Cost-effectiveness of Xpert MTB/RIF and investing in health care in Africa.

www.thelancet.com/lancetgh Vol 3 February 2015 e83 transmission of multidrug-resistant tuberculosis, mainly because we were modelling a setting with low multidrug-resistant tuberculosis. However, our approach could be used in contexts with a higher prevalence of multidrug-resistant tuberculosis than Tanzania because both the eff ect on the level of multidrug-resistant tuberculosis diagnosed and the related costs of multidrug-resistant tuberculosis treatment were included with an assumption that the incidence of multidrug-resistant tuberculosis is in proportion to the number of tuberculosis cases. If this assumption is not applicable, further research would be needed to model effect on multidrug-resistant tuberculosis transmission of alternative diagnostic algorithms. We agree wholeheartedly with Dheda and colleagues’ comment that the sensitivities of the model should be seen as flags that inform programmatic implementation. Indeed, we ran many simulations testing several of the assumptions, which we described in our appendix. Finally, an important element of our study is that, rather than focusing on the eff ect of one diagnostic test, our modelling deliberately models whole diagnostic algorithms and alternative placements of new diagnostic tests within those algorithms. We think that this is an important element of our approach that helps to compare whole algorithms involving microscopy (eg, same-day front-loaded fluorescence light-emitting diode microscopy), targeted use of Xpert MTB/RIF (eg, to HIV-positive or smear-negative patients), empirical clinical diagnosis, and any combination of the above.