An ultrasensitive method for early detection of tuberculosis in asymptomatic high-risk individuals

Summary

Background

The dynamic spectrum of tuberculosis (TB) often results in underdiagnosis warranting the need for better diagnostics to accurately detect Mycobacterium tuberculosis (Mtb) in persons with asymptomatic, paucibacillary or extrapulmonary TB, and to identify individuals at high risk of developing TB disease early. This study aimed to evaluate a dual target-based digital droplet PCR (ddPCR) assay to detect circulating cell-free Mtb DNA in plasma of individuals at high risk of developing TB disease and in those lacking a clear diagnosis of TB (asymptomatic or clinically diagnosed TB).

Methods

Forty-six healthy household contacts (HHCs) of patients with pulmonary TB who developed TB within two years of follow-up (Progressors), and 92 HHCs who did not progress to TB (Non-progressors) were included in the study. Plasma was obtained and subjected to testing using a ddPCR assay targeting two M. tuberculosis (Mtb)-specific insertion sequences, IS6110 and IS1081. Sensitivity, specificity, and ROC curves were used to assess the diagnostic performance of the test.

Findings

IS6110 and IS1081 targets were detected in 10/11 asymptomatic TB cases giving a sensitivity of 90.9% (95% CI: 62.3–99.5), and in 9/11 unconfirmed/possible TB cases giving a sensitivity of 81.8% (95% CI: 52.3–96.8). Further, the test detected Mtb-ccfDNA in 15/19 Progressors even at six months prior to TB diagnosis (79.0% sensitivity, 95% CI: 56.7–91.5; 97.8% specificity, 95% CI: 92.4–99.6). Detection rates at 12- and 18-months prior to onset of TB were 55.0% (95% CI: 34.2–74.2) and 50.0% (95% CI: 29.9–70.1), respectively. Notably, the test showed 100% sensitivity (95% CI: 43.9–100.0) in detecting extrapulmonary TB up to six months prior to clinical diagnosis.

Interpretation

The study highlights the potential of ddPCR-based detection of Mtb ccfDNA as a valuable tool for early identification of individuals at risk of developing active TB disease and for clarifying the diagnosis of TB in diagnostically challenging cases. Further validation in a larger cohort will confirm the findings of the study and endorse the utility of the test in clinical practice.

Funding

This work was supported by the Indian Council of Medical Research (Grant number:5/8/5/45/Adhoc/2022/ECD-1). The CTRIUMPh cohort study which provided samples for this study was supported by the NIH/DBT Indo-US Vaccine Action Programme.

Research in context.

Evidence before this study

Tuberculosis (TB) remains a major global public health challenge, with approximately one-quarter of the world's population and nearly 40% of individuals in TB-endemic countries infected with Mycobacterium tuberculosis (Mtb). In spite of the large population of latently infected individuals, only 5–10% go on to progress to active TB disease. Identifying individuals at high risk of progression and targeting them for preventive therapy is one of the critical priorities for reducing TB incidence. However, existing latent TB diagnostics, including the interferon-gamma release assay (IGRA) and tuberculin skin test (TST), have low positive predictive value, resulting in high numbers needed to treat. Recent TB prevalence surveys from high-burden countries in Asia and Africa also indicate that subclinical/asymptomatic TB, defined as disease caused by viable Mtb in the absence of clinical symptoms, accounts for a substantial proportion of undiagnosed TB cases (36.1–79.7%; median, 50.4%). In India, the 2019–2021 National TB Prevalence Survey reported that 42.6% of newly diagnosed TB cases were subclinical, and would likely have remained undetected without active case finding and chest radiography. Individuals with asymptomatic TB thus constitute a persistent reservoir that would sustain ongoing transmission, underscoring the need for improved tools to detect the disease early.

To identify emerging molecular techniques that can be adapted for ultrasensitive detection of pathogen-derived biomarkers, we conducted a comprehensive literature search in Google Scholar for articles published between January 1, 2001, and May 6, 2025, using search terms like “latent TB”, “subclinical TB”, “asymptomatic TB”, “cell-free DNA”, and “blood”, and identified prior studies that had evaluated Mtb-ccfDNA detection in patients with active pulmonary TB, but found no studies assessing its utility for detecting paucibacillary or asymptomatic TB, or for identifying individuals at high risk for progression to active disease.

Added value of this study

This study evaluated the utility of Mtb-derived ccf-DNA detection employing a highly sensitive method for the early diagnosis of TB, particularly the difficult-to-diagnose forms, and for identification of individuals at high risk of developing active TB several months ahead of the onset of symptomatic disease, using serially collected plasma samples from a retrospective cohort of asymptomatic healthy household contacts of newly diagnosed patients with TB, some of whom subsequently progressed to active disease during longitudinal follow-up for two years.

