Derivation and validation of a clinical prediction model for assessing the risk of lower extremity amputation in patients with type 2 diabetes

Aim This study aims to develop and validate a lower extremity amputation (LEA) risk score system in persons with type 2 diabetes. Methods A retrospective population-based cohort study was conducted among eligible 21,484 participants in the derivation set and 10,742 participants in the validation set who were enrolled in the Taiwan National Diabetes Care Management Program. The risk score system was developed following the steps proposed by the Framingham Heart Study with a Cox proportional hazards model algorithm. Discrimination ability was assessed by the receiver operating characteristic curve, and calibration was performed by Hosmer–Lemeshow test. Results A total of 504 patients developed LEA at an average follow-up of 7.4 years. The point scores were derived from 15 predictors as follows: age, gender, duration of type 2 diabetes, body mass index, HbA1c, triglyceride, eGFR, variation of fasting blood glucose, comorbidities of stroke, diabetes retinopathy, hypoglycemia and foot ulcer, anti-diabetes medication, and use of diuretics and nitrates. The c-statistics for predicting 3-, 5-, and 8-year LEA risks were 0.80 [95% confidence interval (CI) 0.76–0.83], 0.78 (0.75–0.81), and 0.76 (0.74–0.79) in the derivation set, respectively, and 0.81 (0.76–0.85), 0.77 (0.73–0.81), and 0.74 (0.71–0.77) in the validation set, respectively. Conclusions A new risk score for LEA was developed and validated in the clinical setting with good discriminatory ability. Poor glycemic control, glucose variation, comorbidities, and medication use were identified as predictive factors for LEA in patients with type 2 diabetes.