The draft ICH E9(R1) Addendum describes "treatment policy" as one of the strategies for addressing an intercurrent event. It is aimed to capture the effect attributable to assignment to treatment by using all data, regardless of whether or not the event occurred. The Addendum recognizes that this strategy might not always target the treatment effect of greatest relevance to regulatory and decision making. However this estimand remains of great regulatory importance and is being requested by regulatory agencies across therapeutic areas. This presentation is based on a simulation exercise performed for an Alzheimer long-term prevention trial. Treatment discontinuation is one of the intercurrent events and it is addressed through a treatment policy strategy. Different scenarios are considered for the on-treatment versus the post-treatment mean efficacy trajectory. Several estimators that could account or not for the intercurrent event are investigated. This presentation will show how, for each scenario, the selection of the estimators as well as the retrieval rates of post-treatment data could have a strong impact on the estimates of power, mean treatment effect and its variability.