This interesting paper uses the best available data examining the association between cannabis use and schizophrenia to estimate a metric called Number Needed to Prevent (NNP). Clinicians will be familiar with the related concept of Number Needed to Treat (NNT)—the smaller the number, the stronger the effect size of an intervention. The NNP metric for cannabis and psychosis is very large. For example, for young people who are heavy users, we would need to stop use in about 2,800 individuals in order to prevent one member of this cohort from going on to develop a case of schizophrenia (that would not otherwise have occurred due to other factors).

The NNP is even higher in other groups. In other words, because cannabis does not appear to be a potent (large-effect) causal factor, the NNP is large and unimpressive. Taking into account the second-order issue that interventions to stop people from using cannabis themselves have a high NNT, leveraging cannabis use in order to stop psychosis does not look like an appealing prospect. However, cannabis use is also linked to...  Read more

This interesting paper uses the best available data examining the association between cannabis use and schizophrenia to estimate a metric called Number Needed to Prevent (NNP). Clinicians will be familiar with the related concept of Number Needed to Treat (NNT)—the smaller the number, the stronger the effect size of an intervention. The NNP metric for cannabis and psychosis is very large. For example, for young people who are heavy users, we would need to stop use in about 2,800 individuals in order to prevent one member of this cohort from going on to develop a case of schizophrenia (that would not otherwise have occurred due to other factors).

The NNP is even higher in other groups. In other words, because cannabis does not appear to be a potent (large-effect) causal factor, the NNP is large and unimpressive. Taking into account the second-order issue that interventions to stop people from using cannabis themselves have a high NNT, leveraging cannabis use in order to stop psychosis does not look like an appealing prospect. However, cannabis use is also linked to other adverse acute and chronic health outcomes (Hall and Degenhardt, 2009).

While the NNP is a large number, such estimates can still have important public health consequences. When considering options for public health interventions, factors such as 1) the efficacy, cost, and safety of the proposed intervention, and 2) the nature of the outcome, need to be taken into account. A large number of people would be required to buckle up their seat belts in order to prevent one death related to motor vehicle accidents. However, seat belts are relatively cheap and safe; thus, from a public health perspective, the risk-benefit calculus favors this simple preventive measure.

In the absence of universal (population-based) interventions, we need to think about selective interventions. Can we predict which individuals are at high risk of cannabis-related psychosis? Can genes like COMT help? As the genetic architecture of psychosis is slowly unraveled, we may catch glimpses of new candidates that can help predict which individuals are at risk of cannabis-related psychosis.

References:

Hall W, Degenhardt L. Adverse health effects of non-medical cannabis use. Lancet. 2009 Oct 17;374(9698):1383-91. Abstract

This important study demonstrates the need for evidential epidemiology and associated risk estimates as an essential balance to the sort of knee-jerk responses that all too often follow studies linking drug use and mental health problems.

View all comments by David J. Porteous