Implications of all the available evidence

Our findings support the diagnostic utility of ddPCR-based detection of circulating Mtb-derived cell-free DNA in plasma of individuals at high risk for progressing to active TB several months prior to clinical diagnosis. At the time of TB diagnosis, the assay showed higher sensitivity than that of the currently employed routine diagnostic tests considered together. The method is also useful for clarifying the diagnosis of TB in diagnostically challenging cases, such as in possible and asymptomatic TB, and achieved complete detection of extrapulmonary TB. Collectively, these findings address important unmet diagnostic needs and indicate the potential of plasma-based Mtb ccfDNA detection to contribute to improved TB case detection and progress towards the WHO End TB goals.

Introduction

The “End TB strategy” of the World Health Organization envisions the reduction of tuberculosis (TB) incidence by 80% and TB deaths by 90% by 2030.¹ The key strategy for achieving this goal is the systematic screening of individuals at high risk to identify those who will progress to active TB disease, and providing appropriate intervention to those with higher susceptibility.² Contrary to the long-established notion of TB manifesting either in the asymptomatic latent form or the symptomatic active form of disease, it has become evident now that TB infection is a dynamic continuum. The spectrum includes two additional clinical states—incipient TB, where Mtb is slowly replicating but does not exhibit any detectable abnormalities; and subclinical TB, where Mtb is actively replicating and may be detected by radiological and microbiological tests, but with either mild/unrecognisable or no clinical symptoms.³ This pool of individuals with subclinical TB may remain infectious in the community for years without exhibiting symptoms, and therefore represent a critical pool for TB transmission.⁴ This serious state calls for prioritising active case finding using non-symptom based, point-of-care tests with superior performance characteristics. Despite technological advances, the major challenge in devising such a test is the low bacillary load in subclinical TB.⁵

Detection of pathogen-derived cell free DNA (cfDNA) has been gaining much attention in recent years for the diagnosis of several clinical conditions. Microbial DNA fragments, derived from pathogens or dying host cells or tissues, are released into the bloodstream and remain in circulation for a number of days.⁶ Since 70% of cfDNA in the plasma are <300 bp, they can easily cross the kidney barrier and be detected in urine. Several recent studies have harnessed the detection of cfDNA for various pathological conditions. A recent meta-analysis showed that ddPCR targeting the Mtb targets IS6110 and IS1081 gave an AUC up to 0.9 in patients with pulmonary TB.⁷

The insertion sequence IS6110 is a mobile transposable element found solely in the Mtb complex and is widely used as a target for TB diagnosis.⁸ Multiple copies of IS6110 have been identified at different genomic locations in Mtb (up to 25 per genome).⁹ IS1081 is a relatively stable element with 5–7 copies per genome.¹⁰ The combined use of the two insertion sequences has been reported to substantially improve the sensitivity of the ddPCR assay.^11,12

Effective prevention and treatment of tuberculosis necessitates early diagnosis of individuals with TB, underscoring the pivotal role of laboratories in swiftly and precisely identifying cases for appropriate management. Early detection of cases is hampered by the limitations of the currently available diagnostic tests. With a higher turnaround time, high risk of contamination and requirement of trained microbiologists, the gold standard Mtb culture test does not meet the requirements of a rapid test.¹³ Though quick and cost-effective, AFB smear does not differentiate viable from dead Mtb, and Mtb from Mtb complex.¹⁴ It also has a compromised sensitivity as it typically requires 5000–10,000 bacilli per mL of sputum for a positive result.¹⁵ The widely endorsed Gene Xpert test requires a standard curve to determine the copy number. Hence, the search for non-sputum based, host/pathogen-derived biomarker tests that are both rapid as well as sensitive continues.

The ability of ddPCR to detect very low copy numbers of the target and its potential to detect Mtb-derived ccfDNA prompted us to explore its utility in early identification of high-risk individuals who subsequently progressed to active TB (Progressors), including individuals with asymptomatic/subclinical disease, with an assumption that if the test is able to detect Mtb DNA in Progressors much earlier than the development of active infectious TB disease, the implications of the finding would have far reaching benefits for TB control and elimination.

Methods

Ethics

The parent study protocol received ethical approval from the Institutional Ethics Committees of the ICMR-National Institute for Research in Tuberculosis (ICMR-NIRT) in Chennai, India, Byramjee Jeejeebhoy Government Medical College (BJGMC) in Pune, India, and Johns Hopkins University (JHU) in Baltimore, USA (NIRT-IEC No. 2020021, BJGMC-IRB00003631, JHU-IRB3). Written informed consent was obtained from all participants to participate in the parent protocol and to have their samples stored in the Biorepository for future research.

