Baseline IL-6 is a biomarker for unfavorable tuberculosis treatment outcomes: a multi-site discovery and validation study

BACKGROUND: Biomarkers of unfavorable tuberculosis treatment outcomes are needed to accelerate new drug and regimen development. Whether plasma cytokine levels can predict unfavorable tuberculosis treatment outcomes is unclear. METHODS: We identified and internally validated the association between 20 a-priori selected plasma inflammatory markers and unfavorable treatment outcomes of failure, recurrence and all-cause mortality among adults with drug-sensitive pulmonary tuberculosis in India. We externally validated these findings in two independent cohorts of predominantly diabetic and HIV coinfected tuberculosis patients in India and South Africa, respectively. RESULTS: Pre-treatment IFN-γ, IL-13 and IL-6 were associated with treatment failure in the discovery analysis. Internal validation confirmed higher pre-treatment IL-6 concentrations among failure cases compared to controls. External validation among predominantly diabetic tuberculosis patients found an association between pre-treatment IL-6 concentrations and subsequent recurrence and death. Similarly, external validation among predominantly HIV coinfected tuberculosis patients found an association between pre-treatment IL-6 concentrations and subsequent treatment failure and death. In a pooled analysis of 363 tuberculosis cases from the Indian and South African validation cohorts, high pre-treatment IL-6 concentrations were associated with higher risk of failure (adjusted odds ratio [aOR]=2.16, 95%CI 1.08-4.33, p=0.02), recurrence (aOR=5.36, 95%CI 2.48-11.57, p<0.001) and death (aOR=4.62, 95%CI 1.95-10.95, p<0.001). Adding baseline IL-6 to a risk-prediction model comprising of low BMI, high smear grade and cavitation improved model performance by 15 percent (C-statistic of 0.66 versus 0.76, p=0.02). CONCLUSION: Pre-treatment IL-6 is a biomarker for unfavorable tuberculosis treatment outcomes. Future studies should identify optimal IL-6 concentrations for point-of-care risk prediction.