A type of selection bias due to systematic differences between study groups in the number and the way participants are lost from a study.