Horizontal mixture model for competing risks: a method used in waitlisted renal transplant candidates

Area of Science:

• Nephrology
• Transplantation Medicine
• Biostatistics

Background:

• Patients on renal transplant waiting lists need clear prognosis information.
• Competing risks (e.g., death) complicate traditional survival models.
• Existing models are often difficult for non-specialists to interpret.

Purpose of the Study:

• To develop an interpretable model for predicting long-term kidney transplant probability.
• To estimate time in dialysis for transplanted patients.
• To address the challenge of competing risks in transplantation prognosis.

Main Methods:

• Utilized a horizontal mixture model approach.
• Extracted data from French dialysis and transplantation registries.
• Employed internal and external validation strategies, including the "Ile-de-France" region.

Main Results:

• Identified seven variables decreasing long-term transplant probability: age >40, comorbidities, prolonged pre-registration dialysis, and blood groups O/B.
• Found longer transplantation times for recipients <50, overweight, blood groups O/B, and those with pre-transplant immunization.
• Model demonstrated good discriminative capacity (AUC at 5 years = 0.72) but highlighted calibration needs for international use.

Conclusions:

• The horizontal mixture model provides easily interpretable predictions for transplant probability and dialysis time.
• This approach offers a practical alternative to sub-hazard or cause-specific competing risk models.
• Further validation and practical testing with clinicians and patients are recommended for broader application in chronic diseases.