Study cohort

The study population included a subgroup of individuals from a cohort of healthy household contacts (HHCs) of patients with pulmonary TB who were longitudinally followed up for 2 years as part of an earlier cohort study called the Cohort for Tuberculosis Research by the Indo-US Medical Partnership (CTRIUMPh) study carried out at the ICMR-NIRT and BJGMC, as well as from other assorted studies of ICMR-NIRT.¹⁶ All participants had consented to have a part of their samples stored for future research. The sex data was self-reported by study participants. All participants in our study were enrolled from two geographic regions in India, namely, Chennai in the Southern part of India, and Pune in the Western part of India. The study population belonged to the same race but different ethnicities. All HHCs underwent thorough assessment including clinical examination, chest X-ray and sputum tests at baseline, as well as at 4–6 months, 12 months and 24 months to rule out TB. TB diagnosis was confirmed by one or more of the following conventional diagnostic tests: sputum smear microscopy for acid-fast bacilli (AFB), culture on Lowenstein–Jensen (LJ) or MGIT medium, and Cartridge-Based Nucleic Acid Amplification Test (CBNAAT/Xpert MTB/Rif). All X-rays were read by two independent trained readers, and disparities if any between the two readings was settled by an adjudicator. HHCs who developed active TB during follow-up were categorised as Progressors and included in the present analysis. The time point when the HHCs developed active TB was defined as TB breakdown and they were referred for treatment initiation. Supplementary Table S1 provides the definition of the different categories of HHCs classified based on their outcome. HHCs who did not progress to active TB and remained healthy during the entire follow-up period were categorised as ‘Non-progressors’. Whole blood was collected from all the HHCs and processed immediately to separate the plasma. The plasma was stored at −80 °C until use.

Extraction of circulating cell free DNA

To determine the optimal plasma input volume for ccfDNA detection, standardisation experiments were conducted using three input volumes—300 μL, 600 μL, and 1 mL, and the yield and reproducibility were evaluated in the ddPCR assay. We observed that 1 mL of plasma consistently yielded the highest detection rate (copies/20 μL reaction; Supplementary Table S2). Based on these findings, 1 mL of plasma was selected as the standard input volume for all study samples. ccfDNA was extracted from 1 mL of plasma collected at baseline, month two/four, month six/twelve, and at the time of TB breakdown (as available) using the QIAamp MinElute ccfDNA Kit (Qiagen, Germany; Cat no: 55204) according to the manufacturer's protocol. Briefly, 1 mL of plasma was mixed with 30 μL of magnetic bead suspension, 55 μL of Proteinase K and 150 μL of bead binding buffer and incubated for 10 min at room temperature, while shaking end-over-end. The tube was then placed in a magnetic rack and the supernatant was discarded. After this, 200 μL of bead elution buffer was added to the pellet. The tube was positioned in the magnetic rack, and the supernatant was carefully transferred to a fresh tube. 300 μL of buffer ACB was added to the supernatant and mixed well. The mixture was then applied to the QIAamp UCP MinElute column and centrifuged at 6000×g for 1 min. Following this, 500 μL of buffer ACW2 was added and centrifuged again at 6000×g for 1 min ccfDNA was eluted from the column by adding 30 μL of ultraclean water. The extracted DNA was stored at −80 °C for further use.

Detection of Mtb-specific multicopy insertion sequences, IS6110 and IS1081, using ddPCR

Primers and probes targeting the Mtb insertion sequences IS6110 and IS1081 were synthesised based on a previously published study.¹² The sequence of the primers and probes used are provided in Supplementary Table S3.

The ddPCR platform (RRID:SCR_019714) includes an Automated Droplet Generator (autoDG; Cat no: 1864001), PX1 PCR Plate Sealer (Cat no: 1814000), C1000 Touch Thermal Cycler (Cat no:1851196) and the QX200 Droplet Reader (Cat no: 1864003), all manufactured by Bio-Rad (Hercules, CA, USA). The initial ddPCR reaction mixture consisted of the following components: 11 μL of 2× ddPCR Supermix for Probes (No dUTP; cat no: 1863024), 0.88 μL of 10 mM probe, 1 μL of 10 mM each of forward and reverse primers, and DNA template. A reaction pool of slightly over 20 μL (that is 22 μL) was established to guarantee the transfer of precisely 20 μL of the mixture into the DG32 automated droplet generator cartridge (Cat no: 1864108). The reaction plate was covered with a pierceable PCR plate heat seal foil (Cat no: 1814040; Bio-Rad, Hercules, CA, USA) and sealed using the PX1 PCR Plate Sealer at 180 °C for 5 s. The sealed plate was loaded onto the autoDG for droplet generation. The new plate containing the generated droplets was again heat sealed. The sealed reaction plate containing ∼40 μL reaction volume was transferred to the C1000 Touch Thermal Cycler for end point amplification.

