|
Abstract:
|
In clinical trials, placebo response is a beneficial effect arises from the patients’ expectation regarding the treatment. The sequential parallel comparison design (SPCD) is proposed for addressing placebo response. The SPCD consists of two stages typically with equal duration where the first stage is a standard parallel design of placebo versus drug followed by another parallel design among placebo nonresponders identified from the first stage. The treatment effect in SPCD is commonly defined as a weighted average of observed treatment difference of two stages. However, such estimand is lacking interpretability and can be biased due to its dependence on design parameters. Under principal stratification framework, we propose a casual estimand for the treatment effect under three clinically important principal strata: Always Responders, Never Responders, and Drug Only Responders. To estimate such effect and account for the lack of identifiability, we introduce two sensitivity parameters which can be informed by data due to the design feature of SPCD. Simulation studies are conducted to validate our method. We also apply our method to an actual SPCD trial of antidepressant therapy.
|
