This study's main findings can be summarized as follows. Among adults in the USA, about half of the cannabis smokers have used other illegal or internationally regulated drugs, whereas this was somewhat less common in Colombia and Mexico, where 'cannabis only' use was more common. We also found that the cannabis problems were experienced rather infrequently by the CO users, which followed a similar pattern for all three countries. However, there was an exception to that general rule: CO users in Colombia were more likely to experience cannabis-related social and legal problems as compared to CO users in the USA general population. Regrettably, one limitation of this study involves data on cannabis-related 'legal problems,' which were assessed in the NLAAS, but the NLAAS principal investigators did not release these values due to concern about inadvertent violation of confidentiality and incomplete privacy protection (e.g., if the survey data were matched to publicly available data on cannabis offenses). For this reason, we were unable to compare Colombia's estimates with those of the US Hispanic subgroups surveyed for the NLAAS, which would have been especially interesting, given observed greater occurrence of cannabis-associated legal problems in Colombia.
Another finding of potential interest involves recurrent hazard-laden cannabis smoking, and the observation that USA Latino CO users tended to engaged in recurrent hazard-laden use more often than those in Mexico. To the best of our knowledge, there is neither strong theory nor prior evidence on recurrent hazard-laden use from prior cross-national studies of this type, which might have been used to establish a set of Bayesian priors for this contrast, or to calibrate statistical power for a more complete balance of Type I and Type II error. In this context, readers may wish to note that p = 0.051 is not too distant from the conventional standard of 0.05. A slightly larger sample might well have produced p < 0.05 with an effect estimate of this size; this is a contrast that may deserve future investigation and confirmation in future epidemiological studies.
Before detailed discussion, several other limitations deserve to be mentioned. Despite the size of the overall samples in each place, there were relatively small numbers of CO users who had experienced cannabis problems, according to the CIDI-WMH assessment. In consequence, no more than a handful of covariate terms could be introduced in the statistical models; statistical power and precision also were limited. Moreover, there were too few CO users in the samples to permit a focused study on the recent-onset cannabis smokers as has been done elsewhere . This constraint also thwarted any detailed probing of male-female differences, or research on issues of migration and acculturation, as one might wish to examine in relation to the experience of specific subgroups of Spanish-speakers in these samples (e.g., Mexico, Colombia, or other country of origin for the USA Latinos). In addition, the self-report character of the assessment of cannabis-related problems can be expected to introduce some measurement error, which we suspect might be manifest as 'under-reporting' of the clinical features if not cannabis smoking per se, as discussed elsewhere [4, 11].
One other possible limitation has to do with design issues as described in our methods section, especially the CIDI-WMH focus on 'clinically significant' cannabis problems. Readers interested in more details about this issue are referred to original work by Narrow and colleagues  and to more recent contributions by Degenhardt and colleagues [10–12], in which it has been found that this particular CIDI-WMH approach seems to have had no appreciable influence when the task is to estimate the occurrence of DSM-IV cannabis dependence and related problems.
A last limitation is the narrow range of cannabis experiences covered in the standardized diagnostic assessments. In future global health research on cannabis-associated experiences, it should be possible to extend the coverage of cannabis experiences in two directions: (a) toward positive and possible health-enhancing experiences associated with cannabis consumption (e.g., of the type discussed by Griffiths et al.(2006) , as well as appetite-promoting or nausea-reducing effects discussed in the context of medicinal use of cannabis products); and (b) a more complete coverage of craving and the obsession-like or compulsion-like experiences associated with cannabis dependence syndromes. The global prevalence of cannabis smoking provides ample justification for this type of focused inquiry.
Despite caveats such as these, this study has the strength of generally comparable data gathering with a standard design and implementation protocol across countries as part of the WMHS initiative. Each population sample in the three countries was obtained through similar multi-stage sampling methods with good to excellent participation levels. Each participant completed standardized survey assessments of high quality.
Against this background of limitations and strengths, this study shows that CO users represent in general a substantial majority of all cannabis users in the studied countries, which justify a focused look at these CO users with respect to clinical features and problems associated with the drug disorders. Separately, and with evidence from more countries, WMHS collaborators are following up this investigation with a look at all cannabis users, and are estimating the influence of polydrug use on the occurrence of these clinical features, but these analyses introduce a good bit of complexity in the cross-national research context, as there is a good deal of heterogeneity in the profile of internationally regulated drugs used, plus large between-country variations in the police responses to the different drug compounds.
Also, this study found that as a general rule the estimated experience of CO users with respect to 'clinically significant' cannabis problems did not vary appreciably across the samples, even though there is a wide variation in the occurrence of cannabis use across these national boundaries within the Americas. The exception to the rule was found in the Colombia-USA contrast, with an excess occurrence of cannabis-related legal and social problems in Colombia. We speculate that the observed variation might be traced to differences in the context in which cannabis is consumed in each country, such as whether alcohol is being consumed concurrently. Concurrent use of cannabis with alcohol has been described in the epidemiological literature as the most common pattern of polydrug use, and has been related to excess occurrence of social and psychological consequences [6, 22], and also to behavioral patterns that might promote adverse consequences, such as attending parties more often [6, 18]. Our use of AUD-matching and the exact conditional logistic regression was motivated by an effort to probe into this possibility, but AUD does not encompass the concept of concurrent or simultaneous alcohol-cannabis use, so this issue must be left for future investigations, with deliberate assessment of concurrency and simultaneity of alcohol and other drug use, as recommended elsewhere .
When introducing our study in this paper's 'Background' section, we also mentioned the possibility that the THC content of smoked cannabis might vary considerably across countries. We also mentioned other cross-country and within-country variations in cannabis ingestion practices. For example, our colleagues from western Europe, Turkey, and Egypt were surprised to learn that in the Americas and in New Zealand, the cannabis typically is smoked by itself, and not in cannabis-tobacco formulations (with the exception of American 'blunts' - namely, hollowed out tobacco cigars into which the cannabis is inserted and smoked). Our colleagues from India were surprised at the narrow range of cannabis preparations generally available in the Americas, and brought to our attention many different teas, cakes, and other formulations for cannabis in their country. The World Mental Health Surveys were not focused specifically upon cannabis, and gathered essentially no information about these details of cannabis consumption. In future cross-national research focused specifically upon cannabis consumption, it may be possible to study cannabis-related experiences with more careful probing into possibilities that different cannabis experiences depend upon these cannabis ingestion practices.
The evidence of this study also draws attention to some unanswered questions about the experience of Latinos living in the USA. In the WMH research, with similar survey methods used for all places, it was found that 30% of US Latinos had smoked cannabis (unpublished National Latino and Asian American Study, NLAAS, estimate derived for this study report), which is a value about halfway between what was found for Colombia and Mexico (10-11% and 7-8%, respectively) and the USA (42-43%) ; nonetheless, there was no appreciable variation in the experience of clinically significant cannabis problems in the USA general population versus US Latino contrast. In future research with larger sample sizes and a focus on these research issues, it should be possible to examine these interesting cannabis patterns in relation to (a) birth cohort variations, (b) time elapsed since migration into the USA (including the USA-born Latinos), other issues of acculturation, assimilation, and adaptation, as has been done in relation to the epidemiology of cardiovascular diseases [24, 25] and cancer [26, 27]. Indeed, Borges and colleagues studied use of alcohol and other psychoactive drugs among residents of Mexico who had been in the USA at least once and then had returned to Mexico, discovering that this subgroup had greater drug involvement when compared to the subgroup of Mexicans without USA experience stratified or holding constant whether they had close family members who had migrated to the USA .