Abstract:
|
Although data in drug development are often collected longitudinally, frequently primary and key secondary analyses rely on a single post-randomization time point (e.g., change from baseline, response at Week X, etc.). Longitudinal modeling utilizing all observed data offers the opportunity to achieve substantial efficiencies in data analysis compared to the current prevailing practices, such as faster and more accurate decision making (e.g., interim analysis adaptation and futility rules, adaptive dose selection), increased statistical power and/or estimation precision, and combinations of these. This presentation will discuss and illustrate some of the opportunities associated with the broader use of parametric longitudinal modeling, but also the caveats associated with it, such as the need for additional assumptions about the data.
|