Predicting Kidney Transplant Survival Using Tree-Based Modeling
ABSTRACT Predicting the outcome of kidney transplantation is clinically important and computationally challenging. The goal of this project was to develop the models predicting probability of kidney allograft survival at 1, 3, 5, 7, and 10 years. Kidney transplant data from the United States Renal Data System (January 1, 1990, to December 31, 1999, with the follow-up through December 31, 2000) were used (n = 92,844). Independent variables included recipient demographic and anthropometric data, end-stage renal disease course, comorbidity information, donor data, and transplant procedure variables. Tree-based models predicting the probability of the allograft survival were generated using roughly two-thirds of the data (training set), with the remaining one-third left aside to be used for models validation (testing set). The prediction of the probability of graft survival in the independent testing dataset achieved a good correlation with the observed survival (r = 0.94, r = 0.98, r = 0.99, r = 0.93, and r = 0.98) and relatively high areas under the receiving operator characteristic curve (0.63, 0.64, 0.71, 0.82, and 0.90) for 1-, 3-, 5-, 7-, and 10-year survival prediction, respectively. The models predicting the probability of 1-, 3-, 5-, 7-, and 10-year allograft survival have been validated on the independent dataset and demonstrated performance that may suggest implementation in clinical decision support system.
- SourceAvailable from: Ahmed Akl[Show abstract] [Hide abstract]
ABSTRACT: An artificial neural networks (ANNs) model was developed to predict 5-year graft survival of living-donor kidney transplants. Predictions from the validated ANNs were compared with Cox regression-based nomogram. Out of 1900 patients with living-donor kidney transplant; 1581 patients were used for training of the ANNs (training group), the remainder 319 patients were used for its validation (testing group). Many variables were correlated with the graft survival by univariate analysis. Significant ones were used for ANNs construction of a predictive model. The same variables were subjected to a multivariate statistics using Cox regression model; their result was the basis of a nomogram construction. The ANNs predictive model and the nomogram were used to predict the graft survival of the testing group. The predicted probability(s) was compared with the actual survival estimates. The ANNs sensitivity was 88.43% (95% confidence interval [CI] 86.4-90.3), specificity was 73.26% (95% CI 70-76.3), and predictive accuracy was 88% (95% CI 87-90) in the testing group, whereas nomogram sensitivity was 61.84% (95% CI 50-72.8) with 74.9% (95% CI 69-80.2) specificity and predictive accuracy was 72% (95% CI 67-77). The positive predictive value of graft survival was 82.1% and 43.5% for the ANNs and Cox regression-based nomogram, respectively, and the negative predictive value was 82% and 86.3% for the ANNs and Cox regression-based nomogram, respectively. Predictions by both models fitted well with the observed findings. These results suggest that ANNs was more accurate and sensitive than Cox regression-based nomogram in predicting 5-year graft survival.Transplantation 12/2008; 86(10):1401-6. DOI:10.1097/TP.0b013e31818b221f · 3.83 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: We have previously demonstrated that biomarkers of inflammation and immune activity detected within intraoperative renal transplant allograft biopsies are linked to adverse short-term post-transplantation clinical outcomes. Now we provide a post hoc analysis of our earlier data in the light of longer clinical follow-up. A total of 75 consecutively performed renal allografts were analyzed for gene expression of proinflammatory molecules, inflammation-induced adhesion molecules, and antiapoptotic genes expressed 15 minutes after vascular reperfusion to determine whether this analysis can aid in predicting long-term quality of renal function, proteinuria, graft loss, and death-censored graft. We have built predictive models for proteinuria (area under the curve = 0.859, p = 0.0001) and graft loss (area under the curve = 0.724, p = 0.027) 2 years post-transplantation using clinical variables in combination with intragraft gene expression data of tumor necrosis factor-alpha, interleukin-6, CD40, CD3, and tumor necrosis factor-alpha, Bcl-2, and interferon-gamma, respectively. This post hoc analysis demonstrates that hypothesis-driven, targeted polymerase chain reaction profiling of gene expression in the donor kidney at the time of engraftment can predict 2-year post-transplantation clinical outcomes.Human immunology 02/2010; 71(5):451-5. DOI:10.1016/j.humimm.2010.02.013 · 2.14 Impact Factor
- Nephrology Dialysis Transplantation 04/2010; 25(4):1039-47. DOI:10.1093/ndt/gfp782 · 3.58 Impact Factor