Comparing Risk Profiles of Nicotine Pouch and Smokeless Tobacco Use in U.S. Adolescents: A Machine Learning Analysis.

Background: This study identified and compared the risk profiles associated with nicotine pouch (NP) and smokeless tobacco (ST) use among adolescents (12-17 years). Methods: We used publicly available data from the most recent wave (January 2022-April 2023) of the Population Assessment of Tobacco and Health (PATH) Study (N = 10,604 [48.9% female, 51.2% 12-14 years]). A penalized logistic regression framework was used to promote parsimony and reduce overfitting when separately predicting ever use of NPs and STs. Each model included over 70 variables across conceptual domains and identified the top 10 predictors based on penalized coefficients. Results: Overall, 0.9% (n = 105) and 1.4% (n = 168) reported ever-use of NPs and STs, respectively. NP and ST use were the strongest reciprocal predictors of each other, and both models shared key correlates, including other tobacco and substance use behaviors. Exposure to NP-related social media content was influential in the NP model, but no analogous variable appeared in the ST model. Model performance was strong-area under the receiver operating characteristic curve (AUROC) values were 0.946 (NP) and 0.890 (ST). Conclusions: Given that NP-related social media exposure emerged as a key correlate of adolescent NP use, enhanced monitoring of social media platforms, stricter enforcement of digital advertising restrictions, and the development of counter-marketing initiatives may be warranted.