Statistical inference methods for cumulative incidence function curves at a fixed point in time

TL;DR

This paper develops statistical tests for comparing cumulative incidence function curves at specific time points in competing risks data, especially when proportional hazards assumptions are violated or curves cross.

Contribution

It extends existing tests using pseudo-value regression and transformation functions to compare CIFs at fixed times, addressing limitations of traditional methods.

Findings

New tests accommodate crossing CIF curves and non-proportional hazards.

Methods improve accuracy of CIF comparisons at fixed time points.

Extensions based on Gaynor's and Aalen's variance approaches.

Abstract

Competing risks data arise frequently in clinical trials. When the proportional subdistribution hazard assumption is violated or two cumulative incidence function (CIF) curves cross, rather than comparing the overall treatment effects, researchers may be interested in focusing on a comparison of clinical utility at some fixed time points. This paper extend a series of tests that are constructed based on a pseudo-value regression technique or different transformation functions for CIFs and their variances based on Gaynor's or Aalen's work, and the differences among CIFs at a given time point are compared.