PET/CT imaging of tuberculosis lung lesions in marmosets treated with different drug regimens aligns with human clinical outcomes.

Early bactericidal activity and time to sputum conversion are well-established study end points in both preclinical animal models and clinical trials for testing drug regimens for pulmonary tuberculosis (TB). The development and optimization of treatment-shortening drug regimens for TB have been challenged by disparities between these study end points and nonrelapsing cure. We hypothesized that using lung lesions measured by 2-deoxy-2-[18F]fluoro-d-glucose-positron emission tomography/computed tomography (PET/CT) imaging in infected marmosets could help to interpret treatment efficacies and better understand clinical treatment outcomes. Radiographic changes in lung lesions were measured using PET/CT imaging in a cohort of infected marmosets, which were divided into 22 treatment arms (including monotherapies and combination drug treatments) for 2 months. We used unsupervised clustering to define multivariate treatment response profiles that combined quantitative changes in radiographic pathology and terminal bacterial burden per lung lesion to inform lesion-level responses to drug treatments. These drug response profiles not only aligned with known clinical outcomes but also provided lesion-level insights into clinical successes and failures. We found that the inferiority of the 4-month moxifloxacin-rifampicin-pyrazinamide-ethambutol regimen compared with the 6-month standard of care for individuals with lung cavitary TB could be predicted. The marmoset response profiles were matched to their respective histopathological classifications at necropsy and successfully distinguished cavitary granulomas that responded to treatment from cavitary granulomas that failed to improve or worsened after the first month of treatment. Our findings indicate that a combination of quantitative PET/CT measures is more informative of TB treatment outcomes than bacterial burden.