Skip to main content
Download PDF

Sensitive diagnosis of paucibacillary tuberculosis with targeted next-generation sequencing: a molecular diagnostic study

BMC Medicine volume 23, Article number: 178 (2025) Cite this article

Abstract

Background

Targeted next-generation sequencing (tNGS) enables high-performance tuberculosis (TB) diagnosis and drug resistance prediction directly from clinical samples. However, its applicability to paucibacillary TB, including pediatric TB and extrapulmonary TB (EPTB), has been less explored. We aimed to evaluate the performance of tNGS in these challenging TB presentations.

Methods

We prospectively and consecutively enrolled children (< 18 years) with suspected TB and adults with suspected EPTB. All participants underwent a comprehensive clinical examination, laboratory tests, and tNGS analysis. The diagnostic performance of tNGS was evaluated against composite reference standards, while resistance prediction capabilities were assessed with GeneXpert MTB/RIF and phenotypic drug susceptibility testing.

Results

A total of 85 children and 228 adults were enrolled. In children, tNGS showed a sensitivity of 74% (95% CI, 61–84%) and a specificity of 97% (95% CI, 84–100%) for microbiologically and clinically confirmed TB, whereas in adults with microbiologically and clinically confirmed EPTB, it demonstrated 77% sensitivity (95% CI, 68–83%) and 98% specificity (95% CI, 94–100%). For drug resistance prediction, tNGS exhibited variable sensitivity, peaking at 88% for rifampicin (95% CI, 47–100%) and bottoming out at 38% for streptomycin (95% CI, 9–76%), alongside a consistently acceptable specificity ranging from 89% (95% CI, 76–96%) to 100% (95% CI, 93–100%).

Conclusions

tNGS is a potentially promising test that enables rapid and sensitive diagnosis of TB in children and individuals with extrapulmonary TB. However, the variability in its accuracy for predicting drug resistance in these populations needs to be validated and addressed before its clinical application.

Peer Review reports

Background

Globally, tuberculosis (TB) has once again become the leading cause of death from infectious diseases, with an estimated 10.8 million incident cases and 1.25 million deaths occurring in 2023 [1]. The decline in incident cases and deaths deviates from the anticipated reduction rates required to meet the “End TB” schedule [2]. Moreover, the escalating issue of drug-resistant TB is alarming: 3.7% of incident cases in 2023 were multidrug-resistant TB (MDR-TB), totaling 400,000, with 160,000 deaths [1], which represents only a mere fraction of the 19 million individuals with latent MDR-TB infections [3].

Diagnosis is the weakest link in TB management and control [4], and reports indicate that over 40% of incident cases are underdiagnosed [2]. TB diagnostics remain widely dependent on traditional sputum microbiology tests, which have poor performance for children, individuals living with HIV, and those with extrapulmonary TB (EPTB) [5,6,7], where sample collection is challenging and often results in paucibacillary samples. Identifying drug resistance in TB is crucial but remains challenging, with more than two-thirds of global MDR- or rifampicin (RIF)-resistant TB (MDR/RR-TB) undetected [1]. Current methods for detecting drug resistance, such as GeneXpert MTB/RIF (Xpert), are incomplete and miss most RIF-susceptible but isoniazid (INH)-resistant TB and undervalue the 20% prevalence of fluoroquinolone resistance in MDR/RR-TB patients [1], and accurate resistance profiling is lacking for some key drugs (e.g., bedaquiline and linezolid) for short-course regimens [8]. These oversights hinder effective treatment and increase community transmission risk [9].

The use of molecular assays for the sensitive detection of Mycobacterium tuberculosis (Mtb)-specific genes in clinical samples provides a chance for rapidly and accurately identifying TB [10], and many studies have indicated that mutations in Mtb drug resistance genes are strongly correlated with phenotypic drug resistance [11]. Consequently, the WHO has recommended the prioritization of molecular diagnostics for TB diagnosis [10] and has established a Mtb drug resistance gene mutation catalog to facilitate drug resistance profiling [12]. Rapid molecular tests, though beneficial, have limitations. These methods underperform in paucibacillary TB detection and have difficulty with non-respiratory samples [7]. Moreover, their scope is restricted to a few genetic regions linked to drug resistance, leading to incomplete profiles and an inability to detect resistance to newer and repurposed drugs such as bedaquiline, linezolid, delamanid, and primaquine [8], which impedes a comprehensive understanding of drug resistance and effective management of TB cases.

Targeted next-generation sequencing (tNGS), an adjunct tool recommended by the WHO [8], combines nucleic acid preamplification with downstream sequencing analysis. tNGS targets multiple genes, minimizing the risk of false negatives associated with single-gene assays and thus potentially increasing the detection rate of TB in children and EPTB [13,14,15], The preamplification of genes of interest allows for sensitive and specific identification of resistance mutations from clinical samples, eliminating the need for time-consuming Mtb culture. Moreover, unbiased sequencing analysis results in a wider drug resistance profile. A recent meta-analysis encompassing clinical specimens and culture isolations, along with laboratory tests and clinical validation studies, revealed that tNGS has a pooled sensitivity and specificity of up to 94.1% and 98.1%, respectively, for all drugs tested [16].

tNGS shows great promise in improving TB diagnosis and drug resistance prediction, yet there are limited data concerning paucibacillary cases, such as children and people with EPTB. In this study, we prospectively evaluated the diagnostic performance of tNGS in children and adults with EPTB, along with its accuracy in identifying drug resistance, in a real-world clinical scenario.