The annealing temperature of the primers was first optimised using the CFX96 Touch Real-Time PCR Detection System (Cat no: 1845097; Bio-Rad, Hercules, CA, USA) and an annealing temperature of 56 °C was found to be optimum based on a temperature gradient experiment ranging from 54 to 60 °C. The following cycling conditions were used for IS6110 and IS1081 amplification: initial denaturation at 95 °C for 10 min, followed by 40 cycles of denaturation at 94 °C for 30 s and annealing at 56 °C for 1 min, and a final extension step at 98 °C for 10 min for signal stabilisation. A ramp rate of 2 °C per second was employed to ensure that every droplet reached the precise temperature at each stage of the cycle. The plate was incubated at 12 °C for 4 h and then at room temperature for 45 min prior to droplet reading in order to stabilise the droplets. The droplets were then read using the QX200 Droplet Reader and the number of positive droplets were analysed. The droplet reader uses the QX Manager regulatory edition software version 1.2.345 (Cat no: 12012172; Bio-Rad, Hercules, CA, USA) to measure the fluorescence signal within each droplet. The software counts the number of positive and negative droplets and calculates the number of copies per microlitre (copies/μL) based on Poisson statistics. Wells with fewer than 10,000 detected droplets were excluded from subsequent analysis. Inter-and-intra assay variability were ruled out by setting up the assay in duplicate and repeating the samples in different assays (Supplementary Figure S3). The lower and upper limit of detection were determined using extracted DNA from H37Rv, the M. tuberculosis reference genome, as positive control. Starting with an initial concentration of 300 ng/mL, the H37Rv DNA was diluted in nuclease free water to make serial 10-fold dilutions of the DNA ranging from 10¹ (30 ng) to 10⁸ (3 fg). The 10⁵ and 10⁸ dilutions of H37Rv DNA were used as positive control in every assay. A positive ddPCR result was defined as the detection of either IS6110 and/or IS1081.

Statistics

All statistical analyses and visualisations were performed using GraphPad Prism (version 10.1.1; GraphPad Software, La Jolla, CA, USA). Continuous variables are presented as mean ± standard deviation (SD) unless otherwise specified. Since ddPCR copy number data did not follow a normal distribution, non-parametric statistical methods were applied throughout to ensure appropriate and robust comparisons. Comparisons of IS6110 and IS1081 copy numbers among healthy controls, individuals with latent tuberculosis infection (LTBI), and active TB cases were performed using the Kruskal–Wallis test, followed by Dunn's post-hoc test with adjustment for multiple comparisons to identify pairwise differences. Differences between pulmonary and extrapulmonary TB cases were evaluated using the Mann–Whitney U test, which is suitable for comparisons between two independent groups with non-normally distributed data.

The diagnostic performance of the ddPCR assay was assessed by constructing Receiver Operating Characteristic (ROC) curves to estimate sensitivity and specificity across different threshold values. Diagnostic accuracy was summarised using standard biomarker evaluation metrics, including sensitivity, specificity, area under the ROC curve (AUC), and Youden index. Given the relatively small sample sizes in certain subgroups, uncertainty was quantified using bootstrap-derived 95% confidence intervals (1000 resamples) for sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Exact confidence intervals for proportions were reported where appropriate. To compare paired diagnostic performance between the IS6110 & IS1081 ddPCR assay and routine diagnostic tests, the McNemar test was applied, as it is appropriate for assessing differences in sensitivity between two correlated diagnostic methods evaluated on the same individuals. All statistical tests were two-sided, and statistical significance was defined as p <0.05.

Regarding sample size considerations, no formal sample size calculation was performed because participants were drawn from an existing, well-characterised longitudinal cohort of healthy household contacts of patients with TB followed for two years under a parent protocol. Eligibility for this analysis was strictly defined based on pre-specified outcome definitions. All samples were de-identified and randomly assigned, and laboratory personnel were blinded to group allocation. Only the statistician was unblinded during data analysis. Inclusion criteria comprised all cohort participants meeting outcome definitions during follow-up, while exclusion criteria included failure to meet these definitions or incomplete sample or data availability.

Role of funders

The funders did not play a role in study design, data collection, data analyses, interpretation or writing of report.

Results

Characteristics of the study population

The clinical and demographic profile of the study participants are provided in Table 1 and participant selection from the study cohort is shown in Supplementary Figure S1. The participants were all healthy household contacts of index TB cases who were recruited along with the index case and kept on longitudinal follow-up for a period of two years. We identified 46 HHCs who were asymptomatic at the time of enrolment but subsequently developed active TB disease. Among the 46 incident TB cases, 24 (52.2%) were categorised as confirmed incident TB, 11 (23.9%) as subclinical TB, and 11 (23.9%) as clinically diagnosed TB. Twice the number of HHCs (92) who remained healthy throughout the entire follow-up period (Non-progressors) were randomly selected and used as controls. The mean age of the 46 incident TB cases (18 males and 28 females) was 28.1 ± 1.9 years, and that of the 92 Non-progressors (31 males and 61 females) was 33.6 ± 1.5 years. Among the 46 Progressors, 36 (76.6%) were diagnosed to have for pulmonary TB (PTB) as evidenced by either positive for culture, smear or Cartridge Based Nucleic Acid Amplification Test (CBNAAT), and/or had radiological evidence of TB. Ten individuals (21.3%) were diagnosed with extrapulmonary TB (EPTB) by CBNAAT and/or had clinical/radiological evidence of TB. Five (50%) of the 10 EPTB cases had pleural effusion, two (20%) had cold abscess, two (20%) had lymphadenitis and one (10%) had genitourinary TB. One Progressor was diagnosed based on clinical/radiological evidence as having both PTB and EPTB. Supplementary Table S4 provides the detailed clinical characteristics of all the Progressors. Among the 46 Progressors, only one was diagnosed as HIV positive at baseline. All 46 incident TB cases had stored plasma samples from at least two time points (baseline and at the time of TB breakdown). In addition, 13 individuals had plasma stored at four times points, and three had samples stored at three time points (Supplementary Table S5).

