Kidney Week

Abstract: PUB029

Beta-2 Microglobulin Accumulation and Dialysis-Related Amyloidosis in Patients on Maintenance Hemodialysis: Insights from an Artificial Intelligence-Driven Clinical Analysis

Session Information

• Publication Only

Category: Artificial Intelligence, Digital Health, and Data Science

• 300 Artificial Intelligence, Digital Health, and Data Science

Authors

• Bommireddi, Akshay, Harbor-UCLA Medical Center, Torrance, California, United States

Akshay Bommireddi, MD

• Elali, Ibrahim, Harbor-UCLA Medical Center, Torrance, California, United States

Ibrahim Elali, MD

• Shah, Anuja P., Harbor-UCLA Medical Center, Torrance, California, United States

Anuja P. Shah, MD

• Dukkipati, Ramanath B., Harbor-UCLA Medical Center, Torrance, California, United States

Ramanath B. Dukkipati, MD

• Dai, Tiane, Harbor-UCLA Medical Center, Torrance, California, United States

Tiane Dai, MD, PhD

• Shen, Jenny I., Harbor-UCLA Medical Center, Torrance, California, United States

Jenny I. Shen, MD, MS, FASN

• Kopple, Joel D., Harbor-UCLA Medical Center, Torrance, California, United States

Joel D. Kopple, MD, FASN

• Kalantar-Zadeh, Kamyar, Harbor-UCLA Medical Center, Torrance, California, United States

Kamyar Kalantar-Zadeh, MD, DrMed, PhD, MPH, FASN

Group or Team Name

• Harbor UCLA Division of Nephrology, Hypertension, and Transplantation

Background

Dialysis-related amyloidosis (DRA), driven by beta-2 microglobulin (B2M) accumulation, remains an under recognized complication of long-term hemodialysis. We evaluated relationships among dialysis vintage, residual renal function (RRF), serum B2M levels, and DRA-associated symptoms using an artificial intelligence (AI)-assisted clinical research workflow integrated into a nephrology continuous quality assurance program.

Methods

This cross-sectional study included 156 adults receiving maintenance hemodialysis at a U.S. academic dialysis center. Serum B2M and clinical variables were extracted from electronic medical records, while DRA-associated symptoms were assessed using a standardized questionnaire. Descriptive statistics, regression modeling, and correlation analyses, were performed using Hermedion AI clinical research platform. Statistical verification, and interpretation were independently reviewed by study investigators.

Results

Mean dialysis vintage was 6.9 ± 6.8 years and mean serum B2M was 21.8 ± 7.0 mg/L. DRA-associated symptoms were present in 28.1% of patients. Patients with preserved RRF had lower B2M levels (19.4 ± 6.3 vs 24.1 ± 6.9 mg/L; P<0.001) and lower prevalence of DRA-associated symptoms (17.5% vs 36.8%; P=0.02). Longer dialysis vintage correlated with higher B2M levels (ρ=0.32, P<0.001). Absence of RRF was the strongest independent predictor of elevated B2M levels (β=5.16; 95% CI, 1.93–8.38; P=0.002).

Conclusion

AI-assisted analytical workflows can support efficient and reproducible clinical research in nephrology with physician oversight. In this cohort, preservation of RRF was the strongest modifiable factor associated with lower B2M accumulation and reduced DRA symptoms, supporting strategies aimed at preserving kidney function in hemodialysis patients.