Can machine learning predict pharmacotherapy outcomes? An application study in osteoporosis.

Area of Science:

• Computational biology
• Medical informatics
• Machine learning in healthcare

Background:

• Osteoporosis treatment requires personalized approaches for optimal outcomes.
• Predicting treatment response is crucial for managing osteoporosis effectively.

Purpose of the Study:

• To develop and compare machine learning models for predicting osteoporosis treatment outcomes.
• To assess the performance of artificial neural network (ANN), random forest (RF), support vector machine (SVM), and logistic regression (LR) models.

Main Methods:

• Retrospective analysis of electronic clinical data from osteoporosis patients (2011-2018).
• Utilized ANN, RF, SVM, and LR models with features selected by a genetic algorithm.
• Evaluated models using accuracy, precision, sensitivity (recall), F1 score, and area under the receiver operating characteristic curve (AUC).

Main Results:

• No significant differences in accuracy, precision, or AUC were observed among the four models.
• Logistic regression (LR) demonstrated significantly higher recall in the main analysis.
• Artificial neural network (ANN) showed significantly higher recall in the subgroup analysis.

Conclusions:

• Machine learning models show promise for forecasting osteoporosis treatment outcomes.
• These models can support clinical decisions for personalized treatment adjustments, preventing disease progression or treatment failure.
• Early intervention and timely treatment modification are facilitated by predictive modeling.