A bias of randomisation must also be suspected when comparing the hazard ratios of the mammography arm in the prevalence round versus subsequent rounds: 1.47 versus 0.9 [20].