Comparison of the Cox Semi-Parametric Model and Parametric Models in Analyzing the Effective Diagnostic Factors in Kidney Transplantation Survival

  • Mohsen Askarishahi
  • Abdolamir Atapoor
  • Roya Hemayati
  • Shahrzad Shahidi
  • Sajedeh Zeynali
Keywords: Survival Analysis, parametric and semi-parametric models, Akaike information criterion (AIC), Kidney transplant.

Abstract

Introduction: Kidney transplantation is the best treatment for patients with advanced kidney diseases. The aim of this study was to determine the rate of transplanted kidney survival and compare the efficiency of Cox semi-parametric model with the parametric models in determination of survival effective factors.

Method: This is a historic cohort study including the information of 381 ESRD patients, who underwent kidney transplant surgery from December 2007 to March 2016 in Noor hospital of Isfahan, Iran. In order to identify the effective factors in transplantation survival, the parametric and semi-parametric models were fitted with data and Akaike informationcriterion was used for detecting the most efficient model. Data analysis was carried out with R software, Version 3.1.0 at thesignificance level of 0.05.

Results: According to the Kaplan-Mayer method, 1-, 3-, 5-, and 8-year survival rates of transplanted kidney were estimated as 0.987, 0.933, 0.869, and 0.839, respectively. Multi-variable analysis of all fitted models indicated that the duration of dialysis before transplantation (P ≤0.05) and the level of creatinine at the time of discharge from hospital (p≤0.05) had significant relationship with survival of transplanted kidney. Akaike values of Cox, Weibull, exponential, lognormal, and log-logistic models were calculated as 484, 484, 482, 484, and 356, respectively.

Conclusion: Based on the Akaike information criterion, the Cox semi-parametric model was selected and proposed as the superior model.

Published
2019-01-13
Section
Articles