Translational assessment instruments for preclinical to first-in-human decision-making: a scoping review

Preclinical evidence used to support first-in-human (FIH) decisions is often assembled through isolated assessment tools that improve individual parts of translational practice but do not preserve assumptions, uncertainty, and decision intent across discovery-to-clinic handoffs. We therefore reviewed structured translational assessment instruments designed to inform preclinical-to-human decision-making and examined whether the published landscape already contains the components needed for decision-grade evidence integration.

We conducted a scoping review informed by PRISMA-ScR principles to identify peer-reviewed publications describing structured assessment instruments intended to support preclinical-to-clinical translation decisions. Searches were performed in PubMed and complemented by citation snowballing. Nineteen distinct instruments were identified and mapped across the preclinical-to-FIH pathway. The landscape showed clear strengths in model-selection support, reporting quality, risk-of-bias control, and mechanistic structuring of translational claims. However, recurrent limitations were also evident. Few instruments explicitly benchmark performance against downstream human outcomes. Uncertainty associated with evidence generated at one stage is often not explicitly propagated to later stages of the research and decision-making process. Outputs are rarely designed as reusable governance artefacts with explicit context of use, decision owner, breach conditions, and update triggers. The ecosystem is therefore stronger in decision-grade evidence production than in decision-grade evidence integration. Based on this synthesis, we derive a conceptual interoperability framework in which evidence-facing outputs remain source-traceable and uncertainty-aware, while decision-facing outputs are released only within defined governance gates. Within such a framework, AI is most defensible as a constrained support layer for retrieval-grounded extraction, normalization, and interoperability, rather than as an autonomous decision-maker.

Current translational practice is limited less by the absence of assessment instruments than by weak interoperability between them. A staged, uncertainty-aware workflow that preserves assumptions and decision logic across handoffs could improve the auditability and reuse of preclinical evidence for FIH governance. Whether such workflows improve predictive performance or downstream clinical decision quality remains to be tested prospectively.