Authors' reply

Among the sociodemographic variables we reported in Table 1, it is correct to point out that unemployment rates are statistically significantly higher in the cases compared with controls (P<0.001). This difference has already been reported in previous epidemiological studies and there is no evidence that this arises from a bias in the sample selection. However, it is rather a potential confounder. In our paper we did not discuss if or how employment status might have influenced our findings, because, together with other relevant variables, we controlled for it in the statistical analyses. Thus, the higher rate of unemployment in cases than controls might partially account for the drop of the crude odds ratio (OR) of 8.1 (95% CI 4.6–13.5) to the adjusted one (OR=6.8, 95% CI 2.6–25.4), which occurred when we controlled for confounders including unemployment. However, the odds ratio still remains strikingly high and statistically significant (P<0.05), indicating that our findings cannot be explained by the effect of employment status or by any of the other social variables listed.

Lastly, we wish to comment on the suggestion that controls' preference for low-potency cannabis might be consequent to their need to continue being able to work. Would this not indicate that high-potency cannabis is more likely to negatively affect social functioning perhaps via its detrimental effect on mental health? Exactly what our findings suggest.