From AI X-Rays to Ultra-Short Drug Regimens: 2025’s Tuberculosis Breakthroughs

Tuberculosis still kills more people than almost any other single infectious agent on Earth, yet the global response to it has been moving at a pace that public health experts describe as dangerously slow. Incidence is falling by only about 1.5 to 2 percent each year, far short of the trajectory needed to meet the World Health Organization’s End TB Strategy milestones for 2030, which call for a 90 percent reduction in incidence and a 95 percent reduction in deaths compared with 2015 levels. Against this sobering backdrop, a comprehensive new review published in Frontiers in Tuberculosis synthesizes an unusually productive year of scientific progress, arguing that 2025 delivered advances across diagnostics, therapeutics, vaccines, and post-disease care that could, if implemented well, reshape the path toward elimination.

The review, authored by Patrick K. Moonan of the U.S. Centers for Disease Control and Prevention’s Division of Global HIV and Tuberculosis and Timothy H. Holtz of the George Washington University Milken Institute School of Public Health, draws on a structured PubMed search that identified 12,351 tuberculosis-related records published between January 1 and December 31, 2025. Rather than a formal systematic review, the authors describe their work as a narrative synthesis guided by scientific rigor, novelty, translational relevance, and policy impact, prioritizing randomized trials, systematic reviews, high-impact mechanistic studies, and major policy updates. Their central conclusion is both encouraging and cautionary: the science accelerated dramatically in 2025, but health-system limitations and inequitable access continue to constrain real-world impact, and without deliberate strategies to bridge the gap between discovery and delivery, breakthroughs may fail to generate population-level change.

Nowhere was the progress more visible than in diagnostics. Artificial intelligence–assisted chest radiography emerged as one of the year’s most consequential tools, improving diagnostic sensitivity and reader consistency while reducing time to triage in high-volume screening programs across India, Eastern Europe, and Southeast Asia. A systematic review of 152 studies found that AI-based methods, particularly deep learning applied to radiographic data, achieved consistently high diagnostic performance, with pooled sensitivity estimates commonly exceeding 85 percent depending on study design and reference standard. The WHO subsequently recognized computer-aided detection tools as critical for lowering the proportion of undetected disease. In Nigeria, researchers found that ultraportable chest radiography paired with AI interpretation was a cost-effective approach for active case finding, improving diagnostic yield while reducing reliance on scarce expert readers.

Equally transformative were advances in noninvasive specimen collection. Tongue-swab polymerase-chain-reaction tests demonstrated diagnostic accuracy comparable to sputum-based GeneXpert MTB/RIF assays when evaluated against the same reference standards. The multi-country SMART4TB study, which enrolled 1,168 participants with presumptive tuberculosis in the Philippines, South Africa, Nigeria, and Zambia, found that tongue-swab testing achieved remarkably high specificity of 99.6 percent and moderate sensitivity of 66.0 percent against a culture-based reference standard, with low rates of non-actionable results at most sites. These characteristics make the approach particularly valuable for individuals unable to produce sputum, including children, older adults, and people living with HIV. The authors caution, however, that moderate sensitivity in adults means these tests are best deployed within diagnostic algorithms rather than as standalone tools for treatment initiation, with confirmatory testing remaining essential for clinical decision-making.

Stool-based molecular testing also matured significantly. Optimization studies of the Simple One-Step stool method for GeneXpert Ultra showed that real-world variation in stool quantity, transport time, and ambient temperature did not meaningfully affect error rates or positivity, reinforcing suitability for decentralized, child-friendly testing. Programmatic evaluations demonstrated high feasibility and acceptability among health workers and increased bacteriologic confirmation of pediatric tuberculosis when stool testing complemented respiratory sampling. Beyond diagnosis, a multi-country study found that stool-based quantification of Mycobacterium tuberculosis correlated closely with sputum culture, provided results more rapidly, and identified individuals at elevated risk for drug resistance and persistent two-month culture positivity, though no single biomarker reliably predicted symptom resolution or mortality. The WHO responded by issuing guidance early in 2025 emphasizing concurrent testing using multiple specimen types in children, including parallel collection of stool and respiratory samples, with urine lipoarabinomannan testing recommended for children living with HIV to improve microbiologic confirmation and reduce diagnostic delay.

On the therapeutic front, 2025 cemented the shift toward shorter, all-oral regimens. For drug-susceptible tuberculosis, randomized trials established the effectiveness of a four-month regimen combining rifapentine, moxifloxacin, isoniazid, and pyrazinamide, with Study 31 demonstrating noninferiority to the traditional six-month standard. A 2025 subgroup analysis among participants with diabetes reported favorable efficacy and safety, with the rifapentine-moxifloxacin arm showing the lowest proportion of unfavorable outcomes. For rifampin-resistant and multidrug-resistant tuberculosis, a large multinational Phase 3 trial evaluated five all-oral regimens lasting six to nine months, finding that four were noninferior to existing standards and achieved favorable outcomes of approximately 80 percent with improved tolerability compared with injectable-containing regimens. These regimens deliberately combine mechanistically orthogonal agents, including ATP synthase inhibition through bedaquiline, protein-synthesis inhibition through linezolid, and DNA gyrase inhibition through fluoroquinolones, to maximize sterilization across bacterial subpopulations. Updated clinical practice guidelines from the American Thoracic Society, CDC, European Respiratory Society, and Infectious Diseases Society of America formally endorsed these shortened approaches, signaling broad consensus that will shape national policy adoption throughout the decade.