Table 1.

Clinical and demographic characteristics of the study participants.

Participant Characteristics	Progressors (N = 46)	Non-progressors (N = 92)
Age in years Mean ± SEa	28.1 ± 1.9	33.6 ± 1.5
Sexb
Male (%)	18 (40.4%)	31 (33.7%)
Female (%)	28 (59.6%)	61 (66.3%)
Pulmonary TB	36 (78.3%)	–
Extrapulmonary TB	9 (19.5%)	–
Pleural effusion	4 (44.4%)	–
Cold abscess	2 (22.2%)	–
Lymphadenitis	2 (22.2%)	–
Genitourinary TB	1 (11.1%)	–
Pulmonary + Extrapulmonary TB	1 (2.2%)	–
TB diagnosis at breakdown, n (%)
Confirmed TB	24 (52.2%)
Asymptomatic/subclinical TB	11 (23.9%)
Clinically diagnosed TB	11 (23.9%)
Median time to TB breakdown (months, IQR)	13 (9–18)

^ap-value calculated using Mann Whitney test = 0.02.

^bp-value calculated using Fisher's exact test = 0.5.

Detection of circulating Mtb cell-free DNA in plasma at the time of TB breakdown and across longitudinal time points prior to TB breakdown

We determined the sensitivity of detection of ccf Mtb DNA using ddPCR in plasma samples collected at various longitudinal time points. At the time point of TB diagnosis, ddPCR could detect ccf DNA in 40/46 Progressors, giving a sensitivity of 87% (95% CI: 74.3–93.9). Individually, ddPCR detected IS6110 in 38/46 (82.6%) Progressors and IS1081 in 23/46 (50.0%) Progressors. Two out of 92 Non-progressors (2.2%) tested positive for ccf DNA (IS1081) giving a specificity of 97.8% (95% CI: 92.4–99.2). We tested longitudinally collected follow-up samples of these two individuals, but did not detect Mtb DNA in any of the subsequent visit samples. A representative image of the amplitude plot and event count of the ddPCR assay is provided in Supplementary Figure S2.

Mtb ccfDNA was detected in 15/19 Progressors up to six months prior to the development of active TB disease, yielding a sensitivity of 79% (95% CI: 56.7–91.5). Sensitivity declined with increasing time prior to TB breakdown; however, Mtb ccfDNA was still detectable in 50% (10/20) of Progressors as early as 18 months before TB diagnosis. Table 2 summarises the sensitivity of the ddPCR assay for detecting Mtb ccfDNA in Progressors at different time points preceding TB breakdown. Fig. 1 shows the mean copy number of the target detected in 1 mL of plasma in Progressors at various time points prior to TB breakdown.

Table 2.

Sensitivity and specificity of ddPCR assay in detecting Mtb ccfDNA in Progressors at various time points prior to TB breakdown.

Time point of sample collection	No of samples	IS6110 & IS1081				IS6110				IS1081
		Positives detected (N)	Sensitivity % (95% CI)	Specificity % (95% CI)	AUC (95% CI)	Positives detected (N)	Sensitivity % (95% CI)	Specificity % (95% CI)	AUC (95% CI)	Positives detected (N)	Sensitivity % (95% CI)	Specificity % (95% CI)	AUC (95% CI)
At the time of TB breakdown	46/46	40/46	87.0 (74.3–93.9)	97.8 (92.4–99.6)	0.92 (0.86–0.98)	38/46	82.6 (69.3–90.9)	100.0 (96.0–100.0)	0.91 (0.85–0.98)	23/46	50.0 (36.1–63.9)	97.8 (92.4–99.6)	0.74 (0.64–0.84)
≤6 months prior to TB breakdown	19/46	15/19	79.0 (56.7–91.5)	97.8 (92.4–99.6)	0.88 (0.77–0.99)	12/19	63.2 (41.0–80.9)	100.0 (96.0–100.0)	0.82 (0.68–0.95)	5/19	26.3 (11.8–48.8)	97.8 (92.4–99.6)	0.62 (0.46–0.77)
7–11 months prior to TB breakdown	15/46	9/15	60.0 (35.8–80.2)	97.8 (92.4–99.6)	0.79 (0.63–0.94)	8/15	53.3 (30.1–75.2)	100.0 (96.0–100.0)	0.77 (0.60–0.93)	4/15	26.7 (10.9–52.0)	97.8 (92.4–99.6)	0.62 (0.45–0.79)
12–18 months prior to TB breakdown	20/46	11/20	55.0 (34.2–74.2)	97.8 (92.4–99.6)	0.76 (0.62–0.90)	8/20	40.0 (21.9–61.3)	100.0 (96.0–100.0)	0.70 (0.55–0.85)	7/20	35.0 (18.1–56.7)	97.8 (92.4–99.6)	0.66 (0.51–0.81)
>18 months prior to TB breakdown	20/46	10/20	50.0 (29.9–70.1)	97.8 (92.4–99.6)	0.74 (0.60–0.88)	8/20	40.0 (21.9–61.3)	100.0 (96.0–100.0)	0.70 (0.55–0.85)	4/20	20.0 (8.1–41.6)	97.8 (92.4–99.6)	0.59 (0.44–0.74)

Note: Specificity estimates were calculated based on plasma samples obtained from 92 Non-progressors.