Methods

Study design and participants

We performed a diagnostic accuracy study comparing tNGS (Shanghai Bingyuan Medical Technology Co., Ltd.) against microbiological and composite reference standards of TB at Shenyang Chest Hospital, a TB-designated hospital in Northeast China. We consecutively approached and enrolled children under 18 years of age with suspected TB and adults with suspected EPTB from January 1, 2021, to March 1, 2024 (Fig. 1). Suspected TB in children was defined according to the National Institutes of Health criteria [17], including symptoms such as a cough persisting for more than two weeks, weight loss, malnutrition, HIV infection, a history of TB exposure, and abnormal imaging results suggesting TB. Suspected EPTB was identified according to WHO guidelines [18] by site-specific signs and symptoms, including lymphadenopathy, arthralgia with swelling, neck stiffness, and cognitive disturbances, along with general symptoms of active TB.

Fig. 1
figure 1

Flow chart of the study population. The participants were post hoc classified as critical (Additional file 1: Table S2) based on their clinical data. a Continuous inclusion of outpatients with self-reported TB-related symptoms. b Children (< 18) with suspected TB and adults with suspected EPTB. c Children lacked detectable sputum samples and were ineligible for bronchoscopy to obtain alveolar lavage fluid. d Adults without indications for puncture or surgery for collecting specimens from extrapulmonary lesion sites. Abbreviations: EPTB extrapulmonary tuberculosis, TB tuberculosis, tNGS targeted next-generation sequencing

Full size image

Procedures

At the initial consultation, all participants underwent clinical, imaging, and immunological (TST or IGRA) examinations. Sputum samples were collected for microbiological tests, including acid-fast bacillus smear microscopy (AFB smear), Mtb culture, Xpert, loop-mediated isothermal amplification (LAMP), and tNGS. Bronchoalveolar lavage fluid was collected and tested for those with suspected pulmonary TB but without detectable sputum when clinically feasible. Non-respiratory samples were obtained and tested from the corresponding sites in individuals with suspected EPTB (Additional file 1: Supplementary methods). Considering the limited sample volume, the collected samples were tested in a prioritized sequence: tNGS, Xpert, Mtb culture, LAMP, and smear. The culture isolates were tested via MGIT 960-based phenotypic drug susceptibility testing (pDST) (Additional file 1: Table S1). Clinicians make treatment decisions for TB patients based on assessment results, excluding tNGS findings. Participants were followed for > 6 months after enrollment for more definitive diagnosis and to assess disease progression and/or treatment response. Baseline and follow-up data (this may include extra test results) were used for the final classification according to the adapted criteria (Additional file 1: Table S2).

tNGS workflow and quality control

Collected samples are transported with dry ice to the center laboratory for tNGS on the same day by a specialized clinical sample transport service. The laboratory initiates the testing workflow according to its schedule, which includes sample preprocessing and nucleic acid extraction, pre-amplification and library construction, high-throughput sequencing, and bioinformatics analysis. To prevent cross-contamination between samples and the environment, tests are conducted by experienced technicians following standard operating procedures in a negative-pressure partitioned laboratory. If quality control references fail, the test results are deemed invalid, and retesting is conducted (Additional file 1: Supplementary Methods).

Outcomes and case classification

The primary outcome of the study was the diagnostic accuracy of tNGS compared with composite reference standards. The secondary outcome was the drug resistance prediction accuracy of tNGS compared with that of the pDST and Xpert. In addition, the diagnostic delay of tNGS relative to other microbiologic and molecular tests was evaluated.

Owing to the diagnostic challenges inherent in the participants, the post hoc clinical expert review board classified the included children and adults into categories of confirmed TB/EPTB, unconfirmed TB/EPTB, unlikely TB/EPTB, and non-TB, on the basis of adapted criteria (Fig. 1, Additional file 1: Table S2) [17,18,19] that incorporate clinical and laboratory findings as well as treatment response, to accurately evaluate the performance of tNGS. In this study, we categorized cases younger than 18 years with pulmonary and/or extrapulmonary TB evidence as “children with suspected TB” for analysis. For patients over 18 years, those with EPTB evidence regardless of pulmonary TB were classified as “adults with suspected EPTB.” Adults with only pulmonary TB signs and symptoms were excluded (Fig. 1).

Statistical analysis

We hypothesized that the pooled sensitivity of tNGS would be higher than the Xpert for pediatric and extrapulmonary TB. To estimate the required sample size, we used a sensitivity of 70–80% for tNGS, alongside 40–60% for Xpert based on preliminary data and previous studies [13, 14, 20, 21]. Assuming a proportion of 60% confirmed or unconfirmed cases in enrolled participants, and 30% discordant result between tNGS and Xpert result, a sample size of 105 in each group would yield a power above 0.8 at the level of a type I error (α) of 0.025.

Continuous variables are summarized as medians and interquartile ranges (IQRs), and categorical variables are summarized as counts and percentages. Wilcoxon rank-sum tests were used to compare the delay between tNGS and other assays. The diagnostic and drug resistance prediction performance of tNGS was assessed by calculating the sensitivity and specificity against microbiological and composite reference standards. All proportions were calculated and reported with 95% confidence intervals (95% CIs). All tests were two-sided with an α of 0.05. All analyses were performed with Stata 14 (version 17.0). GraphPad Prism (version 10.2.3) was used for data visualization.

Results

We ultimately enrolled 313 eligible participants, including 85 children with suspected TB and 228 adults with suspected EPTB (Fig. 1). Among the children [median 15 (IQR 9–17) years], 88% (75/85) reported TB-related symptoms, all of whom had abnormal radiological findings, 70% (48/69) had positive TST/IGRA results (Table 1, Additional file 1: Table S3), and they were typically first seen at a median of 17 (IQR, 7–56) days post onset. All the children had pulmonary lesions, with a minority (15/85) also having extrapulmonary lesions such as thoracic, soft tissue, and lymph node involvement. Among all the children, 63.5% (54/85) received anti-tuberculosis therapy (ATB), with 93% (50/54) showing clinical improvement. Adults (median 55 (IQR 35–63) years) reported higher proportions of TB-related symptoms (98%, 223/228) and immunosuppressive conditions (20%, 46/228), comparable proportions of abnormal radiographic findings (99.6%, 227/228), and lower percentages of positive immunologic test results (59%, 116/228) (Table 1, Additional file 1: Table S4). The median delay to the first visit for suspected EPTB in adults was 65 (IQR, 28–140) days. Approximately half (52%, 119/228) of the adults had pulmonary lesions, and all adults reported at least one extrapulmonary involvement, including 134, 53, and 44 adults who reported bone, chest, and lymph node and soft tissue lesions, respectively, while seven adults reported lesions in other sites (head, urinary system, etc.). A total of 51% (117) of the adults received ATB, with 89% (100) responding favorably.

