Leen Alhassan, Bassel Alrabadi, Suhel F Batarseh, Yousef Alghzawi, Natalie Bandak, Aseel Badwan, Hadeel Al Kayed
AI-based models show promising potential for predicting immunotherapy response using gut microbiota data. However, limited evidence, substantial methodological heterogeneity, and restricted patient populations limit the reliability and generalizability of current estimates. Larger multicenter studies with standardized AI pipelines and external validation are needed before clinical implementation.
BACKGROUND: Immune checkpoint inhibitors (ICIs) have transformed cancer treatment, but response varies across patients. Gut microbiota may influence immunotherapy outcomes, while artificial intelligence (AI) can help characterize complex microbiome-host interactions. This systematic review and meta-analysis evaluated AI models for predicting immunotherapy outcomes using gut microbiota data.
METHODS: PubMed, Scopus, and Cochrane Library were searched through January 2025. The primary outcome was pooled area under the curve (AUC).
RESULTS: Nine studies were included in the systematic review, with eight eligible for meta-analysis. The pooled AUC was 0.85 (95% CI: 0.78-0.91), indicating strong predictive performance. Several studies identified microbial taxa, including Bacteroides and Porphyromonadaceae, associated with treatment outcomes.
CONCLUSION: AI-based models show promising potential for predicting immunotherapy response using gut microbiota data. However, limited evidence, substantial methodological heterogeneity, and restricted patient populations limit the reliability and generalizability of current estimates. Larger multicenter studies with standardized AI pipelines and external validation are needed before clinical implementation.