SUMMARYPhilip Morris International has developed a heat-not-burn tobacco heating system (THS 2.2) that produces an aerosol without combustion. Adult smokers are anticipated to use the product with differing behaviors, such as puffing volume or puffing frequency, therefore it was important to find an easy way to study how users are exposed to the aerosol constituents. Thus, the intended outcome of this study was to propose and assess a simple approach for the estimation of THS users' exposure to harmful and potentially harmful constituents (HPHCs). THS operates using tobacco sticks (HeatSticks) that include a mouthpiece and a tobacco plug which, when heated, generates an aerosol. The analysis of nicotine retained in the mouthpiece of the HeatSticks during use was identified as a potential approach to estimate users' mouth level exposure (MLE) to HPHCs. Consequently, the following study was conducted with the objectives 1.) to assess the correlation between the quantity of retained nicotine in the mouthpiece (Nicotine MP) of the HeatSticks and the nicotine delivered in the aerosol of machine-smoked products, 2.) to verify the practical range for Nicotine MP based on the analysis of used HeatSticks left by THS users, and 3.) to develop models describing the relationship between Nicotine MP and specific aerosol constituents measured in the aerosol of machine-smoked products. The regular non-mentholated HeatSticks variant was machine-smoked under various smoking regimens to cover the range of anticipated human puffing behaviors. The suitability of this practical range of machine-smoking conditions was verified by collecting used HeatSticks from two different trials conducted with THS users. The determined Nicotine MP distribution indicated that the machinesmoked regimens encompassed the range observed for users. Multiple Linear Regression (MLR) combined with a stepwise approach was used for selecting models describing the relationship between Nicotine MP and specific aerosol constituents. The stepwise approach interactively explores which amongst various tested predictors provides a good fit. The developed models showed good adjusted coefficients of determination (i.e., R 2 adj. $ 0.75) for 28 out of the 43 investigated HPHCs. Previously published studies showed that actual MLE can be estimated from cigarette filter analysis. This study demonstrated that the analysis of nicotine in THS mouthpiece (filter section) corresponded to an estimation of the upper limits of MLE, in line with maximum possible usage conditions. [Beitr. Tabakforsch. Int. 27 (2017) 42-64]