Table 1 Baseline characteristics of children with suspected TB and adults with suspected EPTB
Full size table

We first assessed the reporting delay, the interval from sample submission to result reporting, among different assays (Fig. 2a). On median, all three molecular assays presented significantly shorter reporting times than did Mtb culture [42 (IQR 42–43) days], whereas tNGS (3 (IQR 3–4) days) reported results 2 days later than Xpert and LAMP. This difference in reporting delay was independent of disease type and participant age (Additional file 1: Fig. S1). Notably, Mtb culture typically requires a longer time to obtain negative results.

Fig. 2
figure 2

The diagnostic results of participants with different assays. a Time delay in the diagnosis of tuberculosis (TB) via Mtb culture (culture), GenenXpert MTB/RIF (Xpert), TB-LAMP (LAMP), and targeted next-generation sequencing (tNGS). The time delay was defined as the time from specimen submission to the reporting of results. Two-sided p values from the Wilcoxon rank-sum test. b Results profile of the detection of 313 participants with different TB classifications via culture, Xpert, LAMP, and tNGS. c Venn diagram of tNGS and the indicated assays, including culture, Xpert, or LAMP, in participants post hoc classified as confirmed TB/EPTB (n = 96). d Distribution of tNGS-positive results for different types of specimens from participants with different TB classifications. NA indicates that the participant did not take the test or failed the test. a Assessed post hoc on the basis of adapted critical (Additional file 1: Table S2). b Seven cerebrospinal fluid samples, four blood samples, and two urine samples were included. Abbreviations: BALF bronchoalveolar lavage fluid, EPTB extrapulmonary tuberculosis, NA not amplicabale

Full size image

Given the diagnostic challenges of childhood TB and EPTB, we then evaluated the diagnostic performance of tNGS via a composite reference (Additional file 1: Table S2) that post hoc classified children and adults as confirmed, unconfirmed, unlikely, and non-TB cases (Fig. 1, Table 1, Additional file 1: Table S3–S4). Among all the included paucibacillary cases, tNGS correctly identified 83% (80/96) of the confirmed cases, 65% (45/69) of the unconfirmed cases, and all the unlikely cases (13/13) while incorrectly identifying three non-TB cases (Figs. 1 and 2b, Table 2). Subgroup analysis revealed no significant differences in the diagnostic performance of tNGS between the younger (0–14 years) and older (15–18 years) children, as well as between the children TB and the adult EPTB (Table 2 Additional file 1: Table S5). In summary, tNGS detected microbiologically and clinically confirmed TB in children, with a sensitivity of 74% (95% CI, 61–84%) and a specificity of 97% (95% CI, 84–100%), and in adult EPTB, with a sensitivity of 77% (95% CI, 68–83%) and a specificity of 98% (95% CI, 94–100%) (Additional file 1: Table S6). By predominantly identifying 65% of unconfirmed cases (Fig. 2b, Table 2), tNGS demonstrated significantly greater diagnostic sensitivity than other tests, regardless of age or disease type, with comparable specificity (Additional file 1: Table S6). However, the tNGS gave false-negative results in 16 confirmed cases, 11 of which were negative at baseline and further identified by additional follow-up testing; and five (one “medium” and four “very low” graded) of which were baseline Xpert positive but tNGS negative, which may be due to the heterogeneity of solid and/or paucibacillary samples (Additional file 1: Fig. S2).

Table 2 tNGS results in children with suspected TB and adults with suspected EPTB
Full size table

The subgroup analysis of confirmed cases provided further evidence of the superior diagnostic sensitivity of tNGS, with an additional 11% to 29% of cases diagnosed compared with the other three tests (Fig. 2c). tNGS offers marked diagnostic advantages over Mtb culture for both children and adults (Additional file 1: Fig. S3 and Table S7), which may be due to the predominance of non-respiratory samples with low bacterial loads in this study (Additional file 1: Tables S2 and S3). Therefore, we subsequently evaluated the performance of tNGS across specimen types. In diagnosing confirmed and unconfirmed cases, tNGS demonstrated the highest sensitivity (87–92%) in bone and soft tissue and lymph nodes, moderate and variable sensitivity (60–100%) in sputum and alveolar lavage, and comparatively lower sensitivity (56–67%) in pleural fluid (Fig. 2d). Despite not always being optimal, side-by-side comparisons of other assays revealed that tNGS was superior for a broad spectrum of samples (Additional file 1: Table S8).

A major advantage of tNGS is its ability to rapidly identify multiple drug-resistant TB genotypes [16]. Thus, we further assessed the accuracy of tNGS in predicting drug resistance in paucibacillary TB. We observed that only a minority (54 cases) of individuals underwent pDST testing due to extremely low (22.4%) Mtb culture positivity (Table 1). Similar to the delay in diagnosis, a significant delay [median 59.5 (IQR, 57–69) days] in the pDST was observed (Fig. 3b). Xpert was more widely used, covering 89% (279/313) of the cases, except for 34 adults with suspected EPTB limited by specimen type. All participants underwent tNGS, which identified 84 mutations across 13 resistance genes (Fig. 3c, Additional file 1: Table S9). The most frequently mutated genes were rpoB, pncA, and rpsL, suggesting potential resistance to RIF, pyrazinamide, and streptomycin (STM), respectively. The predominant mutations were rpsL K43R (11 cases), rpoB S450L (9 cases), and katG S315T (7 cases). Among the three resistance tests performed, 40 showed resistance in at least one test (Additional file 1: Fig. S4 and S6).

