The relationship between craving and tobacco use behavior in laboratory studies: a meta-analysis

Published laboratory studies from the last 50 years that included measures of craving and tobacco-consumption or tobacco-seeking measures were included in a meta-analysis in order to assess the relationship between craving and tobacco use. Seeking measures were further subdivided into those that reflected control by nonautomatic and automatic cognitive processes. Of 2,498 articles identified by the initial literature review, 204 analyses from 50 studies were deemed eligible. Overall, the relationship between craving and outcome behaviors was modest (r = .20, p < .001). Studies that imposed abstinence during data collection showed a stronger relationship between craving and outcome (r = .24, p < .001) than studies that did not (r = .18, p < .001). Further, of those studies that reported dependence, the overall association between craving and outcome was stronger for smokers who were less dependent. Separate meta-analyses revealed that the type of outcome measure moderated the omnibus effect, with the relationship between craving and nonautomatic seeking measures (r = .34, p < .001) being stronger than the relationship between craving and automatic seeking/consumption measures (both rs = 0.15, p < .001). These findings suggest that craving may play a role in, but does not fully account for, tobacco-use behaviors; furthermore, the extent to which craving predicts behavior may be increased when the behavior is under nonautomatic cognitive control.