Competing-risk Cumulative Incidence Function (CIF)
When competing events exist (e.g. a "target endpoint" vs "death from other causes"), using 1−KM overestimates the incidence of the target event. The CIF (cumulative incidence function) correctly removes the effect of competing events and is the standard method for estimating cumulative incidence under competing risks. Computed locally in your browser; data are not uploaded.
① Input
Each row: time status [group]. Status: 0=censored, 1=target event, 2 (or higher)=competing event. Group is optional; with multiple groups, the target-event CIF curve is drawn for each.
How to use & methodology
Why can't I use 1−KM?
Kaplan-Meier treats competing events as censoring, implicitly assuming the censored may still experience the target event later; but someone who had a competing event (e.g. died of another cause) can never have the target event. So 1−KM systematically overestimates the target-event cumulative incidence, and under competing risks the CIF should be used instead.
How do I code the status column?
0=censored (follow-up ended without any event), 1=target event (the endpoint you care about), 2 or higher=competing event (an event after which the target event can no longer occur, e.g. death from another cause).
What do the CIF curves sum to?
At any time point, target-event CIF + each competing-event CIF + event-free survival S = 1; these three parts completely partition the population.
How do I do between-group comparison?
For a formal comparison of cumulative incidence under competing risks, use Gray's test; for multivariable analysis, use Fine-Gray subdistribution hazard regression — this site's 'Fine-Gray subdistribution hazard regression' tool does it directly: use the grouping as a covariate, and its subdistribution HR and P value give the between-group comparison (the regression form of Gray's test). This tool provides CIF curves and τ-time-point estimates for an intuitive comparison.