Physical activity level and its determinants in patients with cancer: a secondary analysis using a decision-tree approach.

PURPOSE: Low physical activity is associated with poorer outcomes in patients with cancer. Understanding the determinants of physical activity is essential to designing targeted interventions. This study aimed to identify the determinants of total, aerobic, and resistance physical activity among patients with cancer using a decision-tree machine learning approach. METHODS: This secondary analysis pooled data from three harmonized cross-sectional studies conducted between 2018 and 2024, including adult patients (aged ≥ 18 years) with any cancer type and stage. Physical activity was assessed through validated self-report questionnaires and classified according to current guidelines. Sociodemographic and clinical variables were considered as potential correlates. Decision-tree models using the Classification and Regression Trees method were applied to identify predictors of adherence to aerobic and resistance exercise recommendations. RESULTS: A total of 1132 patients were included. Only 17.5% met aerobic activity recommendations (i.e., 150 min/week or ACSM guidelines), and 12.7% met resistance training recommendations (i.e., 2 times/week); 2.6% met both. cancer site emerged as the primary determinant of both aerobic and resistance activity. Occupational status, body mass index, and educational level further influenced adherence patterns. Neither cancer stage nor treatment timing significantly predicted physical activity behavior. Distinct determinants were identified for aerobic versus resistance exercise. CONCLUSIONS: Among the factors included in the decision-tree models, physical activity was mainly influenced by cancer site and sociodemographic characteristics. Aerobic and resistance activities are governed by different predictors, underscoring the need for modality-specific assessment and intervention strategies. IMPLICATIONS FOR CANCER SURVIVORS: Tailored, cancer-specific supports that account for sociodemographic context and address barriers to both aerobic and resistance exercise are urgently needed to increase participation in physical activity and optimize survivorship outcomes.