Re-evaluating animal models reveals heterogeneity in diabetic neuropathic pain research

A new review underscores the complexity of modelling diabetic neuropathic pain, highlighting variability in animal models and calling for more nuanced phenotyping to improve translation to human conditions.

A new review in Dove Press says the central challenge in diabetic neuropathic pain research is not whether animal models can reproduce diabetes-linked nerve damage, but how reliably they capture the painful and painless forms of the condition seen in patients. The paper argues that current models remain valuable for studying mechanisms and testing therapies, yet the field has often blurred two different questions: whether neuropathy is present, and whether it produces pain.

The distinction matters because diabetic neuropathy and painful diabetic neuropathy are not the same thing. The review notes that some patients develop burning, stabbing or electric-shock-like pain, while others experience numbness, reduced sensation or little pain at all. Earlier reviews indexed in PubMed have similarly stressed that diabetic neuropathic pain models are useful but imperfect, and that careful model choice is essential if preclinical findings are to translate to human disease.

Much of the debate centres on streptozotocin, or STZ, the most widely used chemical method for inducing diabetes in rodents. The Dove Press review says STZ models reliably produce hyperglycaemia and nerve injury, but the pain response is far from uniform: some studies report clear hypersensitivity, while others describe sensory loss or reduced pain responses. A PubMed review on diabetes-induced neuropathic pain and another overview of painful diabetic neuropathy models both reached similar conclusions, noting that dose, protocol, strain and study design can all alter the outcome.

The same problem appears in genetic models such as db/db mice and Zucker diabetic fatty rats. These animals may mirror the metabolic features of type 2 diabetes more closely than chemically induced models, but their sensory phenotype can change with age, background strain and disease stage. One study summarised in the review found that db/db mice could show early mechanical allodynia that later gave way to higher sensory thresholds, while other work reported progressive loss of sensation rather than overt hypersensitivity. That variability, the review says, should not automatically be read as failure.

Researchers have also increasingly used high-fat diet plus low-dose STZ combinations to mimic type 2 diabetes. These hybrids may better reflect insulin resistance and metabolic dysfunction, but the review says they still do not produce a single, predictable pain phenotype. Studies differ in diet composition, induction methods and timing, making direct comparison difficult. Earlier assessments of diabetic neuropathy models, including one review in ScienceDirect, likewise warned that no single system fully captures the human condition.

The paper argues that some of the inconsistency reflects experimental noise, but some may represent real biological diversity. Sex, genetic background, age and disease duration all appear to shape whether neuropathy becomes painful or remains largely painless. That has important implications for behavioural testing, because commonly used assays such as von Frey filaments and thermal withdrawal tests capture only a narrow slice of the pain experience and can miss spontaneous or ongoing pain altogether. A recent systematic review on non-evoked pain in preclinical neuropathic pain research reached a similar conclusion, urging broader behavioural assessment.

The authors say the field should move towards more transparent phenotyping rather than treating every non-hypersensitive animal as a failed model. They call for prespecified subgroup analysis, longitudinal follow-up, objective measures such as nerve conduction and structural markers, and clearer reporting of negative or mixed outcomes. The paper also points to human surrogate models of neuropathic pain and broader translational reviews as reminders that preclinical research works best when it is tied to clinically defined phenotypes rather than to a single laboratory read-out. In that sense, the review’s core message is not to abandon current models, but to use them more carefully: to separate neuropathy from pain, methodological variation from biological heterogeneity, and model utility from the mistaken idea that any one animal system can stand in for the full complexity of diabetic neuropathic pain.