Abstract: SA-PO176
Establishing a Hazard Predictive Model of Renal Outcomes for Diabetic Nephropathy
Session Information
- Diabetic Kidney Disease: Clinical - II
October 27, 2018 | Location: Exhibit Hall, San Diego Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Diabetic Kidney Disease
- 602 Diabetic Kidney Disease: Clinical
Authors
- Feng, Qiqi, The Third Affiliated Hospital of Sun Yat-sen University, Guangzhou, China
- Lou, Tan-qi, The Third Affiliated Hospital of Sun Yat-Sen University, Guangzhou, GUANGDONG, China
Background
The present study aims to find out the independent risk factors associated with loss of renal functions in diabetic nephropathy (DN) and to further establish a hazard predictive model to evaluate prognosis of DN patients.
Methods
We enrolled patients with biopsy-proven DN in our hospital between January 2004 to June 2014. They were followed up until reaching the renal end-points including renal replacement therapy and doubling of serum creatinine before 31 December 2014. Univariate and multivariate Cox regression models were used to determine the independent variables associated with prognosis. A hazard predictive model was established based on the prognostic index (PI) determined by the regression coefficients in the multivariate Cox model.
Results
The mean follow-up duration was 25.9 months. Of the 57 people enrolled in this study, 25 reached the renal outcomes. The clinical and pathological parameters associated with renal outcomes in the univariate Cox regression models were introduced into the multivariate Cox models which showed that estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN), hemoglobin and scoring of interstitial fibrosis and tubular atrophy (IFTA) were independent risk factors for reaching renal end-points. The independent risk factors were included in the PI calculations to establish the hazard predictive model of renal outcomes. Participants were randomly divided into the training set (n = 38) and the validation set (n = 19). Patients with PI≤0.807 were assigned to the low-risk group while patients with PI＞0.807 were assigned to the high-risk group based on the cut-off point of the ROC curve (AUC 0.901, 95%CI 0.800-1.000, P = 0.000) with the renal endpoints as the state variable and PIs as the test variable in the training set. The survival analysis showed that there was significant difference in the renal survival curves between the low-risk group and the high-risk group in both sets. The AUCs of ROCs in both sets were greater than the AUCs of ROCs of any single risk factors.
Conclusion
The hazard predictive model combining clinical and pathologic independent variables were able to predict prognosis of DN with more accuracy so as to help clinicians act early to patients with high risk of developing ESRD.