This study showed that heavy cannabis users, but not controls, have an approach-bias specifically for cannabis-related images (not for neutral images), as measured with the AAT [9
]. In line with our hypothesis, the approach-bias predicted changes in cannabis use 6 months later in heavy cannabis users: stronger approach-biases were related to increases in weekly cannabis use. In contrast to our hypothesis, the approach-bias was related negatively to craving for relief from negative affect and anticipation of positive outcome (i.e. MCQ emotionality and expectancy factor). No associations were found between the approach-bias and measures of cannabis-related problems and dependence.
The approach-bias found here in heavy cannabis users supports the idea that an approach-bias for substance-related stimuli is a common phenomenon in cannabis users as well as in alcohol users and tobacco smokers [9
]. The most important finding is that the approach-bias predicted changes in cannabis use after 6 months. Heavy cannabis users with stronger approach-biases were more likely to increase weekly cannabis use, while lower approach (or even avoid) biases were related to decreases in use. To our knowledge, this study is the first to find a prospective predictive relation between an approach-bias and course of drug use. This predictive relationship may have clinical implications. Even after prolonged drug use, some heavy using individuals develop abuse and dependence while others do not. The approach-bias could be a predictor of the course of drug use. It might be used for identifying individuals especially at risk for increasing cannabis use for targeted interventions. An advantage of using implicit measures such as the AAT is that they do not rely upon self-report. Insight into severity of drug dependence and self-awareness might be compromised in dependent individuals, thereby influencing the reliability of self-reports [46
]. A second clinical implication could be using a modified AAT to retrain heavy cannabis users to avoid cannabis. Recent studies showed that approach action tendencies in heavy alcohol drinkers and alcohol-dependent patients can be modified [30
]. Successful training to avoid alcohol was related to decreased subsequent alcohol use and improved treatment outcome. Future research is needed to verify this in heavy cannabis users and clinical cannabis users, as has been shown recently for alcohol-dependent patients [47
In contrast to our hypothesis, no associations were found between approach-bias and changes in measures of cannabis dependence. This could be due to methodological issues. Inherent to the sample, cannabis-related problems were relatively low and a 6-month follow-up might have been too short to detect changes in measures of dependence. An alternative explanation is that cognitive biases such as the approach-bias mainly play a role in the course of drug use in the earlier
stages of addiction. The approach-bias may predict who will use more, but not who will progress to problematic drug use. This appears to disagree with the incentive sensitization theory of addiction [48
], and seems more in line with theories where incentive sensitization is mainly important during escalation of drug use and less when subsequent compulsive drug use progresses [50
]. To test this hypothesis, associations between approach-bias and prospective cannabis use needs to be assessed in larger samples of dependent, heavy and sporadic cannabis users compared to non-using controls.
Also in contrast to our hypothesis, the approach-bias was associated negatively with pre-test levels of craving for relief from negative affect and craving for anticipation of positive outcome. Post-test craving was not associated with the approach-bias. Compulsive craving predicted cannabis use after 6 months: higher craving was related to increased use. However, the effect disappeared when the approach-bias entered the regression model. Most theories predict a bidirectional positive association between approach-biases and craving [6
]. A recent meta-analysis showed weak positive relations between craving and attentional bias for alcohol [52
]. In cannabis users a positive relation between post-test craving and attentional bias has been reported [25
], although no relationships between post-test craving, attentional bias and approach-bias were found in a different study [26
]. Clearly, more research is needed to assess relationships between cognitive biases and craving. Further, the MCQ can differentiate reliably between craving factors [42
], but a theoretical framework should be developed further, which is beyond the scope of this paper. However, our findings emphasize the relevance of measuring both pre- and post-test craving and using factorial decomposition of self-reported craving.
Finally, some limitations must be taken into account. First, there were more tobacco smokers among heavy cannabis users compared to controls, and almost all cannabis users smoked cannabis cigarettes combined with tobacco (most common use-form in the Netherlands [53
]). Tobacco might increase the effects of cannabis [54
], and the resemblance between tobacco and cannabis cigarettes possibly activates approach actions towards tobacco in tobacco users. Neither in heavy cannabis users nor in controls was tobacco use associated with the approach-bias. However, our sample prevents discrimination between cannabis and tobacco effects. Secondly, in the course towards dependence, increased sensitivity to general rewards might precede incentive salience of drugs over natural rewards [5
]. Indeed, it has been reported that heavy drinking male carriers of the OPRM1 G-allele had an approach-bias for both alcohol and other appetitive stimuli [9
]. However, with the present design it cannot be determined if the approach-bias in heavy cannabis users generalizes to other rewarding stimuli. Thirdly, the results should be interpreted bearing in mind that the approach-bias reflects the relative difference between approaching and avoiding cannabis images. Although the group comparison suggests that strong approach tendencies for cannabis, rather than weak avoid tendencies for cannabis, predict changes in prospective cannabis use in heavy users, the present findings with a relative measure are not conclusive regarding this issue. Alternatively, the interplay or conflict between approach and avoid tendencies may predict changes in cannabis use. This is an important question that needs to be addressed in future research. Fourthly, the AAT is a relatively new measure and its temporal stability is unknown. Finally, the absence of a relation between approach-bias and baseline levels of cannabis use might suggest a limitation in the construct validity of the task.
In conclusion, heavy cannabis-smoking young adults automatically activate approach action tendencies in response to cannabis-related stimuli (approach-bias), and the extent to which they do so predicts further escalation of their use.