++
This JAMA Guide to Statistics and Methods article discusses accounting for competing risks in clinical research.
++
Survival analyses are statistical methods for the analysis of time-to-event outcomes.1 An example is time from study entry to death. A competing risk is an event whose occurrence precludes the occurrence of the primary event of interest. In a study whose outcome is time to death due to cardiovascular causes, for instance, death due to a noncardiovascular cause is a competing risk. Conventional statistical methods for the analysis of survival data typically aim to estimate the probability of the event of interest over time or the effect of a risk factor or treatment on that probability or on the intensity with which events occur. These methods require modification in the presence of competing risks. A key feature of survival analysis is the ability to properly account for censoring, which occurs when the outcome event is not observed before the end of the study participant’s follow-up period.
++
In a study published in JAMA, Roumie and colleagues2 compared the rates of major adverse cardiovascular events (MACE), including cardiovascular death, in patients with diabetes and reduced kidney function who were treated with metformin vs those treated with sulfonylurea. In their analyses, the authors considered 2 competing risks: changes in therapy and noncardiovascular deaths.
+++
EXPLANATION OF THE CONCEPT
++
When study participants are followed up over time, to determine the time to an event of interest, other types of events can also occur. In studies in which the outcome of interest is death from a particular cause, the other types of events can be deaths due to other causes. Competing risks are present if the different types of events are mutually exclusive so the occurrence of one type of event precludes the occurrence of any of the other types of events. Thus, if the causes of death were classified as cardiovascular or noncardiovascular, competing risks would be present because one can die of only a single cause. Other common instances of competing risks include settings in which an event of interest is not fatal, such as acute myocardial infarction, with all-cause mortality being the competing risk. Because the occurrence of acute myocardial infarction does not prevent death, but death does prevent subsequent acute myocardial infarction, this setting can be referred to as having semicompeting risks.3
++
In the setting of competing risks, a common goal of the statistical analysis is to estimate the cumulative incidence function (CIF). The CIF is the probability that the event of interest will occur before a specific time, expressed as a proportion of the original population at risk. The CIF is a nondecreasing curve over time, starting at zero, with a more rapid increase indicating a higher incidence of the event of interest. Competing risks, because they preclude the possibility that the event of interest will ...