Fig. 3
figure 3

Drug resistance prediction of participants with different assays. a Distribution of phenotypic drug susceptibility test (pDST), GeneXpert MTB/RIF (Xpert), and targeted next-generation sequencing (tNGS) results among participants. b The time delay for drug susceptibility testing with the indicated three assays. Two-sided p values from the Wilcoxon rank-sum test. c The distribution of genes associated with 84 instances of drug resistance identified in the tNGS results. d Evaluation of the performance of tNGS in predicting rifampicin (RIF) resistance with Xpert as the reference assay. e Evaluation of the performance of tNGS in predicting resistance to first-line and key drugs, including isoniazid (INH), RIF, ethambutol (EMB), pyrazinamide (PZA), fluoroquinolones (FQs), linezolid (LZD), and bedaquiline (BDQ), with the pDST as the reference assay. Abbreviations: R drug resistant, S drug susceptible, U unclear (MIC between sensitivity and resistance thresholds), NA not applicable

Full size image

Given the limited pDST coverage, we initially assessed the ability of tNGS to predict RIF resistance compared with Xpert. tNGS missed nine Xpert RIF-positive cases but identified three Xpert RIF-negative cases, indicating a low sensitivity of 59% (95% CI: 39–79%) but a high specificity of 99% (95% CI: 97–100%) (Fig. 3d). However, a further evaluation against the scarce pDST data revealed that both tNGS and Xpert had suboptimal accuracy, with Xpert having a higher rate of false positives (Fig. 3d).

Finally, in an extensive assessment of resistance prediction accuracy using pDST as a benchmark, tNGS demonstrated varied performance across different drugs (Fig. 3e, Additional file 1: Fig. S6-S7, and Table S10). tNGS demonstrated robust and high specificity (90–100%) for the majority of the 14 drugs with pDST results (Fig. 3e, Additional file 1: Table S10) but fell short for predicting RIF resistance, with a specificity of 89% (95% CI, 76–96%). The sensitivity estimates for resistance prediction were inconsistent and suboptimal, possibly limited by the small number of resistant cases and the paucibacillary nature of the participants. For the first-line drugs INH, RIF, and ethambutol, tNGS had sensitivities of 88% (95% CI, 47–100%), 70% (95% CI, 35–93%), and 67% (95% CI, 9–99%), respectively (Fig. 3e, Additional file 1: Table S10). Notably, tNGS correctly identified one out of two fluoroquinolone-resistant cases, two out of three STM-resistant cases, and one out of two levofloxacin-resistant cases. Unfortunately, tNGS identified five potential pyrazinamide-resistant cases and one bedaquiline-resistant case, but the lack of corresponding pDST results prevented validation of these findings (Fig. 3e).

Discussion

In this study, we conducted a prospective evaluation of the utility of tNGS in diagnosing TB and predicting drug resistance in children and individuals with EPTB, populations often underdiagnosed in clinical practice [5,6,7]. Our findings indicate a significant advantage of tNGS in diagnostic performance within these participants, albeit with suboptimal performance in drug resistance prediction.

Timely and accurate diagnosis is critical for children with TB and those with EPTB, as they are at increased risk of mortality [22, 23]. Children (aged 0–14 years) account for an estimated 12% of incident TB cases and account for 16% of TB-related deaths annually, yet only 8% of notifications in 2022 involved children [1], a proportion that is even lower in some high-burden countries. The challenge of collecting sputum samples and the paucibacillary nature of pediatric TB contribute to suboptimal detection rates [5, 6]. Despite advances such as the detection of swallowed Mtb in stool samples, the diagnosis of TB in this population remains a significant challenge, with an unsatisfactory sensitivity of 67% [24]. The diagnosis of EPTB is particularly challenging due to the diverse clinical presentations that depend on the affected organ, disease stage, and immune response of the host [25]. Direct evidence of EPTB is a positive microbiological result from an invasive specimen from a suspected lesion; however, these tests often yield negative results owing to the paucibacillary nature of EPTB [26]. Despite the introduction of rapid molecular tests such as Xpert, the sensitivity for diagnosing EPTB varies significantly, from 21 to 81% [7], compared with composite reference criteria. Consequently, empirical therapy is commonly employed in clinical practice.

Encouragingly, tNGS demonstrates comparable diagnostic performance to Xpert for paucibacillary populations with positive microbiological evidence and significantly outperforms smear, culture, and LAMP-TB tests. For clinically diagnosed populations with typically lower bacillary loads, tNGS exhibited unparalleled superiority in this study. The high diagnostic performance of tNGS likely stems from the simultaneous amplification of multiple genes of Mtb, significantly increasing the likelihood of detecting Mtb-DNA even at trace levels in clinical samples [16]. Additionally, tNGS was specifically designed with a variety of diagnostic scenarios, including non-respiratory specimens [13,14,15], and offers higher diagnostic efficacy than Xpert and LAMP for extrapulmonary specimens. However, we recognize that single-specimen-based tNGS does not fully address the diagnostic challenges of paucibacillary cases. Adhering to the recommendations from the WHO for multi-specimen testing may yield enhanced diagnostic efficiency for these populations. Nonetheless, given the limited availability of precious specimens obtained through invasive sampling for localized diseases, it remains crucial to optimize and develop highly sensitive diagnostic tools, such as CRISPR-based assay [27], to ensure accurate diagnosis.

