• Luluel Khan, Gemma Cramarossa, Madeline Lemke, Janet Nguyen, Liying Zhang, Emily Chen, and Edward Chow.
• Rapid Response Radiotherapy Program, Odette Cancer Centre, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, ON, Canada.
• Support Care Cancer. 2013 Feb 1;21(2):467-73.

PurposeThe purpose of this paper is to determine whether the use of three different statistical methods influences the composition of symptom clusters derived from patients with brain metastases.Materials And MethodsThe dataset previously compiled from 129 brain metastases patients who completed the Spitzer Quality of Life Index (SQLI) and an additional study-designed 17-item symptom questionnaire was employed in this study. Symptom clusters extracted using principal component analysis in our previous study were compared to clusters determined using hierarchical cluster analysis and exploratory factor analysis. Clusters were identified using the three statistical methods at baseline, and at 1, 2, and 3 months following whole brain radiotherapy.ResultsThe number and composition of symptom clusters at each time point varied based on the statistical method employed, despite the use of an identical dataset. However, some domains consistently clustered together, such as activity and daily living from the SQLI items. Of the 17 additional symptoms, memory loss, confusion, and trouble concentrating were always present in the same cluster. Nausea and vomiting also occurred in conjunction regardless of the analytical method employed.ConclusionSymptom clusters vary with respect to occurrence, quantity, and composition based on the statistical method utilized to extract them. Further studies should be conducted to determine an ideal statistical method in order to select the optimal method to employ. The use of a single analytical method is essential for consistency and comparison purposes in future symptom cluster research.

18 keywords...

hide…