Abstract
PURPOSE: Diabetic peripheral neuropathy (DPN), a common complication of type 2 diabetes mellitus (T2DM), lacks effective diagnostic tools. This study aimed to develop a nomogram that integrates corneal nerve parameters for individualized DPN risk prediction.
METHODS: A total of 111 patients with T2DM and 110 healthy controls were enrolled. All participants underwent bilateral corneal confocal microscopy (CCM). High-quality images were selected by four blinded investigators. Corneal nerve fiber length (CNFL), corneal nerve branch density (CNBD), and corneal nerve fiber density (CNFD) were quantified using ACCMetrics and AiCCMetrics software. Diagnostic models-including single- and multi-parameter models-and a nomogram incorporating CNFL, CNBD, CNFD, and age were developed. Model performance was evaluated using receiver operating characteristic analysis with 500 bootstrap resamples, calibration curves, decision curve analysis, and clinical impact curves. Sensitivity analyses assessed robustness.
RESULTS: Patients with DPN were significantly older (P = 0.005). CNFL and CNFD were higher in the DPN- group (P < 0.05), whereas CNBD showed no group difference. Single-parameter models yielded area under the curve (AUC) values ranging from 0.495 to 0.727, whereas multivariate models demonstrated improved performance with AUCs between 0.737 and 0.782. In the nomogram, CNFL and CNFD were protective factors, whereas CNBD paradoxically increased DPN risk. The model demonstrated good discrimination, calibration, clinical utility, and robustness.
CONCLUSIONS: A nomogram combining multiple corneal nerve parameters may outperform single-parameter models, thereby representing a potential tool for DPN risk stratification in T2DM.
TRANSLATIONAL RELEVANCE: The corneal nerve-based nomogram may assist in personalized DPN risk prediction and holds potential translational value for individuals with T2DM.