评估心脏移植术后死亡率预测指数(IMPACT)和器官分配联合网络(UNOS)评分模型在中国医学科学院阜外医院(以下简称阜外医院)心脏移植人群中的适用性。

回顾性分析2005年1月至2019年12月在阜外医院行心脏移植的914例受者临床资料。主要结局指标为心脏移植术后1年全因死亡率，次要结局指标为院内死亡率。根据IMPACT和UNOS评分模型计算每例受者风险评分及心脏移植术后1年预测死亡率。将阜外医院心脏移植人群按移植时间分为2005至2009年、2010至2014年及2015至2019年3个亚组。根据IMPACT得分为<5分、5~<10分和≥10分以及UNOS得分为0~2分、3~5分、>5分分别将受者分为低危、中危和高危组。组间正态分布计量资料采用成组t检验或单因素方差分析进行比较，非正态分布计量资料采用Mann-Whitney U检验比较，计数资料采用卡方检验进行比较。采用二元Logistic回归分析IMPACT和UNOS评分与阜外医院心脏移植人群移植后院内及术后1年死亡风险相关性。采用受试者工作特征(ROC)曲线评估IMPACT和UNOS评分模型对阜外医院心脏移植人群及基于移植时间分类的亚组移植术后1年死亡率的预测效能，计算曲线下面积(AUC)。采用Kaplan-Meier法绘制基于IMPACT和UNOS评分模型分组的低、中和高危组阜外医院心脏移植受者生存曲线，并采用log-rank检验进行比较。P<0.05为差异有统计学意义。

阜外医院心脏移植队列中位IMPACT评分模型得分为3分(0~26分)，中位UNOS评分模型得分为3分(0~11分)。Logistic回归分析结果提示，IMPACT和UNOS评分模型得分升高与受者移植后院内死亡风险增加[(OR＝1.25，95%CI：1.18~1.34，P<0.05)和(OR＝1.77，95%CI：1.48~2.12，P<0.05)]和术后1年死亡风险增加[(OR＝1.21，95%CI：1.13~1.28，P<0.05)和(OR＝1.58，95%CI：1.35~1.84，P<0.05)]均相关。IMPACT评分模型预测术后1年死亡率为9.91%，校正系数为0.65。UNOS评分模型预测术后1年死亡率为11.20%，校正系数为0.58。IMPACT评分模型预测阜外医院心脏移植队列1年死亡ROC AUC为0.662(95%CI：0.587~0.736，P<0.05)，当截断值＝0.10，敏感度为49.2%，特异度为72.9%。UNOS评分模型预测阜外医院心脏移植队列1年死亡ROC AUC为0.661(95%CI：0.586至0.736，P<0.05)，当截断值＝0.10，敏感度为81.4%，特异度为40.0%。IMPACT评分模型预测2005至2009年、2010至2014年及2015至2019年时间段阜外医院心脏移植人群1年死亡ROC AUC均<0.7；UNOS评分模型预测2005至2009年时间段阜外医院心脏移植人群1年死亡ROC AUC为0.799，预测2010至2014年及2015至2019年时间段的ROC AUC均<0.7。基于IMPACT评分模型分组的阜外医院低、中和高危组心脏移植受者院内死亡率分别为2.9%、5.3%和34.3%，术后1年死亡率分别为4.6%、7.5%和34.3%，差异均有统计学意义(χ²＝73.2和49.1，P均<0.05)；预测术后1年死亡率分别为7.8%、12.4%和32.1%。基于UNOS评分模型分组的低、中和高危组心脏移植受者院内死亡率分别为1.7%、4.5%和28.0%，术后1年死亡率分别为3.1%、6.7%和28.0%，差异均有统计学意义(χ²＝67.7和45.0，P均<0.05)；预测术后1年死亡率分别为7.0%、13.0%和22.0%。基于IMPACT评分模型分组的阜外医院低、中和高危组心脏移植受者术后10年生存率分别为79.0%、72.5%、52.3%，差异有统计学意义(χ²＝26.7，P<0.05)；基于UNOS评分模型分组的阜外医院低、中和高危组心脏移植受者术后10年生存率分别为82.0%、76.4%、57.3%，差异有统计学意义(χ²＝29.4，P<0.05)。

IMPACT和UNOS评分模型在一定程度上可反映中国心脏移植受者手术风险，但准确性欠佳。亟待建立一个适用于中国心脏移植受者术后生存风险预测模型。

To evaluate the applicability of the index for mortality prediction after cardiac transplantation (IMPACT) and the United Organ Allocation Network (UNOS) scoring model in the heart transplant population of Fuwai Hospital of Chinese Academy of Medical Sciences (hereinafter referred to as Fuwai Hospital).

