Microbial biomarkers of tuberculosis infection and disease in blood: systematic review and meta-analysis

This systematic review and meta-analysis of 137 studies demonstrates that microbial blood biomarkers, particularly cell-associated Mtb DNA and protein/peptide antigens, exhibit high specificity (exceeding 90%) for diagnosing tuberculosis and show promising responses to antimicrobial therapy, though the findings are tempered by a high risk of bias in the majority of included primary studies.

The Gist

Imagine your body is a bustling city, and the bacteria that causes tuberculosis (Mycobacterium tuberculosis, or Mtb) is a group of sneaky intruders trying to break in.

For a long time, doctors have tried to find these intruders by asking the city's security guards (your immune system) if they've seen anything suspicious. This is like asking, "Did you see a burglar?" The guards might say "Yes!" even if the burglar left years ago, or they might say "No!" even if a burglar is currently hiding in the basement. These are the old blood tests (like the IGRA or skin tests), and they can be a bit unreliable.

The New Idea: Finding the Burglar's Trash
Instead of asking the guards, what if we could find the intruders' actual trash? Maybe a dropped glove, a piece of DNA, or a crumpled map they left behind in the bloodstream? This is what this study is all about. It looks at microbial blood biomarkers—tiny pieces of the bacteria (like DNA, proteins, or fats) floating freely in your blood.

The researchers wanted to answer two big questions:

1. Can we find the intruders? (Diagnosis: Does the person have active TB?)
2. Did the police raid work? (Treatment Monitoring: Did the medicine kill the bacteria?)

The Great Detective Hunt (The Study)

The team acted like super-detectives. They didn't just look at one case; they gathered clues from 137 different studies conducted all over the world over the last 30+ years. They looked at four different types of "trash" the bacteria might leave behind:

• Cell-free DNA: The bacteria's genetic code floating freely in the blood (like a lost ID card).
• Cell-associated DNA: The bacteria's genetic code trapped inside your own white blood cells (like a burglar hiding inside a guard's pocket).
• Proteins/Peptides: The bacteria's building blocks or tools (like a dropped wrench).
• Lipids/Glycolipids: The bacteria's fatty outer shell or skin (like a dropped jacket).

What Did They Find?

1. The "Trash" is a Great Clue for Diagnosis
The study found that looking for this bacterial "trash" in the blood is actually a very good way to diagnose active tuberculosis.

• The Good News: If the test says "Yes, we found the trash," it is almost certainly true (about 93–97% accurate). It rarely gives a false alarm.
• The Catch: Sometimes, if the intruders are hiding very well (low bacterial load), the test might miss them (sensitivity is lower, around 44–79%). It's like a metal detector that finds big gold bars easily but might miss a tiny gold ring.
• The Winner: The test for Lipids/Glycolipids (the fatty skin) and Proteins (the tools) seemed to be the most accurate detectives, with a "success rate" (AUC) of 94–96%.

2. The "Trash" Disappears When the Raid Works
This is the most exciting part. The researchers looked at patients before and after they started taking TB medicine.

• The Metaphor: Imagine the bacteria are throwing a party in your blood. When you start taking antibiotics, it's like the police raiding the party.
• The Result: As the medicine worked, the amount of "trash" (biomarkers) in the blood dropped significantly. Specifically, the cell-associated DNA (the DNA inside the cells) showed a very clear drop. This suggests that these blood tests could be used to tell doctors, "Hey, the medicine is working! The party is over!" much faster than waiting for a patient to feel better or coughing up sputum.

The Caveats (Why We Need to Be Careful)

While the results are promising, the "detectives" (the original studies) weren't all perfect.

• The Flawed Clues: Many of the studies they reviewed were a bit messy. Some compared very sick patients to very healthy people, which makes the test look better than it might be in the real world. The authors call this a "high risk of bias."
• The Verdict: It's like finding a great map, but realizing it was drawn by someone who might have exaggerated the treasure's location. We need more high-quality, strict studies to confirm these findings before we can use them as standard tools in every clinic.

