The clinical relevance of smoking cue-induced craving has recently been called into question . Sayette & Tiffany  argue that this is due to methodological and theoretical limitations of current cue-exposure models of craving, and present initial evidence for an alternative ‘peak provoked craving’ paradigm, which incorporates both cue-induced and abstinence-induced cigarette cravings. While any potential methodological advance deserves careful consideration, here we suggest a number of reasons why this approach alone may not be sufficient to address the poor predictive validity of current laboratory measures, including peak provoked craving.
We have argued previously that some of the difficulty with cue-induced craving research is the reliance on acutely abstinent smokers who are not undergoing a quit attempt . Participants are often smokers not interested in quitting, and therefore withdrawal modelled in these participants is unlikely to reflect the withdrawal state of a quit attempt. In particular, as Sayette & Tiffany emphasize, cravings fluctuate considerably over the course of a quit attempt in response to various extrinsic and intrinsic factors, a point we return to below. Laboratory assessments cannot capture this. Moreover, the perceived availability of cigarettes is known to modulate craving and response to various laboratory challenges and tasks . It is plausible that there will be considerable variability in perceived availability across participants within a laboratory study; for example, if it is not clear how long the session will last or whether they will be allowed to smoke within a certain time-frame. This will also contribute to substantial inter-individual variability in measures taken in this context. We appreciate that Sayette & Tiffany are not suggesting that all studies should take place within a laboratory setting, or that peak provoked craving can be applied only in this context. Nevertheless, given that much cue-induced craving research takes place within a laboratory setting, the use of peak provoked craving models alone will not address this problem.
It is also worth noting that there is a long history of novel methods of assessing craving (or closely related constructs) which show initial promise only to ultimately prove disappointing. One example is the use of biobehavioural measures intended to capture cognitive biases associated with substance use, which are predicted by various models of addiction [5, 6]. Early studies suggested that these measures of cognitive bias might be predictive of relapse following a quit attempt , but these do not appear to have been replicated, and certainly have not led to a dramatic improvement in our understanding of the determinants of relapse. Critically, it is now apparent that these measures may have poor internal reliability , although it may be possible to improve this somewhat [9, 10]. It is therefore perhaps unsurprising that they correlate only minimally with self-report measures of craving , despite both ostensibly capturing a common underlying construct. What this illustrates is that ostensibly objective measures of psychological phenomena which have good face validity may lack basic psychometric properties such as acceptable internal reliability.
Recent animal studies have suggested that the drug-seeking response to drug-related cues increases with prolonged drug abstinence [12, 13], and human studies have indicated similarly that cue-induced cigarette craving increases across the time–course of abstinence . This suggests a complex and dynamic relationship between cue-induced craving and abstinence, including potentially abstinence-induced craving. How can this relationship be captured adequately? It is unlikely that laboratory studies will ever be able to do this. While the peak provoked craving method, by capturing both cue-induced and abstinence-induced components of craving, may offer a partial solution, a full understanding of the interplay between these over time, and how they relate to clinically relevant outcomes such as relapse to smoking, will be achieved only through the development of ambulatory monitoring techniques. As smartphones and tablet computers decrease in cost and increase in functionality, the scope to incorporate location data, self-report data and biobehavioural data, all captured in a naturalistic environment with high temporal resolution over the course of an extended quit attempt, increases substantially. These techniques are increasingly sophisticated and developing rapidly , and already beginning to provide important insights .