Prediction of 3-yr cadaveric graft survival based on pre-transplant variables in a large national dataset

Clin Transplant. 2003 Dec;17(6):485-97. doi: 10.1046/j.0902-0063.2003.00051.x.


Pre- and post-transplant predictive factors of graft survival for optimal and expanded criteria grafts have been studied in the past. The goal of our study was to evaluate the recent large set of United Network of Organ Sharing records (1990-1998) to generate a prediction algorithm of 3-yr graft survival based on pre-transplant variables alone. The dataset of patients with end-stage renal disease and cadaveric kidney or kidney-pancreas transplantation (1990-1998) used in the study consisted of 37,407 records. Logistic regression (LM) and a tree-based model (TBM) were used to identify predictors of 3-yr allograft survival and to generate prediction algorithm. Donor and recipient demographic characteristics (age, race, and gender) and body mass index showed non-linear, while human leukocyte antigen match showed strong linear relationships with 3-yr graft survival. Prediction of the probability of graft survival from the model, achieved a good match with the observed survival of the separate dataset, with a correlation of r = 0.998 for LM and r = 0.984 for TBM. The positive predictive value (PV) of allograft survival with LM and TBM was 76.0% and the negative PV was 63 and 53.8% for LM and TBM, respectively. Both LM and the TBM can potentially be used in clinical practice for long-term prediction of kidney allograft survival based on pre-transplant variables.

Publication types

  • Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

  • Algorithms
  • Cadaver
  • Databases, Factual / statistics & numerical data
  • Decision Trees
  • Female
  • Graft Survival*
  • Humans
  • Kidney Failure, Chronic / surgery*
  • Kidney Transplantation* / statistics & numerical data
  • Logistic Models
  • Male
  • Models, Statistical
  • Pancreas Transplantation / statistics & numerical data
  • Predictive Value of Tests
  • Probability
  • Time Factors