Tuberculosis (TB) remains a global public health emergency, ranking as a leading cause of death from an infectious disease (World Health Organization, 2024). The World Health Organization (WHO) reported an estimated 10.8 million new cases (134 per 100,000 population) in 2023, and China ranked the third among high-TB-burden countries with approximately 741,000 annual cases, including 29,000 multidrug-resistant/rifampicin-resistant TB (MDR/RR-TB) cases, accounting for 7.3% of the global burden (World Health Organization, 2024). Therefore, timely and accurate diagnosis of TB and drug resistance is crucial to eradicate tuberculosis in China.

Despite being the gold standard for TB diagnosis, conventional mycobacterial culture is limited by the requirement of biosafety infrastructure and lengthy culture time (3–8 weeks) (Schumacher et al., 2019). As with low sensitivity and inability to distinguishing non-tuberculous mycobacteria (NTM) from Mycobacterium tuberculosis (MTB), the diagnostic utility for smear microscopy is also limited (Steingart et al., 2006). Several nucleic acid amplification tests (NAATs) endorsed by WHO have demonstrated excellent diagnostic performance for MTB over the past decade, particularly the Xpert MTB/RIF (Xpert), which simultaneously detected MTB and rifampicin resistance within 2 h by targeting the rpoB gene (World Health Organization, 2021; World Health Organization, 2013; World Health Organization, 2020; Chakravorty et al., 2017; Nathavitharana et al., 2017; Penn-Nicholson et al., 2021). However, its widespread implementation in resource-limited settings remains constrained by infrastructure requirements and economic considerations.

The iFIND TBR (iFIND) assay, based on microfluidic technology, is an all-in-one assay for simultaneously identification TB and rifampicin resistance by targeting IS6110, IS1081 and rpoB gene, which integrates nucleic acid extraction, amplification, and detection into a single-use cartridge (Ou et al., 2024). It simplified the cumbersome manual operations with entire results delivered only 85 min. Preliminary studies indicated that the iFIND TBR system is user-friendly, highly accurate, and well-suited for resource-limited settings (Ou et al., 2024). However, clinically validated performance data in real-world settings is unavailable for iFIND TBR. In this study, we employed a head-to-head comparative design to evaluate the performance of iFIND-TBR technology at municipal-level TB prevention and control institutions, which will provide essential evidence to support the scaled implementation of this novel diagnostic platform.

To address critical gaps in global tuberculosis control, there is an urgent need for novel diagnostic solutions that can overcome the limitations of current methods by delivering enhanced diagnostic accuracy, operational advantages, and implementation feasibility (Seki et al., 2018; MacLean et al., 2020). iFIND, an automatic molecular point-of care testing (POCT) system, provides an option in accurately and rapidly for TB diagnosis and resistance detection. The specific targets are crucial for diagnosis, and IS6110 was considered as the most commonly target for detecting MTB (Eisenach et al., 1990; Thierry et al., 1990). However, it was reported to be missing in some clinical samples, resulting in false-negative (Nghiem et al., 2015) and false-positive results (Sankar et al., 2011). Therefore, iFIND combined IS6110 with another insertion sequence IS1081 (Vansoolingen et al., 1992; Collins and Stephens, 1991) based on nested real-time fluorescent quantitative PCR technology to eliminate the false-negative results of MTB diagnosis. Moreover, iFIND utilized five distinct molecular beacons specifically designed to the rifampicin resistance-determining region (RRDR) of the rpoB gene, using the melting curve method to detect rifampin resistance. Preliminary research results on frozen clinical samples showed that the iFIND detection system is simple to operate and has high accuracy (Ou et al., 2024).

This in real world head-to-head evaluation demonstrated that iFIND achieved a sensitivity of >98% for MTB detection compared with the solid culture reference and bacteriologically TB standard, respectively. The observed discordance, specifically bacteriology-positive but molecular-negative results, reflects a well-recognized diagnostic challenge in global TB control. A large operational study in Myanmar reported that 4% of smear-positive patients were Xpert-negative, with low bacterial load (scanty/1+) and advanced age (≥65 years) identified as key contributing factors (Phyu et al., 2018), aligning closely with our findings. In our cohort, all culture-positive but iFIND-negative specimens exhibited low or very low bacterial loads upon Xpert quantification. Conversely, 91.04% cases with iFIND-positive but culture-negative outcomes were clinically diagnosed as TB, indicating the comparability sensitivity in MTB detection for iFIND. According to the WHO’s Target Product Profile (TPP) (World Health Organization, 2024), the optimal sensitivity requires≥95% for MTBC detection on sputum-based assays. And the sensitivity of iFIND for the detection of MTB was significantly higher than that of Xpert compared with clinically diagnosis standard. Therefore, iFIND was a robust tool for paucibacillary samples, such as those from HIV/TB co-infected, extrapulmonary TB or pediatric TB patients. Though there was no statistically significant difference, the sensitivity of iFIND for the detection of MTB was slightly higher than that of Xpert compared with solid culture reference. And the discordant results between iFIND and Xpert MTB assays predominantly occurred in specimens with low bacterial loads, which are more susceptible to minor perturbations, such as sample degradation during transport, inadequate homogenization, and processing delays, resulting in false-negative results.

