In survival analysis, informative censoring occurs when the reason for censoring is related to the risk of the event, which can bias hazard ratio estimates.

• A. It has no impact on estimates.
• B. It biases hazard ratio estimates unless addressed with appropriate models.
• C. It only affects odds ratios.
• D. It always biases estimates in favor of the exposure.

Answer: (B)

Informative censoring happens when the reason someone is censored is linked to their risk of the event. In survival analysis, methods like Cox regression assume censoring is non-informative, meaning censored individuals would have the same risk as those who remain, after accounting for covariates. If censoring is informative, the risk sets at each time point become biased because those who are censored early may have different hazards than those who stay in the study. This distortion leads to biased estimates of the hazard function and, therefore, the hazard ratio, unless the analysis uses methods that account for the informative censoring—such as inverse probability of censoring weighting or joint models that incorporate the censoring mechanism. So the effect on hazard ratio estimates is real and depends on how censoring relates to risk, and it can be mitigated with appropriate modeling techniques. This isn’t about having no impact, it isn’t limited to odds ratios, and the direction of bias is not fixed to favor the exposure.