In addition, it is prudent to exercise caution, as our study revealed that tNGS incorrectly identified three participants without TB with low sequencing readings. Two of these individuals presented with TB-related symptoms, while the third presented with positive immunological test results. The post-hoc analysis confirmed that all samples were processed with precise quality control, suggesting that contaminations mainly originate from complicated clinical scenarios, including non-standardized sampling and potential contamination on-site. This highlights the need for standardized sampling protocols to reduce the risk of contamination, including defined sampling rooms, specialized protocols for different sample types, case-specific isolation and decontamination strategies, and staff training. Contamination of low-titer samples by high-titer ones is a well-recognized challenge in viral sequencing and likely impacts TB tNGS practices as well [28]. Adopting sequential processing strategies based on bacterial load—starting with low-load specimens (e.g., cerebrospinal fluid and blood) before high-load ones (e.g., sputum and pus)—is recommended, as it effectively minimizes cross-contamination and ensures diagnostic accuracy. Moreover, setting appropriate thresholds or integrating diverse clinical data for result interpretation could minimize false-positive outcomes [29].

The most compelling aspect of tNGS is to concurrently predict multi-drug resistance in TB, positioning it as a high-priority tool to predict drug-resistant in confirmed TB cases and in cases with high drug resistance risk, and it was endorsed by the WHO [8, 30, 31]. Recent meta-analyses and multicenter clinical evaluation studies have shown that tNGS demonstrates high accuracy and reliability in DR-TB detection [16, 32]. However, our findings indicate that the accuracy of tNGS in drug resistance prediction among the studied populations was suboptimal, likely due to several contributing factors. Firstly, some known low-frequency resistance genes are excluded due to testing costs; furthermore, potential novel mutations may be undetected due to incomplete characterization of genetic resistance profiles for certain drugs [32]. This highlights a critical issue with tNGS: the lack of thorough premarket clinical evaluation, which is standard for other rapid molecular diagnostic methods to ensure clinical applicability. Importantly, the paucibacillary nature of our study population, known to reduce the sensitivity of next-generation sequencing [33], may have further compromised observed performance [34], a factor warranting careful consideration. Unfortunately, limited data hinder the identification of potential sources of population heterogeneity. Therefore, we advise caution in interpreting these results and suggest the following key directions to improve the clinical performance of tNGS: (1) more comprehensive delineation of Mtb resistance gene profiles and establishment of robust mapping between phenotypic and genotypic resistance; (2) strategic selection of target genes guided by the WHO Mtb drug resistance gene catalog [12], taking into account mutation frequency, drug resistance relevance and cost to optimize technical feasibility, diagnostic performance and cost-effectiveness; (3) optimization of amplification and sequencing systems to improve target detection efficiency and precision; and (4) establishment of interpretation frameworks in extensive clinical data to define thresholds for different drugs across populations.

Notably, tNGS also revealed positive results in some pDST-negative cases, with a low proportion of mutant readings. This could indicate heterogeneous resistance and the capacity of tNGS to detect minor variants and heteroresistance [35]. These findings may raise concerns among clinicians regarding the potential for developing drug resistance and prompt discussions on the adoption of targeted therapeutic strategies [35, 36]. Specifically, clinicians may need to consider whether to initiate drug-resistant regimens or adapt the dosages within drug-susceptible regimens. This highlights a significant challenge in refining tNGS, establishing accurate thresholds for various drugs based on genotype‒phenotype correlations to reliably report resistance. Achieving this goal will be a major focus of future work. Furthermore, considering the limited number of participants analyzed for drug resistance and the complexities discussed, we advise caution when interpreting the tNGS resistance prediction outcomes in this study.

tNGS offers a significant advantage by testing clinical samples directly, providing a shorter turnaround time [16] that accelerates the initiation of anti-TB treatment, which is critical for reducing the risk of serious disease and death, particularly for children and those with EPTB. While tNGS has a markedly shorter turnaround time than culture does, it does not provide same-day results such as Xpert and LAMP-TB. Post hoc analyses suggest that the main delays are due to waiting times for transport and pre-testing. Currently, to minimize costs, tNGS is predominantly commercialized through centralized regional testing, which means that transport and batch testing contribute to delays in rapid result reporting. However, we expect these issues to be resolved or mitigated as tNGS adoption becomes more widespread. At this stage, reducing the pre-test waiting time by reducing the sample size tested in each batch may facilitate the release of tNGS reports. However, this will require the formulation of economic impact models for different disease subtypes to assess the trade-off between waiting time and cost.

We notice that tNGS entails higher upfront costs, highlighting potential cost-related challenges in implementing tNGS within routine clinical settings. A WHO cost-effectiveness model analysis of tNGS for detecting DR-TB in three countries (Georgia, India, and South Africa) with varying TB prevalence has corroborated this finding [8]. However, our results also suggest that for certain cases requiring multiple Xpert tests for confirmation, tNGS offers greater cost-efficiency. In particular, the comprehensive drug resistance information provided by tNGS is valuable for the timely initiation of personalized therapies. We therefore recommend that future studies focus on practical applications and long-term economic evaluations to fully assess the benefits of incorporating tNGS into TB diagnostic algorithms.

Our study has limitations that could affect its interpretability. The definition of pediatric TB is debated [37]; here, we define children as individuals < 18 years, aligning with the UN Convention on the Rights of the Child and Chinese Law to align with local clinical practice; however, this definition differs from the definition (< 15 years) used in the WHO Global TB Report, which may affect the comparability of the study results. However, no tNGS performance differences were observed between younger (0–14 years) and older (15–18 years) children in this study. Nevertheless, it remains unclear whether this reflects the broad age tolerance of tNGS. Given the limited number of participants, we recommend interpreting the effect of age on tNGS with caution and specifically addressing this issue in larger, well-designed cohort studies. Second, this study was set in a TB-designated hospital, which may introduce bias due to the characteristics of the visit patient population. In this study, we included a lower proportion of unlikely cases among participants and a higher incidence of bone TB within the EPTB group, which could restrict the comparability and generalizability of findings for these cases. Future multicenter studies should investigate the robustness of tNGS in accommodating population heterogeneity, encompassing various age groups and disease subtypes. When different assays were compared, the complexity of the clinical scenarios prevented consistent use of the same sample for head-to-head validation. Furthermore, although we included immunosuppressed individuals, we did not include HIV-infected individuals, who are the most important paucibacillary population and should be thoroughly investigated in future research. At last, a locally accessible tNGS platform was used, as WHO-endorsed platforms were unavailable at study initiation, reflecting early clinical adoption of tNGS. However, since this platform has not undergone comprehensive evaluation, its robustness across different specimen types and forms of TB remains uncertain, potentially raising concerns about its accuracy. End-to-end comparative studies between the WHO-endorsed and non-endorsed tNGS will be essential to address this transitional challenge [32].

