Mechanistic Basis for the Applicability of Animal Models in Diabetic Wound Therapy: A Comparative Analysis of Animal and Human Data

Abstract

Diabetic foot ulcers (DFUs) are a major clinical and economic burden. Animal models of diabetic wounds are key preclinical tools for developing drugs and therapies. However, their clinical predictive value has long been debated. This review uses leptin receptor-deficient diabetic (db/db) mice and streptozotocin (STZ)-induced diabetes, ischemia, pressure, and infection models as benchmarks. We compared evidence from these models with evidence from human diabetic wound tissue, primary cells, ex vivo skin, and small clinical studies. Evidence was classified as animal causal evidence (A), animal correlative evidence (B), human correlative evidence (C), and human interventional/causal evidence (D). This grading system is a custom evidence-structure framework. It is not a risk-of-bias assessment. Evidence was organized across eight mechanistic axes: advanced glycation end product (AGE)/receptor for AGE (RAGE) and oxidative stress; hypoxia-inducible factor-1α (HIF-1α)/vascular endothelial growth factor (VEGF); inflammation and immunity; cellular senescence; nerve; extracellular matrix (ECM); infection; and epigenetic and metabolic reprogramming. The comparative analysis showed that animal and human evidence was more consistently aligned in direction for AGE/RAGE, HIF-1α/VEGF, NETosis, and CXCR2-related senescence. However, human causal/interventional evidence was limited. No human wound intervention study has targeted AGE/RAGE. Most human HIF-1α/VEGF studies used nonselective interventions. Human CXCR2 evidence came mainly from small ex vivo studies. For macrophage polarization (with conflicting results related to healing stage), nerve repair, reduced fibroblast responsiveness to multiple growth factors, ECM remodeling, infection susceptibility, and epigenetic/metabolic reprogramming, the overall direction of evidence was consistent. However, the strength of evidence varied. Human interventional validation is still lacking for epigenetic/metabolic reprogramming. Therefore, animal models can be used to validate targets and mechanisms of action. However, caution is needed when predicting effect size and clinical endpoints. Model applicability must also be constrained by face validity, construct validity, and predictive validity. On this basis, we propose a tiered validation pathway. This is a narrative mechanistic comparative review. We did not perform a systematic search or quantitative synthesis.