Time-related analysis of nonfatal heart valve complications: cumulative incidence (actual) versus Kaplan-Meier (actuarial).

BACKGROUND: The cumulative incidence of a postoperative event is the percentage of patients who experience the event by postoperative time T. Its complete determination requires all patients to be followed until T. In ongoing series, the Kaplan-Meier method is employed because not all patients have been observed until T. When applied to nonfatal events, however, the Kaplan-Meier estimates probabilities as if the patients who die before they sustain an event continue to be at risk thereafter. It thus estimates risk in the unrealistic situation where death does not occur. METHODS AND
RESULTS: Cumulative incidence can be estimated directly, to provide the probability of actually experiencing an event before death, that is, when death properly eliminates patients from further risk of the event. We compare cumulative incidence and Kaplan-Meier estimates in two series of mitral valve replacement patients: thromboembolism in a completed series of ball valves implanted in relatively young patients and valve explant in an ongoing series of porcine valves in older patients. Kaplan-Meier estimated a higher event percentage than the cumulative incidence, and the difference was greater in the older patients, who had a higher death rate.
CONCLUSIONS: Cumulative incidence, unlike Kaplan-Meier, provides estimates of the percentage of patients who will actually sustain an event. Cumulative incidence is more meaningful for individual patient counseling and more useful for estimating resource utilization in a managed population.