Landmark analysis

Date created: 2021-11-02

Typically used in Survival analysis settings to overcome guarantee-time bias.

Landmark analysis is when you select a time point during follow-up and analyze only the subjects who have survived until then. Any subjects who were lost to follow-up or died prior to the landmark are excluded. You then classify response and apply the usual survival analysis methods.

Choosing the landmark

  • Late landmarks = smaller sample size & lower statistical power
  • Early landmarks = late responders are misclassified as non-responders, low Sensitivity

Usually the landmark is picked from a clinically meaningful period, or you use multiple landmark times for sensitivity analyses.