Cannabis and school dropouts: a reply to Hickman et al.

Abstract

Sir—Dr Hickman and colleagues make a number of valid points about the limitations of covariate adjustment and the importance of estimating causal effects using the attributable risk fraction (ARF). However, these points need to be leavened by the following considerations. For reasons that are not clear, Hickman et al. (2004) chose to estimate the ARF associated with eliminating frequent cannabis use at 16. Because only 1.2% of our cohort were frequent users at this age, the causal effect of frequent cannabis use on school dropout was small. However, it is more realistic to present the ARF for the elimination of all cannabis use. This proves to be not inconsequential and our results suggest that the elimination of all cannabis use could reduce rates of school dropout by nearly 14%. It is also important to recognize that the ARF depends critically on the fraction of the sample exposed (i.e. Pexp) and the estimate is specific to that fraction. Thus, in a situation in which 1.2% of the Christchurch Health and Development Study (CHDS) sample were exposed the ARF is 3%. However, if 50% had been exposed then the ARF would have been 57%. These results highlight the importance of presenting the ARF as a measure of causal effect that applies to a stipulated level of exposure in the population and not as an absolute measure of causal effect. Looked at from this viewpoint the headline message of Dr Hickman's letter (‘eradicating frequent cannabis use . . . could reduce school dropout by 3%’) is misleading. It is easy to show from the formula presented that the ARF (assuming a causal effect and an OR of 3.7) can range from 0% to 72.9%[this follows because as Pexp→1; ARF→(OR-1)/OR], depending on the proportion exposed (Pexp); the ARF happens to have a value 3.0% at an exposure level of 1.2%. Dr Hickman and colleagues note that some may find a reduction of 3% inconsequential. However, such arguments overlook the fact that although the effects of cannabis may not be large as assessed by the ARF, these effects appear to be pervasive and span a wide range of physical, educational and psychosocial outcomes. There is, thus, a need to locate the effects of cannabis on school dropout in the context of the broader literature of the adverse effects of cannabis. As we point out in our paper (Fergusson, Horwood & Beautrais 2003), the findings add to growing evidence suggesting that the use of cannabis by young people may have consequences for a wide range of developmental outcomes, including school dropout.