Clinical predictors of treatment outcome in adult Tuberculous Meningitis patients at the Kundiawa General Hospital

Background

Tuberculous meningitis (TBM) is the most severe form of extrapulmonary Tuberculosis (TB) and accounted for up to 5 % of 10 million TB cases reported globally by WHO in 2018, with a high mortality and morbidity ^{1, 2}. TB meningitis studies have reported mortalities as a high as 19.35% in children and 30% in adults, with an associated high burden of neurological disability after treatment ^3-5. Mortality was as high as 80% in studies from Sub-Saharan Africa and Asia ⁵.

The ongoing challenge for effective treatment with optimal outcome is further compounded in resource-poor settings in which clinicians are often faced with limited diagnostic and therapeutic options that impede timely and appropriate treatment ^6-7. Clinicians often rely on clinical parameters for diagnosis, treatment, monitoring and outcome, to inform management decisions as well as discussions on disease condition and likely outcome with patients and relatives.

This observational study was therefore done to explore the role of certain clinical and cerebrospinal fluid (CSF) parameters in predicting treatment outcome amongst Melanesian adults with TB Meningitis at the Simbu Provincial Hospital in Papua New Guinea. The primary outcome measure assessed was death or discharge at 28 days, with length-of-hospitalization as the secondary outcome measure. Potential predictors of outcome assessed included duration of illness, duration of TB treatment, admission Glasgow Coma Score (GCS), mean arterial pressure (MAP) and plasma sodium on admission, fluid balance status 24 hours after admission, HIV-TBM co-infection and CSF findings (polymorphs, lymphocytes and acid-fast bacilli [AFB]), besides age and gender^{8, 9}.

Aim

To identify potential clinical predictors of i) length of in-hospital stay, and ii) death or discharge at 28 days, amongst Melanesian adults with TB meningitis at the Kundiawa General Hospital in Simbu Province.

Method

A retrospective observational study was conducted on a total of 65 Melanesian adults with TB meningitis at the medical ward of the Sir Joseph Nombri Memorial, Kundiawa General Hospital in Simbu province between 2015 and 2019.

A modified Thwaites Criteria using clinical, radiological and CSF findings was employed for the diagnosis of Tuberculous Meningitis ^{10, 11}, whereby a patient with at least four of the following thematic features was diagnosed as a case of TBM:

i) meningitis symptoms (headache, fever, photophobia, vomiting, fitting) ≥ 2 weeks, ii) personal or family history or symptoms of TB, iii) neck stiffness ± Kernig’s or Brudzinski’s sign, iv) altered conscious status, v) focal neurological symptoms, vi) regional lymphadenopathy or other extrapulmonary TB features, vii) consistent chest radiograph and viii) CSF findings (AFB positive, lymphocyte predominance, elevated glucose with low protein). Of the four diagnostic features, at least one of TB and one of meningitis was required to make a TB meningitis diagnosis. Focal neurological symptoms included cranial nerve palsies and CSF AFB was assessed by polymerase chain reaction (PCR).

Study data was collected retrospectively from admission notes in patients’ information chart using a semi-structured document review form by one investigator at a monthly interval over a 5-year period. Raw data was entered onto an excel spread sheet and imported into Stata 15.1 for analysis. Summary statistics and multivariate regression analysis were performed on variables of interest for possible relationships between predictor and outcome variables. Logistic regression was applied for 28-day mortality given its dichotomous nature, whilst liner regression was applied for length of in-hospital stay as it is numerical. The primary outcome measure assessed was death or discharge at 28 days, with length-of-hospitalization as the secondary outcome measure. Parameters assessed included duration of illness (days), duration of TB treatment (days), admission Glasgow Coma Score (GCS), mean arterial pressure (MAP) and plasma sodium on admission, fluid balance status 24 hours after admission, HIV-TBM co-infection and CSF findings (polymorphs, lymphocytes and acid-fast bacilli [AFB]), besides age and gender ^{8, 9}. The retrospective nature of the study limited its ability to effectively asses neurological deficit (neuro-motor, cognitive) as an outcome measure, owing to data insufficiency.

65 patients, over 13 years of age and diagnosed with TB meningitis (per criteria) with or without HIV infection were included for this study. All study subjects received a course of dexamethasone (2-8 mg iv/oral/NGT) during in-hospital treatment.

Results

Discussion

The high case fatality (48%) and mortality rates (2.22 per 100,000 per year) for TB Meningitis in this study concur with similar studies from resource-constrained settings around the world ^13-15. Apart from the diagnostic and therapeutic limitations, parameters inherent to the biology of mycobacterium tuberculosis including virulence and persistence, together with the site and nature of disease condition, may account for similar higher mortalities for TB meningitis observed even in resource-rich settings ^16-18.

At the estimated mortality rate of 2.22 per 100,000 adults per year in the province, 9 adults would have died from TB meningitis every year for a population of around 400,000 adults for the 5-year study period in the province.

