Developing a Risk Score to Predict Long-term Adverse Outcomes After Kidney Transplantation.

Abstract

With great interest, we read the recent article by Molnar and colleagues,1 developing and validating a risk score for prediction of long-term adverse outcomes (including mortality, graft failure and combined outcome of death or allograft loss) after kidney transplantation. By the Cox regression analysis, a risk score including 10 predictors (recipients’ age, cause and length of end-stage renal disease, hemoglobin, albumin, selected comorbidities, race and type of insurance as well as donors’ age, diabetes status, extended criterion donor kidney, and number of HLA mismatches), was developed. By internal validation with the concordance (c-statistic), the authors conclude that this risk score can predict relevant long-term adverse outcomes and performs better than currently used models to predict patients’ graft survival. The main strength of this study is the use of the Scientific Registry of Transplant Recipients database including a relatively large sample and consistent surgery patients. Furthermore, authors had rightly used statistical methods to develop the models and validate model discrimination for postoperative long-term adverse outcomes. However, we noted that regardless of this new risk score or other currently used risk scores for performance comparison in their study, the models only included preoperative variables, but not intraoperative and postoperative factors that significantly affect the occurrence of mortality and graft failure after kidney transplantation. Indeed, risk prediction models based on preoperative variables are often used to determine eligibility for transplant. However, available evidence indicates that preoperative variables commonly used for risk prediction model may not completely explain postoperative outcomes, especially long-term outcomes. Some dynamic intraoperative and postoperative parameters not related to preoperative health status and preexisting comorbidities are also associated with postoperative adverse outcomes. It has been shown that intraoperative hypotension, prolonged operative time, whole blood, and fresh frozen plasma transfusions, fluid volume, and albumin infusion are significantly associated with postoperative short- and long-term graft and patient outcomes.2,3 Most importantly, each additional hour of cold ischemia time can significantly increase the risk of graft failure and mortality after kidney transplantation.4 Furthermore, a number of acute rejection episodes, anemia, infectious complications, acute respiratory failure, gastrointestinal bleeding and posttransplantation glomerulonephritis in postoperative period have been shown as independent predictors of inferior graft and patient outcomes after kidney transplantation.5 Thus, we argue that no inclusion of intraoperative and postoperative risk factors affecting postoperative mortality and graft failure would have decreased sensitivity, specificity, and predictive value of this risk score developed in this study. In this study, model predictive discrimination for postoperative long-term adverse outcomes was assessed with c-statistic, which ranges from 0.5 (no discriminative value) to 1 (perfect discrimination), with a c-statistic of 0.5 indicating a discriminative value equivalent to a coin toss. Discrimination is generally considered adequate when the c-statistic exceeds 0.7 and strong when the c-statistic exceeds 0.8.6 Although the risk score developed in this study outperforms the currently used risk scores, its c-statistic indexes for mortality, allograft loss, and combined event only are 0.7, 0.63, and 0.63, respectively. These results suggest that this model only based on available data before surgery is a risk predictive score with the poor discrimination for long-term adverse outcomes after kidney transplantation. Thus, we argue that discriminative ability of this risk score would have been improved, if the mode design had included preoperative, intraoperative, and postoperative variables associated with long-term adverse outcomes after kidney transplantation.