Important steps to resolve addiction vs. exposure dependent alterations in cannabis users include the characterization of the neural, behavioral, polygenic risk markers, and epigenetics (Sherva et al., 2016; Walters and Owen, 2016) that identify (i) which recreational cannabis users or cannabis naïve individuals are vulnerable to develop a CUD; (ii) the recreational non-problem cannabis users en route to develop a severe CUD; (iii) which users have persistent severe CUDs (Van der Pol et al., 2013a); (iv) which users transit to a lower severity CUD/non-problem cannabis use/abstinence; and (v) sex differences in the development of CUDs, as females are underrepresented in the existing literature, while presenting a distinct profile from males (e.g., faster transition to dependence, more treatment resistant; Lorenzetti and Carter, 2015).

Twin studies are required to address the above research question (i) and disentangle epigenetic factors that confer risk to develop CUDs in some but not others (Gillespie et al., 2009). Large-scale longitudinal studies are warranted to address research questions (i)–(v) and track young adolescents yet naïve to cannabis through to late adulthood. Such studies should combine diagnostic cutoffs that inform on the clinical significance of cannabis use (i.e., DSM 5, ICD-10); self-reported detailed information on cannabis use (daily/almost daily use, duration, and age of onset); objective (i.e., biological specimens) quantification of cannabinoids (Lorenzetti et al., 2016); cognitive and brain anatomy/function (using MRI); and recruitment approaches that maximize sample representativeness (e.g., the general community, coffee shops; Van der Pol et al., 2011). Notable examples include the Netherlands XTC Toxicity (NeXT) study (De Win et al., 2005), the USA ABCD study http://addictionresearch.nih.gov/abcd-study and the Dutch Cannabis Dependence Study (CanDep; Van der Pol et al., 2011).

As longitudinal studies are very expensive and time consuming, we can address relevant research questions (i)–(vi) in a timely fashion, via re-examination, online data sharing, and follow up from already collected neuroscientific datasets on cannabis using cohorts with varying levels of problems with use. Problem vs. recreational users may be segregated (to enable their comparison) using information available from each study's instruments, on key addiction phenotypes (e.g., craving, withdrawal, difficulties in controlling use, persistent use despite harmful consequences on mental health, and socio-occupational functioning, higher priority given to drug use than to other activities and obligations; APA., 2013; WHO, 2016). A notable example of this includes the internationally coordinated initiative ENIGMA Addiction Working Group (Mackey et al., 2016).

Finally, we propose the development of an agreed-upon instrument to objectively assess key features of cannabis addiction vs. non-problem use in neuroscientific settings, as most measures of cannabis related problems are borrowed from those used for other substances, the features of which may not fit those relevant to cannabis dependence (Lorenzetti et al., 2016). To this end, a Delphi review (Dalkey and Helmer, 1963) with world-class experts in the clinical/neuroscientific aspects of cannabis and substance dependence may be conducted and may identify additional hot topics in CUDs.

In conclusion, we call for a greater, systematic and coordinated research effort internationally, to identify pathways in and out of CUDs and fill the existing gap between the limited knowledge on CUDs and the increasing availability of cannabis products.