Abstract: TH-PO1124
External Validation of Predictive Score for Post-Transplantation Outcome in US Deceased Kidney Transplant Recipients
Session Information
- Transplantation: Clinical - Predictors of Outcomes - Biomarkers and Beyond
November 07, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Transplantation
- 1902 Transplantation: Clinical
Authors
- Yazawa, Masahiko, University of Tennessee Health Science Center, Memphis, United States
- Thomas, Fridtjof, University of Tennessee Health Science Center, Memphis, Tennessee, United States
- Cseprekál, Orsolya, Semmelweis University, Budapest, Hungary
- Kar, Suryatapa, University of Tennessee Health Science Center, Memphis, Tennessee, United States
- Abdelaal, Ahmad, University of Tennessee Health Science Center, Memphis, Tennessee, United States
- Talwar, Manish, University of Tennessee Health Science Center, Memphis, Tennessee, United States
- Balaraman, Vasanthi, University of Tennessee Health Science Center, Memphis, Tennessee, United States
- Kovesdy, Csaba P., University of Tennessee Health Science Center, Memphis, Tennessee, United States
- Eason, James D., University of Tennessee Health Science Center, Memphis, Tennessee, United States
- Molnar, Miklos Zsolt, University of Tennessee Health Science Center, Memphis, Tennessee, United States
Group or Team Name
- METEOR (MEthodist Transplant EpidemiOlogy Research) Group
Background
We previously published a prediction model (www.transplantscore.com (TS)) for allograft and patient survival, which consisted of only predictors available at the time of kidney transplant (KT). Here, we aimed to perform external validation to assess the robustness, reliability, and applicability of our model.
Methods
Five hundred eleven patients who underwent first deceased KT in our Institute between 2010 to 2017 were included. We computed the original prediction score for these patients and compared the results with the observed outcome in terms of the score’s calibration (goodness of fit) and discrimination (AUC: Area Under the Curve). We also assessed the predictive performance in terms of re-classification (NRI: Net Reclassification Improvement) when compared with a binary classifier based on the EPTS raw score.
Results
In the entire cohort, the mean age was 51.2±11.8 years old, 83% were African-American, most of the patients were on hemodialysis (81%) before KT and mean time on dialysis was 5.4 years. The TS-predicted mortality probabilities clearly separate patients as demonstrated in the Kaplan-Meier curves using all available follow-up (Figure, panel A). The AUCs based on TS for 1- and 2-year mortality (panel B) were 0.737 and 0.682, respectively. These were higher than those for the classifier based on the EPTS score (AUC of 0.649 and 0.623 for 1 and 2 year mortality, respectively) and the NRI computed to 0.302 and 0.149 for 1 and 2 year classifications in favor of TS. However, the differences in the AUCs were not statistically significant (p = 0.138 and p = 0.149 for 1 and 2 year comparisons). The Hosmer and Lemeshow goodness of fit test of TS indicated some inadequate fitting (p = 0.015 and 0.038, respectively) apparently especially an overestimation for higher-score patients.
Conclusion
TS appears to broadly correctly classify patients with respect to their 1 and 2 year mortality rate.
Figure 1