Abstract: FR-PO0352
Optimal Timing for Dialysis Preparation in Diabetic Nephropathy
Session Information
- Diabetic Kidney Disease: Progression, Predictive Tools, Therapeutics, and Outcomes
November 07, 2025 | Location: Exhibit Hall, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Diabetic Kidney Disease
- 702 Diabetic Kidney Disease: Clinical
Author
- Choi, Yoonwon, Soonchunhyang University Hospital Bucheon, Bucheon-si, Gyeonggi-do, Korea (the Republic of)
Background
End-stage kidney disease (ESKD) is increasing, with diabetic nephropathy as the leading cause. It has poor prognosis and progresses rapidly among chronic kidney disease(CKD), yet current guidelines lack specific recommendations for dialysis preparation timing. The 2019 KDOQI guidelines favor individualized decision-making over fixed GFR thresholds. We conducted a retrospective cohort study to develop a model for predicting appropriate dialysis preparation timing in diabetic nephropathy.
Methods
We included patients with diabetic CKD followed in division of nephrology for ≥1 year who initiated hemodialysis (HD) between January 2011 and June 2024. Patients with prior renal replacement therapy (RRT) or who discontinued HD within 3 months were excluded. Data included demographics, comorbidities, labs, echocardiography, and chest X-ray findings. A multiple linear regression (MLR) model was developed to predict estimated glomerular filtration rate 6 months prior to HD (eGFR_6M), considered the optimal time to start dialysis preparation.
Results
A total of 296 patients (186 males; mean age 62 ± 14 years) were included. The average time from first nephrology visit to HD initiation was 59.4 ± 45.4 months. Variables significant in univariate analysis were included in the MLR model. The model intercept was 17.140 (95% CI: 11.376–22.905, p < 0.001). Higher ejection fraction (EF) or blood urea nitrogen (BUN) predicted dialysis preparation at a lower eGFR, whereas higher hemoglobin predicted the need for earlier preparation. The final model was:
eGFR_6M = 17.140 − 0.074 × (EF) + 0.847 × (hemoglobin) − 0.150 × (BUN)
The adjusted R2 was 0.325 and The Durbin-Watson statistic was 2.222, indicating residual independence. Scatter and Q-Q plots confirmed assumptions of normality, homoscedasticity, and linearity.
Conclusion
This study presents a predictive model for estimating the optimal timing for dialysis preparation in patients with diabetic CKD. The model may assist nephrologists in improving the management of diabetic nephropathy through timely vascular access planning.
Table 1. Multiple linear regression of eGFR_6M in diabetic nephropathy.
| Variable | B (95% CI) | SE | p-value |
| Echocardiography_EF | -0.074(-0.124, 0.024) | 0.025 | 0.004 |
| Hemoglobin | 0.847(0.380, 1.313) | 0.237 | <0.001 |
| BUN | -0.150 (-0.180, -0.119) | 0.015 | <0.001 |
95% CI, 95% confidence interval; EF, Ejection fraction; BUN, Blood urea nitrogen