


Figure 3

Figure 3. ROC analysis for the discrimination ability of CSF NC and blood NLR on bacterial vs. viral meningitis. The analysis was based on 2815 blood samples.


The above results provide important clues to both the questions posed. Firstly, CSF samples reflect more accurately the underlying meningeal infection than blood samples. Secondly the current findings confirm that the CSF neutrophil count is an indispensable biomarker for the DD between bacterial and viral meningitis; however, they also advocate that introducing a collateral biomarker, the NLR, significantly improves the above diagnostic questions. It seems that these two biomarkers – NC and NLR – operate in conjunction, i.e., the concurrent presence of both parameters is highly symptomatic with bacterial infection, while the combined absence suggests viral meningitis. It then follows that the presence of only one of the two biomarkers, which happens in a comparatively small proportion of patients, cannot lead to a concrete diagnosis.

The above pinpoint to the fact that the CSF NLR provides independent and supplementary information to the CSF NC for the DD between bacterial and viral meningitis. The practical importance of this finding mainly lies in the fact that the NLR can be directly calculated from already measured CSF parameters. Therefore, its introduction comes at no additional costs, an issue particularly important for resource-poor countries. Furthermore, using the calculated cutoffs for CSF NC and NLR (287 cells and 2, respectively), the clinician can arrive at the DD in a very simple manner.



Figure 4

Figure 4. Distribution of meningitis cases based on the combination of the two biomarkers in the CSF. Inside or above the columns, the percentage of cases of bacterial meningitis depending on the presence of positive or negative predictors in the CSF (odds ratios [OR] with 95% confidence intervals [CI]).



  1. GBD 2013 Mortality and Causes of Death Collaborators. Global, regional, and national age-sex specific all-cause and cause-specific mortality for 240 causes of death, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2015; 385(9963): 117-71.
  2. Gray LD, Fedorko DP. Laboratory diagnosis of bacterial meningitis. Clin Microbiol Rev. 1992; 5(2):130-45.
  3. Giulieri S, Chapuis-Taillard C, Jaton K, Cometta A, Chuard C, Hugli O, Du Pasquier R, Bille J, Meylan P, Manuel O, Marchetti O. CSF lactate for accurate diagnosis of community-acquired bacterial meningitis. Eur J Clin Microbiol Infect Dis. 2015; 34(10):2049-55.
  4. Tzanakaki G, Tsopanomichalou M, Kesanopoulos K, Matzourani R, Sioumala M, Tabaki A, Kremastinou J. Simultaneous single-tube PCR assay for the detection of Neisseria meningitidis, Haemophilus influenzae type b and Streptococcus pneumoniae. Clin Microbiol Infect. 2005; 11(5):386-90.
  5. Xirogianni A, Tzanakaki G, Karagianni E, Markoulatos P, Kourea-Kremastinou J. Development of a single-tube polymerase chain reaction assay for the simultaneous detection of Haemophilus influenzae, Pseudomonas aeruginosa, Staphylococcus aureus, and Streptococcus spp. directly in clinical samples. Diagn Microbiol Infect Dis. 2009; 63(2):121-6.



Dr. Georgina Tzanakaki

Head, National Meningitis Reference Laboratory

Department of Public Health

National School of Public Health.

196 Alexandras Avenue,

115 21 Athens, Greece.

Tel.: + 30 213 20 10 267, + 30 213 20 10 268

FAX: + 30 210 64 23 041




Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann