Introduction

Tuberculosis (TB) is a major global health issue, affecting approximately 10 million individuals annually and serving as the leading cause of death from a single infectious disease.¹ Tuberculous meningitis (TBM) is acknowledged as the most severe form of TB infection,^2,3 accounting for nearly 40% of TB-related mortality⁴ and substantially increasing the likelihood of neurological disability.⁵ Therefore, it is of great importance to identify individuals at high risk of neurological disability in TBM patients, as they may require more intensive therapeutic interventions to improve clinical outcomes.

Recent studies have challenged the traditional view of the brain as an immune-privileged organ due to the presence of the blood-brain barrier.^6–8 These studies demonstrate that peripheral immune cells can penetrate the blood-brain barrier during neurological disorders.^9,10 Previous research has suggested that dead neuronal cells release inflammatory mediators within brain tissue,¹¹ triggering the infiltration of peripheral immune cells into local tissue and leading to neurological deterioration.¹² To mitigate the overactive immune response, national guidelines recommend early glucocorticoid (GC) treatment in patients with TBM.^13,14 GCs have been shown to be effective anti-inflammatory agents, significantly reducing TBM-related mortality rates.¹⁵ However, concerns have been raised regarding the clinical efficacy of GCs in TBM patients co-infected with HIV. A recent clinical trial indicated that dexamethasone did not provide substantial benefits in terms of survival or functional improvement for individuals with HIV.¹⁶ It is speculated that the effectiveness of GCs may vary among specific subgroups of TBM patients; however, identifying those who may benefit from GC treatment remains unclear.¹⁶ A retrospective cohort study of COVID-19 patients revealed a significant association between elevated levels of systemic inflammation on admission and an increased risk of mortality, with GC therapy demonstrating a reduction in mortality specifically among high-risk individuals identified by systemic inflammation level.¹⁷ Furthermore, another study focused on patients with acute respiratory distress syndrome also found a positive correlation between elevated systemic inflammation and a favorable response to GC treatment.¹⁸ This finding suggests that systemic inflammation levels may serve as a novel marker for predicting treatment outcomes.¹⁸ Nevertheless, no prior studies have investigated whether systemic inflammation can predict clinical outcomes and the efficacy of GC therapy in TBM patients.

In this multi-center, retrospective cohort study of TBM from China, we hypothesized that elevated levels of systemic inflammation are associated with early neurological deterioration and subsequent disability, and may also influence the outcomes of GC treatment.

Materials and Methods

Study Population

We conducted a retrospective cohort study across five hospitals in China from January 2011 to December 2021. We investigated patients who received a laboratory-confirmed or clinical diagnosis of TBM according to established guidelines.¹⁹ The diagnostic criteria for TBM were categorized into definite, probable, and possible TBM.

Definite TBM was confirmed by the presence of acid-fast bacilli in cerebrospinal fluid (CSF), a positive culture for mycobacterium, or detection of mycobacterial nucleic acids via polymerase chain reaction in CSF. Probable TBM was defined as patients who fulfilled clinical criteria with a total diagnostic score of 10 or more points (when brain imaging was not available) or 12 or more points (when brain imaging was available). For possible TBM, patients needed to meet clinical criteria with a total diagnostic score ranging from 6 to 9 points without brain imaging or 6 to 11 points with brain imaging.¹⁹ This study was approved by the ethics committee at Beijing Chest Hospital (Approval number: YJS-2022-07). Informed consent was waived due to the retrospective nature of the study, all patient data was anonymized or maintained with confidentiality. The study protocol was reported in accordance with the “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) guideline.²⁰

Data Collection

Demographic information, clinical manifestations, vital signs (temperature, heart rate, and respiratory rate), laboratory results, radiological findings, and treatment were collected. The clinical severity at admission was staged according to the British Medical Research Council (BMRC) grade.²¹ The BMRC grade criteria included two main dimensions: impaired consciousness status and neurological deficits. A clinical stage between grade I and III was assigned for diagnosis. To assess neurological deficit, we utilized the Expanded Disability Severity Score (EDSS) at baseline and 7 days after therapy. The EDSS score, ranging from 0 (no disability) to 10 (death),^22,23 was used to identify early neurological deterioration, defined as an increased score≥1. Baseline peripheral blood cell parameters, including white blood cell count, neutrophils, monocytes, lymphocytes, and platelets, were collected within 24 hours after admission before treatment initiation. Follow-up parameters were obtained at 3 days post-treatment. Additionally, cerebrospinal fluid (CSF) findings, such as white blood cell count, mononuclear cells, multinuclear cells, glucose and chloride concentration, and protein levels, were collected based on the first examination after admission.

