Strategies to investigate and mitigate collider bias in genetic and Mendelian randomisation studies of disease progression