Based a Machine Learning Approach to Investigate the Factors Influencing Nirmatrelvir/ritonavir Exposure in Human Plasma: A Multicenter, Observational Study

Objectives: Nirmatrelvir/ritonavir (N/R) is an effective drug for treating COVID-19. However, there is currently a lack of evidence for therapeutic drug monitoring of N/R, which may increase the risk of adverse drug reactions and compromise its efficacy. Methods: In this study, we retrospectively analyzed data from 139 patients in two center who were prescribed N/R. We collected baseline data from all patients and monitored nirmatrelvir and ritonavir concentrations on the third day of medication. We conducted a logistic regression analysis to investigate the relationship between drug concentration and prognosis. We also analyzed the correlations among features and used a random forest model to select significant factors that affect drug exposure. Subsequently, we constructed an XGBoost model to predict drug concentration using the selected features. Results: Our findings indicated that the concentration of N/R could not predict patient outcomes. We also identified potential factors that affect N/R concentration, including estimated glomerular filtration rate, creatine kinase, aspartate aminotransferase, alanine aminotransferase, lymphocytes, and platelet count. Ultimately, the evaluation of the predictive model resulted in a mean absolute error (MAE) of 0.717, mean squared error (MSE) of 1.328, root mean squared error (RMSE) of 1.152, and coefficient of determination (R-squared) of 0.779. The prediction model performs well and can provide risk prediction for medication management for N/R, as well as assist in personalized medication. Conclusions: We identified a set of variables that affect the treatment of N/R through therapeutic drug monitoring and established a machine learning model to identify drug risks. This provides a reference for clarifying the significance of therapeutic drug monitoring for N/R treatment and the subsequent development of multivariate prognostic models.