Tuberculosis (TB) remains a major global public health challenge, particularly in high-burden, resource-constrained settings [1,2]. In 2023, TB re-emerged as the leading cause of death among infectious diseases, surpassing COVID-19, with approximately 1.25 million deaths—a substantial proportion linked to drug-resistant TB (DR-TB) [1,3]. The emergence of pre-extensively drug-resistant TB (pre-XDR-TB) and extensively drug-resistant TB (XDR-TB) has further complicated treatment strategies and increased the risk of poor health outcomes [1,4,5].

An estimated 400,000 DR-TB cases were reported globally in 2023, yet a significant diagnostic gap persists, with only 42% of cases diagnosed and treated [1]. The burden of DR-TB-related mortality and disability is especially pronounced in low- and middle-income countries (LMICs), where socioeconomic challenges, limited healthcare infrastructure, and HIV co-infection exacerbate the situation [1,2,6].

Early and accurate detection of DR-TB is essential for effective management and reducing transmission [7,8]. Phenotypic drug susceptibility testing (pDST)—the current gold standard for determining TB drug resistance—is hindered by long turnaround times, technical complexity, and limited accessibility [4,8,9]. Targeted next-generation sequencing (tNGS) has emerged as a promising alternative for rapid drug resistance detection and timely treatment guidance [4,10,11].

By amplifying selected genes, tNGS can simultaneously identify mycobacterial species, genotype strains, and predict resistance to multiple anti-TB drugs from a single sample [4,12]. Clinical studies show tNGS results concordant with pDST, offering advantages in speed and elimination of bacterial culture [4,11].

Integrating tNGS into routine practice in high-burden settings raises important health and economic questions [13]. A recent WHO-commissioned analysis by Shrestha et al. [14] provided foundational evidence on tNGS cost-effectiveness in South Africa, India, and Georgia but did not explicitly model the mortality impact of diagnostic delay—a critical mechanism by which rapid diagnostics may improve outcomes [15].

Our study advances the evidence by directly quantifying the survival benefit of earlier treatment initiation—a key pathway missing from prior analyses—while incorporating updated 2022 WHO guidelines and regimen-specific treatment pathways [4,16]. We therefore evaluated whether tNGS—alone or in combination with pDST—is cost-effective for diagnosing pre-XDR/XDR-TB in RR-TB patients across South Africa, India, and Georgia and quantified the mortality reduction attributable to earlier treatment initiation enabled by rapid molecular diagnostics.

By focusing on these high-burden settings, we provide actionable insights into the economic viability and health benefits of integrating tNGS into national TB programs, filling a critical evidence gap for policy decisions.

This study evaluated the cost-effectiveness of tNGS for diagnosing pre-XDR and XDR-TB in RR-TB patients across South Africa, India, and Georgia. tNGS-based strategies improved early treatment initiation and reduced TB-related mortality in all settings, but cost-effectiveness varied substantially by context. In South Africa, tNGS alone was cost-effective (ICER, $2805/QALY) below the national WTP threshold ($6253). In Georgia, the combination strategy (tNGS+pDST) was cost-effective (ICER, $6361/QALY) below the national WTP threshold ($8120). In India, neither strategy met the WTP threshold ($2485) at current prices; tNGS would require cost reduction to ≤$116/test to achieve cost-effectiveness. Probabilistic sensitivity analysis confirmed robustness, with tNGS cost-effective in 58% of simulations in South Africa and the combination strategy cost-effective in 42% of simulations in Georgia.

Our findings complement and extend the WHO-commissioned analysis by Shrestha et al. [14], which evaluated tNGS cost-effectiveness in the same three countries but reached different conclusions—notably, that tNGS was more costly and less effective than universal pDST in the base case, with cost-effectiveness observed only under specific conditions (e.g., at three times GDP per capita in South Africa) [4,14]. The primary distinction lies in model structure: whereas Shrestha et al. [14] focused on diagnostic accuracy, our model explicitly incorporates time-to-treatment effects by linking diagnostic delay to mortality risk, aligns regimen assignment with 2022 WHO guidance (BPaL, BPaLM, and longer oral regimens) [16], and includes a broader health-state structure (surgery, palliative care, and LTFU). These additions demonstrate that tNGS improves outcomes not merely through resistance detection but fundamentally by enabling earlier initiation of effective therapy [13,15]. Consequently, tNGS becomes cost-effective in South Africa and Georgia when this mechanism is accounted for, whereas India remains not cost-effective at current prices. We also provide implementation thresholds (e.g., tNGS ≤$116/test in India) and alternative WTP perspectives to support policy decisions. A structured comparison is provided in Appendix S3 and Table S5.

The HR, for delayed treatment (HR, 1.53) and tNGS cost were the primary drivers of cost-effectiveness across all settings. Threshold analyses identified tNGS price as a modifiable leverage point: cost-effectiveness in South Africa and Georgia was maintained at up to $356 and $311 per test, respectively, whereas India required a price of ≤$116 to achieve cost neutrality. These findings highlight the importance of volume-based pricing negotiations, technology transfer, and pooled procurement to reduce costs in resource-constrained settings. The predominance of the HR, underscores that interventions to reduce diagnostic delays—such as optimized sample transport systems and rapid result communication—are essential to realize the full survival benefits of tNGS.