Definition of Systemic Inflammation and Follow-Up Outcome

Systemic inflammation was assessed using the neutrophil-to-lymphocyte ratio (NLR), with NLR_baseline and NLR_after representing the baseline and post-treatment systemic inflammation, respectively. The NLR_change was computed by subtracting NLR_baseline from NLR_after, while the NLR_ratio was determined by dividing NLR_after by NLR_baseline. The baseline NLR values were classified into four categories: Q1 (NLR≤2.8), Q2 (2.8< NLR≤4.5), Q3 (4.5<NLR≤8.0), and Q4 (NLR>8.0).

We also evaluated the criteria of the systemic inflammatory response syndrome (SIRS) within 24 hours after admission to define systemic inflammation. SIRS were diagnosed if at least two of the following criteria were met: temperature over 38°C, respiratory rate over 20 cycles per minute, heart rate over 90 bpm, and leukocytosis/leukopenia/bandemia (leukocytes >12×10⁹/L, <4×10⁹/L, or bandemia ≥10%).²⁴

Following discharge, patients underwent follow-up assessments at 3, 6, and 12 months at local centers to evaluate their clinical status. The outcome measure was determined using the modified Rankin Scale (mRS), with a poor functional outcome (mRS score of 3–6 at 12 months) serving as the primary endpoint. Information on mRS scores was collected through telephone interviews or face-to-face follow-up sessions with patients or their family members.

Statistical Analyses

The sample size was calculated based on the number of outcome events. We estimated that 10–15 potential variables are significantly associated with poor outcome in TBM. The minimum sample size required approximately 100–150 events of poor outcome to avoid violating the principle of approximately ten outcome events per variable in generalized linear mixed models.²⁵ Descriptive characteristics were reported as percentages for categorical variables or mean with standard deviation for continuous variables. Chi-square or Fisher exact test, Student t-test, or the Mann–Whitney U-test were performed for statistical analysis when appropriate. To account for the expected correlation in outcomes within different centers, a two-level generalized linear mixed model was fitted using a logit link. This model was utilized to analyze potential predictors associated with clinical outcomes, with patients at the first level and hospitals at the second level. Additionally, the net reclassification index (NRI) and integrated discrimination index (IDI) were employed to quantify the improvement in correct reclassification and sensitivity by adding different NLR parameters to outcome models.

To mitigate the influence of confounding factors, propensity score matching (PSM) was conducted. Subgroup analysis was performed to identify factors associated with dexamethasone treatment. All tests were two-tailed, with a significance level set at P<0.05. All analyses were conducted using R version 4.2.0.

Results

Baseline Characteristics

A total of 1203 patients were included in the analysis, as depicted in the study flowchart (Figure S1). The baseline characteristics of the participants are detailed in Table 1, revealing that 636 (52.8%) were male, with a mean age of 38.6 (18.2) years. Based on the diagnostic criteria, 1024 (85.1%) patients had a clinical diagnosis of TBM, while 179 (14.9%) were laboratory-confirmed TBM. At admission, 603 (50.1%) patients were classified as BMRC grade I, 444 (36.9%) as grade II, and 156 (12.9%) as grade III.

Table 1 Baseline Characteristics of Study Participants

Analysis of the Association Between NLR and Early Neurological Deterioration

Following a 7-day follow-up after admission, a total of 144 (13.6%) participants experienced early neurological deterioration. Univariate analysis indicated that several factors such as age, diabetes, BMRC Grade, other coexisting infections, SIRS, meningeal enhancement, hydrocephalus and NLR, were associated with an increased risk of early neurological deterioration (Table S1). To analyze the association between NLR and early neurological deterioration, we employed various adjusted models (Table 2). After adjusting for demographic, clinical and treatment variables, NLR_after (OR=1.25; 95% CI, 1.14–1.33; P<0.001), NLR_change (OR=1.08; 95% CI, 1.03–1.12; P<0.001) and NLR_ratio (OR=1.32; 95% CI, 1.16–1.41; P<0.001) emerged as independent risk factors for early neurological deterioration.

