关键词: 糖尿病，2型, 微量白蛋白尿, 糖尿病并发症, 列线图

Abstract: Background　The early onset of diabetic kidney disease （DKD） is insidious，and most patients have irreversible kidney impairment at the time of diagnosis. Early diagnosis and treatment greatly contribute to the prevention or delay the development of DKD. Hence，construction of a simple and effective personalized risk prediction model will significantly help the early diagnosis and treatment of DKD. Objective　To identify the risk factors independently associated with microalbuminuria（MAU） in type 2 diabetes mellitus （T2DM） patients，and to use them to develop a simple and effective personalized risk prediction model for MAU in T2DM. Methods　T2DM participants（n=1 311） were recruited from Department of Endocrinology，the First Affiliated Hospital of Air Force Medical University，and assigned those who were hospitalized between March 2014 and September 2015，and between October 2015 and March 2016 to a development sub-cohort（n=933），and a validation sub-cohort（n=378），for the convenience of developing and validating a predictive model for MAU. Demographics，results of laboratory and auxiliary examinations，pharmacological treatment，and prevalence of albuminuria（UACR<30 mg/g） or MAU （30 mg/g <UACR≤300 mg/g） for all cases were collected. LASSO regression was applied to screen the optimized variables by running cyclic coordinate descent. Multivariate Logistic regression analyses were applied to build a prediction nomogram incorporating the selected features. The receiver operating characteristic curve （ROC），calibration curves，and Hosmer-Lemeshow test were used to validate and evaluate the discrimination and calibration of the model，while the decision curve analysis was used to evaluate its clinical validity. Results　A multivariable model that included diabetes duration，systolic blood pressure （SBP），fasting plasma glucose（FPG），triglyceride（TG），serum creatinine（Scr），cystatin C（Cys C），and diabetic retinopathy（DR） was represented as the nomogram. The results of multivariate Logistic regression analysis showed that SBP≥140 mm Hg，FPG≥7.0 mmol/L，TG≥1.7 mmol/L，Scr>106 μmol/L，Cys C>1.09 mg/L，and DR were risk factors for MAU in T2DM patients （P<0.05）. The predictive model was constructed by drawing nomogram according to the predictors. The nomogram model demonstrated very well discrimination with the development sub-cohort AUC of 0.762〔95%CI（0.734，0.789）〕，while the internal validation AUC was 0.734〔95%CI（0.686，0.777）〕. The Hosmer-Lemeshow test showed perfect fitting degree （internal validation：P=0.377；external validation：P=0.236）. Decision curve analysis showed a risk threshold of 20% and demonstrated a clinically effective predictive model. Conclusion　The nomogram model containing seven predictors（diabetes duration，SBP，FPG，TG，Scr，Cys C，and DR）could be used to predict the risk of MAU in T2DM patients.

Key words: Diabetes mellitus, type 2, Microalbuminuria, Diabetes complications, Nomograms