The clinical parameters associated with 28-day mortality from this study were admission GCS, HIV-TBM co-infection, positive fluid balance status in the first 24 hours of treatment and male gender.

A multivariate regression analysis (Table 1) showed a statistically significant association between admission GCS (P=0.039, 95% CI), male gender (P=0.057, 95% CI) and 28-day mortality after controlling for possible confounders. The strength of the association between the predictor and outcome variables was however weak (R²=0.3108) in the model, owing largely to the small sample size (65), and therefore limiting the ability of the study to conclusively quantify the observed associations in terms of the relevant measure of association in odds ratio (OR). Hence, whilst it may be statistically sound to state that there were observed associations between predictor and outcomes variables of interest, any meaningful discussion of the OR (strength of these associations) is limited. A regression coefficient of R² = 0.31 in the model means that a unit change in any of the predictor variables will entail only a 31% corresponding change in the outcome variable (28-day mortality), which is far less than at least 75% (R² ≥ 0.75) which is deemed necessary to represent a reliably strong association between predictor and outcome variables. Therefore, larger, prospective cohort studies (RCT) would be ideal to further explore and clearly ascertain the potential associations observed in this study.

55.4% of study subjects had a GCS of 13 or lower on admission, which inadvertently affected their progress on treatment and subsequent outcome of higher mortality at 28 days. Depressed conscious status has long been identified as an independent prognostic factor for poor outcomes in TB, and other forms of chronic meningitis around the world ¹⁹. Delayed presentation to health care facilities with late treatment initiation continue to contribute to disease severity on admission, as reflected in this study by a high proportion of patients with lower admission GCS, and with subsequent poor outcomes in developing countries ^{20, 21}.

The marginally significant association between positive fluid balance (P=0.061, 95% CI) 24 hours after admission and 28-day mortality observed, after controlling for confounding, highlights the role of fluid and electrolyte balance in treatment outcomes. A positive fluid balance after 24 hours of treatment was observed in only 25% of the study subjects but was significantly associated with higher mortality at 28 days, which warrants a closer attention by clinicians to fluid and electrolytes status during in-patient treatment of patients with TB meningitis and associated raised intracranial pressures ^{22, 23}. TB Meningitis and raised intracranial pressure can cause syndromes of salt wasting or excess anti-diuretic hormone action (SIADH) with resulting hyponatremia which in-turn entails functional neurological derangement with poor outcomes ^24-27.

TBM-HIV Co-infection is another independent predictor of mortality reported in several similar studies, with higher mortalities compared to non-HIV subjects with TB meningitis ^28-30. The co-infection rate in this study was relatively low at 16.92% but with a significantly high case fatality of 90.9%. There was no statistically significant association between HIV-TBM co-infection (P=0.199, 95% CI) and 28-day mortality in the multivariate regression after controlling for other variables (Table 1). However, in the model regressing identified predictors (admission GCS, gender, fluid balance 24 hours after admission and HIV-TB con-infection) in table 2, there was a statistically significant association between HIV-TBM co-infection and 28-day mortality (P=0.049, 95% CI). This disparity may be attributable to potential confounding and/or underpowering of the study owing to the small sample size of 65 (type 2 error).

The duration of hospitalization ranged from 1 to 473 days, with a mean duration of 36.62 days. 90% of subjects stayed more than two weeks but less than 60 days (2 months) in the hospital, which is characteristic of chronic diseases and highlights the challenge of considerations in longer in-hospital care.

In the regression model for identified predictors (Table 5), there was a weak (R²=0.25) but statistically significant association between positive CSF AFB (P=0.000, 95% CI), TBM-HIV Co-infection (P=0.036, 955 CI), positive fluid balance (P=0.036, 95% CI) and length of in-hospital stay after controlling for each other. Again, this may be attributable to confounding and the small sample size.

The weak associations observed in regression models (R² < 0.75) limit the value of assessed parameters as reliable predictors of length of hospitalization.

Conclusion

This study showed a relatively high case fatality (48%) and mortality rates (2.22 per 100,000 per year) for TB Meningitis amongst Melanesian adults admitted to the Kundiawa General Hospital. The case fatality for HIV-infected individuals with TB meningitis was much higher at 90.9%. Ninety percent of the subjects stayed for more than 2 weeks but less than 2 months in the hospital.

Admission GCS, positive fluid balance 24 hours after admission, HIV-TB co-infection and male gender were associated with higher 28-day mortality, and therefore can serve as useful clinical predictors of mortality in the care of TB meningitis patients, especially in resource-constrained settings. A positive CSF AFB can also serve as a useful indicator for duration of hospitalization amongst TB meningitis patients.

Prospective cohort studies with larger sample sizes are needed to further explore and ascertain the associations observed in this study.

Ethics

Ethical consideration and approval was granted by the interim Institutional Review Committee of the SiPHA-Clinical Research Center.