Abstract: TH-PO819

Transplantation as a Competing Risk in Dialysis RCTs

Session Information

Category: Dialysis

  • 607 Dialysis: Epidemiology, Outcomes, Clinical Trials - Non-Cardiovascular


  • Argyropoulos, Christos, University of New Mexico , Albuquerque, New Mexico, United States
  • Roumelioti, Maria-Eleni, University of New Mexico, Albuquerque, New Mexico, United States
  • Unruh, Mark L., University of New Mexico , Albuquerque, New Mexico, United States
  • Pankratz, V. Shane, UNM Health Sciences Center, Albuquerque, New Mexico, United States
  • Locatelli, Francesco, Azienda Ospedaliera Della Provincia di Lecco-Ospedale Alessandro Manzoni, Lecco, Italy
  • Gauly, Adelheid, Fresenius Medical Care, Bad Homburg, Germany

Transplantation is a competing risk and potential confounder in the analysis of outcomes in ESRD RCTs. We examined whether results of RCTs are affected by the method to account for these informative censoring events.


We analyzed patient level data from the NIDDK sponsored HEMO and the European MPO RCTs of high flux (HF) membranes. Together these two studies contribute 96% of the evidence for the use of HF in clinical practice. We compared conventional Cox proportional hazards (CPH) models and methods for competing risk events (cumulative incidence functions, CIF) and relevant regressions models.


In HEMO there were 194 transplantation out of 1846 patients; 170 out of 647 participants were transplanted during MPO. In unadjusted analyses of CIF[figure], HF dialysis was not associated with improved survival (p=0.20).
In adjusted CPH analyses, treatment effects differed in both studies (HR 0.82 in MPO, but 0.95 in HEMO). In analyses that accounted for the competing risks of transplant, treatment effects were similar in magnitude [table]. Furthermore, when the studies were analyzed together HF was associated with 16% reduction in the incidence of death (p=0.022).


Effect sizes, and perceived congruency of interventions in RCTs in ESRD, may depend on the methods for handling censoring due to transplantation. Our findings mirror recent reports in hemodiafiltration (HDF) RCTs [Nephrol Dial Transplant (2017) 32: ii31–ii39], raising the question whether many negative (e.g. statins) and/or discrepant results in nephrology (HF or HDF) are due to the statistical methods employed to analyze trials.

 CPHCompeting Risk Model
HEMO0.92 [0.80,1.06]0.84 [0.72,0.99]0.89 [0.64,1.22]
MPO0.82 [0.59,1.12]0.80 [0.56,1.15]0.84 [0.58,1.22]
Both0.92 [0.81,1.04]0.84 [0.73,0.98]0.96 [0.77,1.20]

Adjusted for age, gender, black race, albumin, diabetic status and country


  • Clinical Revenue Support