Prognostic Models for 9-Month Mortality in Tuberculous Meningitis

Abstract

Background: Tuberculous meningitis (TBM) is the most severe form of extrapulmonary tuberculosis. We developed and validated prognostic models for 9-month mortality in adults with TBM, with or without human immunodeficiency virus (HIV) infection.

Methods: We included 1699 subjects from 4 randomized clinical trials and 1 prospective observational study conducted at 2 major referral hospitals in Southern Vietnam from 2001-2015. Modeling was based on multivariable Cox proportional hazards regression. The final prognostic models were validated internally and temporally and were displayed using nomograms and a Web-based app (https://thaole.shinyapps.io/tbmapp/).

Results: 951 HIV-uninfected and 748 HIV-infected subjects with TBM were included; 219 of 951 (23.0%) and 384 of 748 (51.3%) died during 9-month follow-up. Common predictors for increased mortality in both populations were higher Medical Research Council (MRC) disease severity grade and lower cerebrospinal fluid lymphocyte cell count. In HIV-uninfected subjects, older age, previous tuberculosis, not receiving adjunctive dexamethasone, and focal neurological signs were additional risk factors; in HIV-infected subjects, lower weight, lower peripheral blood CD4 cell count, and abnormal plasma sodium were additional risk factors. The areas under the receiver operating characteristic curves (AUCs) for the final prognostic models were 0.77 (HIV-uninfected population) and 0.78 (HIV-infected population), demonstrating better discrimination than the MRC grade (AUC, 0.66 and 0.70) or Glasgow Coma Scale score (AUC, 0.68 and 0.71) alone.

Conclusions: The developed models showed good performance and could be used in clinical practice to assist physicians in identifying patients with TBM at high risk of death and with increased need of supportive care.

Keywords: HIV; mortality; prognostic models; tuberculous meningitis.

Figure 1.

Annual recruitment into the 5 contributing studies by human immunodeficiency virus status. Black lines show estimated 9-month mortality rates (as percentages) for each calendar year, based on the Kaplan-Meier method. Abbreviations: ART, antiretroviral therapy; HIV, human immunodeficiency virus; TBM, tuberculous meningitis.

Figure 2.

Comparison between predicted survival with the final models and observed Kaplan-Meier estimates. Risk groups are defined using cutoff points at the 16th, 50th, and 84th percentiles of the prognostic index as generated by the final models (defining “good,” “fairly good,” “fairly poor,” and “poor” prognostic subgroups). (A) HIV-uninfected TBM population, temporal validation; (B) HIV-infected TBM population, temporal validation; (C) HIV-uninfected TBM population, all multiply imputed data; (D) HIV-infected TBM population, all multiply imputed data. Abbreviations: HIV, human immunodeficiency virus; KM, Kaplan-Meier; TBM, tuberculous meningitis.

Figure 3.

Cumulative/dynamic receiver operating characteristic curves (apparent estimate) for mortality rates evaluated at 9 months for the final prognostic model, Glasgow Coma Scale (GCS) score, and Medical Research Council (MRC) grade in the tuberculous meningitis populations with or without human immunodeficiency virus infection. MRC grade I is defined as GCS score 15 with no focal neurological signs; grade II, GCS score 11–14, or 15 with focal neurological signs; grade III, GCS score ≤0. (A) HIV-uninfected TBM population; (B) HIV-infected TBM population. Abbreviations: GCS, Glasgow Coma Scale; HIV, human immunodeficiency virus; MRC, Medical Research Council; TBM, tuberculous meningitis.

Figure 4.

Nomograms for the prediction of 9-month mortality based on the final prognostic models for the tuberculous meningitis populations without (A) or with (B) human immunodeficiency virus (HIV) infection. To derive a prediction, locate the value of each predictor on the corresponding variable line, read the corresponding points assigned on the 0–100 scale, and sum all of these points to a total point score. Then read the result on the “total points” scale and its corresponding prediction below. For HIV-infected population, the cohort variable was chosen as the most recent trial [3], because this is most relevant for future prediction. Abbreviations: CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; MRC, Medical Research Council; TBM, tuberculous meningitis.