There has been remarkably little demonstration of the deleterious impact of publication bias within addiction science or indeed in wider healthcare policy and practice. An account is provided here of how publication bias was identified in relation to a series of drug education reviews which have been very influential on subsequent research, policy and practice. Later data analyses unpublished by the same review team demonstrated earlier findings to be unreliable. These later findings were not published. The policy context in which evidence on drug education in schools is produced is considered and the need for unbiased evidence is emphasised. A broadened conception of publication bias is proposed which takes account of the environment in which publication decision-making occurs. It is suggested that this is particularly necessary for subjects with such direct policy relevance as the effectiveness of drug education in schools.