The main aim of this study was to investigate the role of cigarette smoking on cannabis dependence, above and beyond the effects of cannabis exposure, in a sample of young cannabis and tobacco co-users. We conducted an exploratory follow-up of these users four years later with a 27% response rate of which 70% of individuals had smoked cannabis and tobacco at baseline. The 65 participants that were followed up were equivalent in demographics and smoking behaviour to those who were not followed up, at baseline. We hypothesised that cigarette smoking would predict CUD, at both time points. We also investigated whether the effects of cannabis use on cannabis dependence were mediated by cigarette smoking.

Cigarette smoking at baseline was predictive of CUD at baseline when controlling for cannabis use variables in young people who smoke cannabis and tobacco. The most efficient model accounted for 30% of the variance in cannabis dependence which is considered to be a large effect size as R > 0.5 (Cohen, 1988). However, this seems no longer the case four years later, where only baseline CUD predicted follow-up CUD, accounting for almost 25% of the variance and replicating previous findings (Swift et al., 2000). When we investigated how cigarette smoking predicted concurrent CUD; we found that cigarette smoking (years of cigarette smoking and days per month cigarette smoking) mediated the relationship between cannabis use and cannabis dependence suggesting a role of tobacco use in the pathogenesis of CUD in cannabis and tobacco users. We also found these effects to be robust when controlling for depression and premorbid IQ (which were found to be comparable to normative values). Although causality cannot be assumed in this cross-sectional analysis, these results suggest that cigarette smoking may enhance the dependence-forming effects of cannabis. Alternatively, our results may suggest that CUD (as measured by the SDS) may capture some aspects of nicotine dependence in a subset of young people with CUD. As such, this research supplements previous epidemiological research that stresses the predictive ability of tobacco smoking in developing CUDs (Coffey et al., 2000; Prince van Leeuwen et al., 2014; von Sydow et al., 2002).

Our results, based in a naturalistic setting, parallel results from a recent controlled lab study that found cannabis users who smoke cigarettes are more likely to relapse in comparison to those who do not smoke cigarettes, perhaps as a result of this indirect pathway (de Dios et al., 2009; Haney et al., 2012). As such, reducing cannabis dependence might be facilitated by helping individuals quit cigarette smoking (Macleod et al., 2004). We were able to account for about 30% of the variance in CUD from four predictor variables. However, CUD is a complex disorder and causality cannot be determined from one factor. There are many other factors that can predict CUD that were beyond the scope of the current study but have interesting implications. For example, a recent study by van der Pol et al. (2013) found that current problems (such as living alone, coping motives for cannabis use and negative life events) were better predictors of cannabis dependence in young adults than cannabis exposure itself. As a result of this study, we included demographics and scores on the CTQ to our most efficient model, however these did not account for a significant proportion of variance to be included in the final model or in the mediation analysis. It is clear that CUD is a complex disorder that has many predictors and vulnerability factors that were not included in the model.

In the past, regular cigarette smoking would precede cannabis use (Kandel et al., 1992). This sequence in drug use seems to be tapering off, for example, around 1 in 5 young cannabis users have never smoked a cigarette (Suris et al., 2007). Interestingly, both cannabis and tobacco smoking were initiated 4.9 and 4.7 years previously, respectively, at the baseline visit, suggesting simultaneous age of onset in the current study. Therefore, these results do not speak to sequential use as on average the sample initiated both substances at the same time.

Stricter tobacco laws in some countries have altered perceptions such that cigarette smoking is considered a more risky behaviour than previously. In 2013, for the first time, tobacco smoking prevalence was estimated to be below 20% in the UK (Brown and West, 2014). In comparison, cannabis use has become normal and perceptions of regular cannabis use as a risky behaviour are at an all-time low (Johnston et al., 2013) with risk perception inversely related to prevalence of cannabis use (Kleber and Dupont, 2012). This may be due to the shifting landscape and debate over legalisation of both medical and recreational marijuana in states such as Colorado, California and Washington in the United States as well as countries such as Uruguay and the Netherlands (Volkow et al., 2014). As a result, whilst tobacco smoking decreases generally, it is possible that tobacco use will also increase indirectly over time due to increased cannabis use (Patton et al., 2005). Our findings are timely because they suggest tobacco may be involved in the pathogenesis of CUD, a possible risk factor of legalisation.

Our results may be a product of the common liability to the use of cannabis and tobacco including such risk factors like shared genetic and temperamental factors (Agrawal et al., 2008, 2010; Brook et al., 2010; Creemers et al., 2009). For example, recent research shows that nicotine dependence was associated stronger with lifetime CUD for females than males (Blanco et al., 2014). Moreover, Cooper and Haney (2014) have recently demonstrated that whilst subjective effects are equal across genders, females report more abuse related effects. Thus, an interesting analysis would be to investigate whether the mediators suggested in the present study, were stronger in females than males however, given that the sample was 71% male, this was not possible. Demographic variables were instead added to the most efficient model and we found that gender and age did not predict cannabis dependence after accounting for cannabis and tobacco use. Our results may also be a product of the common route of administration (Agrawal and Lynskey, 2009) where inhalation of one substance may sensitise an individual to the inhalation of another substance.