Causal inference approaches for dealing with time-dependent confounding in longitudinal studies, with applications to multiple sclerosis research