The choice of cost-effectiveness threshold also influences interpretation. Although we used GDP per capita-based thresholds following WHO guidelines [28], more recent literature advocates more conservative, empirically derived thresholds that reflect true health opportunity costs [30–32]. Applying the Pichon-Rivière framework [30]—which derives thresholds from health expenditure growth and life expectancy gains—yields alternative benchmarks of $4512 for South Africa (0.72×GDP), $487 for India (0.20×GDP), and $4710 for Georgia (0.58×GDP). Against these conservative thresholds, tNGS remains cost-effective in South Africa and Georgia (tNGS alone), whereas the combination strategy in Georgia becomes borderline, and India remains unaffordable. This sensitivity analysis (Appendix S4, Table S6) strengthens confidence in our conclusions for South Africa and India while underscoring the need for context-specific threshold deliberation in Georgia. We therefore interpret our results as upper-bound estimates and recommend that country programs consider local opportunity cost-based thresholds when making adoption decisions.

Our findings were also sensitive to the analytic time horizon, reflecting the lag between diagnostic investment and downstream health gains [33]. Short horizons (e.g., 1 year) systematically undervalue tNGS because the survival benefits of earlier treatment accrue over several years through reduced mortality and relapse. Conversely, very long horizons introduce additional uncertainty without materially altering conclusions. Scenario analyses showed that 5–10 years were sufficient to capture the major health and economic impacts of tNGS, supporting the base-case 10-year horizon as an appropriate balance between clinical realism and model stability.

The superior cost-effectiveness of tNGS in South Africa reflects higher baseline mortality rates and greater capacity to absorb diagnostic costs. In Georgia, favorable outcomes for the combination strategy reflect moderate XDR-TB prevalence and a relatively high WTP threshold. In India, the lack of cost-effectiveness is driven by tNGS pricing exceeding affordability thresholds and by the use of shorter oral regimens that limit incremental QALY gains from mortality reduction. These disparities underscore that a one-size-fits-all approach is impractical. In South Africa, scaling up tNGS should be accompanied by investments in laboratory capacity and supply chains. In India, a phased rollout contingent on cost reductions and pilot studies in high-prevalence regions is advisable. Georgia may benefit from selectively deploying combination strategies in areas with high DR-TB prevalence.

These country-specific findings align with recent policy analyses emphasizing the urgency of strengthening RR-TB diagnostic capacity in LMICs. Yates et al. [34] highlight that progress in RR-TB treatment is threatened by emerging drug resistance and poor political choices, making efficient resource allocation paramount. Cavalcanti et al. [35] demonstrate that reductions in health sector funding can substantially worsen mortality, heightening the importance of investing in cost-effective diagnostics. In this context, the selective adoption of tNGS, guided by local cost-effectiveness evidence, provides a pragmatic strategy for maximizing health gains under constrained budgets.

Several limitations must be acknowledged. First, our findings are limited to three countries and may not be generalizable to other high-burden settings with different epidemiological or economic profiles; although the modeling framework is adaptable, local data would be required for context-specific guidance. Second, the model relies on published literature and secondary data, with assumptions—such as perfect pDST accuracy and immediate tNGS results—that may not reflect real-world constraints, including logistical delays, operational barriers, and systemic challenges such as infrastructure deficits and fragmented healthcare delivery. Third, we focused on direct medical costs, excluding indirect costs (e.g., patient transportation and lost productivity) and transmission dynamics, which may underestimate the full societal benefits of tNGS. Future research should incorporate these elements and extend the analysis to other LMICs scaling up molecular diagnostics, enabling cross-country comparisons and evidence-based adoption globally.

This study provides robust evidence supporting the selective adoption of tNGS for diagnosing pre-XDR/XDR-TB in high-burden settings while advancing mechanistic understanding of how rapid diagnostics generate clinical and economic value. By explicitly accounting for the survival benefit of earlier treatment initiation, we demonstrate that tNGS is cost-effective in South Africa and that the combination strategy is cost-effective in Georgia, with context-specific price thresholds guiding implementation. In India, where current tNGS pricing exceeds affordability thresholds, cost reductions and targeted deployment in high-prevalence regions are essential. These findings move beyond whether tNGS is cost-effective to provide actionable guidance on why, how, and under what conditions it delivers value—information critical for national TB programs designing context-specific strategies to reduce diagnostic delays and improve survival.

GF conceived and designed the study. GF, TT, BGL, and TRF developed the model structure and defined the input parameters. GF, TT, FB, MBF, CA, and MJ retrieved and verified the underlying data estimates. GF, FB, XY, SCC, CA, MS, and WKM conducted the cost-effectiveness analysis and interpreted the results. WKM supervised the statistical analyses and overall project coordination. GF, TT, BGL, MBF, and TRF drafted the initial manuscript. XY, CA, LG, NC, LP, MJ, SC, MS, and WKM reviewed and revised subsequent drafts. All authors had full access to the data, contributed to the final edits, and approved the final version of the manuscript for submission.

Not applicable. This study is a decision-analytic cost-effectiveness modeling study based exclusively on published, publicly available, aggregated, and de-identified data from scientific literature and public health reports. It did not involve direct interaction with human participants, access to individual-level patient data, or identifiable records. Therefore, institutional ethics approval and informed consent were not required.

The initial draft of the graphical abstract was generated using the DALL·E 3 image generation model (accessed via the DeeVid.ai platform) based on text prompts provided by the authors. The final version was substantially modified, edited, and formatted by the authors using Adobe software to ensure scientific accuracy and visual clarity. No AI tools were used for data analysis, interpretation, or manuscript text generation.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

All data used in this study were obtained from publicly available sources, which are fully referenced in the manuscript. No restrictions apply to data access. All data supporting the findings are included within the article and its supplementary materials.

The authors declare that they have no conflicts of interest that may be considered to influence directly or indirectly the content of the manuscript.