What does this research mean for the field?

Standard proportional hazards models are subject to built-in selection bias due to unmeasured heterogeneity from omitted covariates, even in randomized trials, which can be mitigated by alternative approaches such as frailty models or accelerated failure time models. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To evaluate how unmeasured heterogeneity caused by omitted covariates biases treatment hazard ratio estimates in proportional hazards models and to evaluate alternative statistical methods.

September 14, 2026Statistical Methods in Medical ResearchOpen Access

Built-in selection bias in proportional hazards models with omitted covariates: Simulation evidence and alternative approaches

View Full Paper

Ask AI

Bookmark

Authors

ABAyoub BifenziUniversité de Bordeaux HJHélène Jacqmin‐GaddaUniversité de Bordeaux

Discussion

Member takes

Implication

Simulation study reveals built-in selection bias in standard proportional hazards models from omitted covariates, highlighting robust alternative survival analysis methods.

Key Points

To evaluate how unmeasured heterogeneity caused by omitted covariates biases treatment hazard ratio estimates in proportional hazards models and to evaluate alternative statistical methods.
Simulated semi-parametric Cox and parametric proportional hazards models across varied unmeasured heterogeneity scenarios to assess treatment effect bias.
Evaluated alternative estimators including frailty models, accelerated failure time models, non-parametric Kaplan-Meier curves, and Cox models with time-dependent effects.
Applied and compared all models using real-world randomized controlled trial data from the Radiation Therapy Oncology Group (RTOG9202).
Standard proportional hazards models systematically biased the treatment hazard ratio when key covariates were omitted, despite treatment randomization.
Alternative strategies, particularly frailty models and accelerated failure time models, effectively mitigated selection bias caused by unobserved heterogeneity.

Cite This Study

Bifenzi et al. (2026) studied this question.

synapsesocial.com/papers/6aa7b23b0926e14a848b09fc https://doi.org/10.1177/09622802261485825

View Full Paper

Ask AI

Bookmark

Implication

Simulation study reveals built-in selection bias in standard proportional hazards models from omitted covariates, highlighting robust alternative survival analysis methods.

Built-in selection bias in proportional hazards models with omitted covariates: Simulation evidence and alternative approaches

Discussion

Member takes

Implication

Key Points

Cite This Study

Built-in selection bias in proportional hazards models with omitted covariates: Simulation evidence and alternative approaches

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider