• Emily Chen, Janet Nguyen, Gemma Cramarossa, Luluel Khan, Liying Zhang, May Tsao, Cyril Danjoux, Elizabeth Barnes, Arjun Sahgal, Lori Holden, