We know that screening a large number of drugs against 56 cancer sites will lead to many associations that are merely due to chance.  The criteria that we have used have been based on judgment and our long experience in screening drugs.  The requirements of odds ratio of at least 1.5 and some indication of dose-response help to reduce the number of chance associations.  Our current criteria are subject to change as experience accumulates.  We are not wedded to them.