Causal Deep Neural Network-Based Model for First-Line Hypertension Management.

Area of Science:

• Cardiology
• Artificial Intelligence
• Pharmacogenomics

Background:

• Hypertension affects millions globally, necessitating personalized treatment strategies.
• Current treatment selection relies on general guidelines, often leading to suboptimal outcomes.
• Predicting individual response to antihypertensive medications remains a challenge.

Purpose of the Study:

• To develop and validate a causal, deep neural network-based machine learning model.
• To predict the most successful antihypertensive treatment for individual patients.
• To enhance personalized hypertension management.

Main Methods:

• Trained a deep neural network on data from 16,917 newly diagnosed hypertensive patients.
• Included patients with primary hypertension, pre-treatment measurements, and at least 1 year of follow-up.
• Defined treatment success as blood pressure control without moderate/severe adverse effects.

Main Results:

• The model achieved a precision of 51.7%, recall of 44.4%, and F1 score of 47.8% in predicting individualized treatment success.
• Angiotensin-converting enzyme inhibitor-thiazide combination showed the highest average success (44.4%).
• The algorithm aligned with hypertension guidelines 95.7% of the time on the validation set.

Conclusions:

• Machine learning can accurately predict antihypertensive treatment success.
• This AI-driven approach facilitates personalized hypertension management.
• The model shows potential to optimize treatment selection and improve patient care.