Characterizing potential subgroups and predictive factors of cancer-related fatigue among breast cancer patients through latent profile analysis and machine learning algorithms
Related Experiment Videos
Jiashi Zheng1, Xiaofang Zhao2, Yanhe Cui3
1The First Clinical College, China Medical University, No. 77 Puhe Road, Shenyang North New Area, Shenyang, Liaoning Province, 110122, P.R. China.
Purpose:
Breast cancer patients often experience fatigue during treatment, which may adversely affect their health outcomes and quality of life. However, machine learning (ML) research on cancer-related fatigue (CRF) in this population remains limited. Therefore, this study aimed to assess CRF symptoms using the Cancer Fatigue Scale (CFS), identify latent classes of CRF, and apply multiple ML algorithms to identify key predictors, thereby providing a basis for developing targeted interventions.
Methods:
Using a cross-sectional stratified sampling approach, 1005 participants completed a self-administered questionnaire to assess CRF symptoms with the CFS. Latent profile analysis (LPA) was used to identify latent classes of CRF. Five ML algorithms were used to identify predictors of CRF.
Results:
Three latent classes were identified: stable fatigue group (9.8%), low fatigue unstable group (27.4%), and high fatigue unstable group (62.9%). The area under the curve (AUC) values of the five ML models-K-nearest neighbors (KNN), Adaptive Boosting (AdaBoost), Gaussian Naive Bayes (Gaussian NB), Extreme Gradient Boosting (XGBoost), and Random Forest (RF)-ranged from 0.83 to 0.88, with XGBoost performing the best (AUC = 0.88). The final ML model identified perceived stress, depression, and self-compassion as the most important predictors of CRF.
Conclusions:
Breast cancer patients exhibit heterogeneity in fatigue symptoms during treatment. Machine learning may help identify factors associated with CRF, including perceived stress, depression, and self-compassion. For the high fatigue unstable group, interventions may target perceived stress and depression; for the low fatigue unstable group, interventions may focus on self-compassion and stress management; and for the stable fatigue group, interventions may target anxiety reduction and social support enhancement. This stratified approach may provide a practical framework for clinicians to develop individualized fatigue management plans.
