An exploratory study to clarify the cluster of symptoms predictive of chemotherapy-related nausea using random forest modeling

Molassiotis, Alex and Farrell, Carole and Bourne, Kathryn and Brearley, Sarah G and Pilling, Mark (2012) An exploratory study to clarify the cluster of symptoms predictive of chemotherapy-related nausea using random forest modeling. Journal of Pain and Symptom Management, 44 (5). pp. 692-703. ISSN 0885-3924

Abstract

CONTEXT: Chemotherapy-related nausea is experienced by most cancer patients receiving chemotherapy. Although vomiting is managed well with current antiemetics, nausea is difficult to manage and little is understood about its development. OBJECTIVES: The aim was to determine whether nausea exists as part of a symptom cluster and evaluate the symptom cluster's impact on patients' quality of life, psychological distress, and nutritional status. METHODS: A prospective observational design over two cycles of chemotherapy was used. Patients completed the Memorial Symptom Assessment Scale, Hospital Anxiety and Depression Scale, Functional Assessment of Cancer Therapy-General, and Patient-Generated Subjective Global Assessment before chemotherapy and at the end of the first and second cycles of treatment. Random forest modeling, a state-of-the-art prediction method, was used to analyze the data. RESULTS: One hundred four patients participated in the study. Nausea was found to be a dynamic experience, changing over time. "Core" symptoms, predictive of the presence of nausea, were identified and included appetite loss, feeling bloated, vomiting, taste changes, and lack of energy. Although nausea alone did have an impact on patient outcomes, the impact was significantly higher in relation to physical and functional quality of life and nutritional status in those patients who had more than two symptoms from the nausea cluster. CONCLUSION: This exploratory study, using an innovative analytical approach, has shown that nausea is a complex symptom affected by the presence and/or severity of other concurrent symptoms (the symptom cluster). The findings have implications for the measurement of nausea and also to target people for interventions to manage nausea and its cluster of symptoms.