Treatment crossover means patients may switch from the assigned treatment, either to the other trial treatment, a non-trial treatment or to stop receiving treatment altogether. Despite many efforts taken to minimize treatment crossover, it is oftentimes inevitable and sometimes ethical to have treatment crossover. Treatment crossover will mask the true treatment effect, often attenuating it to null, therefore making it difficult to power a study. On the other hand, a study may be incorrectly stopped when the crossover effect is not properly accounted for in constructing the stopping boundary. Moreover, the estimate of the treatment effect is often biased and needs statistical approaches for adjustment. In this talk, I use a multi-state model to characterize treatment crossover, based on which an estimation method is proposed to better estimate the true treatment effect. A visualization tool is presented for study design.