Abstract
INTRODUCTION: Hypertension is a significant public health concern. Several relevant risk factors have been identified. However, since it is a complex condition with broad variability and strong dependence on environmental and lifestyle factors, current risk factors only account for a fraction of the observed prevalence. This study aims to investigate the emerging early-onset hypertension risk factors using a data-driven approach by implementing machine learning models within a well-established cohort in Mexico City, comprising initially 2,500 healthy adults aged 18 to 50 years. METHODS: Hypertensive individuals were newly diagnosed during 6,000 person-years, and normotensive individuals were those who, during the same time, remained without exceeding 140 mm Hg in systolic blood pressure and/or diastolic blood pressure of 90 mm Hg. Data on sociodemographic, lifestyle, anthropometric, clinical, and biochemical variables were collected through standardized questionnaires as well as clinical and laboratory assessments. Extreme Gradient Boosting (XGBoost), Logistic Regression (LG) and Support Vector Machines (SVM) were employed to evaluate the relationship between these factors and hypertension risk. RESULTS: The Random Forest (RF) Importance Percent was calculated to assess the structural relevance of each variable in the model, while Shapley Additive Explanations (SHAP) analysis quantified both the average impact and direction of each feature on individual predictions. Additionally, odds ratios were calculated to express the size and direction of influence for each variable, and a sex-stratified analysis was conducted to identify any gender-specific risk factors. DISCUSSION: This nested study provides evidence that sleep disorders, a sedentary lifestyle, consumption of high-fat foods, and energy drinks are potentially modifiable risk factors for hypertension in a Mexico City cohort of young and relatively healthy adults. These findings underscore the importance of addressing these factors in hypertension prevention and management strategies.