Kidney Week

Abstract: SA-PO369

Dynamic Prediction of Cardiovascular Events in Incident Patients on Peritoneal Dialysis with Multivariate Joint Models Adjusting for Phosphate, Albumin, and Calcium Trajectories and Competing Risks

Session Information

• Hypertension, CVD, and the Kidneys: Clinical Research
  October 26, 2024 | Location: Exhibit Hall, Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Hypertension and CVD

• 1602 Hypertension and CVD: Clinical

Authors

• Damgov, Ivan, Center for Pediatric and Adolescent Medicine, University of Heidelberg, Heidelberg, Germany

Ivan Damgov, MSc

• Kieser, Meinhard, Institute of Medical Biometry, University of Heidelberg, Heidelberg, Germany

Meinhard Kieser, MSc, PhD

• Davies, Simon J., Faculty of Medicine and Health Sciences, Keele University, Stoke-on-Trent, United Kingdom

Simon J. Davies, MD, MA, MBBS

• Wong, Muh Geot, Department of Renal Medicine, Royal North Shore Hospital, St. Leonards, New South Wales, Australia

Muh Geot Wong, MBBS, PhD

• Pollock, Carol A., Kolling Institute, Sydney Medical School, University of Sydney, Royal North Shore Hospital, St Leonards, New South Wales, Australia

Carol A. Pollock, MD, MBBS, PhD

• Johnson, David W., Australasian Kidney Trials Network, University of Queensland, Brisbane, Queensland, Australia

David W. Johnson, DrMed, MBBS, PhD, FASN

• Schmitt, Claus Peter, Center for Pediatric and Adolescent Medicine, University of Heidelberg, Heidelberg, Germany

Claus Peter Schmitt, MD, PhD

Background

Joint models (JM) offer dynamic personalized risk predictions by exploring the link between longitudinal biomarkers and clinical outcomes. We investigated serum phosphate, albumin and calcium trajectories to predict cardiovascular events in peritoneal dialysis (PD) patients for the first time using innovative multivariate JM.

Methods

We evaluated the association of phosphate, albumin and calcium with composite cardiovascular events (CVE) in incident PD patients followed over 8 years in the Initiating Dialysis Early And Late trial. We adjusted time trajectories for non-linearities and considered non-cardiovascular death causes and transfer to hemodialysis as competing events. The complete dataset (N=318 patients) was used for model selection, while dynamic predictive performance was compared using a 4-fold and 10-repeat cross-validation. Individual patient CVE predictions were obtained for forecast horizons until 5 years on PD at cut-offs 1, 1.5 and 2 years utilizing biomarkers trajectory and 4 baseline risk factors.

Results

A median of 10 records per patient and a 28% CVE rate ensured convergence of all 8 JM models investigated. Multivariate JM demonstrated strong positive association of phosphate and strong inverse association of albumin with composite CVE, while calcium adjusted for albumin did not provide additional explanatory nor prognostic value.
All JM demonstrated excellent to outstanding predictive performance in both short- and long-term CVE forecasts, regardless of competing risks adjustment. Multivariate JM with phosphate and albumin outperformed univariate JM (5-year forecast area under the curve (AUC) median values at 1, 1.5 and 2 years: 0.85, 0.84 and 0.80). Prediction performance of all multivariate JM surpassed the classical Cox model with baseline parameters only ignoring biomarker trajectories (median AUC = 0.79, 0.78, 0.77).

Conclusion

This first investigation of multivariate JM in PD patients outlines longitudinal phosphate and albumin as independent predictors of composite CVE with excellent potential in dynamic personized forecasts superior to the classical Cox approach.