Re-assessing prolonged cold ischemia time in kidney transplantation through machine learning consensus clustering

Area of Science:

• Nephrology
• Transplant Surgery
• Machine Learning in Medicine

Background:

• Kidney transplantation is a vital treatment for end-stage renal disease.
• Prolonged cold ischemia time (CIT) is a significant factor impacting kidney graft outcomes.
• Identifying recipient and donor characteristics associated with prolonged CIT is crucial for optimizing transplant success.

Purpose of the Study:

• To apply unsupervised machine learning to cluster deceased donor kidney transplant recipients with prolonged CIT (>24 hours).
• To identify distinct patient and donor characteristics within these clusters.
• To compare post-transplant outcomes between the identified clusters.

Main Methods:

• Utilized consensus cluster analysis on a large dataset of 11,615 deceased donor kidney transplants (2015-2019) from OPTN/UNOS data.
• Analyzed recipient demographics, donor characteristics, cold ischemia time, and machine perfusion use.
• Compared 1-year patient and death-censored graft survival between clusters.

Main Results:

• Two clinically distinct clusters were identified.
• Cluster 1: Younger, non-diabetic recipients with younger, healthier donors (lower KDPI), lower CIT, and better outcomes.
• Cluster 2: Older, diabetic recipients with older donors (higher KDPI), longer CIT, higher delayed graft function, and slightly lower survival rates.

Conclusions:

• Unsupervised machine learning effectively characterized kidney transplant recipients with prolonged CIT into two outcome-differentiated clusters.
• While Cluster 1 showed more favorable characteristics, Cluster 2 also demonstrated acceptable outcomes.
• Findings suggest potential opportunities for transplant centers to safely extend CIT.