May 21, 2025
The authors focused on adjustment for conditional bias in hazard ratios from overall survival (OS) in both interim and final analysis in trial where the overall hierarchical strategy was applied. They first showed a conditional bias (CB) adjusted estimator for overall survival. They wrote it as the difference in the expectation of the truncated normal distribution and the true treatment effect. They also introduced a CMAE which incorporates a mean adjusted estimator for the bias correction. For a trial stopped early for benefit, it had previously been proposed in the literature to use a penalized MLE (pMLE) which was then used to update the CB adjusted estimator. They then led up to a CB adjusted estimator of OS log HR.
They ran two simulation studies to check their methods. In the first one, they applied O’Brien-Fleming type of Pocock type alpha spending functions to compute stopping boundaries for OS analyses. They found their proposed CMAE showed almost the same performance as the pMLE through most of the scenarios except for some differences in different scenarios when say the correlation between endpoints was larger than the true value of zero. They also ran a real data analysis and drew mostly the same conclusions. Their method could potentially be extended to more than two endpoints. There was not much discussion on sequential testing with using adjustment for CB but perhaps they meant that for later discussions.
Written by,
Usha Govindarajulu
Keywords: overall survival, conditional bias, hazard ratio, sequential testing
References:
Izumi S, Nomura S, and Matsuyama (2025) “Adjustment of Conditional Bias in Hazard Ratios for Group Sequential Testing of Progression-Free Survival and Overall Survival” Statistics in Medicine. https://doi.org/10.1002/sim.70112
https://onlinelibrary.wiley.com/cms/asset/20766f0f-7f22-448b-983e-e6ae4076ea55/sim70112-fig-0001-m.jpg