Answering Clinical Questions Using Machine Learning: Should We Look at Diastolic Blood Pressure When Tailoring Blood...

Area of Science:

• Cardiovascular Medicine
• Biostatistics
• Machine Learning in Healthcare

Background:

• Current guidelines advocate for intensive blood pressure control, primarily focusing on systolic blood pressure (SBP) reduction.
• The safety and prognostic significance of reducing diastolic blood pressure (DBP) remain debated, with conflicting evidence from various studies.
• Previous research suggests low DBP should not impede achieving SBP targets, but further investigation is warranted.

Purpose of the Study:

• To investigate the independent predictive value of diastolic blood pressure (DBP) for major adverse cardiovascular outcomes using machine learning.
• To determine if DBP is a significant risk factor for predicting stroke, heart failure (HF), myocardial infarction (MI), and the primary composite outcome in the SPRINT trial.
• To assess the necessity of including DBP in risk prediction models for intensive blood pressure management.

Main Methods:

• Machine learning algorithms including decision trees, random forests, k-nearest neighbors, naive Bayes, multi-layer perceptrons, and logistic regression were employed.
• Models were trained and evaluated with and without DBP as a predictor variable in both the overall SPRINT population and a subgroup with DBP < 70 mmHg.
• Model performance was assessed using metrics such as accuracy, Area Under the Curve (AUC), and F-measure.

Main Results:

• The inclusion of DBP as a risk factor did not significantly improve the predictive performance of the machine learning models.
• Models achieved comparable accuracy, AUC, and F-measure scores whether DBP was included or excluded.
• DBP was found to be unnecessary for accurate prediction of stroke, MI, HF, and the primary outcome in the analyzed SPRINT trial data.

Conclusions:

• Machine learning analysis of the SPRINT trial data indicates that DBP should not be considered an independent risk factor in the context of intensive blood pressure control.
• These findings suggest that clinicians can focus on achieving systolic blood pressure targets without DBP being a primary concern for adverse outcomes.
• The study supports prioritizing SBP management over DBP thresholds when intensifying antihypertensive therapy.