Construction of a Nomogram Prediction Model for Screening of Serum Markers for Lower Extremity Vasculopathy Secondary to Type 2 Diabetes Mellitus.

Objective: To screen serum markers for secondary lower extremity angiopathy (LEAD) in patients with type 2 diabetes mellitus (T2DM) and construct a nomogram prediction model accordingly.

Methods: The clinical data of 200 T2DM patients admitted to the hospital from December 2022 to October 2024 were retrospectively collected. It was also divided into modeling group (n=160) and internal validation group (n=40) in a 4:1 ratio by using the leave-out method. As the external validation group, clinical data from 100 T2DM patients who were admitted to other hospitals within the same time period were also gathered. Combined with previous reports of collecting serum marker data related to LEAD secondary to T2DM, key serum markers were screened using LASSO regression. Moreover, multifactorial analysis helped to clarify independent risk factors, and a nomogram prediction model was built and tested for accuracy.

Results: The incidence of LEAD in 200 T2DM patients in the hospital was 21.00% (42/200). A total of 14 variables were screened by LASSO regression analysis. After multifactorial analysis, it was found that disease duration, history of alcohol consumption, mean platelet volume, fasting blood glucose, fibrinogen, high-sensitivity C-reactive protein, insulin-like growth factor 1, nucleotide binding oligomerization domain like receptor protein 3 were independent risk factors for LEAD secondary to T2DM.The results of model validation showed AUCs of 0.971, 0.900, and 0.959 for the modeling cohort, internal validation cohort, and external validation cohort, respectively. The Hosmer-Lemeshow test was χ²=6.607, 7.962, and 6.585 (p>0.05). Positive net benefits were obtained by intervening with patients using a nomogram model within the high-risk threshold of 0 to 0.9.

Conclusion: In this study, eight risk factors associated with LEAD secondary to T2DM are screened by LASSO regression and multifactorial analysis, and a nomogram prediction model is constructed.