Artificial Intelligence and Machine Learning-based prediction of tuberculosis treatment failure: A systematic review and meta-analysis
Künstliche Intelligenz und Machine Learning zur Vorhersage von Tuberkulose-Behandlungsversagen: systematische Übersicht und Metaanalyse
PLOS One · 17.08.2026 · 21 (8) · e0355124
DOI: 10.1371/journal.pone.0355124 ↗Auf einen Blick
34 Studien untersuchten AI/ML-Modelle zur Vorhersage von TB-Behandlungsversagen. Die gepoolte Trennschärfe war gut, die Heterogenität jedoch hoch.
Fragestellung
Wie gut sagen AI/ML-Modelle ein Tuberkulose-Behandlungsversagen voraus?
Methode
Systematische Suche in PubMed/MEDLINE und Embase von Januar 2000 bis Oktober 2025; random-effects-Metaanalyse von Studien mit AUC und 95 %-KI; Biasbewertung mit PROBAST.
Ergebnisse
34 Studien erfüllten die Kriterien. 19 Studien mit 100.790 Teilnehmenden gingen in die Metaanalyse ein. Gepoolte AUC 0,836 (95 %-KI 0,799–0,868); I² = 97,9 %.
Einschränkungen
Nur 8 Studien mit externer Validierung; nur 1 Studie mit niedrigem Gesamtrisiko für Bias; Egger-Test Hinweis auf Publikationsbias (p = 0,024).
Schlussfolgerung der Autor:innen
AI/ML-Modelle zeigen eine vielversprechende Trennschärfe, sind aber für den Routineeinsatz noch nicht reif. Methodische Limitationen und unzureichende Validierung begrenzen die Aussagekraft.
Kernaussagen
- Gepoolte AUC 0,836 (95 %-KI 0,799–0,868)
- I² = 97,9 %
- Externe Validierung nur in 8 Studien; PROBAST meist nicht niedriges Biasrisiko
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Systematische Übersicht in der Zeitschrift PLOS One, online erschienen am 17.08.2026. Verfasst von Rogers Kamulegeya und 12 weiteren Autor:innen. Beteiligte Einrichtung: Department of Immunology and Molecular Biology, Makerere University College of Health Sciences (MakCHS), Kampala, Uganda. Hauptthemen laut MEDLINE: Antitubercular Agents, Artificial Intelligence, Machine Learning, Tuberculosis. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 64 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: FIC NIH HHS, Fogarty International Center.
Abstract (Originalsprache)
Background
Tuberculosis (TB) remains a leading cause of infectious disease mortality worldwide, and treatment failure contributes to ongoing transmission, drug resistance, and poor clinical outcomes. Artificial intelligence (AI) and machine learning (ML) approaches have attracted growing interest in predicting TB treatment outcomes, but the literature is heterogeneous and lacks a comprehensive synthesis.
Methods
We systematically searched PubMed/MEDLINE and Embase (January 2000–October 2025) for studies developing or validating AI/ML models to predict TB treatment failure. Two reviewers independently screened records and extracted data on study characteristics, predictor modalities, algorithms, validation strategies, and performance metrics, i.e., area under the curve (AUC), sensitivity, specificity, and confidence intervals. Studies reporting AUC with confidence intervals or sufficient data for calculation were included in random-effects meta-analysis. Missing standard errors were estimated from sample sizes and event rates using established methods. Risk of bias was assessed using PROBAST. Subgroup analyses and meta-regression explored heterogeneity, and publication bias was assessed using funnel plots, Egger’s test, and trim-and-fill analysis. The study is registered with PROSPERO (CRD420251101443).
Results
Thirty-four studies met the inclusion criteria. Publications increased markedly from 2019 onwards (91% of studies). Tree-based methods predominated (52.9%), and multimodal models (≥3 data types) were used in 41.2% of the studies. Nineteen studies (100,790 participants) contributed to the meta-analysis. The pooled AUC was 0.836 (95% CI 0.799–0.868), with substantial heterogeneity (I² = 97.9%). In subgroup analyses, studies including HIV-positive participants showed lower discrimination (AUC 0.748) than those excluding them (0.924). Only eight studies (23.5%) performed external validation, and only one study (2.9%) was rated low risk of bias overall (PROBAST), primarily due to analytical domain deficiencies. Egger’s test suggested publication bias (p = 0.024). Major evidence gaps included underrepresentation of high-burden countries, HIV-affected populations, social determinants, pediatric TB, and extrapulmonary disease.
Conclusions
AI/ML models for predicting TB treatment failure show promising discrimination but are not yet ready for routine clinical implementation. Performance varies substantially across populations and settings, and methodological limitations, including inadequate validation, poor calibration assessment, and high risk of bias, limit confidence in current estimates. Future research should prioritize rigorous external validation, calibration assessment, and development in underrepresented populations, particularly HIV-affected and high TB burden settings.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Substanzen
Autor:innen und Institutionen
- Department of Immunology and Molecular Biology, Makerere University College of Health Sciences (MakCHS), Kampala, Uganda.
- Department of Global Health and Amsterdam Institute for Global Health and Development, Amsterdam University Medical Centers, Amsterdam, Netherlands.
- Department of Computer Science, College of Computing and Information Sciences, Makerere University (CoCIS), Kampala, Uganda.
- Masaka Regional Referral Hospital, Masaka, Uganda.
- Makerere University Lung Institute (MLI), Kampala, Uganda.
- Department of Electrical and Computer Engineering, College of Engineering, Art and Design, Makerere University (CEDAT), Kampala, Uganda.
ORCID-Profile: Willy Ssengooba, Florence Nameere Kivunike, David Patrick Kateete
Förderung
- FIC NIH HHS
- Fogarty International Center
Zitieren
APA
Kamulegeya, R., Nabatanzi, R., Semugenze, D., Mugala, F., Takuwa, M., Nasinghe, E., Musinguzi, D., Namiiro, S., Katumba, A., Ssengooba, W., Nakatumba-Nabende, J., Kivunike, F. N., & Kateete, D. P. (2026). Artificial Intelligence and Machine Learning-based prediction of tuberculosis treatment failure: A systematic review and meta-analysis. PLOS One, 21(8), e0355124. https://doi.org/10.1371/journal.pone.0355124
Vancouver
Kamulegeya R, Nabatanzi R, Semugenze D, Mugala F, Takuwa M, Nasinghe E, et al. Artificial Intelligence and Machine Learning-based prediction of tuberculosis treatment failure: A systematic review and meta-analysis. PLOS One. 2026;21(8):e0355124. doi:10.1371/journal.pone.0355124
BibTeX
@article{Kamulegeya2026,
author = {Kamulegeya, Rogers and Nabatanzi, Rose and Semugenze, Derrick and Mugala, Faridah and Takuwa, Mercy and Nasinghe, Emmanuel and Musinguzi, Denis and Namiiro, Sharon and Katumba, Andrew and Ssengooba, Willy and Nakatumba-Nabende, Joyce and Kivunike, Florence Nameere and Kateete, David Patrick},
title = {Artificial Intelligence and Machine Learning-based prediction of tuberculosis treatment failure: A systematic review and meta-analysis},
journal = {PLOS One},
year = {2026},
volume = {21},
number = {8},
pages = {e0355124},
doi = {10.1371/journal.pone.0355124}
}
EUVIMED