The Continuing Conundrum of Craving

Every few years some addiction researchers rediscover the irritating fact that craving simply does not operate in the way it is supposed to—i.e. measures of craving and drug use are often not associated strongly. Reactions to the aggravating realities of craving have been diverse. Some people simply ignore any bothersome data and proceed as if craving and drug use are parallel manifestations of the same drug-motivational process. Others interpret an absence of a relationship between a particular craving assessment and an isolated measure of drug use as license to question whether craving is at all relevant to our understanding of addictive processes. A third view is to embrace the mysteries of drug craving and to pursue research and theory that might explain the mercurial relationships between self-reported desire and drug-use behaviors. We believe this last approach is the most scientifically fruitful and realistic: fruitful, because a comprehensive science of addiction will have to account for both craving and drug use; realistic, because craving, despite earnest attempts to wish it away, will not disappear from the researcher's or clinician's ledger. People addicted to drugs will continue to agonize over craving, and they will demand that scientists acknowledge its intrusive, nocuous qualities. Perkins [1] argues that a form of craving—cue-specific craving—is suspect because there is scant evidence that this craving predicts relapse and that relapse is the most important feature of addiction. He recommends that we conduct further research to establish the predictive validity of cue-specific craving vis-à-vis relapse. What Perkins does not make clear is what we should do if we continue to find that cue-specific craving displays little or no relationship with relapse. Presumably, under that scenario, he might recommend that we turn our back on cue–reactivity studies, particularly those that focus on craving. As noted by Perkins, associations between craving and drug use are often not strong—in fact, one of us has been making that point for many years (e.g. [2]). We also agree that the field would benefit from much more programmatic research on the validity of cue–reactivity measures, although we believe that it would be sensible to select criterion measures that are more broadly representative of the drug-addiction construct than the narrow domain suggested by Perkins. Equally importantly, we need to examine the network of relationships across the panoply of predictor variables that can be assessed in the cue–reactivity paradigm. That is, cue–reactivity manipulations influence much more than just craving—drug-cue manipulations influence self-report, physiological and behavioral measures representing emotional, cognitive and motivational constructs. The possibility that just one of these variables (e.g. laboratory-based cue-induced craving) might display a weak association with one feature of addiction (e.g. relapse) says little about the overall validity of the cue–reactivity paradigm as a model for studying addictive processes. Indeed, it is hard for us to imagine a mature science of addiction that does not acknowledge the important role of drug-related cues in ongoing drug use and relapse. Regardless of the outcome, it would be extremely enlightening if we had clearer understanding of the associations between cue-specific craving and drug relapse but, for this research to be maximally informative, it must address a large number of critical psychometric and methodological issues. These include consideration of the reliability, stability, sensitivity, specificity, distribution and dimensionality of both the predictor and criterion variables; the content validity (i.e. representativeness) of the evocative cues; the structure of the cue–reactivity procedure (e.g. the optimum number of cue trials, the spacing of trials, the extent to which drug use is possible in the presence of cues, the suitability of the control-cue conditions); the mode of cue presentation (e.g. in vivo, photographs, imagery, virtual reality); the context of the cue–reactivity assessment (e.g. laboratory-based, clinic-based, in the natural environment); the timing of the assessment relative to a quit attempt (i.e. whether the assessment trials are conducted before, during or after a quit attempt); the temporal relationships between cue–reactivity assessment and relapse episode; the nature of the treatment intervention (if any); and key characteristics of the research sample (e.g. abstinence status of the participants, history of drug use, level of drug addiction, stage of change) (see [3,4], for a review of these issues). Any of these factors can have a profound influence on cue-specific craving and/or relapse with a corresponding impact on the strength of the associations across these variables. Indeed, the question of whether or not craving (in any of its manifestations) is correlated with relapse strikes us as unnecessarily bounded. A more penetrating exploration of the associations between craving and drug use would take into account the potential moderating and/or mediating effect of the factors listed above. Accordingly, generating meaningful answers to these questions will be neither simple nor attainable in a single study or even a handful of studies. Nevertheless, mapping the intricacy of the relationships between cue–reactivity variables (including craving) and other features of drug addiction will certainly teach us a great deal about the fundamental nature of addictive processes. No doubt many of the answers uncovered from this research program will surprise us. We hope those surprises will not deter addictions researchers from continuing to seek the truth about the full complexity of addictive disorders. S. Tiffany is supported by an NIH Grant R01 CA120412. None.