Despite the availability of good-quality data and robust models, the idea of a gateway is riddled with a number of theoretical problems. When investigating possible gateways between substances, the first objection would be that there is no consensus on the definition of «soft» and «hard» drugs. For example, many would disagree with the categorisation of cannabis as «harder» than alcohol (23). Similarly, it is unclear what should be seen as the initial gateway substance, and this leads to an infinite regression. If alcohol and tobacco smoking both are gateways, why not sugar?
A related limitation is that people do not necessarily start using «soft» drugs before turning to «hard» ones, but that the order of use reflects local patterns of use and the local market situation (24). In addition, if a substantial gateway effect exists, we should find a positive correlation between the use of different substances over time. Often, this is not the case. For example, we find that the use of cocaine among young people increased relatively strongly in the late 1990s, while the prevalence of smoking decreased (25). This does not lend support to the idea of predisposition.
Furthermore, the arrow may point in the opposite direction. For example, the increasing use of snus in Norway in recent decades has been accompanied by a marked decrease in tobacco smoking. If there is an association between these two, it appears more likely that snus displaces tobacco smoking than that snus helps recruit tobacco smokers. Nor has the prevalence of snus helped resurrect the use of old-fashioned dry snuff, to provide a historical example.
As regards the idea of biochemical predisposition, this may in theory explain why some people become addicted to a new addictive substance, but it cannot explain why somebody chose to try the substance in the first place. Naturally, this is not a problem in studies of mice.
Nor can predisposition explain a possible transition between different nicotine and tobacco products, since the intoxicant in any case is nicotine. To the extent that nicotine increases the tolerance and urge for more nicotine, users of snus or e-cigarettes/nicotine vaporisers can simply increase their nicotine intake by using snus or e-cigarettes/nicotine vaporisers, without having to resort to tobacco smoking. A possible causal association, if any, must be wholly or partly attributable to other conditions.
Perhaps the strongest objection is that in studies that investigate the use of «hard» drugs, the observed correlation is a direct consequence of examining a selected group, i.e. persons who have used «hard» drugs. If we instead investigate the proportion of users of «hard» drugs among those who at any point in time have used «soft» drugs, the association will naturally be negligible (26). Similarly, this will be a problem if we investigate the effect of snus or e-cigarettes/nicotine vaporisers on smoking.
Does this mean that there can be no possible causal associations between different intoxicants and nicotine and tobacco products? Obviously not. My concern is that in the absence of unambiguous empirical observations, such associations must be based on sound reasons. This is also required in order to adequately model any causal associations. To date, no convincing explanations exist of a causal association between the use of snus and e-cigarettes and smoking. In this respect, the notion of a «gateway» is redundant.