Machine learning approaches for the identification of real-world predictors of mepolizumab response

Authors:

Abstract
Background: Mepolizumab improves control and reduces clinically significant exacerbations in severe asthma patients. Further research is needed to understand factors influencing patient response. Objective: Identify predictors of response to mepolizumab using machine learning techniques. Methods: Thirty clinically relevant features from the REALITI-A dataset were analyzed. Response to mepolizumab was defined as change in clinically significant exacerbations at Year 1 post-exposure, being free from clinically significant exacerbations at Year 1, and being free from clinically significant exacerbations at both Year 1 and Year 2 post-exposure for primary, secondary, and exploratory analyses, respectively. Linear/logistic regression, random forest, and XGBoost models were tested. The best-performing model based on R 2 or classification accuracy was used for feature selection. A SHapley Additive exPlanations post-hoc analysis was conducted. Results: The primary and secondary analyses included 685 patients; the exploratory analysis included 529 patients. Using XGBoost for the primary analysis, higher pre-exposure clinically significant exacerbations rate, lower body mass index, total immunoglobulin E, lower baseline activity impairment, and lower baseline Asthma Control Questionnaire-5 score were top predictors. For the secondary analysis, the random forest model identified lower pre-exposure clinically significant exacerbations rate, residency outside the United Kingdom, and lower baseline Asthma Control Questionnaire-5 score, body mass index, and baseline activity impairment as top predictors. For the exploratory analysis, XGBoost identified residency outside the United Kingdom, and lower pre-exposure clinically significant exacerbations rate, baseline Asthma Control Questionnaire-5 score, eosinophil levels, and body mass index as top predictors. Conclusion: Machine learning models identified key predictors of response to mepolizumab, which could guide clinical decision-making and improve patient outcomes in severe asthma patients.
Funding Information
  • GSK (215038)

This publication has 30 references indexed in Scilit: