Department of Medicine, Hennepin County Medical Center, University of Minnesota, Minneapolis, MN 55415, USA.
Am J Kidney Dis. 2010 Nov;56(5):947-60. doi: 10.1053/j.ajkd.2010.06.020.
Surprisingly few tools have been developed to predict outcomes after kidney transplant.
Retrospective observational cohort study.
SETTING & PARTICIPANTS: Adult patients from US Renal Data System (USRDS) data who underwent deceased donor kidney transplant in 2000-2006.
Full and abbreviated prediction tools for graft loss using candidate predictor variables available in the USRDS registry, including data from the Organ Procurement and Transplantation Network and the Centers for Medicare & Medicaid Services End-Stage Renal Disease Program.
Graft loss within 5 years, defined as return to maintenance dialysis therapy, preemptive retransplant, or death with a functioning graft.
We used Cox proportional hazards analyses to develop separate tools for assessment (1) pretransplant, (2) at 7 days posttransplant, and (3) at 1 year posttransplant to predict subsequent risk of graft loss within 5 years of transplant. We used measures of discrimination and explained variation to determine the number of variables needed to predict outcomes at each assessment time in the full and abbreviated equations, creating simple user-friendly prediction tools.
Although we could identify 32, 29, and 18 variables that predicted graft loss assessed pretransplant and at 7 days and 1 year posttransplant ("full" models), 98% of the discriminatory ability and >80% of the variability explained by the full models could be achieved using only 11, 8, and 6 variables, respectively.
Comorbidity data were from the Centers for Medicare & Medicaid Medical Evidence Report, which may significantly underreport comorbid conditions; C statistic values may indicate only modest ability to discriminate risk for an individual patient.
This method produced risk-prediction tools that can be used easily by patients and clinicians to aid in understanding the absolute and relative risk of graft loss within 5 years of transplant.
令人惊讶的是，目前仅有少数工具可用于预测肾移植后的结果。
回顾性观察队列研究。
来自美国肾脏数据系统（USRDS）数据的成年患者，他们在 2000 年至 2006 年期间接受了已故供体的肾移植。
使用 USRDS 登记处中可用的候选预测变量，包括来自器官采购和移植网络以及医疗保险和医疗补助服务中心终末期肾病计划的数据，为移植物丢失开发完整和缩写的预测工具。
5 年内移植物丢失，定义为返回维持透析治疗、抢先再次移植或带功能移植物死亡。
我们使用 Cox 比例风险分析为评估（1）移植前、（2）移植后 7 天和（3）移植后 1 年分别开发单独的工具，以预测移植后 5 年内移植物丢失的后续风险。我们使用区分度和解释变异的测量值来确定在完整和缩写方程的每个评估时间点预测结果所需的变量数量，创建简单易用的预测工具。
尽管我们可以确定 32、29 和 18 个变量可用于评估移植前、移植后 7 天和 1 年的移植物丢失（“完整”模型），但 98%的区分能力和>80%的完整模型解释的变异性可以仅使用 11、8 和 6 个变量分别实现。
合并症数据来自医疗保险和医疗补助医疗证据报告，可能显著低估合并症情况；C 统计值可能仅表示对个体患者风险区分的适度能力。
这种方法产生的风险预测工具可以由患者和临床医生轻松使用，以帮助理解移植后 5 年内移植物丢失的绝对和相对风险。
