Classification and Change in the Development of Alcohol, Nicotine, and Cannabis Use: A Systematic Review

Background. Substance use is associated with significant public health and economic consequences and a rich literature exists on the development of substance use. However, no systematic review has attempted to synthesize this literature across alcohol, nicotine, and cannabis, nor attempted to characterize the wide variety of samples and methodology employed in modeling change in use over time. Methods. We conducted a systematic literature review to evaluate the samples, measures, and analytical approaches in publications reporting random effects or mixture modeling in longitudinal substance use across the lifespan. Results. Across 317 included publications, the mean number of assessment waves was 5.5 (range=2-52), spanning an average of 7.3 years from baseline (range=1-38). Almost all publications included some coverage between ages 14-30, few studied mid-adulthood or older age. Most publications (72%) evaluated only one substance and 14% of publications evaluated all three. The average number of classes found via mixture modeling was four (range=2-12). Continuous random effects models of change primarily returned linear or quadratic solutions as best fitting. We found no significant differences between substances for number of classes (hybrid p=.50, class p=.13), nor curve shape identified (curve p=.98, hybrid p=.95).Discussion. No one statistical method was predominant in the literature, nor was there evidence of growing consensus on preferred model as a function of publication date. Studies of middle and older age were largely absent from the literature. We offer recommendations in reporting to assist in the replication, extension, and interpretation of this body of work.