How can we prove that a preventive measure in elite sport is effective when the prevalence of the injury (eg, ACL tear in alpine ski racing) is low? A case for surrogate outcomes