Researchers of a study focused on creating a model to predict the imatinib plasma trough concentration (IM Cmin) in patients with advanced gastrointestinal stromal tumors (GISTs) in China. With the understanding that IM Cmin below 1100 ng/ml could lead to reduced drug efficacy, the researchers aimed at effective Cmin monitoring. Using least absolute shrinkage and selection operator regression and forward stepwise binary logistic regression, key variables impacting IM Cmin were identified. Nine machine learning classification models were then constructed and evaluated using various statistical methods to find the most suitable model.
The study’s results identified six essential variables for the model: gender, daily IM dose, metastatic site, red blood cell count, platelet count, and percentage of neutrophils. The Extreme Gradient Boosting (XGBoost) model emerged as the top performer in the validation set, exhibiting the largest AUROC, the lowest Brier score, and the highest values in the decision and precision-recall curves. The researchers concluded that the XGBoost model, using initial patient data and laboratory indicators, could accurately predict if IM Cmin is below 1100 ng/ml in patients with advanced GISTs, making it a useful tool for clinicians in optimizing IM treatment strategies.
Reference: Ran P, Tan T, Li J, Yang H, Li J, Zhang J. Advanced gastrointestinal stromal tumor: reliable classification of imatinib plasma trough concentration via machine learning. BMC Cancer. 2024;24(1):264. doi: 10.1186/s12885-024-11930-6.
Link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10894477/