Conclusions

Our findings provide strong preliminary evidence for the utility of tNGS for the sensitive diagnosis of paucibacillary TB. However, further validation and refinement are necessary to improve its predictive accuracy for drug resistance in paucibacillary TB.

Data availability

The main data supporting the results of this study are available within the paper and its Supplementary Information. The raw datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request following publication.

Abbreviations

AFB:

Acid-fast bacillus

BDQ:

Bedaquiline

Cls:

Confidence intervals

EMB:

Ethambutol

EPTB:

Extrapulmonary tuberculosis

FQs:

Fluoroquinolones

IGRA:

Interferon-gamma release assay

INH:

Isoniazid

IQRs:

Interquartile ranges

LAMP:

Loop-mediated isothermal amplification

LZD:

Linezolid

MDR-TB:

Multidrug-resistant tuberculosis

Mtb:

Mycobacterium tuberculosis

pDST:

Phenotypic drug susceptibility testing

PZA:

Pyrazinamide

RIF:

Rifampicin

RR-TB:

Rifampicin-resistant tuberculosis

References

  1. WHO. Global tuberculosis report 2024. Geneva: WHO; 2024. https://www.who.int/publications/i/item/9789240101531. Accessed 19 March 2025.

  2. Ledesma JR, Ma J, Zhang M, Basting AVL, Chu HT, Vongpradith A, Novotney A, LeGrand KE, Xu YY, Dai X, et al. Global, regional, and national age-specific progress towards the 2020 milestones of the WHO End TB Strategy: a systematic analysis for the Global Burden of Disease Study 2021. Lancet Infect Dis. 2024;24(7):698–725.

    Google Scholar 

  3. Knight GM, McQuaid CF, Dodd PJ, Houben RM. Global burden of latent multidrug-resistant tuberculosis: trends and estimates based on mathematical modelling. Lancet Infect Dis. 2019;19(8):903–12.

    PubMed  PubMed Central  Google Scholar 

  4. Pai M, Dewan PK, Swaminathan S. Transforming tuberculosis diagnosis. Nat Microbiol. 2023;8(5):756–9.

    CAS  PubMed  Google Scholar 

  5. Thomas TA. Tuberculosis in Children. Pediatr Clin North Am. 2017;64(4):893–909.

    PubMed  PubMed Central  Google Scholar 

  6. Gupta RK, Lucas SB, Fielding KL, Lawn SD. Prevalence of tuberculosis in post-mortem studies of HIV-infected adults and children in resource-limited settings: a systematic review and meta-analysis. AIDS. 2015;29(15):1987–2002.

    PubMed  Google Scholar 

  7. Denkinger CM, Schumacher SG, Boehme CC, Dendukuri N, Pai M, Steingart KR. Xpert MTB/RIF assay for the diagnosis of extrapulmonary tuberculosis: a systematic review and meta-analysis. Eur J Respir Med. 2014;44(2):435–46.

    Google Scholar 

  8. WHO. Use of targeted next-generation sequencing to detect drug-resistant tuberculosis: rapid communication. Geneva: WHO; 2023. https://www.who.int/publications/i/item/9789240076372. Accessed 19 March 2025.

  9. Farhat M, Cox H, Ghanem M, Denkinger CM, Rodrigues C, Abd El Aziz MS, Enkh-Amgalan H, Vambe D, Ugarte-Gil C, Furin J, et al. Drug-resistant tuberculosis: a persistent global health concern. Nat Rev Microbiol. 2024;22(10):617–35.

    CAS  PubMed  Google Scholar 

  10. WHO. WHO standard: universal access to rapid tuberculosis diagnostics. Geneva: WHO; 2023. https://www.who.int/publications/i/item/9789240071315. Accessed 19 March 2025.

  11. Zhang Y, Yew WW. Mechanisms of drug resistance in Mycobacterium tuberculosis: update 2015. Int J Tuberc Lung Dis. 2015;19(11):1276–89.

    CAS  PubMed  Google Scholar 

  12. WHO. Catalogue of mutations in Mycobacterium tuberculosis complex and their association with drug resistance. Geneva: WHO; 2023. https://www.who.int/publications/i/item/9789240082410. Accessed 19 March 2025.

  13. Chen S, Wang C, Zou Y, Zong Z, Xue Y, Jia J, Dong L, Zhao L, Chen L, Liu L, et al. Tuberculosis-targeted next-generation sequencing and machine learning: an ultrasensitive diagnostic strategy for paucibacillary pulmonary tuberculosis and tuberculous meningitis. Clin Chim Acta. 2024;553: 117697.

    CAS  PubMed  Google Scholar 

  14. Zheng H, Yang H, Wang Y, Li F, Xiao J, Guo Y, Chen H, Wang X, Li H, Shen C. Diagnostic value of tNGS vs Xpert MTB/RIF in childhood TB. Heliyon. 2024;10(1): e23217.

    CAS  PubMed  Google Scholar 

  15. Zhang G, Zhang H, Zhang Y, Hu X, Tang M, Gao Q. Targeted next-generation sequencing technology showed great potential in identifying spinal tuberculosis and predicting the drug resistance. J Infect. 2023;87(6):e110–2.

    CAS  PubMed  Google Scholar 

  16. Schwab TC, Perrig L, Göller PC, Guebely De la Hoz FF, Lahousse AP, Minder B, Günther G, Efthimiou O, Omar SV, Egger M, et al. Targeted next-generation sequencing to diagnose drug-resistant tuberculosis: a systematic review and meta-analysis. Lancet Infect Dis. 2024;24(10):1162–76.

