Digital twins in medicine: Future horizons

Abstract

Virtual patients have advanced from proof-of-concept to pragmatic tools in precision medicine, yet their role must be recalibrated. This Perspective argues that virtual patients should function not as substitutes for human participants, but as ethically governed auxiliary systems for addressing data scarcity, high-dimensional multi-omics complexity, and clinical trial constraints. Their value lies less in statistical mimicry than in mechanistically interpretable links between interventions and physiological responses. This requires hybrid architectures that integrate biological knowledge, pathway logic, PK/PD principles, and adaptive learning. A major bottleneck remains validation: retrospective consistency is insufficient, and prospective testing against independent real-world cohorts with pre-specified endpoints and equivalence margins is needed. Virtual patients’ highest near-term utility resides in pre-trial simulation—optimizing dosing, refining stratification, and mitigating late-stage failure risks—not in supplanting phase III RCTs. Ultimately, virtual patients should work symbiotically with empirical studies, exploring counterfactuals in silico while reserving definitive validation for real-world evidence.