Artificial intelligence in colorectal cancer multidisciplinary decision-making: concordance, predictive support and clinical translation

Abstract

Multidisciplinary teams (MDTs) are central to colorectal cancer management, where treatment decisions increasingly depend on the integration of tumor stage, molecular characteristics, patient fitness, and multimodal treatment strategies. However, MDT workflows are time-consuming, subject to inter-team variability, and influenced by differences in expertise, local practices, and resource availability. Artificial intelligence (AI) has emerged as a potential support tool for data integration, standardization, and risk stratification. This review examines AI in colorectal cancer multidisciplinary decision-making, focusing on AI–MDT concordance, predictive models with potential relevance to MDT discussions, and clinical translation. Reported concordance between large language models and MDT decisions varies substantially and appears to be influenced by disease context, age, performance status, case complexity, input quality, and the inclusion of clinically important variables. Predictive models may provide additional prognostic information relevant to treatment planning. Prospective evidence of AI integrated into colorectal cancer MDT workflows remains very limited. Nevertheless, the literature remains limited by predominantly retrospective designs, small or selected cohorts, heterogeneous endpoints, and unresolved issues related to transparency, reproducibility, regulation, and accountability. Current evidence is insufficient to establish improvements in MDT decision quality or patient outcomes, and AI should therefore be regarded as a supervised support tool rather than a replacement for expert multidisciplinary judgment.