Machine learning-based models for predicting clinical outcomes after surgery in unilateral primary aldosteronism.

Area of Science:

• Endocrinology
• Surgical Hypertension Management
• Machine Learning in Medicine

Background:

• Unilateral primary aldosteronism (PA) is a surgically curable cause of endocrine hypertension.
• Persistent hypertension after adrenalectomy affects over half of PA patients, potentially deterring surgical treatment.
• Predicting postoperative outcomes is crucial for patient selection and management.

Purpose of the Study:

• To develop and validate machine learning models for predicting hypertensive remission post-unilateral adrenalectomy in PA patients.
• To identify readily available preoperative predictors of successful surgical outcomes.
• To aid in clinical decision-making for surgical intervention in PA.

Main Methods:

• Retrospective cross-sectional study of 107 PA patients with biochemical success post-adrenalectomy.
• Development of supervised machine learning models using training and test datasets.
• Selection of key predictors: hypertension duration, antihypertensive medication use, plasma aldosterone concentration (PAC), sex, BMI, and age.

Main Results:

• The developed model achieved 77.3% accuracy and an AUC of 0.884 in the test dataset.
• External validation in an independent cohort showed comparable performance (80.4% accuracy, AUC 0.867).
• Hypertension duration, medication use (DDD), PAC, and BMI were significant non-linear predictors.

Conclusions:

• Machine learning models can accurately predict hypertensive remission after adrenalectomy for PA.
• Preoperative clinical factors are valuable for assessing surgical benefit and guiding treatment decisions.
• The model supports personalized treatment strategies for PA patients undergoing adrenalectomy.