BACKGROUND-AIM:
The cumulative effects of repeated e-cigarette use and the long-term health consequences remain unclear. This study aimed to explore associations of aerosol components, systemic exposure and metabolic profiles of e-cig compounds using an integrated analytical, pharmacokinetic and computational strategy.

METHODS:
A controlled clinical experiment utilizing 30 healthy volunteers (15 e-cig users, 15 never-smokers) assessed exposure during a standardized 60 min vaping session which was divided into 3periods of vaping (with washout periods) at controlled device and puffing conditions. Blood samples were obtained at baseline, during exposure and at postexposure. Urine samples were collected at baseline and at post-exposure. Specific assessment of nicotine, cotinine, trans-3'-hydroxycotinine, and nornicotine was conducted using validated LC–MS/MS, while additional exposure-related compounds were studied utilizing LC–HRMS metabolomic screening. Pharmacokinetic characterization included non-compartmental (Cmax, Tmax, AUClast, t½) and population pharmacokinetic modelling. Covariate search evaluated demographic profile, refill liquid composition (PG/VG ratio, nicotine concentration), and vaping behavior parameters. Simultaneously machine learning and deep learning models were implemented to unveil latent patterns from vaping profiles. Participants underwent pulmonary function tests (PFTs) prior to exposure and after completion of the vaping session, including oscillometry.

RESULTS:
Early detections of nicotine and nicotine metabolites in blood and urine were tested in dense time points during exposure and at post-exposure. The nicotine concentration–time profiles indicated both a rapid increase during the vaping phase and a biexponential decline during the washout phase, indicative of inhalation pharmacokinetics. Initial pharmacokinetic modelling showed that nicotine kinetics can be well characterized by oneor two-compartment models with first-order absorption and elimination, whereas metabolite dynamics adhered to parent–metabolite template structures. The trans-3'-hydroxycotinine/cotinine ratio is known to model betweenpatient differences in nicotine metabolism. Having piloted data, we identified an optimum sparse sampling scheme for the pivotal study consisting of blood samples collected at 0, 10, 20, 40 and 60 min during exposure and +30 min, +2 h and +5 h for the study to allow for a reliable estimation of pharmacokinetic parameters with minimal sampling burden. Preliminary modeling results indicated that differences in pharmacokinetic parameters, based on vaping schedules, refill fluid compositions and individual metabolic, can vary across patients.

CONCLUSIONS:
This study presented a framework of integrated exposure analysis by chemical methods, clinical biomonitoring, pharmacokinetic modelling, and machine learning to characterize exposure to e-cigarette components.