Abstract Body

We integrated blood-based pathogen antigenic detection and host proteomic biomarkers to predict culture conversion status and TB recurrence.

A total of 628 clinical trial and cohort participants were included in the analysis: 538 from two Phase 2B trials (TBTC Study 29/29X) that collected serum at 2 time points (pretreatment and week 8) and 90 (60 cured, 30 recurrences) from a Phase 3 trial (REMoxTB) and cohort studies, provided by the Markedly Accelerating Research with Knowledge of TB Biomarkers Biobank, that collected serum at 7 time points (pretreatment and weeks 2, 4, 8, 17, 26, and 52). Levels for 54 host proteins and one pathogen antigenic detection marker (lipoarabinomannan, LAM) were available for analysis. Nonlinear mixed effect modeling and machine learning algorithms were used to search for biomarker signatures that predict culture conversion status at 8 weeks (all participants), time to positivity in the MGIT assay during treatment (N=538), and TB recurrence up to 18 months after start of treatment (N=90). A random 75% of the population was used for training and 25% for testing. Area under the receiver operating characteristic curve (ROC AUC) was used to assess model discrimination.

With a ROC AUC of 0.77 (95% CI, 0.68-0.85), week 8 serum amyloid A1 (SAA1) and regulated on activation, normal T cell expressed and secreted (RANTES) levels were found to predict longitudinal time to positivity and culture conversion status at 8 weeks. Week 8 SAA1 and RANTES poorly predicted TB recurrence with a ROC AUC < 0.50. A reanalysis of all biomarkers for prediction of TB recurrence resulted in biomarker signatures with ROC AUC of 0.75 (0.52-0.98) using pretreatment levels, 0.86 (0.63-1.00) using pretreatment to week 2 levels, 0.88 (0.63-1.00) using pretreatment to week 4 levels, and 0.94 (0.85-1.00) using pretreatment to week 8 levels (Figure). Host neopterin, tumor necrosis factor-beta (TNF-?) and triggering receptor expressed on myeloid cells 1 (TREM1) levels and pathogenic LAM levels in blood were the top biomarkers that predict TB recurrence, suggesting that biomarkers for clinical outcomes may be different from biomarkers for culture conversion.

Integrated host and pathogen blood-based signatures have potential to provide an alternative, non-culture based tool to project long term relapse-free outcomes and to help inform both clinical care and the design of TB clinical trials.