The mean survival time for participants in the cannabis dependence model was 13.4 years (s.d. 4.9) (time from opportunity to dependence, or for those who did not develop dependence, time from opportunity to age at interview). This figure is higher than may be expected from the mean dependence age as individuals who have not developed dependence by the point of interview are also included in the survival time, with their time from opportunity to age at interview in place of time to dependence. All covariates were tested for breach of the proportional hazards assumption, as outlined in the methods section. The following did not satisfy the proportional hazards assumption for the opportunity to use model and therefore the interaction term between the factor and analysis time was modelled in the cannabis use opportunity analysis (Bellera et al., 2010): conduct disorder, parental drug problems, weekly tobacco use and monthly alcohol use. Similarly, for the cannabis dependence analysis the following variables had the interaction with analysis time modelled in the analysis: parental drug problems, alcohol dependence and other drug use.