Table 2 Association of Baseline, Follow-Up and Dynamic Change of NLR with Clinical Outcomes

Analysis of the Association Between NLR and Poor Functional Outcome

With a follow-up of 12 months, 345 (28.7%) participants exhibited poor functional outcome. Univariate analysis showed that several factors were associated with an elevated risk of poor functional outcomes, including age, diabetes, drug-resistant tuberculosis, BMRC grade, other coexisting infections, SIRS, cerebral infarction, hydrocephalus, and NLR (Table S2). To further investigate the relationship between NLR levels and poor functional outcome, we utilized various adjusted models (Table 2). After adjusting for confounding factors, NLR_baseline (OR=1.16; 95% CI, 1.08–1.20; P<0.001), NLR_after (OR=1.34; 95% CI, 1.26–1.45; P<0.001), NLR_change (OR=1.09; 95% CI, 1.03–1.11; P<0.001) and NLR_ratio (OR=1.16; 95% CI, 1.05–1.27; P=0.006) were identified as independent risk factors significantly predicting poor functional outcome. Sensitive analysis was performed in patients with different BMRC grades and those who received dexamethasone treatment (Table S3).

Discrimination and Calibration Analysis of NLR to Predictive Models

The inclusion of NLR_after to the basic model significantly improved the accuracy in predicting early neurological deterioration and poor functional outcome compared to other parameters. Specifically, the addition of the NLR_after did resulted a significant increase in NRI (12.26%, P=0.003) and IDI (3.52%, P<0.001) for early neurological deterioration, as well as NRI (14.83%, P<0.001) and IDI (7.99%, P<0.001) for poor functional outcome, respectively. Detailed results of NRI and IDI can be found in Table 3.

Table 3 Performance of Predictive Models with NLR for Clinical Outcomes

Propensity Score Matching to Explore Whether NLR Stratification Influences the Treatment Effect of Dexamethasone

After employing PSM to mitigate the impact of confounding factors, the patient groups were found to be well-balanced (Table S4). The findings indicate that dexamethasone treatment did not demonstrate an advantage in reducing the risk of early neurological deterioration (OR=0.83; 95% CI, 0.42–1.66; P=0.610; Figure 1) and poor functional outcome (OR=1.22; 95% CI, 0.49–2.11; P=0.490; Figure 2) in the highest NLR_after quartile. The treatment effect of dexamethasone on clinical outcomes did not differ across varying disease severity (Tables 4 and S5–S7) and other subgroups (Figures S2 and S3).

Table 4 Effect of Dexamethasone on Clinical Outcomes in Post-Treatment NLR Category According to Severity Stratification

Figure 1 Forestplot for NLR subgroup difference in dexamethasone treatment for early neurological deterioration. NLR baseline: Q1 (NLR≤2.8), Q2 (2.8< NLR≤4.5), Q3 (4.5<NLR≤8.0), and Q4 (NLR>8.0); NLR after: Q1 (NLR≤3.4), Q2 (3.4< NLR≤6.4), Q3 (6.4<NLR≤8.4), and Q4 (NLR>8.4); NLR change: Q1 (change≤-1.9), Q2 (−1.9< change≤0.8), Q3 (0.8<change≤3.6), and Q4 (change>3.6); NLR ratio: Q1 (ratio≤0.7), Q2 (0.7< ratio≤1.2), Q3 (1.2<ratio≤2.2), and Q4 (ratio>2.2). Abbreviation: NLR, neutrophil-lymphocyte-ratio.

Figure 2 Forestplot for NLR subgroup difference in dexamethasone treatment for poor functional outcome. NLR baseline: Q1 (NLR≤2.8), Q2 (2.8< NLR≤4.5), Q3 (4.5<NLR≤8.0), and Q4 (NLR>8.0); NLR after: Q1 (NLR≤3.4), Q2 (3.4< NLR≤6.4), Q3 (6.4<NLR≤8.4), and Q4 (NLR>8.4); NLR change: Q1 (change≤-1.9), Q2 (−1.9< change≤0.8), Q3 (0.8<change≤3.6), and Q4 (change>3.6); NLR ratio: Q1 (ratio≤0.7), Q2 (0.7< ratio≤1.2), Q3 (1.2<ratio≤2.2), and Q4 (ratio>2.2). Abbreviation: NLR, neutrophil-lymphocyte-ratio.

Discussion

The present multi-center study investigated the associations between system inflammation levels and clinical outcomes in patients with TBM. Our results indicate that the elevated levels of systemic inflammation serve as predictor of early neurological deterioration and subsequent poor outcome. However, we found that dexamethasone treatment was not associated with improvements in neurological disability, and its treatment effect did not vary significantly among patients with different levels of system inflammation.