It is important to note that this study employed multiple reference standards to ensure a comprehensive and clinically relevant evaluation of iFIND’s diagnostic performance. In clinical practice, especially in high-TB-burden settings, no single test is perfect or universally available. The traditional “gold standard” of solid culture, while highly specific, may yield false negatives in paucibacillary cases or be inaccessible in resource-limited settings. By adopting a bacteriological standard, we aimed to maximize sensitivity for case detection, reflecting the real-world clinical strategy of utilizing all available laboratory evidence. This approach provides a performance estimate most relevant to clinicians, who must act upon any positive diagnostic result. Conversely, the use of clinical diagnosis as a comparator enabled us to evaluate test performance in the challenging subgroup of culture-negative patients who still require treatment, allowing us to assess the utility of iFIND across the full spectrum of presumptive TB. Our analytical focus was directed toward comparing newer, more sensitive molecular assays against the most relevant and rigorous benchmarks available. Consequently, directly comparing with smear positivity holds limited utility, given that smear microscopy exhibits low sensitivity and is frequently negative in paucibacillary and extrapulmonary TB, which constitute a substantial proportion of clinically diagnosed cases. A separate comparison would primarily underscore the well-established limitations of smear microscopy in these populations, offering little novel insight into the performance of the iFIND assay.

In this study, the diagnostic efficiency of iFIND was evaluated using the DOR. As a comprehensive metric that integrates both sensitivity and specificity, the DOR reflects the ratio of the odds of a positive result in diseased individuals to the odds in non-diseased individuals, thereby providing a single statistical measure of overall test performance. A higher DOR generally indicates better diagnostic accuracy. Regardless of the reference standard applied, iFIND consistently yielded high DOR values, demonstrating its robust diagnostic efficiency. Moreover, shorter diagnostic turnaround time (<2 h) for iFIND enable clinical decision-making at municipal TB control centers on the same-day of patient visit, making it particularly advantageous in resource-limited settings. And the automated computer-based interpretation eliminates subjective errors associated with manual result analysis. Although the iFIND assay exhibited a slightly higher failure rate (2.51%, 12/478) compared to Xpert (1.46%, 7/478), this difference was not statistically significant (p > 0.05). The primary reasons for test failure may contribute to technical errors during initial processing steps, which are unlikely to pose a major barrier to the deployment of iFIND in primary healthcare or general hospital laboratories.

The iFIND TBR assay also facilitates detection of RIF resistance mutations within RRDR. In this study, iFIND demonstrated sensitivities of 98.10% for RIF resistance, meeting the WHO TPP targets of ≥95% for RIF (World Health Organization, 2024), which is effective in diagnosing clinical rifampicin resistance. However, the sensitivity of iFIND for rifampicin resistance detection is lower than that of MTB diagnosis, indicating that MTB-positive with indeterminate rifampicin resistance cases may occurred. Moreover, our data revealed a strong bacterial load-dependent performance pattern (Supplementary Table 1), where specimens with low-level MTB positivity (1+) exhibited lower resistance detection rates compared to those with a higher bacterial load (≥2+). This observation aligns with established literature that samples with low bacterial load are likely to result in inconsistent molecular RIF susceptibility results (Deng et al., 2023; Huo et al., 2020). However, there are some samples with discordant results between iFIND TBR and phenotypic DST, but consistent with the Xpert assay. The 2 iFIND-sensitive/phenotypically-resistant cases were confirmed as sensitive by sequencing, which may be due to mutations outside the RRDR region, potentially leading to undetected mutations for iFIND. And 3 iFIND-resistant/phenotypically-sensitive cases (Supplementary Table 2) included one hetero-resistance and two cases with mutations in rpoB_L511P and L533P, respectively, suggesting that these two mutations may be associated with low-level resistance to RIF, which falls within or below the critical value for the proportional method of RIF. Therefore, iFIND offers clinically guidance for tuberculosis management, particularly in treatment-experienced patients where resistance patterns may be complex. Moreover, two iFIND-sensitive/Xpert-resistant cases were confirmed as heteroresistant mutations by sequencing analysis. This discrepancy may be attributed to the methodological and interpretation criteria differences, with Xpert using PCR fluorescence probe method, while iFIND uses melting curve method. In addition, uneven distribution of mutant and wild-type bacilli in clinical specimens might lead to differences in the results of the two tests.

This study has several limitations. First, this study is limited to the Chengde, Hebei Province, which may not fully represent the genetic diversity of Mycobacterium tuberculosis across China. To ensure broader applicability, future studies should be validated through multicenter studies and geographically diverse cohorts. Second, while this study provides valuable data on iFIND performance for pulmonary tuberculosis diagnosis, its applicability to extrapulmonary TB and pediatric cases remains to be established. Third, the exclusion of tests that failed due to technical or sample-related reasons, though low in number, may lead to an underestimation of real-world operational challenges. Fourth, we did not include a formal cost-effectiveness analysis or an assessment of long-term health system impact, which are crucial for guiding policy and implementation. Finally, inter-operator variability in test performance was not systematically evaluated.

This study demonstrated that the iFIND assay is a rapid and automated assay with high sensitivity and specificity comparable to Xpert for early TB diagnosis and drug resistance screening, making it particularly suitable for implementation in primary healthcare settings and general hospitals.