Background. No reliable scoring system is available to predict organ quality and outcomes. We aimed to identify a reliable composite scoring system to predict organ quality and outcomes post-kidney transplantation (KT).
Methods. 192 deceased donor KT recipients (KTRs) were evaluated. Genome-wide gene expression (GE) profiling of pre-implantation donor kidney biopsies was performed. Clinicopathological and transcriptome data available at KT time was integrated for developing the composite score models. Short-term KT outcome endpoints were evaluated with delayed graft function (DGF) and estimate glomerular filtration rate (eGFR) at 1 month post-KT, while 6- and 12-mo graft function were used as long-term KT outcomes. Using two different patient subsets, GE biomarkers predictive of short- and long-term KT outcomes were first identified and confirmed by a non-parametric Spearman correlation test. Combing these predictive biomarkers with patients clinical variables, generalized multivariate regression models were developed on training patient data. The performance of the final multivariate prediction models was validated on independent patient subsets by evaluating overall prediction concordance with a rank-based correlation test and the C statistics.
Results. The obtained and independently validated composite predictive models for both short-term and long-term outcomes of KT are shown in Table 1.
|Month||Predictive biomarkers, variables||Model p value||Independent test p-value|
|1||Donor age, 203767_s_at, 206254_at, 218108_at||1.25e-05||0.012|
|6||Donor age, 208891_at, 208892_s_at, 216834_at, 217028_at||8.01e-06||0.032|
|12||Donor age, race, recipient age, 201141_at, 204319_s_at, 213364_s_at||2.38e-04||0.046|
Donor age, race and recipient age and several molecular expression patterns were found to be important for the prediction of graft outcomes. ROC analyses as well as PPVs/NPVs for stratifying KT outcomes measured with eGFR were evaluated at different time points. The 6-mo predictive composite model was highly predictive with AUC (area under the curve) = 0.78 (Wilcoxen rank-sum test p-value =0.003) from the ROC analysis on an independent patient set. At the optimal cutoff point PPV reached ∼90% with the corresponding NPV ∼55%. Similar results were obtained for the models at different tested times.
Conclusions Composite predictive models for organ quality and outcome in KT were established and validated using a rigorous statistical study based on independent dataset validation.
To cite this abstract in AMA style:Mas V, Xin W, Suh J, Gerhau R, King A, Brayman K, Lee J, Maluf D. Composite Score at Time of Kidney Transplantation Predicts Graft Function [abstract]. Am J Transplant. 2013; 13 (suppl 5). https://atcmeetingabstracts.com/abstract/composite-score-at-time-of-kidney-transplantation-predicts-graft-function/. Accessed July 29, 2021.
« Back to 2013 American Transplant Congress