ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: SA-PO175

Impact of Targeting Bone Mineral Flow on Achievement of KDIGO Guidelines for CKD-MBD

Session Information

Category: Bone and Mineral Metabolism

  • 402 Bone and Mineral Metabolism: Clinical

Authors

  • Lederer, Eleanor D., VA North Texas Health Care System, Dallas, Texas, United States
  • Brier, Michael E., VA Robley Rex Medical Center, Louisville, Kentucky, United States
  • Gaweda, Adam E., University of Louisville School of Medicine, Louisville, Kentucky, United States
Background

KDIGO guidelines for CKD-MBD management in patients on dialysis focus on achieving calcium (Ca), phosphate (Pi), and parathyroid hormone (PTH) targets. We have developed a Quantitative Systems Pharmacology (QSP) model of CKD-MBD that estimates mineral flow out of bone and into soft tissue. We used an Artificial Intelligence method called Reinforcement Learning (RL) to discover treatment strategies directly aimed at minimization of Ca and Pi flow out of bone and into vascular smooth muscle. We hypothesized that direct optimization of Ca and Pi flux instead of biochemical proxies would result in improved bone mineral metabolism and reduction in soft tissue calcification.

Methods

We used a QSP model of CKD-MBD to simulate a virtual cohort of 80 ESRD subjects. The RL Agent representing the treatment strategy was implemented as a Deep Neural Network and trained for 10,000 episodes, each episode representing a 5 year treatment period. The RL Agent observed the Ca, Pi, PTH trajectories and adjusted the doses of Pi binder, Calcitriol, and a Calcimimetic on a monthly basis, to minimize the mineral flow out of bone and into soft tissue estimated by the model (Flux RL). All simulations were performed in Matlab / Simulink (Natick, MA). We compared the results to those achieved by a simulated physician (SP) and a RL Agent trained to achieve KDIGO guidelines for Ca, Pi, and PTH (KDIGO-RL).

Results

The Flux-RL Agent achieved 23.9% reduction in Ca flux into the soft tissue, compared to 15.1% (p=0.001) (SP) and 15.6% (p<0.001) (KDIGO-RL). Ca flux out of the bone was reduced by 42.6% (Flux-RL), compared to 28.9% (p<0.001) (SP) and 39.1% (p=0.322) (RL-KDIGO). Mean Pi level was 5.1 (Flux-RL) mg/dL, compared to 5.3 (p<0.001) (SP) and 5.2 (p=0.100) (RL-KDIGO). Mean Ca level was 9.2 mg/dL (Flux-RL), compared to 8.9 (p<0.001) (SP) and 9.0 (p=0.068) (RL-KDIGO). Mean PTH level was 226 pg/mL (Flux-RL), compared to 276 (p=0.012) (SP) and 238 (p=0.779) (RL-KDIGO). Flux-RL Agent utilized greater amounts of Calcitriol than SP or KDIGO-RL.

Conclusion

CKD-MBD treatment designed to optimize bone mineral flow significantly reduces the undesired Ca flow into soft tissue compared to targeting conventional biochemical parameters. These findings suggest that Pi control can be achieved even with higher vitamin D use and that the resultant higher Ca levels do not promote soft tissue calcification.

Funding

  • Veterans Affairs Support