medRxivJClub
Competing event regression on the relative subdistribution and cumulative-incidence scales
medRxiv · · Preprint · Open access
Mell, Overgaard
Abstract ↗AI summary
The abstract is read at the publisher; the summary is JClub's.
A stacked Lunn-McNeil extension estimates competing-event ratios on subdistribution and cumulative-incidence scales, including omega-tilde+(t) and rho(t), with efficiency tradeoffs versus the hazard-scale contrast omega+(t).
- Why it matters: Standard cause-by-cause tests can miss clinically important differences in the ratio between the event of interest and the competing event. The paper targets inference for these realized-incidence ratios and clarifies how scale choice changes interpretation and efficiency.
- What they did: It extends the generalized competing event framework by estimating omega-tilde+(t) via a risk-set-weighted stacked model and rho(t) via stacked per-cause pseudo-observation regression of the Aalen-Johansen estimator. It derives variance formulas, links rho to omega+ under proportional cause-specific hazards, and links rho to omega-tilde+ only in the small-time limit under proportional subdistribution hazards.
- The result: The hazard-scale ratio omega+ is exactly tied to rho under proportional cause-specific hazards, while omega-tilde+ matches rho only at early times and then drifts toward 1. The authors derive covariance terms, assess operating characteristics by simulation, and provide implementation in the gcemod R package.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.
- Open access
Free. We’ll email you when a new paper cites this one or any other paper you follow on JClub — at most one email a day. Every email has a one-tap unsubscribe.