Differences between Very Highly Sensitized Kidney Transplant Recipients as Identified by Machine Learning Consensus...

Area of Science:

• Nephrology
• Transplantation immunology
• Data science in medicine

Background:

• Kidney transplant recipients with panel reactive antibody (PRA) ≥ 98% have inferior clinical outcomes despite allocation priority.
• Identifying subgroups within this population is crucial for tailored management strategies.
• Vulnerable patient populations require precise risk stratification for improved post-transplant success.

Purpose of the Study:

• To categorize highly sensitized kidney transplant recipients (PRA ≥ 98%) using unsupervised machine learning.
• To identify distinct subgroups with varying clinical outcomes.
• To inform individualized management strategies for improved patient and graft survival.

Main Methods:

• Analysis of the Organ Procurement and Transplantation Network (OPTN)/United Network for Organ Sharing (UNOS) database (2010-2019).
• Consensus cluster analysis of 7458 kidney transplant patients with pre-transplant PRA ≥ 98%.
• Comparison of post-transplant outcomes between identified clusters based on recipient, donor, and transplant characteristics.

Main Results:

• Two distinct clusters of highly sensitized kidney transplant recipients were identified.
• Cluster 1: Younger, male, prior transplant history, less diabetic kidney disease; lower death-censored graft survival, higher acute rejection.
• Cluster 2: Older, female, first-time transplant; comparable patient survival, better death-censored graft survival, lower acute rejection.

Conclusions:

• Unsupervised machine learning effectively categorized highly sensitized kidney transplant patients into two clinically distinct subgroups.
• These subgroups exhibit differing post-transplant outcomes, particularly regarding graft survival and acute rejection.
• Understanding these distinct clusters can guide the development of individualized care strategies to improve outcomes for this vulnerable population.