JUPITER-MARS: a reporting framework and preliminary assessment score for pathophysiology-driven clinical digital twin studies. (Preprint)

BACKGROUND Clinical digital twins are emerging as promising tools to support personalized medicine by integrating multimodal biological, clinical and computational data into virtual representations of patients or organs. However, despite increasing methodological development in digital twin modelling, standardized reporting frameworks specifically addressing pathophysiology-driven clinical digital twin studies remain limited. Existing validation approaches, such as the Verification, Validation and Uncertainty Quantification (VVUQ) framework, primarily focus on model evaluation rather than structured reporting and methodological transparency. OBJECTIVE This study aims to introduce JUPITER-MARS, a methodological framework designed to support transparent reporting and preliminary assessment of pathophysiology-driven clinical digital twin studies. Specifically, it is a credibility-oriented framework integrating reporting (JUPITER), assessment (MARS), and validation (VVUQ) to address the current methodological gap in clinical digital twin research. METHODS We developed JUPITER, a structured reporting framework aligned with conventional scientific manuscript architecture, covering seven key domains: justification, underlying pathophysiology, parameter selection, integration modelling, testing (VVUQ), expected outputs and reproducibility. JUPITER is designed not only as a reporting checklist but as a design-oriented framework to guide the development of pathophysiology-driven digital twins. In parallel, we propose MARS, a complementary domain-based preliminary assessment score intended to support structured evaluation of reporting completeness in digital twin studies. To facilitate literature screening and methodological auditing, we additionally developed a prototype computational tool implemented in Python within a Google Colab environment. The tool uses heuristic keyword-based text mining and biomedical language models to identify key methodological reporting elements in digital twin publications. The JUPITER-MARS-VVUQ tool was benchmarked with a corpus of papers (digital twin and computational models using VVUQ) extrapolated from the review by Tudor et al., 2025. Given the limited formal adoption of VVUQ in current clinical digital twin literature, the benchmarking corpus was intentionally small and was used to support an initial methodological validation rather than definitive performance estimation. RESULTS Our findings suggest that the lack of VVUQ implementation represents a systemic limitation across both mechanistic and AI-driven digital twin paradigms. This clustering represents a novel methodological taxonomy of clinical digital twins based on physiological grounding and VVUQ maturity. JUPITER provides a structured checklist to guide the design and reporting of pathophysiology-driven digital twin protocols, while MARS enables a preliminary evaluation of reporting transparency across seven domains including data description, multimodal integration, outcome reporting and model performance. The associated prototype computational tool demonstrates the feasibility of supporting structured literature screening and methodological signal detection within the emerging digital twin literature. CONCLUSIONS JUPITER-MARS offers a structured methodological framework to improve transparency, reproducibility and comparability in clinical digital twin research. By complementing existing validation approaches such as VVUQ, the framework may support both prospective digital twin development and systematic evaluation of the growing digital twin literature. Future studies may adopt JUPITER-MARS to standardize reporting practices and facilitate methodological rigor in digital twin research across medical domains. CLINICALTRIAL Not applicable.