The Bottom Line

This study suggests that blood tests looking for the bacteria's own parts are a powerful new tool. They are highly specific (they don't cry wolf often) and they seem to fade away when the bacteria are killed by medicine.

If we can refine these tests and make them cheaper and easier to use, we could move away from the difficult "cough into a cup" sputum tests and toward a simple blood draw. This would make diagnosing and monitoring tuberculosis much easier, especially for children and people who can't produce sputum, potentially saving millions of lives.

In short: We found a new way to track the TB bacteria by looking for its footprints in the blood. The footprints are clear and disappear when the bacteria are gone, but we need to build better cameras to make sure we never miss a single footprint.

Technical Summary

1. Problem Statement

Tuberculosis (TB) remains the leading infectious cause of mortality globally. Current diagnostic tools rely heavily on sputum, which is difficult to obtain from children, HIV-positive individuals, and those with extrapulmonary TB. Furthermore, existing blood tests (like IGRA and TST) detect host immune responses rather than the pathogen itself, leading to issues with sensitivity (false negatives in immunocompromised patients) and specificity (persistence of memory responses after resolved infection).

While numerous studies have investigated microbial blood biomarkers (direct detection of Mycobacterium tuberculosis components in blood), there has been a lack of up-to-date, comprehensive systematic reviews covering all classes of these biomarkers. Specifically, there was no meta-analysis synthesizing data on cell-associated Mtb DNA in blood, and existing reviews on cell-free DNA and antigens were outdated.

2. Methodology

The study followed a rigorous systematic review and aggregate data meta-analysis protocol registered with PROSPERO.

• Search Strategy: A comprehensive search of MEDLINE, EMBASE, and Scopus was conducted from January 1, 1990, to October 22, 2025.
• Eligibility Criteria:
  • Population: Humans of any age with active TB, TB contacts, asymptomatic individuals, or non-TB illness.
  • Intervention/Exposure: Blood-derived samples (whole blood, plasma, serum, PBMCs, etc.) tested for direct Mtb components.
  • Biomarker Classes:
    1. Cell-free Mtb DNA (cfDNA).
    2. Cell-associated Mtb DNA (caDNA).
    3. Protein/peptide Mtb antigens.
    4. Lipid/glycolipid Mtb antigens.
  • Exclusions: Host-response assays, non-blood samples, animal studies, and studies not reporting proportions of positive tests.
• Data Synthesis:
  • Prevalence: One-proportion meta-analysis using Freeman-Tukey double arcsine transformation and multilevel random-effects models to estimate baseline positivity rates across clinical subgroups.
  • Diagnostic Accuracy: Bivariate Hierarchical Summary Receiver Operating Characteristic (HSROC) models were used to calculate pooled sensitivity, specificity, and Area Under the Curve (AUC) for active TB diagnosis.
  • Longitudinal Analysis: Risk Difference (RD) meta-analysis was performed to evaluate changes in biomarker positivity before vs. after antimicrobial therapy.
• Bias Assessment: A modified QUADAS-3 tool was used, with AI assistance (Gemini 3 pro) for initial domain judgments, followed by manual verification.

3. Key Contributions

• Comprehensive Scope: This is the most extensive meta-analysis to date, incorporating 137 eligible studies (109 for diagnostic accuracy, 13 for longitudinal monitoring).
• Novelty: It is the first meta-analysis to synthesize data specifically for cell-associated Mtb DNA in blood.
• Granular Stratification: The analysis stratified results by biomarker class, HIV status, clinical phenotype (active TB vs. non-TB vs. asymptomatic), and geographic region.
• Therapeutic Monitoring: It provides the first pooled estimates of how different microbial biomarker classes respond to antimicrobial therapy.

4. Key Results

A. Diagnostic Accuracy (Active TB vs. Non-TB)

The study evaluated 109 studies (159 investigations) involving over 28,000 samples. All biomarker classes showed high specificity but variable sensitivity.

