Background
Direct bronchial spread of tuberculosis was extensively described in pre-antibiotic human pathology literature but this description has been overlooked in the post-antibiotic era, in which most pathology data come from animal models that emphasise the granuloma. Modern techniques, such as [18F]2-fluoro-2-deoxy-D-glucose (FDG) PET-CT scans, might provide further insight. Our aim was to understand normal early tuberculosis resolution patterns on pulmonary PET-CT scans in treated patients with tuberculosis who were subsequently cured.

Methods
In this observational analysis, we analysed data from PredictTB, an ongoing, prospective, randomised clinical trial that examined sequential baseline and week 4 FDG-PET-CT scans from participants successfully treated (sputum culture negative 18 months after enrolment) for drug-susceptible pulmonary tuberculosis in South Africa and China. Participants who were aged 18–75 years, GeneXpert MTB/RIF positive for tuberculosis and negative for rifampicin resistance, had not yet started tuberculosis treatment, had not been treated for active tuberculosis within the previous 3 years, and met basic safety laboratory criteria were included and participants with diabetes, HIV infection, or with extrapulmonary tuberculosis including pleural tuberculosis were excluded. Scans were assessed by two readers for the location of tuberculosis lesions (eg, cavities and consolidations), bronchial thickening patterns, and changes from baseline to week 4 of treatment.

Findings
Among the first 124 participants (enrolled from June 22, 2017, to Sept 27, 2018) who were successfully treated, 161 primarily apical cavitary lesions were identified at baseline. Bronchial thickening and inflammation linking non-cavitary consolidative lesions to cavities were observed in 121 (98%) of 124 participants' baseline PET-CT scans. After 4 weeks of treatment, 21 (17%) of 124 participants had new or expanding lesions linked to cavities via bronchial inflammation that were not present at baseline, particularly participants with two or more cavities at baseline and participants from South Africa.

Interpretation
In participants with pulmonary tuberculosis who were subsequently cured, the location of cavitary and non-cavitary lesions at baseline and new lesions at week 4 of treatment suggest a cavitary origin of disease and bronchial spread through the lungs. Bronchial spread from cavities might play a larger role in the spread of pulmonary tuberculosis than has been appreciated. Elucidating cavity lesion dynamics and Mycobacterium tuberculosis viability within cavities might better explain treatment outcomes and why some patients are cured and others relapse.

Funding
Bill & Melinda Gates Foundation, European and Developing Countries Clinical Trials Partnership, China Ministry of Science and Technology, National Natural Science Foundation of China, and National Institutes of Health.

Translations
For the Chinese, Afrikaans and Xhosa translations of the abstract see Supplementary Materials section.