Abstract
BACKGROUND: Total cerebral small vessel disease (CSVD) burden score is an important predictor of vascular cognitive impairment (VCI). However, few predictive models of VCI in type 2 diabetes mellitus (T2DM) patients have included the total CSVD burden score, especially in the early stage of VCI. OBJECTIVE: To develop and validate a nomogram that includes the total CSVD burden score to predict mild VCI in patients with T2DM. METHODS: A total of 322 eligible participants with T2DM who were divided into mild and normal cognitive groups were enrolled in this retrospective study. Demographic data, laboratory data and imaging markers of CSVD were collected. The total CSVD burden score was calculated by combining the different CSVD markers. Step-backward multivariable logistic regression analysis with the Akaike information criterion was applied to select significant predictors and develop a best-fit predictive nomogram. The performance of the nomogram was assessed in terms of discriminative ability, calibrated ability, and clinical usefulness. RESULTS: The nomogram model consisted of five variables: age, education, hemoglobin A1c level, serum homocysteine level, and total CSVD burden score. A nomogram with these variables showed good discriminative ability (area under the receiver operating characteristic curve was 0.801 in internal verification). In addition, the Hosmer-Lemeshow test (χ(2) =9.226, P=0.417) and bootstrap-corrected calibration plot indicated that the nomogram had good calibration. The Brier score of the predictive model was 0.178. Decision curve analysis demonstrated that when the threshold probability ranged between 16% and 98%, the use of the nomogram to predict mild VCI in patients with T2DM provide a greater net benefit. CONCLUSIONS: The nomogram, composed of age, education, stroke, HbA1c level, Hcy level, and total CSVD burden score, had good predictive accuracy and may provide clinicians with a practical tool for predicting the risk of mild VCI in T2DM patients.