Non-experimental data, such as electronic medical records, are often used for causal inference to estimate the effect of an exposure on an outcome variable. However, these data do not come from a study design ensuring a balance of patient characteristics between exposure groups. Patients are also observed irregularly over time. These imbalances can bias the estimation of causal effects. Methods have recently been proposed to address these challenges, but they mostly focused on acute treatment effects. In this presentation, we propose a methodology to consistently estimate the causal effect of a cumulative exposure over time on a continuous response. It allows for the consideration of delayed treatment effects and the causal effect is estimated from irregularly measured responses. Confounding and bias induced by irregular observation times are addressed through weighting methods.