Predicting nicotine emissions and plasma nicotine boost in E-cigarette users using machine learning.

Area of Science:

• Computational toxicology and pharmacology
• Data science in public health

Background:

• Global rise in e-cigarette use necessitates accurate nicotine exposure quantification.
• Existing methods for measuring nicotine exposure are often complex and not scalable.
• Puff topography and device characteristics are key determinants of nicotine delivery.

Purpose of the Study:

• To develop and evaluate supervised machine learning (ML) models for predicting nicotine emissions and plasma nicotine boost.
• To utilize puff topography, user characteristics, and e-liquid consumption as input features.
• To identify the most effective ML algorithms for this predictive task.

Main Methods:

• Collected 259 measurements from adult e-cigarette users (21-35 years).
• Combined human puff topography data with laboratory toxicant emission tests using a puffing robot.
• Trained six regression models (e.g., XGBoost, Random Forest, Neural Networks) and evaluated using R², RMSE, and MAE with bootstrap resampling.

Main Results:

• XGBoost model demonstrated superior performance in predicting both nicotine emissions and plasma nicotine boost.
• Nicotine emission prediction accuracy was high using puff number (R² = 0.778) or liquid consumption (R² = 0.747) combined with puff duration and device features.
• Plasma nicotine boost models using XGBoost showed strong explanatory power based on puffing parameters (R² = 0.613) and liquid consumption (R² = 0.699).

Conclusions:

• Supervised machine learning models, particularly XGBoost, can effectively estimate nicotine emissions and plasma nicotine boost from e-cigarette usage patterns.
• These ML models offer a scalable and reliable approach for quantifying nicotine exposure.
• The findings support regulatory efforts, addiction risk assessment, and public health surveillance related to e-cigarette use.