The burden of neurological disability in TBM is substantial.⁵ Previous studies have shown that TBM patients often presented with multiple brain lesions, including tuberculoma, cerebral infarction, and hydrocephalus.^26–28 These lesions, which involve the brain parenchyma, lead to various clinical symptoms; thus, it is crucial to assess early neurological deficits in patients with TBM.²¹ Prior research has emphasized that functional assessment are vital in neurological disorders, and should be more closely monitored during treatment.^29,30 However, no previous studies have reported early neurological deterioration in TBM patients. Our study suggests a significant association between elevated NLR levels and early neurological deterioration as well as functional disability at follow-up. A previous study reported an increased NLR in pulmonary TB patients but did not examine its association with treatment outcomes.³¹ In our TBM cohort, we observed a higher proportion of pulmonary TB patients exhibited elevated NLR at baseline, this may be attributed to the combined effect of chronic lung infection and the activation of inflammatory mediators released by the brain on the peripheral immune system during acute meningitis onset.^11,12 Recent studies have shown conflicting results regarding the role of NLR as a prognostic biomarker in neurological disorders.^32–34 In a cohort study of multiple sclerosis, the author did not identify NLR as a marker of disease activity and disability;³³ however, this conclusion was drawn without considering different follow-up time points. A recent study on ischemic stroke patients found that the NLR after reperfusion therapy was associated with the severity of cerebral edema and long-term functional outcome.³⁴ In this study, our results also indicate that NLR post-treatment can serve as a significant marker for early warning signals of neurological deterioration and poor outcome in TBM patients. These findings stress the importance of monitoring NLR during treatment.

Dexamethasone has been used as an anti-inflammatory agent to reduce early mortality rates in TBM,¹⁵ however, limited evidence supports that dexamethasone treatment is associated with long-term neurological disability in TBM or bacterial meningitis.^35,36 Systemic inflammation markers are low-cost and readily available parameters that could to be useful in identifying patients who may benefit from GC,^17,18 especially when compared to the application of genotype to guide GC treatment.³⁷ In our cohort, we found that dexamethasone did not improve poor outcomes across different NLR stratifications among TBM patients, suggesting that system inflammation may be not sufficient to evaluate the efficacy of anti-inflammatory treatment. In HIV-negative TBM adults treated with dexamethasone, elevated levels of inflammatory cytokines in the CSF can predict survival response to dexamethasone. This suggests that local inflammation may play a crucial role in clinical settings.³⁸ Meningeal enhancement is frequently observed on brain MRI in patients with TBM, indicating a more severe inflammatory process in certain individuals. Meningeal enhancement is frequently observed on brain MRI in TBM patients, indicating a more severe inflammatory process in certain individuals. In our study, we examined the treatment effect of dexamethasone on subgroups classified by brain imaging (Figures S2 and S3). However, the results did not demonstrate any benefit to support dexamethasone treatment in patients with meningeal enhancement. Future studies should investigate more inflammation stratification makers and verify the effect of adjunctive dexamethasone in patients with TBM.

Strengths and Limitations

The strengths of this study include a large sample of participants recruited from multiple centers and a robust analytical approach that selects optimal models based on objective statistical assessments. However, we acknowledged several limitations. First, as an observational study, there is a potential for residual confounding of unmeasured variables. Second, we recognized the low rate of laboratory-confirmed TBM in our cohort (14.9%), consistent with a previous study (13.2%).³⁹ This highlights the insufficient diagnosis of definite TBM in the real world when compared to clinical trials.¹ Third, paradoxical reaction is a frequent phenomenon during treatment in TBM patients,^40,41 we did not design this outcome because the retrospective nature of our study and we were unable to collect repeated brain imaging and CSF data from all patients after discharge. Fourth, our dataset does not include patients co-infected with HIV, so caution should be exercised when generalizing the results to broader populations. Fifth, half of the TBM patients included in our study were classified as BMRC Grade I. Future studies with large sample sizes are needed to explore the effect of dexamethasone on neurological outcomes in a more severely affected population to confirm these findings.

Conclusion

Our study demonstrated that NLR is associated with early neurological deterioration and poor outcome in TBM patients. Further research should investigate NLR in more severely affected population to predict the clinical outcomes following anti-inflammatory therapies.