Elevated cerebrospinal fluid cytokine levels in tuberculous meningitis predict survival in response to dexamethasone

Abstract

Adjunctive treatment with antiinflammatory corticosteroids like dexamethasone increases survival in tuberculosis meningitis. Dexamethasone responsiveness associates with a C/T variant in Leukotriene A4 Hydrolase (LTA4H), which regulates expression of the proinflammatory mediator leukotriene B₄ (LTB₄). TT homozygotes, with increased expression of LTA4H, have the highest survival when treated with dexamethasone and the lowest survival without. While the T allele is present in only a minority of the world's population, corticosteroids confer modest survival benefit worldwide. Using Bayesian methods, we examined how pretreatment levels of cerebrospinal fluid proinflammatory cytokines affect survival in dexamethasone-treated tuberculous meningitis. LTA4H TT homozygosity was associated with global cytokine increases, including tumor necrosis factor. Association between higher cytokine levels and survival extended to non-TT patients, suggesting that other genetic variants may also induce dexamethasone-responsive pathological inflammation. These findings warrant studies that tailor dexamethasone therapy to pretreatment cerebrospinal fluid cytokine concentrations, while searching for additional genetic loci shaping the inflammatory milieu.

Keywords: Bayesian analysis; corticosteroids; cytokines; inflammation; tuberculous meningitis.

Fig. 1.

CSF cytokine levels grouped by LTA4H genotype. (A) Cytokine levels in CSF from CC (n = 148), CT (n = 142), and TT (n = 16) patients. (B) Cytokine levels compared between non-TT (n = 290) and TT (n = 16) genotypes. Magenta lines indicate geometric means. Asterisks indicate probability that right-hand group values are significantly greater than the left (* ≥ 0.95, ** ≥ 0.99, *** ≥ 0.999, **** ≥ 0.9999). Unspecified comparisons are not significant.

Fig. 2.

Cytokine levels by grade and by genotype. (A) All patients (grade 1 n = 109, grade 2 n = 141, grade 3 n = 55); (B) Non-TT patients (grade 1 n = 105, grade 2 n = 133, grade 3 n = 51), and (C) TT patients (grade 1 n = 4, grade 2 n = 8, grade 3 n = 4). (D) Grade 1 patients; (E) grade 2 patients; (F) grade 3 patients. Magenta lines indicate geometric means. Asterisks indicate probability that right-hand group values are significantly greater than the left (* ≥ 0.95, ** ≥ 0.99, *** ≥ 0.999, **** ≥ 0.9999). Hash symbols indicate probability that left-hand group values are significantly greater than the right (^# ≥0.95, ^## ≥0.99). Unspecified comparisons are not significant.

Fig. 3.

Cytokine levels in survivors and nonsurvivors. (A) Cytokine levels in patients who survived (S, n = 248) versus those who died (D, n = 57); (B) separated by genotypes into non-TT (deaths, n = 57, survivors, n = 232) and TT (all survived, n = 16). Magenta lines indicate geometric means. Asterisks indicate probability that right-hand group values are significantly greater than the left (* ≥ 0.95, ** ≥ 0.99, *** ≥ 0.999, **** ≥ 0.9999). Comparisons performed for each cytokine: Non-TT dead vs. survived and non-TT survived vs. TT survived. Cytokine levels between non-TT dead and TT survived were not compared.

Fig. 4.

Effect of dexamethasone and LTA4H rs17525495 genotype on survival probability of patients from Tobin et al. (8) and Thuong et al. (7). Mean posterior survival probability curves; inset plots represent mean posterior hazard rates for the first 30 d. Shaded areas represent the 95% Bayesian confidence limits for posterior probability. (A) In Tobin et al. (8) (No Dex), TT survival was nonsignificantly reduced compared to non-TT (maximum probability 0.946). (B) In Tobin et al. (8) (Dex), TT survival was significantly greater than non-TT from day 40 onwards (maximum probability 0.976, survival gap 17%). Probability that TT hazard rate (Inset) is lower than non-TT is >0.95 from day 2 to day 252 (maximum probability 0.972, peak ratio 7.5 at day 97). CT survival was significantly greater than CC from day 3 onwards (maximum probability 0.999, survival gap 23%), and CT hazard rate significantly lower than CC from day 1 onwards (maximum probability 0.996, ratio peaks at 12 on day 3 and remains >3 throughout). (C) In Thuong et al. (7) patients (Dex), CC and CT survival comparisons do not differ significantly (maximum probability 0.91). TT survival was significantly greater than CC from day 42 onwards (maximum probability 0.987, survival gap 12%). Probability that TT hazard rate is lower than CC is >0.95 days 15 to 138 (maximum probability 0.991, ratio peaks at 3.4 on day 62 and remains >1 until day 250). TT survival was also significantly greater than CT from day 53 to day 254 (maximum probability 0.964, survival gap 9%). Probability that TT hazard rate is lower than CT is >0.95 from day 7 to day 73 (maximum probability 0.979, ratio peaks at 2.9 on day 22 and remains >1 to day 234). (D) In CC (+Dex) patients, survival was nonsignificantly greater in the Thuong et al. (7) cohort (maximum probability 0.939, survival gap 9%). (E) In CT (+Dex) patients, survival was significantly greater in the Tobin et al. (8) cohort from day 5 onwards (maximum probability 0.993, survival gap 11%). Tobin CT (+Dex) hazard rate was significantly lower than Thuong CT from day 2 to day 45 (maximum probability 0.997, peak ratio 9.8 on day 4). (F) In TT patients (+Dex), survival was nonsignificantly greater in the Tobin et al. (8) cohort (maximum probability 0.90, survival gap 6%). (G) Tobin CC patient survival did not differ significantly with and without Dex treatment (maximum probability 0.80). (H) Tobin CT patient survival was significantly greater with Dex from day 7 to day 88 (maximum probability 0.964, survival gap 11%). CT (+Dex) hazard rate was significantly lower than CT (No Dex) from day 3 to day 18 (maximum probability 0.967, peak ratio 9 on day 2 and remains >1 throughout). (I) Tobin TT patient survival was significantly greater with Dex from day 1 onwards (maximum probability 0.997, survival gap 41%). TT (+Dex) hazard rate was significantly lower from day 1 onwards (maximum probability 0.996, peak ratio 35 on day 2). See SI Appendix, Supplementary Box S2 for explanation of definitions and abbreviations used.

Fig. 5.

Proposed model of dexamethasone-mediated effects on survival probability interacting with LTA4H rs17525495 genotypes.