    CAS  PubMed  PubMed Central  Google Scholar 

  17. Graham SM, Cuevas LE, Jean-Philippe P, Browning R, Casenghi M, Detjen AK, Gnanashanmugam D, Hesseling AC, Kampmann B, Mandalakas A, et al. Clinical case definitions for classification of intrathoracic tuberculosis in children: an update. Clin Infect Dis. 2015;61Suppl 3:S179-187.

    PubMed  Google Scholar 

  18. WHO. Improving the diagnosis and treatment of smear-negative pulmonary and extrapulmonary tuberculosis among adults and adolescents. Geneva: WHO; 2007. https://www.who.int/publications/i/item/WHO-HTM-TB-2007.379. Accessed 19 March 2025.

  19. Weber SF, Saravu K, Heller T, Kadavigere R, Vishwanath S, Gehring S, Bélard S, And Pocus Eti Study G. Point-of-care ultrasound for extrapulmonary tuberculosis in India: a prospective cohort study in HIV-positive and HIV-negative presumptive tuberculosis patients. Am J Trop Med Hyg. 2018;98(1):266–73.

    PubMed  Google Scholar 

  20. Xia L, Liu X, Qian X, Li T, Xi X, Fan X, Liu P, Lu S. Performance of Xpert/MTB/RIF assay for childhood pulmonary tuberculosis among HIV negative children with real world evidence in China. J Infect Public Health. 2020;13(11):1762–7.

    PubMed  Google Scholar 

  21. Li TX, Wang J, Yang YS, Wang PS, Zhou G, Liao CY, Zhang HZ, Luo M, Zeng XG, Yang GQ, et al. Evaluation of Xpert MTB/RIF assay for the diagnosis of extrapulmonary tuberculosis in Southwest China. PLoS Negl Trop Dis. 2023;17(6):e0011403.

    CAS  PubMed  PubMed Central  Google Scholar 

  22. Jenkins HE, Yuen CM, Rodriguez CA, Nathavitharana RR, McLaughlin MM, Donald P, Marais BJ, Becerra MC. Mortality in children diagnosed with tuberculosis: a systematic review and meta-analysis. Lancet Infect Dis. 2017;17(3):285–95.

    PubMed  Google Scholar 

  23. Qian X, Nguyen DT, Lyu J, Albers AE, Bi X, Graviss EA. Risk factors for extrapulmonary dissemination of tuberculosis and associated mortality during treatment for extrapulmonary tuberculosis. Emerg Microbes Infect. 2018;7(1):1–14.

    CAS  Google Scholar 

  24. MacLean E, Sulis G, Denkinger CM, Johnston JC, Pai M, Ahmad Khan F. Diagnostic accuracy of stool Xpert MTB/RIF for detection of pulmonary tuberculosis in children: a systematic review and meta-analysis. J Clin Microbiol. 2019;57(6):e02057–118.

    PubMed  PubMed Central  Google Scholar 

  25. Baykan AH, Sayiner HS, Aydin E, Koc M, Inan I, Erturk SM. Extrapulmonary tuberculosıs: an old but resurgent problem. Insights Imaging. 2022;13(1):39.

    PubMed  PubMed Central  Google Scholar 

  26. Purohit M, Mustafa T. Laboratory diagnosis of extra-pulmonary tuberculosis (EPTB) in resource-constrained setting: state of the art, challenges and the need. J Clin Diagn Res. 2015;9(4):EE01.

    PubMed  PubMed Central  Google Scholar 

  27. Huang Z, LaCourse SM, Kay AW, Stern J, Escudero JN, Youngquist BM, Zheng W, Vambe D, Dlamini M, Mtetwa G, et al. CRISPR detection of circulating cell-free Mycobacterium tuberculosis DNA in adults and children, including children with HIV: a molecular diagnostics study. Lancet Microbe. 2022;3(7):e482–92.

    CAS  PubMed  PubMed Central  Google Scholar 

  28. Jurasz H, Pawłowski T, Perlejewski K. Contamination issue in viral metagenomics: problems, solutions, and clinical perspectives. Front Microbiol. 2021;12: 745076.

    PubMed  PubMed Central  Google Scholar 

  29. Tengyingzi L, Nanxi L, Xiaomin Z, Dandan N, Yongchao G, Dan W, Dachuan L, Tong L, Manyu H, Yingluan Z. A case of comprehensive analysis and diagnosis of mycobacterium tuberculosis complicated with multi-pathogen pneumonia by pathogens targeted next-generation sequencing. Electronic Journal of Emerging Infectious Diseases. 2024;9(3):73.

    Google Scholar 

  30. WHO. WHO consolidated guidelines on tuberculosis. Module 3: diagnosis–rapid diagnostics for tuberculosis detection, 3rd ed. Geneva: WHO; 2024. https://www.who.int/publications/i/item/9789240089488. Accessed 19 March 2025.

  31. WHO. Target product profile for next-generation drug-susceptibility testing at peripheral centers. Geneva: WHO; 2021. https://www.who.int/publications/i/item/9789240032361. Accessed 19 March 2025.

  32. Colman RE, Seifert M, De la Rossa A, Georghiou SB, Hoogland C, Uplekar S, Laurent S, Rodrigues C, Kambli P, Tukvadze N, et al. Evaluating culture-free targeted next-generation sequencing for diagnosing drug-resistant tuberculosis: a multicentre clinical study of two end-to-end commercial workflows. Lancet Infect Dis. 2024;25(3):325–34.

  33. Dookie N, Khan A, Padayatchi N, Naidoo K. Application of next generation sequencing for diagnosis and clinical management of drug-resistant tuberculosis: updates on recent developments in the field. Front Microbiol. 2022;13: 775030.

