Abstract:
|
The gold standard for evaluating the effects of medical interventions is the randomized trial with concurrent control. Unfortunately, in some situations, especially for rare diseases, randomized controlled trials are not feasible. In these cases, historical controls, such as natural history data from registries or literature, might be considered in assessing the efficacy of a new drug or biologic. A major concern related to the application of historical controls is whether the populations studied are comparable at baseline. A potential cause for the observed differences between intervention and historical control arms might be variation in the assessments used to measure patients' performance, resulting, at times, in many issues associated with the interpretation of the missing observations in the natural history datasets. To overcome challenges in using historical controls, we must examine and assess the "totality of the evidence" presented, and thoroughly analyze the study data, applying a number of different assumptions. In my presentation, a complicated case example will be used to illustrate these potential issues and solutions, including: ordinal regression; survival analyse
|