Optimizing drug-resistant tuberculosis diagnosis: cost-effectiveness of rapid molecular and phenotypic assays in South Africa

BACKGROUND: Timely detection of drug-resistant tuberculosis (DR-TB) is essential for effective treatment and preventing poor outcomes. Rapid molecular diagnostics are promising alternatives to conventional phenotypic drug susceptibility testing (pDST), offering faster and more accessible detection of resistance. This study evaluated the cost-effectiveness of rapid molecular assays, alone or combined with pDST, for detecting resistance to isoniazid, rifampicin, and fluoroquinolones from a South African healthcare provider perspective. METHODS: A decision-analytic model was developed to simulate TB-related outcomes for a hypothetical cohort of microbiologically confirmed TB patients. Nine diagnostic strategies were evaluated: pDST alone; four rapid molecular tests (line probe assays [LPAs], Xpert MTB/RIF [Xpert] followed by Xpert MTB/XDR [Xpert XDR], Xpert MTB/RIF Ultra [Xpert Ultra] followed by Xpert XDR, and targeted next-generation sequencing [tNGS]); and combinations pairing each molecular test with pDST. Outcomes included early treatment rates, mortality, direct medical costs, disability-adjusted life-years (DALYs), and incremental cost-effectiveness ratios (ICERs). Base-case, sensitivity, and scenario analyses were performed. RESULTS: In the base-case analysis, 'Xpert followed by Xpert XDR + pDST' was the preferred cost-effective strategy, with an ICER of USD 6,554/DALY averted-below South Africa's GDP per capita threshold. While 'tNGS + pDST' yielded the greatest health benefits-lowest DALYs (1.9877), highest early treatment rate (995.54/1,000 tested), and lowest mortality (90.22/1000 tested)-its ICER (USD 25,918/DALY averted) exceeded three times the GDP per capita, rendering it not cost-effective. Sensitivity analyses highlighted the impact of diagnostic accuracy and treatment timing on cost-effectiveness outcomes. Probabilistic sensitivity analysis showed 'tNGS + pDST' had the highest probability of being cost-effective when the willingness-to-pay threshold exceeded USD 10,500/DALY averted. Diagnostic replacement scenario analysis revealed that tNGS alone could be a cost-effective alternative (ICER = USD 1712 per DALY averted) when pDST was unavailable. An extended two-year time horizon analysis confirmed base-case robustness. CONCLUSIONS: Combining rapid molecular diagnostics with pDST offers a cost-effective and clinically beneficial approach for DR-TB detection in high-burden settings. The Xpert-based strategy provides an optimal balance of diagnostic yield, early treatment, and economic efficiency in South Africa. tNGS represents a feasible alternative in settings where pDST is inaccessible, warranting further evaluation for broader implementation.