Abstract: TH-PO740

Development of a Differential Diagnostic Model of Diabetic Kidney Disease and Non-Diabetic Kidney Disease in Patients with Type 2 Diabetes Mellitus

Session Information

Category: Diabetes

  • 502 Diabetes Mellitus and Obesity: Clinical

Authors

  • Tan, Hui Zhuan, Singapore General Hospital, Singapore, Singapore
  • Kwek, Jia Liang, Singapore General Hospital, Singapore, Singapore
  • Fook, Stephanie Chong man chung, Singapore General Hospital, Singapore, Singapore
  • Choong Meng, Chan, Singapore General Hospital, Singapore, Singapore
  • Choo Chon Jun, Jason, Singapore General Hospital, Singapore, Singapore
Background

Renal biopsy is the gold standard for distinguishing diabetic kidney disease (DKD) from non-diabetic kidney disease (NDKD) but is invasive. We aimed to identify predictive factors of DKD in diabetic patients with kidney disease in our population and develop a quantitative differential diagnostic model to guide decision for kidney biopsy.

Methods

Clinical and laboratory data from 102 patients with Type 2 Diabetes Mellitus (T2DM) who underwent kidney biopsy from 2007 to 2016 in our tertiary hospital were analyzed. Univariate analysis and multivariate logistic regression were performed to identify independent predictors of DKD and generate a differential diagnostic model. Model discrimination and calibration were evaluated using the area under the receiver operating characteristic curve (AUROC) and Hosmer-Lemeshow goodness-of-fit test respectively.

Results

The cohort included 64 males (62.7%), mean age 55.7 years (±11.0) and duration of T2DM 10.4 years (±7.2). Mean serum creatinine at time of biopsy was 121.7 μmol/L (±46.4) with mean estimated glomerular filtration rate (eGFR) (MDRD) of 61.7 ml/min/1.73m2 (±30.8). Sixty-five patients (63.7%) were diagnosed with diabetic nephropathy (DN) on kidney biopsy. In multivariate analysis, presence of diabetic retinopathy (Dr), higher Hba1c (Gh), absence of hematuria defined as urinary red blood cells <10 per high power field (Hu), and absence of positive systemic markers were revealed to be independent predictors of DN. A differential diagnostic model was constructed as follows: exp (-4.857 + 0.623 Gh – 1.208 Hu + 2.742 Dr -1.402 SM) / [1 + exp ((-4.857 + 0.623 Gh – 1.208 Hu + 2.742 Dr -1.402 SM)]. The model showed excellent discrimination (AUC=0.886; 95% CI 0.815-0.956) and calibration (Hosmer-Lemeshow p=0.242, good calibration plot of observed vs predicted probability, close to equality line).

Conclusion

The differential diagnostic model may be useful in the clinical differentiation of DKD and NDKD in patients with T2DM. Further validation of the model is required to determine its clinical utility.