Background: Diabetic sensorimotor polyneuropathy (DSPN) is often asymptomatic and remains undiagnosed. The ability of clinical and anthropometric variables to identify individuals likely to have DSPN might be limited. Here, we aimed to integrate protein biomarkers for reliably predicting present DSPN.
Methods: Using the proximity extension assay, we measured 135 neurological and protein biomarkers of inflammation in blood samples of 423 individuals with recent-onset diabetes from the German Diabetes Study (GDS). DSPN was diagnosed based on the Toronto Consensus Criteria. We constructed (i) a protein-based prediction model using LASSO logistic regression, (ii) an optimised traditional risk model with age, sex, waist circumference, height and diabetes type and (iii) a model combining both. All models were bootstrapped to assess the robustness, and optimism-corrected AUCs (95% CI) were reported.
Results: DSPN was present in 16% of the study population. LASSO logistic regression selected the neurofilament light chain (NFL) and fibroblast growth factor-19 (FGF-19) as the most predictive protein biomarkers for detecting DSPN in individuals with recent-onset diabetes. The protein-based model achieved an AUC of 0.66 (0.59, 0.73), while the traditional risk model had an AUC of 0.66 (0.61, 0.74). However, combined features boosted the model performance to an AUC of 0.72 (0.67, 0.79).
Conclusion: We developed a prediction model for DSPN in recent-onset diabetes based on two protein biomarkers and five standard anthropometric, demographic and clinical variables. The model has a fair discrimination performance and might be used to inform the referral of patients for further testing.
Keywords: diabetes; diabetic neuropathy; inflammation; machine learning; nerve conduction study; neurological biomarkers; peripheral nervous system; peripheral neuropathy; quantitative sensory tests.
© 2024 The Author(s). Diabetes/Metabolism Research and Reviews published by John Wiley & Sons Ltd.