Over the last few years the World Health Organization (WHO) has endorsed several tests for the rapid detection of multidrug-resistant tuberculosis (MDR-TB) in resource-poor settings. The objective of this study was to compare the cost-effectiveness of a line probe assay test (less than one week for results) to conventional (bacterial culture) drug susceptibility testing (one month for results) for the detection of MDR-TB in an HIV-positive South African population by estimating the incremental cost-effectiveness ratio (ICER) per disability-adjusted life-year (DALY) averted. Costs of testing, drug treatment, hospitalization, as well as estimates for mortality, treatment success, and failure rates were obtained from literature sources, the South African Department of Health, the WHO, the Foundation of Innovative Diagnostics (FIND), and expert opinion. The willingness-to-pay threshold for a DALY averted was pre-set at 3 times the 2009 GDP per capita (about $17,400) for South Africa. In the base-case scenario for a prevalence of 30% of MDR-TB among HIV-positive patients, the average cost per person for the line probe assay testing strategy was $3,539/0.458 DALY averted and the conventional testing approach was $3,011/0.430 DALY averted. The base-line ICER was about $18,800 per DALY averted – about $1,400 above the pre-set threshold. In sensitivity analyses, the model was robust to changes in prevalence (+ 50%); costs (+ 10%), and probabilities of death, success and failure (+ 20%). However, when the treatment success rate for the line probe assay test was increased to 60% (one of the targets set by WHO in TB treatment) the ICER was below the willingness-to-pay level (i.e., cost-effective). The probabilistic sensitivity analysis showed there is a 70% chance that the additional cost of the line probe assay, compared with conventional testing, was less than $30,000 per DALY averted. However, the model may have underestimated the benefits of the line probe assay because it did not account for a decrease in the transmission of the disease due to earlier treatment nor did it measure any benefits more than a year after testing. / text