Exhaled Mycobacterium tuberculosis predicts incident infection in household contacts

Background Halting transmission of Mycobacterium tuberculosis (Mtb) by identifying infectious individuals early is key to eradicating Tuberculosis (TB). Here we evaluate face mask sampling as a tool for stratifying infection risk in household contacts of pulmonary TB (pTB). Methods Forty-six sputum positive pTB patients in The Gambia (Aug 2016-Nov 2017) consented to mask sampling prior to commencing treatment. Incident Mtb infection was defined in their 181 household contacts as QuantiFERON (QFT) conversion or an increase in Interferon-ƴ release of ≥ 1IU/ml, 6 months after index diagnosis. Multilevel mixed-effects logistical regression analysis with cluster adjustment by household was used to identify predictors of incident infection. Findings Mtb was detected in 91% of pTB mask samples with high variation in IS6110 copies (5.3 x102 to 1.2 x107). A high mask Mtb level (≥20,000 IS6110 copies) was observed in 45% of cases and independently associated with increased likelihood of incident Mtb infection in contacts (AOR (95%CI) 3.20 (1.26 - 8.12), p=0.01), compared with cases having low/negative mask Mtb levels. Mask Mtb level was a better predictor of incident Mtb infection than sputum bacillary load, chest radiographic characteristics or sleeping proximity. Interpretation Mask sampling offers a highly sensitive and non-invasive tool to support both diagnosis of pTB and stratification of individuals who are most infectious. Our findings have the potential to revolutionise contact screening strategies and outbreak management in high TB burden settings and is of urgent public health importance.