Biomarker Class	AUC (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)	Samples / Investigations
Cell-free Mtb DNA	0.87 (0.84–0.89)	61.5% (51.0–71.0%)	93.0% (88.1–96.1%)	4,878 / 34
Cell-associated Mtb DNA	0.93 (0.90–0.95)	43.9% (29.4–59.4%)	97.1% (94.5–98.5%)	3,589 / 32
Protein/Peptide Antigens	0.94 (0.92–0.96)	78.9% (73.2–83.6%)	92.9% (90.7–94.5%)	10,260 / 61
Lipid/Glycolipid Antigens	0.96 (0.94–0.97)	68.6% (54.1–80.3%)	97.0% (94.0–98.5%)	3,287 / 22

• Observation: Specificity consistently exceeded sensitivity across all classes. Protein/peptide antigens demonstrated the highest sensitivity, while lipid/glycolipid and cell-associated DNA showed the highest specificity.
• Heterogeneity: Moderate heterogeneity was observed ($I^2$ 25–50% for most).
• Subgroup Analysis:
  • Sample Type: For protein/peptide antigens, plasma samples yielded higher sensitivity (90.8%) than serum (76.9%).
  • Geography: For cell-free DNA, sensitivity varied significantly by WHO region (highest in the Americas, lowest in the Western Pacific).

B. Response to Antimicrobial Therapy

Longitudinal data from 13 studies (14 investigations) showed that biomarker positivity decreased after treatment initiation:

• Cell-associated Mtb DNA: Pooled Risk Difference (RD) = -0.46 (95% CI -0.88 to -0.03). This was the only class with a statistically significant decline.
• Cell-free Mtb DNA: RD = -0.44 (95% CI -0.89 to 0.01).
• Protein/Peptide Antigens: RD = -0.24 (95% CI -0.75 to 0.28).
• Lipid/Glycolipid: No longitudinal studies were identified.

C. Prevalence in Asymptomatic Individuals

• Biomarker positivity was generally low in asymptomatic IGRA/TST-negative individuals (0.03–0.13).
• In asymptomatic IGRA/TST-positive individuals, positivity was higher for cell-associated DNA (0.42) and protein antigens (0.60) compared to cell-free DNA (0.07), suggesting that host memory responses (IGRA+) do not always correlate with the presence of circulating microbial material.

D. Risk of Bias

A critical finding was the high risk of bias in the primary literature:

• 98/109 studies in the diagnostic accuracy analysis were assessed as high risk (mostly due to multi-gate/case-control designs introducing spectrum bias).
• 11/13 longitudinal studies were high risk.
• Sensitivity analysis excluding high-risk studies for cell-associated DNA still yielded a high AUC (0.94), suggesting robustness despite bias, but similar analyses were not possible for other classes due to insufficient low-risk studies.

5. Significance and Implications

• Clinical Utility: Microbial blood biomarkers offer a promising alternative to sputum-based testing, particularly for patients who cannot produce sputum. The high specificity (>90%) suggests they are excellent for ruling in active TB when positive.
• Treatment Monitoring: The significant decline in cell-associated DNA and trends in other markers post-treatment support their potential use as biomarkers for therapeutic response, a capability currently lacking in host-response tests (which often remain positive despite cure).
• Limitations & Future Directions: The high prevalence of bias in primary studies (spectrum bias from case-control designs) necessitates caution in interpreting absolute sensitivity values. The authors call for high-quality, single-gate, prospective studies to validate these findings.
• Research Gaps: There is a critical lack of data on the prognostic value of these biomarkers for predicting disease progression in asymptomatic infected individuals and for predicting relapse after treatment completion.

In conclusion, while molecular and biochemical microbial blood biomarkers show comparable and promising diagnostic accuracy for TB, their widespread clinical adoption awaits validation through higher-quality, prospective studies designed to minimize bias.