Skip to main content

Table 3 The impact of mis-specifying p% on the design and analysis. At the design stage, the sample size required to achieve the target power is calculated using the conventional design ignoring response dichotomy and the p-embedded design recognizing response dichotomy but mis-specifying the true p% to be pm%. Given the sample size calculated, the empirical power (EP) is evaluated under the true response dichotomy p%. Target power is 80%. Lag duration is 1 month. Hazard ratio for responding patients is 0.3. Total study duration is 5 years

From: Treating non-responders: pitfalls and implications for cancer immunotherapy trial design

P-embedded designConventional design
DesignAnalysisDesignAnalysis
Mis-specifying p*Over-specifying p* = pm − 10Under-specifying p* = pm + 10Ignoring p*
pm (%)Np* (%)EPp* (%)EPNp* (%)EPp* (%)EP
202691044.90%3095.20%27106.61%3013.39%
301372059.30%4088.90%208.81%4018.58%
40893063.70%5089.20%3013.39%5025.73%
50684070.30%6089.00%4018.58%6032.81%
60525071.40%7088.00%5025.73%7040.93%