|
Abstract:
|
Observational studies are common in medical device clinical studies. In those studies, patients are not randomly selected for each treatment, so the benefit from randomized studies such as balance in covariates between two treatment groups no longer exists, and treatment groups may not be comparable at baseline. Propensity score analysis can be used for adjusting for imbalance in observed covariates, and improving comparability between treatment groups. An application of the propensity methodology in a medical device study is presented with the objective to compare survival curves between ICD and non-ICD patients. To balance significant differences at baseline between the two groups, a propensity score analysis is employed under Cox's proportional hazard framework.
|
