Optimal treatment sequence and mortality: Application of landmark analysis to real-world data
Objectives
The objective is to present a pragmatic framework for implementing landmark analysis in order to assess the associations between treatment sequences and subsequent all-cause mortality, and to illustrate its application through a study of the impact of treatment on mortality, using real-world data from the French National Health Data System (SNDS).
Conclusion
This work shows how landmark analysis makes it possible to estimate associations between the evolution of treatment sequences and subsequent mortality, while accounting for time-dependent exposures. In this illustration, single-treatment sequences were generally associated with lower risks compared with treatment A before 3 years, whereas combinations showed higher risks, with contrasts becoming more pronounced after 3 years. Although inference is limited to patients who survived up to each predefined time, landmark analysis is a simple, robust and widely accessible approach, offering clear and interpretable results, for handling time-dependent covariates in studies, as well as various potential outcomes.