Fig. 1.

Mean copies of IS6110 and IS1081 detected in Progressors: Comparison of mean copy numbers of IS6110 and IS1081 across risk groups stratified by estimated time to disease: NP (n = 92), >18 M (n = 20), 12–18 M (n = 20), 7–11 M (n = 15), ≤6 M (n = 20), and TB (n = 46). Statistical significance was determined using Kruskal–Wallis test with Dunn's post hoc correction for multiple group comparisons. p-values were denoted as: p < 0.05 (∗), p < 0.01 (∗∗), p < 0.001 (∗∗∗), and p < 0.0001 (∗∗∗∗). Error bars represent standard deviation (SD) of the mean copies/μ. PTB: Pulmonary Tuberculosis; EPTB: Extrapulmonary Tuberculosis.

Sensitivity of ddPCR in detecting circulating cell-free Mtb DNA in extra pulmonary tuberculosis cases

Progressors diagnosed with extra pulmonary TB (EPTB) had significantly higher copies of IS6110 (8.3 ± 5.2) (p < 0.001) than those with pulmonary TB (PTB) (2.8 ± 0.9). Similarly, higher copy numbers of IS1081 (3.5 ± 2.1) were detected in patients with EPTB who progressed to active TB disease as compared to patients with pulmonary TB (1.1 ± 0.5) (p < 0.001). The sensitivity of detecting both IS6110 and IS1081 in EPTB progressors at the time of TB breakdown and up to six months prior to development of active TB was 100.0% (Supplementary Table S6).

Sensitivity of ddPCR in detecting circulating cell-free Mtb DNA in asymptomatic and clinically diagnosed TB cases

Out of the 11 subclinical TB cases, ccf DNA was detected in 10 individuals giving a sensitivity of 90.9% (95% CI: 62.3–99.5). Among the 11 clinically diagnosed TB cases who were diagnosed based only on clinical/radiological evidence, 9 tested positive for ccf DNA, giving a sensitivity of 81.8% (95% CI: 52.3–96.8). The sensitivity, specificity and AUC of the ddPCR assay in detecting subclinical and clinically diagnosed PTB and EPTB are provided in Table 3.

Table 3.

Sensitivity of ddPCR assay in detecting Mtb ccfDNA in subclinical and clinically diagnosed TB cases.

Outcome	No of samples at the time of TB breakdown	IS6110&IS1081				IS6110				IS1081
		Positives detected (N)	Sensitivity % (95% CI)	Specificity % (95% CI)	AUC (95% CI)	Positives detected (N)	Sensitivity % (95% CI)	Specificity % (95% CI)	AUC (95% CI)	Positives detected (N)	Sensitivity % (95% CI)	Specificity % (95% CI)	AUC (95% CI)
Confirmed PTB	21	18/21	85.7 (65.4–95.0)	97.8 (92.4–99.6)	0.92 (0.83–1.00)	16/21	76.2 (54.9–89.4)	100.0 (96.0–100.0)	0.88 (0.77–0.99)	10/21	47.6 (28.3–67.6)	97.8 (92.4–99.6)	0.73 (0.58–0.87)
Confirmed EPTB	3	3/3	100.0 (43.9–100.0)	98.9 (94.1–99.9)	1.00 (0.99–1.00)	3/3	100.0 (43.9–100.0)	100.0 (96.0–100.0)	1.00 (1.00–1.00)	3/3	100.0 (43.9–100.0)	98.9 (94.1–99.9)	1.00 (0.99–1.00)
Subclinical PTB	11	10/11	90.9 (62.3–99.5)	97.8 (92.4–99.6)	0.95 (0.84–1.00)	10/11	90.9 (62.3–99.5)	100.0 (96.0–100.0)	0.95 (0.85–1.00)	4/11	36.4 (15.2–64.6)	97.8 (92.4–99.6)	0.67 (0.47–0.87)
Clinically diagnosed PTB	5	3/5	60.0 (23.1–92.9)	98.9 (94.1–99.9)	0.79 (0.52–1.00)	3/5	60.0 (23.1–92.9)	100.0 (96.0–100.0)	0.80 (0.53–1.00)	3/5	60.0 (23.1–92.9)	97.8 (92.4–99.6)	0.79 (0.52–1.00)
Clinically diagnosed EPTB	6	6/6	100.0 (61.0–100.0)	97.8 (92.4–99.6)	0.99 (0.97–1.00)	6/6	100.0 (61.0–100.0)	100.0 (96.0–100.0)	1.00 (1.00–1.00)	3/6	50.0 (18.8–81.2)	97.8 (92.4–99.6)	0.73 (0.48–0.99)

Diagnostic performance of Mtb ccfDNA detection by IS6110–IS1081 ddPCR versus the routine TB diagnostic tests at the time of time onset among the progressors

We compared the sensitivity of routine tuberculosis diagnostic tests with the concurrent detection of IS6110 and IS1081 by ddPCR using samples collected at the time of TB breakdown. Detection of Mtb ccfDNA by ddPCR showed a diagnostic sensitivity of 87.0%, which was higher than the combined sensitivity of all routine diagnostic tests (76.1%). Table 4 summarises the sensitivity of the individual routine diagnostic tests and IS6110–IS1081 ddPCR at the time of TB breakdown in this cohort.