    PubMed  PubMed Central  Google Scholar 

  34. Murphy SG, Smith C, Lapierre P, Shea J, Patel K, Halse TA, Dickinson M, Escuyer V, Rowlinson MC, Musser KA. Direct detection of drug-resistant Mycobacterium tuberculosis using targeted next generation sequencing. Front Public Health. 2023;11: 1206056.

    PubMed  PubMed Central  Google Scholar 

  35. Koch A, Cox H, Mizrahi V. Drug-resistant tuberculosis: challenges and opportunities for diagnosis and treatment. Curr Opin Pharmacol. 2018;42:7–15.

    CAS  PubMed  PubMed Central  Google Scholar 

  36. Baquero F. Low-level antibacterial resistance: a gateway to clinical resistance. Drug Resist Updat. 2001;4(2):93–105.

    CAS  PubMed  Google Scholar 

  37. Basu Roy R, Whittaker E, Seddon JA, Kampmann B. Tuberculosis susceptibility and protection in children. Lancet Infect Dis. 2019;19(3):e96–108.

    PubMed  Google Scholar 

Download references

Acknowledgements

We thank Dr. Wenpan Huang and Dr. Qing Ye for their technical support.

Funding

This study was supported by the National Natural and Science Foundation of China (82302614), the Medical Scientific Research Foundation of Guangdong Province (No. A2023170), the Natural Science Foundation of Liaoning (2022-MS-432), and the Shanghai 2020 “Science and Technology Innovation Action Plan” Technological Innovation Fund (20Z11900500).

Author information

Author notes

  1. Yu Chen, Lichao Fan, and Zhong Ren contributed equally to this work.

Authors and Affiliations

  1. Department of Tuberculosis, Shenyang Tenth People’s Hospital/Shenyang Chest Hospital, Shenyang, 110044, China

    Yu Chen & Lichao Fan

  2. Department of Epidemiology, School of Public Health, China Medical University, Shenyang, 110122, China

    Yu Chen

  3. Department of Orthopedics General Surgery, Changchun Infectious Disease Hospital, Changchun, 130123, China

    Zhong Ren

  4. Tuberculosis Laboratory, Shenyang Tenth People’s Hospital, Shenyang Chest Hospital, Shenyang, 110044, China

    Yanhong Yu & Jiao Sun

  5. Department of Plastic Surgery, China Medical University, Shenyang, 110001, China

    Miaoran Wang

  6. Department of Thoracic Surgery, Shenyang Chest Hospital, Shenyang, 110044, China

    Chang Liu

  7. The First Laboratory of Cancer Institute, The First Hospital of China Medical University, Shenyang, 110001, China

    Chang Liu

  8. The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310003, China

    Ying Zhang

  9. National Clinical Research Center for Infectious Disease , Shenzhen Third People’s Hospital, Shenzhen, 518112, China

    Shuihua Lu & Zhen Huang

  10. Department of Tuberculosis, Shanghai Public Health Clinical Center, Fudan University, Shanghai, 201508, China

    Xuhui Liu

Authors
  1. Yu Chen
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Lichao Fan
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Zhong Ren
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Yanhong Yu
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Jiao Sun
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Miaoran Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Chang Liu
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Ying Zhang
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Shuihua Lu
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Xuhui Liu
    View author publications

    You can also search for this author inPubMed Google Scholar

  11. Zhen Huang
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

YC, XHL, and ZH conceived and designed the study. YC and LCF reformed the clinical trial. YC, LCF, ZR, YHY, JS, MRW, and CL contributed to the sample and data collection. WPH and QY conducted the tNGS testing. YC, LCF, XHL, and ZH performed the data analysis. ZH drafted the manuscript. YC, LCF, YZ, and SHL provided critical revision. All the authors approved the final manuscript.

Corresponding authors

Correspondence to Shuihua Lu, Xuhui Liu or Zhen Huang.

Ethics declarations

Ethics approval and consent to participate

The study protocol was reviewed and approved by the ethics committee review board of Shenyang Chest Hospital (KYXM-2022–007-02). All adult participants provided written informed consent, and parents or guardians provided consent on behalf of the minors.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

12916_2025_3996_MOESM1_ESM.docx

Supplementary Material 1: Supplementary methods, Table S1-S10, and Fig. S1-S7. Table S1-Criteria for results of phenotypic drug susceptibility testing of Mycobacterium tuberculosis complex. Table S2-Criteria for case definitions. Table S3-Baseline characteristics of children (<18) at different TB classification. Table S4-Baseline characteristics of adults at different TB classification. Table S5-tNGS Results in children with suspected TB: comparison between age groups 0–14 and 15–18. Table S6-Mtb-culture, Xpert, TB-LAMP, and tNGS diagnosis performance in children and adults. Table S7-Mtb-culture, Xpert, TB-LAMP, and tNGS diagnosis performance on confirmed cases. Table S8-Mtb-culture, Xpert, and TB-LAMP positive rate in different specimens. Table S9-The gene mutation information detected by tNGS. Table S10-Drug resistance prediction performance of tNGS by comparing with pDST (n54). Fig. S1-Reporting delays in diagnosis of TB using culture, Xpert, LAMP, and tNGS. Fig. S2-Distribution of baseline tNGS results in adults and children by Xpert grade. Fig. S3-Venn diagram of tNGS and indicated assays for children and adults post-hoc classified as confirmed TB/EPTB. Fig. S4- The Venn diagram of drug resistance events defined by the indicated assays. Fig. S5-Drug resistance profile of participants with Xpert, pDST, or tNGS suggesting resistance to any drug. Fig. S6-Evaluation of the performance of tNGS in predicting resistance of different drugs. Fig. S7-Drug-resistance prediction results of participates who only received tNGS.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Chen, Y., Fan, L., Ren, Z. et al. Sensitive diagnosis of paucibacillary tuberculosis with targeted next-generation sequencing: a molecular diagnostic study. BMC Med 23, 178 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12916-025-03996-1

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12916-025-03996-1

Keywords

BMC Medicine

ISSN: 1741-7015

Contact us