Progression and Death as Competing Risks in Ovarian Cancer
Abstract: Background: Progression of a cancer disease and dying without progression can be understood as competing risks. The Cause-Specific Hazards Model and the Fine and Gray model on cumulative incidences are common statistical models to handle this problem. The pseudo value approach by Andersen and Klein is also able to cope with competing risks. It is still unclear which model suits best in which situation.
Methods:For a simulated dataset and a real data example of ovarian cancer patients who are exposed to progression and death the three models are examined. We compare the three models with regards to interpretation and modeling requirements.
Results:In this study,the parameter estimates for the competing risks are similar from the Cause-Specific Hazards Model and the Fine and Gray model. The pseudo value approach yields divergent results which are heavily dependent on modeling details.
Conclusions:The investigated approaches do not exclude each other but moreover complement one another. The pseudo value approach is an alternative that circumvents proportionality assumptions. As in all survival analyses, situations with low event rates should be interpreted carefully.Keywords: Multistate Models, pseudo values, cause-specific hazards, cumulative incidence, Fine and Gray model.
Download Full Article