Optimizing the program-specific reporting of pancreas transplant outcomes.

Area of Science:

• Transplantation Science
• Medical Statistics
• Health Services Research

Background:

• Monitoring pancreas transplant program outcomes is challenging due to distinct procedure types (SPK, PAK, PTA) and limited statistical power when analyzed separately.
• The Scientific Registry of Transplant Recipients requires accurate program-specific outcome reports for US organ transplant programs.

Purpose of the Study:

• To develop and validate predictive models for pancreas transplant outcomes by combining different transplant types.
• To assess the utility of pooled data for improving the statistical power and reliability of program-specific outcome reporting.

Main Methods:

• Combined two consecutive 2.5-year cohorts of pancreas transplant recipients.
• Developed Cox proportional hazards models for 1- and 3-year graft and patient survival, analyzing SPK, PAK, and PTA separately initially.
• Pooled data across transplant types within centers to compare observed versus predicted outcomes and tested models on the second cohort.

Main Results:

• Models for 1-year pancreas graft and patient survival achieved C statistics of 0.65 and 0.66, respectively, comparable to other organ transplant survival models.
• Model calibration using the Hosmer-Lemeshow method was acceptable, indicating good predictive accuracy.
• Pooling SPK, PAK, and PTA data demonstrated potential for generating useful program-specific outcome reports.

Conclusions:

• Pooling data from simultaneous pancreas-kidney (SPK), pancreas after kidney (PAK), and pancreas transplant alone (PTA) procedures can create effective predictive models.
• This approach enhances statistical power, enabling more robust reporting of program-specific pancreas transplant outcomes.
• The developed models offer a viable solution for the previously problematic monitoring of pancreas transplant graft survival.