The grassy knoll ... and an elephant

Abstract

"... as we know, there are known knowns; there are things we know we know.We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns - the ones we don't know we don't know." -- Secretary of Defense Donald H. Rumsfeld, 12 February 2002. Copeland suggests that there are currently enough known knowns about cannabis use and its consequences for us to develop suitable evidence-based and targeted cannabis-related policies, treatment guidelines, and prevention and intervention strategies. As is apparent from the other commentaries, this is not a universally held opinion. For example, there are known unknowns that are essential for evidence-based and targeted clinical practice and public health strategies which are discussed by Hammersley (e.g. lack of a theory of dependence that fits cannabis using behaviour), Patton (e.g. lack of knowledge of the predictive value of different indicators of use for health and developmental outcomes), and Andreasson (e.g. lack of knowledge about the hidden population of cannabis users). All of the commentaries support the known known status of methodological limitations in the cannabis use literature. Copeland, citing longitudinal studies, implies the association between cannabis use and psychosis is a known known. However, as pointed out by Hammersley, the lack of any theory explaining how cannabis use might cause psychosis suggests that the nature of this relationship is a known unknown. Additionally, while longitudinal studies contribute to the determination of causality, this is only possible if confounding is adequately controlled, the cannabis data collected is appropriately detailed, and the assumptions on which the studies are based are accurate. Meta-analytic studies are only as accurate as the studies on which they are based. Similarly, the inference that the relationship between high frequency of use and adverse outcomes is a known known does not stand up once the methodological issues associated with frequency of use variables are considered.