Time-to-event analysis when the event is defined on a finite time interval

Area of Science:

• Hematopoietic Cell Transplantation
• Immunology
• Biostatistics

Background:

• Acute graft-versus-host disease (GVHD) is a significant complication post-hematopoietic cell transplantation (HCT).
• Existing research often overlooks high post-transplant mortality rates and the limited diagnosis window for acute GVHD (within 100 days).
• Standard time-to-event models do not adequately address these clinical realities.

Purpose of the Study:

• To propose a novel multi-state model for simultaneously analyzing acute GVHD and mortality.
• To explicitly account for the finite time interval during which acute GVHD can be diagnosed.
• To develop methods for estimating the absolute risk of both acute GVHD and death concurrently.

Main Methods:

• Development of a novel multi-state model incorporating joint modeling of acute GVHD and death.
• Derivation of the observed data likelihood and estimation/inference via maximum likelihood.
• Simulation studies to compare the proposed framework with alternative approaches.

Main Results:

• The proposed multi-state model effectively accounts for mortality and the finite diagnosis interval of acute GVHD.
• Methods for simultaneous estimation of absolute risks for acute GVHD and death were established.
• Simulations demonstrated the advantages of the novel framework over existing methods.

Conclusions:

• The novel multi-state model provides a more accurate framework for understanding acute GVHD and mortality risks in HCT.
• This approach improves risk assessment by jointly considering GVHD, death, and the specific diagnostic timeframe.
• The findings are illustrated through an analysis of HCT data, highlighting clinical relevance.