Is it time to use observational data to estimate treatment effectiveness in multiple sclerosis?