Commentary: Understanding pathways to cannabis use and from use to harm

apparently robust association between maternal partner change and reports of child cannabis use may have arisen. First, 'reverse causation' argues that it was the child's cannabis use that led to the break-up of their mother's relationship rather than vice versa. This seems unlikely as the relationship changes were before the child was 14 years old and most of the cannabis use reported was subsequent to this. Second, 'reporting bias' is a possibility. This bias is most relevant when self-reported exposures influenced by notions of social desirability are related to similarly influenced self-reported outcomes.15 It may also be important when individuals whose outcome status is already known report exposure retrospectively.16 Neither of these situations applies to the present study since partner change was reported by mothers when their children were aged 14 and cannabis use was reported by children themselves aged 21. The third possibility is confounding by common antecedents of both maternal partner change and children's cannabis use. The obvious candidate for a confounding factor in this context is general material deprivation. Partnerships may be more fragile and prone to change in difficult circumstances and children growing up in these circumstances may be at greater risk of using cannabis (particularly at an earlier age) through mechanisms independent of their mother's partner changes. To some extent, this question has been addressed in the analysis, which adjusted for: maternal use of drugs, low income, prior psychological problems, and lone parenthood itself. The issue of paternal drug use was not considered and it may have been important. As with maternal drug use, paternal drug use predicts offspring drug use and maternal relations with drug using fathers or father figures may be more volatile. The general point about confounding, however, is its pervasive nature and the difficulty in overcoming