Distinguishing drug-target effects from natural lipid variation and incorporating imaging outcomes in Mendelian randomisation studies of multiple sclerosis.