Cost analysis of rapid diagnostics for drug-resistant tuberculosis

Background: Growth-based drug susceptibility testing (DST) is the reference standard for diagnosing drug-resistant tuberculosis (TB), but standard time to result (TTR) is typically >= 3 weeks. Rapid tests can reduce that TTR to days or hours, but accuracy may be lowered. In addition to the TTR and test accuracy, the cost of a diagnostic test may affect whether it is adopted in clinical settings. We examine the cost-effectiveness of rapid diagnostics for extremely drug-resistant TB (XDR-TB) in three different high-prevalence settings. Methods: 1128 patients with confirmed TB were enrolled at clinics in Mumbai, India; Chisinau, Moldova; and Port Elizabeth, South Africa. Patient sputum samples underwent DST for first and second line TB drugs using 2 growth-based (MGIT, MODS) and 2 molecular (Pyrosequencing [PSQ], line-probe assays [LPA]) assays. TTR was the primary measure of effectiveness. Sensitivity and specificity were also evaluated. The cost to perform each test at each site was recorded and included test-specific materials, personnel, and equipment costs. Incremental cost-effectiveness ratios were calculated in terms of $/day saved. Sensitivity analyses examine the impact of batch size, equipment, and personnel costs. Results: Our prior results indicated that the LPA and PSQ returned results in a little over 1 day. Mean cost per sample without equipment or overhead was $23, $28, $33, and $41 for the MODS, MGIT, PSQ, and LPA, respectively. For diagnosing XDR-TB, MODS was the most accurate, followed by PSQ, and LPA. MODS was quicker and less costly than MGIT. PSQ and LPA were considerably faster but cost more than MODS. Batch size and personnel costs were the main drivers of cost variation. Conclusions: Multiple factors must be weighed when selecting a test for diagnosis of XDR-TB. Rapid tests can greatly improve the time required to diagnose drug-resistant TB, potentially improving treatment success, and preventing the spread of XDR-TB. Faster time to result must be weighed against the potential for reduced accuracy, and increased costs.