Marginal structural models might overcome confounding when analyzing multiple treatment effects in observational studies | Publicación