Table 4.

Sensitivity of ddPCR assay and routine diagnostic tests in detecting TB cases at the time of TB onset.

Method	No. of positives detected (N)	Sensitivity
Routine diagnostic testsd	35/46	76.1% (61.2, 87.4%)a
Culture	13/46	28.3% (16.0, 43.5%)c
AFB smear	10/46	21.7% (11.0, 36.4%)c
NAAT	23/46	50.0% (34.9, 65.1%)b
IS6110 & IS1081 ddPCR	40/46	87.0% (73.7, 95.1%)a

ddPCR and routine diagnostic test results presented in this table were performed using samples collected at the time of TB breakdown.

Methods sharing the same superscript letter are not significantly different from each other.

^aIS6110 & IS1081 ddPCR = Routine diagnostic tests (p > 0.05).

^bNAAT is significantly different from both ddPCR and culture/AFB smear.

^cCulture = AFB smear (p > 0.05), both significantly lower than ddPCR and routine diagnostics.

^dRoutine diagnostic tests include AFB smear, culture, and NAAT combined.

Discussion

The present study demonstrates the presence of Mtb–derived circulating cell-free DNA in the blood of individuals at high risk of progression to active TB disease several months prior to development of clinical disease. Using ddPCR, we were able to detect circulating Mtb DNA in a significant proportion (79.0%) of Progressors six months prior to actual TB diagnosis, and in approximately half of at-risk individuals even up to 18 months ahead of TB breakdown. These findings clearly highlight the potential utility of this approach for early identification of individuals at high risk for developing infectious TB disease so that they can be accurately targeted for intervention to prevent disease development and transmission. Notably, in individuals who subsequently went on to develop extrapulmonary TB, Mtb ccfDNA was detected in all individuals 6 months prior to actual diagnosis. The diagnostic performance of the ddPCR test thus meets the Target Product Profile (TPP) framed by the WHO for tests that predict TB progression.¹⁷

Delayed and missed diagnosis of TB has remained one of the major barriers to achieving global TB elimination targets, particularly in high-burden settings where a substantial proportion of transmission is thought to arise from individuals with asymptomatic or minimally symptomatic disease. We therefore went on to analyse the diagnostic utility of the ddPCR test in clarifying the diagnosis of TB in various other clinically significant groups defined in Supplementary Table S1. We were able to detect Mtb-ccfDNA in the plasma of 9/11 clinically diagnosed TB cases and 10/11 asymptomatic/subclinical TB cases using the ddPCR test, demonstrating a remarkable sensitivity of 81.8% and 90.9% respectively. This clearly demonstrates the usefulness of the test in supporting accurate diagnosis of TB in instances where diagnosis cannot be established with confidence using the currently available routine diagnostic tests.

The ddPCR test also demonstrated great potential for the early detection of EPTB, which has always been challenging given the insidious clinical presentation of the disease and the difficulty in collecting an appropriate sample.¹⁸ The current available TB diagnostics are largely optimised for sputum-positive, symptomatic TB, and are poorly suited for early disease states, extrapulmonary TB, or individuals who are unable to produce sputum.^12,19 The dual-target IS6110–IS1081 ddPCR assay evaluated in this study detected all EPTB cases, giving a sensitivity of 100.0%.

Previous studies on detection of ccfDNA in the plasma of patients with TB had yielded much lower detection rates of 42.6–65.0%.^11,20 The higher sensitivity observed in our study is attributable to the meticulous optimisation of our protocol. Key steps in the optimised methodology included the use of a larger volume (1 mL) of sample for the assay, selection of magnetic bead-based method for cfDNA extraction and the use of two multicopy insertion sequences in the M. tuberculosis genome as targets for amplification. A point worth mentioning is that the samples used for the present analysis were retrospectively collected from well-characterised cohorts and stored at ideal conditions in a centralised Biorepository. Our ability to detect Mtb-ccfDNA with high sensitivity in these samples shows that appropriately collected and properly stored plasma can indeed yield reliable results by this test, even after extended periods of storage.

