LAUSR

Drug resistance is a known risk factor for poor tuberculosis (TB) treatment outcomes, but the contribution of other bacterial factors to poor outcomes in drug susceptible TB is less understood. Here, we generate a population-based dataset of drug-susceptible Mycobacterium tuberculosis (MTB) isolates from China to identify factors associated with poor treatment outcomes. We sequenced the whole genome of 3496 MTB strains and linked genomes to patient epidemiological data. A genome-wide association study (GWAS) was performed to identify bacterial genomic variants associated with poor outcomes. Risk factors identified by logistic regression analysis were used in clinical models to predict treatment outcomes and their associations were assessed with structural equation models (SEM). GWAS identified fourteen MTB variants (24.2% vs 7.5%, P<0.001) and a de novo reactive oxygen species (ROS) mutational signature (26.3%±18.2% vs 22.9%±13.8%, P=0.027) that were more frequent in patients with poor treatment outcomes. Patient age, sex, and duration of diagnostic delay were also independently associated with poor outcomes. The best clinical prediction model, with an AUC of 0.74, incorporates both host and bacterial risk factors, and host factors are more important. Together, our results reveal that although host factors are the most important determinants for poor treatment outcomes, the genomic characteristics of the infecting MTB strain may also contribute significantly to poor treatment outcomes. Fourteen genetic variants were statistically associated with poor TB treatment outcomes, but the optimal model for predicting treatment outcomes includes both patient characteristics and bacterial genomic determinants.