The clinical data of 914 recipients underwent heart transplant in Fuwai Hospital from January 2005 to December 2019 were retrospectively analyzed. The primary outcome was post-transplant 1-year all-cause mortality, and the secondary outcome was the in-hospital mortality. The risk score of each recipient and the 1-year predicted mortality were calculated according to the IMPACT and UNOS scoring models. Total 914 recipients were divided into three subgroups according to the time of transplantation: 2005-2009, 2010-2014 and 2015-2019. The 914 recipients were also divided into low-risk, intermediate-risk and high-risk groups based on IMPACT scores of <5, 5-<10, and ≥10 points, and UNOS scores of 0-2, 3-5, and >5 points. Normally distributed measurement data between groups were compared by group t test or one-way analysis of variance, non-normally distributed measurement data were compared by Mann-Whitney U test, and count data were compared by Chi-square test. Binary Logistic regression was used to analyze the correlation between IMPACT and UNOS score values and the in-hospital and 1-year postoperative mortality. The receiver operating characteristic (ROC) curve was used to evaluate the predictive performance of IMPACT and UNOS scoring models on the 1-year mortality rate of 914 recipients and the subgroups, and the area under curve (AUC) was calculated. The Kaplan-Meier method was used to draw survival curves of heart transplant recipients in the low-, intermediate- and high-risk groups, and the log-rank test was used for comparison. P<0.05 indicates that the difference is statistically significant.

The median IMPACT score of the Fuwai Hospital heart transplant cohort was 3 points (0-26 points), and the median UNOS score was 3 points (0-11 points). Logistic regression analysis results suggest that the increased IMPACT and UNOS scores are correlated with increased risk for in-hospital mortality [(OR=1.25, 95%CI: 1.18-1.34, P<0.05) and (OR=1.77, 95%CI: 1.48-2.12), P<0.05)] and 1-year mortality [(OR=1.21, 95%CI: 1.13-1.28, P<0.05) and (OR=1.58, 95%CI: 1.35-1.84, P<0.05)]. The 1-year mortality rates of 914 recipients predicted by IMPACT and UNOS score were 9.91% and 11.20% respectively, and the correction coefficients were 0.65 and 0.58 respectively. The ROC AUC of the IMPACT score for prediction of 1-year mortality was 0.662 (95% CI: 0.587-0.736, P<0.05), and the sensitivity was 49.2% and the specificity was 72.9% when the cut-off value was 0.10. The ROC AUC of the UNOS score for prediction of 1-year mortality was 0.661 (95% CI: 0.586-0.736, P<0.05), and the sensitivity was 81.4% and the specificity was 40.0% when the cut-off value was 0.10. The prediction efficacies of the IMPACT and UNOS scores were both poor, with the ROC AUC was less than 0.7 in population of different periods except for the UNOS score used in the population underwent heart transplant in the period of 2005-2009 (AUC=0.799). Based on the IMPACT score, the in-hospital mortality rates in the low-, intermediate-, and high-risk groups were 2.9%, 5.3%, and 34.3%, and the 1-year mortality rates were 4.6%, 7.5%, and 34.3%, the differences were both statistically significant (χ²=73.2 and 49.1, all P<0.05); the predicted 1-year mortality rates were 7.8%, 12.4%, and 32.1% respectively. Based on the UNOS scoring model, the in-hospital mortality rates in the low-, intermediate-, and high-risk groups were 1.7%, 4.5%, and 28.0%, and the 1-year mortality rates were 3.1%, 6.7%, and 28.0%, the differences were both statistically significant (χ²=67.7 and 45.0, all P<0.05); the predicted 1-year mortality rates were 7.0%, 13.0%, and 22.0%, respectively. Based on the IMPACT scoring model, the 10-year survival rates of the low-, intermediate-, and high-risk groups were 79.0%, 72.5%, and 52.3%, respectively, and the difference was statistically significant (χ²= 26.7, P<0.05); Based on the UNOS scoring model, the 10-year long-term survival rates of low-, intermediate- and high-risk groups were 82.0%, 76.4%, and 57.3%, respectively, and the difference was statistically significant (χ²= 29.4, P<0.05).

The IMPACT and UNOS scoring models can reflect the surgical risk of Chinese heart transplant recipients to a certain extent, but the accuracy is not good. It is urgent to establish a survival risk prediction model suitable for Chinese heart transplant recipients.