The mechanistic foundation underlying these therapeutic advances is increasingly sophisticated. Antimycobacterial drugs act on three principal domains of M. tuberculosis physiology: cell-wall biosynthesis, energy metabolism and oxidative phosphorylation, and protein or nucleic acid synthesis. Bedaquiline and next-generation diarylquinolines collapse intracellular ATP stores by binding the c-subunit of ATP synthase, while cytochrome bc1 inhibitors such as telacebec disrupt the electron transport chain. Benzothiazinones including BTZ-043 target DprE1, an essential epimerase in arabinogalactan synthesis, and oxazolidinones block peptide bond formation at the 50S ribosomal subunit. Emerging compounds reported in 2025 include sorfequiline (TBAJ-876), which in the multi-country Phase 2 NC-009 trial demonstrated greater early bactericidal activity in combination with pretomanid and linezolid than standard first-line therapy, and TBAJ-587, now in Phase 1 evaluation with preclinical data showing higher potency and improved safety margins relative to bedaquiline. A Phase 2a trial of telacebec demonstrated dose-dependent reductions in bacterial load among patients with drug-susceptible pulmonary tuberculosis, and a Phase 1b-2a trial of BTZ-043 in South Africa reported dose-proportional bactericidal activity with favorable tolerability.

Prevention also advanced meaningfully. Two randomized controlled trials evaluated fluoroquinolone-based tuberculosis preventive treatment for household contacts of patients with multidrug-resistant tuberculosis. The VQUIN trial in Vietnam randomized 2,041 household contacts to receive six months of daily levofloxacin or placebo, finding tuberculosis incidence of 0.6 percent in the levofloxacin group compared with 1.1 percent in the placebo group at 30 months. The TB-CHAMP trial in South Africa randomized 922 children exposed to multidrug-resistant tuberculosis to levofloxacin or placebo. Although neither trial reached statistical significance individually due to lower-than-expected incidence, a preplanned individual participant data meta-analysis demonstrated a 60 percent relative reduction in tuberculosis incidence among adult and pediatric contacts receiving levofloxacin, though musculoskeletal adverse events were more frequent among participants aged 10 years and older. Ongoing trials including PHOENIx, evaluating six months of delamanid, and BREACH-TB, evaluating six months of bedaquiline, are expected to further inform preventive strategies for contacts exposed to fluoroquinolone-resistant disease.

Vaccine development progressed on multiple fronts. MTBVAC, a live-attenuated M. tuberculosis vaccine, demonstrated a favorable safety profile and elicited significantly stronger Th1-restricted CD4-positive T-cell responses than BCG in a Phase 1b-2a trial, supporting advancement to Phase 3 efficacy evaluation. The M72/AS01E-4 vaccine completed enrollment of approximately 20,000 participants in its Phase 3 trial in 2025, a major milestone toward generating definitive evidence on protection against tuberculosis. A Phase 2 trial in virally suppressed adolescents and adults living with HIV showed durable antigen-specific T-cell and antibody responses, addressing a critical evidence gap. Not all results were positive: H56:IC31 failed to reduce recurrence among adults who recently completed treatment despite strong immunogenicity, and a Phase 2b evaluation of BCG revaccination did not reduce sustained interferon-gamma release assay conversion in adolescents, underscoring continued uncertainty regarding immunologic correlates of protection. Meanwhile, advances in immunometabolism revealed that tuberculosis infection induces strong upregulation of indoleamine 2,3-dioxygenase-1, shifting tryptophan metabolism toward kynurenine and generating an immunosuppressive granuloma microenvironment, with plasma kynurenine-to-tryptophan ratios emerging as a potential biomarker for infection activity and treatment response.

Perhaps most conceptually significant, 2025 saw a fundamental challenge to how tuberculosis is classified and described. Analyses demonstrated that commonly used symptom-based dichotomies, such as asymptomatic versus symptomatic or subclinical versus clinical tuberculosis, are inadequate classifiers of disease state lacking both biological and operational validity. Nearly half of individuals with bacteriologically confirmed tuberculosis in prevalence surveys do not report symptoms at screening, and a sizeable proportion of reported symptoms in both tuberculosis and non-tuberculosis populations likely arise from unrelated causes. The authors of the review argue that symptoms lie on a continuum rather than a categorical boundary and should not serve as the foundation for disease definitions, epidemiologic estimates, or treatment algorithms, calling instead for objective, reproducible indicators including bacteriologic burden, radiographic extent, and functional impairment. This shift parallels progress in characterizing post-tuberculosis lung disease, for which a 2025 Lancet Infectious Diseases Personal View proposed the field’s first research case definition anchored in lung function, respiratory symptoms, and imaging abnormalities. Despite this remarkable scientific momentum, the WHO Global TB Report documented persistent gaps in case detection, suboptimal treatment success for drug-resistant tuberculosis, and chronic underfinancing, with global funding still far below the estimated US$22 billion needed annually. As the late Dr. Paul Farmer consistently argued, tuberculosis is not only a biomedical problem but a political and moral one, and scientific advances will translate into population-level impact only if matched by sustained commitments to equity, strong health systems, and the social conditions required to deliver care.

Subject of Research: 2025 advances in tuberculosis diagnostics, treatment, prevention, and elimination strategy

Article Title: Research overview—2025: integrating scientific progress to accelerate global tuberculosis elimination

Article References: Moonan, P. K., & Holtz, T. H. (2026). Research overview—2025: integrating scientific progress to accelerate global tuberculosis elimination. Frontiers in Tuberculosis, 4, Article 1805962. https://doi.org/10.3389/ftubr.2026.1805962

Image Credits: AI Generated

DOI: 10.3389/ftubr.2026.1805962

Keywords: tuberculosis, diagnostics, drug-resistant TB, vaccines, AI chest radiography, tongue swab testing, stool-based testing, bedaquiline, host-directed therapy, post-tuberculosis lung disease, WHO End TB Strategy, global health