Labeled nicotine strength is widely used as a proxy for oral nicotine pouch performance and abuse liability; however, Controlled Use pharmacokinetic (PK) data show that labeled strength alone does not consistently predict nicotine delivery. This disconnect raises an important question for product developers and regulators: which product attributes are most strongly associated with Cmax, Tmax, and AUC under Controlled Use conditions? To address this question, we applied machine-learning methods to a dataset of Controlled Use PK studies across oral nicotine products. Product-level inputs included labeled and measured nicotine content, pH,nicotine form (free base vs. polacrilex), estimated unprotonated nicotine fraction, matrix composition, moisture content, in vitro nicotine release behavior, pouch mass, and physical design characteristics expected to influence hydration, diffusion, and oral residence time. Feature-importance and model-comparison analyses were used to evaluate the relative contribution of these attributes to observed PK outcomes. Across analyses, nicotine delivery was more strongly associated with combinations of formulation and design attributes than with labeled nicotine strength alone. Attributes related to nicotine form and availability, including pH, estimated unprotonated nicotine fraction, moisture content, release kinetics, and matrix composition, contributed meaningfully to prediction of Cmax and AUC, while features influencing hydration, diffusion, and oral residence time were relevant to Tmax. These findings support a mechanistic interpretation of nicotine pouch PK, in which delivery reflects the combined effects of chemical form, matrix-controlled release, and product design. This analysis provides a quantitative framework for identifying the product attributes most likely to influence Controlled Use PK outcomes and supports more rational formulation screening, study design, and bridging across related nicotine pouch products.

TSRC 2026