While the ddPCR test demonstrates remarkable diagnostic capabilities, particularly for early detection of TB, its feasibility and scalability in high-burden, resource-limited settings warrant careful consideration. The estimated cost per test is ₹1600, nearly the same as that of the other nucleic acid-based TB tests in India. However, compared to the existing screening tools, ddPCR presents a unique set of trade-offs. Chest X-ray with Computer-Aided Detection (CAD) offers a rapid and relatively low-cost screening option, particularly for pulmonary TB, and can be deployed in mobile units, but the primary limitation lies in its inability to directly detect the pathogen, differentiate between active and inactive disease, and identify extrapulmonary forms of TB with high accuracy. GeneXpert Ultra, a widely endorsed NAAT, has significantly improved turnaround time and reduced technical complexity compared to traditional culture, making it more accessible at the point of care. However, as noted in our study, its sensitivity for detecting paucibacillary cases, especially subclinical TB, is still compromised (51.1%). In contrast, the ddPCR test targeting IS6110 alone gave a sensitivity of 82.6%, while the use of IS1081 along with IS6110 further increased the sensitivity to 87.0% among all categories of patients with TB at the time of clinical diagnosis.

The unique advantages of ddPCR include its exceptionally high sensitivity for low copy number targets coupled with the use of non-sputum samples, making it an invaluable tool for specific use cases within resource-limited settings. For instance, in the confirmatory diagnosis of diagnostically challenging cases (e.g., clinically diagnosed TB, subclinical TB or EPTB) or for targeted screening of high-risk contacts, where the benefits of early and accurate detection outweigh the logistical complexities, ddPCR could play a crucial role. Its ability to detect Mtb DNA in plasma, particularly in asymptomatic individuals at risk of progression, opens up possibilities for predictive testing that current tools cannot achieve.

Limitations

This study has a few limitations. The case–control design is appropriate for proof-of-concept evaluation but does not allow estimation of population-level predictive performance. The subgroup sensitivity estimates should be interpreted with caution, as the small sample size limits statistical power to detect modest differences in sensitivity between subgroups and methods. To address this, we report bootstrap-derived and exact binomial 95% confidence intervals and present subgroup analyses as descriptive findings, acknowledging that larger studies are needed to confirm these differences. In addition, the use of retrospectively collected and stored plasma samples introduces potential pre-analytical variability, although stringent standardisation of collection, processing, and storage was applied to minimise degradation. Finally, the requirement for relatively large plasma input volumes may pose practical challenges in certain clinical settings.

In addition, the current ddPCR technology typically requires skilled personnel and a dedicated laboratory setup, which might pose a significant logistical barrier in many peripheral healthcare facilities where such infrastructure and trained human resources are scarce. To circumvent these issues and enhance the practicality of such biomarkers as widely applicable screening tools, our future efforts will focus on translating the detection of critical biomarkers to portable, field-deployable biosensor-based detection platforms, such as Fiberoptic Absorbance Biosensors (FABs). This strategic shift will significantly improve the feasibility and scalability of early TB detection in high-burden, resource-limited environments, bringing advanced diagnostic capabilities closer to the point of need.

Conclusion

In conclusion, the present study demonstrated that a dual-target IS6110–IS1081 ddPCR assay for detecting circulating Mtb-derived cell-free DNA can help identify individuals at high risk of progression to active tuberculosis several months before clinical diagnosis and can provide microbiological evidence in diagnostically challenging TB presentations. Our findings also suggest the potential of plasma-based Mtb ccfDNA detection as a complementary tool to existing diagnostics, particularly for paucibacillary conditions such as subclinical and extrapulmonary disease where sputum-based tests perform sub-optimally. With further validation in a larger, prospective, multicentric study, this diagnostic tool may contribute to earlier case detection and improved targeting for interventions, and thereby contribute to the global efforts to reduce diagnostic delays and advance progress towards the WHO End TB Strategy.

Contributors

All authors have read and approved the final version of the manuscript. EAD and LEH have accessed and verified the underlying data.

Evangeline Ann Daniel–data curation, formal analysis, methodology, investigation, visualisation, writing—original draft.

Manohar Nesakumar–data curation, formal analysis, methodology, investigation.

Hemalatha Haribabu–data curation, methodology, investigation.

Nancy Hilda–data curation, resources.

Kannan Thiruvengadam–data curation, formal analysis, software.

Umashankar Vetrivel–conceptualisation, review & editing.

Anbalagan Selvaraj–project administration, resources.

Sathyamurthi Pattabiraman–project administration, resources.

Brindha Bhanu–project administration, resources.

Amsaveni Sivaprakasam–project administration, resources.

Sudhakar Natarajan–project administration, resources.

Vandana Kulkarni–project administration, resources.

Rajesh Karyakarte–project administration, resources.

Mandar Paradkar–project administration, resources, review & editing.

Shri Vijay Bala Yogendra Shivakumar–project administration, resources.

Vidya Mave-project administration, resources, review & editing.

Padmapriyadarsini Chandrasekaran–project administration, resources.

Amita Gupta–project administration, resources, supervision, writing—review & editing.

Luke Elizabeth Hanna–conceptualisation, supervision, funding acquisition, project administration, review & editing.

Data sharing statement

All individual datapoints and raw data relevant to the study and required to reproduce the figures are available from the corresponding author upon reasonable request. Any queries about data can be directed to hannatrc@yahoo.com.

Declaration of interests

The authors declare no potential conflict of interests.