Observational data, from electronic health records, claim databases or disease registries, are increasingly used to understand the causal effect of treatments or exposures on health outcomes. When the start of follow-up, the time of eligibility assessment and the time of treatment initiation do not coincide, standard analysis methods may lead to biased estimates of treatment effects. This bias is known as immortal-time bias. In this presentation, we will first describe situations in which this bias occurs, and then introduce a few solutions to prevent or handle this issue, with a focus on the clone-censor-weight approach.