Lehr Carli J, Dalton Jarrod E, Dewey Elizabeth N, Gunsalus Paul R, Rose Johnie, Valapour Maryam
Department of Pulmonary Medicine, Cleveland Clinic, Cleveland, Ohio.
Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, Ohio.
JHLT Open. 2024 Jul 1;5:100122. doi: 10.1016/j.jhlto.2024.100122. eCollection 2024 Aug.
Predicting post-transplant (PT) survival in lung allocation remains an elusive goal. We analyzed the impact of donor factors on PT survival and how these relationships vary among transplant recipients.
We studied primary bilateral lung transplant recipients ( = 7,609) from the US Scientific Registry of Transplant Recipients (19 February 2015-1 February 2020). Main and interaction effects were evaluated and adjusted across candidate age, sex, and diagnosis. Models predicting PT survival were compared to the PT Composite Allocation Score model (PT-CAS): (1) Cox regression donor multivariable model (COX), (2) COX + PT-CAS, (3) random forest model (RF), and (4) RF + PT-CAS. Model discrimination and calibration measures were compared.
Interactions between donor and recipient factors emerged by age: lower survival for donation after circulatory death organs for recipients aged 55 to 69 years, donor smoking for recipients aged 30 to 54 and 70+, Hispanic donor for recipients <30, non-Hispanic Black donor for recipients aged 30+; sex: cytomegalovirus mismatch for males; diagnosis: higher donor recipient weight ratio for diagnosis group C (e.g., cystic fibrosis), donor diabetes for diagnosis group D (e.g., idiopathic pulmonary fibrosis). COX and RF models performed similarly to PT-CAS; however, the combined COX + PT-CAS model had improved discrimination (1-year area under the receiver operator characteristic curve [AUC] PT-CAS 0.609 vs 1-year AUC COX + PT-CAS 0.626) and improved calibration across a broader range of predicted risk.
The influence of donor factors on recipient PT survival differed by age, sex, and diagnosis. The addition of donor factors to existing models predicting PT survival led to only modest improvement in prediction accuracy. Future efforts may focus on optimizing matching strategies to improve donor utilization.
预测肺移植后的生存率仍然是一个难以实现的目标。我们分析了供体因素对移植后生存率的影响，以及这些关系在移植受者中的差异。
我们研究了美国移植受者科学登记处（2015年2月19日至2020年2月1日）的原发性双侧肺移植受者（n = 7609）。评估并调整了候选者年龄、性别和诊断方面的主要和交互作用。将预测移植后生存率的模型与移植后综合分配评分模型（PT-CAS）进行比较：（1）Cox回归供体多变量模型（COX），（2）COX + PT-CAS，（3）随机森林模型（RF），以及（4）RF + PT-CAS。比较了模型的辨别力和校准措施。
供体和受者因素之间的相互作用按年龄出现：55至69岁受者接受循环死亡后器官捐赠的生存率较低，30至54岁和70岁以上受者的供体吸烟，年龄小于30岁受者的西班牙裔供体，30岁以上受者的非西班牙裔黑人供体；性别：男性的巨细胞病毒不匹配；诊断：C诊断组（如囊性纤维化）的供受者体重比更高，D诊断组（如特发性肺纤维化）的供体糖尿病。COX和RF模型的表现与PT-CAS相似；然而，COX + PT-CAS组合模型的辨别力有所提高（接受者操作特征曲线下1年面积[AUC]，PT-CAS为0.609，而1年AUC COX + PT-CAS为0.626），并且在更广泛的预测风险范围内校准得到改善。
供体因素对受者移植后生存率的影响因年龄、性别和诊断而异。在预测移植后生存率的现有模型中加入供体因素仅使预测准确性略有提高。未来的努力可能集中在优化匹配策略以提高供体利用率。
