|
Abstract:
|
Sputum culture suggests that tuberculosis (TB) treatment is biphasic: an initial 7 day bactericidal phase kills 99% of Mycobacterium tuberculosis (Mtb) and a months-long sterilizing phase slowly kills drug-tolerant "persisters." Recently, it has been shown that treatment-naïve sputum may be dominated by Mtb phenotypes that do not readily grow in culture. If these phenotypes are also more drug tolerant, the initial rate of killing may be much slower than measured in culture; this could have important implications for tuberculosis drug development and treatment. Newly developed methods to measure Mtb DNA in sputum provide a culture-independent measure of Mtb burden; however, these methods measure the presence of DNA from both viable Mtb and dead Mtb that have been killed by the treatment, thus over estimating the amount of Mtb present. We develop a non-linear mixed effects model to estimate Mtb DNA over the course of treatment, while accounting for the presence of DNA from dead bacteria using a convolution integral. This model allows the amount of dead and viable Mtb to be estimated separately over time. We apply this model to longitudinal data from 41 TB patients in Uganda.
|
