Artificial Intelligence and Machine Learning for Athletes Biological Passport Analytics and Anti-Doping Intelligence: A Systematic Review

Abstract

Abstract Introduction: The Athlete Biological Passport (ABP) is an anti-doping tool that monitors athletes’ biomarkers over time in an effort to detect atypical variations associated with potential doping. However, the classical ABP analysis relies on adaptive Bayesian models that may have limitations in detecting subtle changes, microdosing and newly emerging doping practices. Artificial intelligence (AI) and machine learning (ML) can complement these methods by identifying patterns in longitudinal data. Objective: The systematic review aims at summarising evidence of the application of AI and ML in ABP-based anti-doping analytics, including types of models, biomarkers, performance, comparison with traditional approaches, validation methods and operational use. Methods: This review was conducted following PRISMA guidelines. We searched PubMed, Scopus, IEEE Xplore and SPORTDiscus for studies applying AI/ML to haematological, steroidal or endocrine biomarkers relevant to ABP analysis. Data on algorithms, biomarkers, study populations, performance measures, comparator methods, and validation strategies were abstracted. Results: The studies reported use of supervised learning, unsupervised anomaly detection, and hybrid statistical-ML approaches. Several reported enhanced or complementary detection of atypical biological profiles and microdosing patterns compared with conventional statistical methods. However, heterogeneity of methods was observed. Many studies used simulated or retrospective data sets, with no confirmed cases of doping. Generalisability was hampered by limited external validation, small samples, variation in biomarker selection, and poor reporting of model interpretability. Conclusion: AI and ML show promising potential to improve the pattern recognition in ABP-based anti-doping surveillance. However, evidence is insufficient for operational use. Future research should focus on prospective validation, confirmed doping cases, transparent and reproducible methods, explainable AI, standardised performance measures, and alignment with anti-doping regulations. Keywords: Athlete Biological Passport, Machine Learning, Anti-Doping Intelligence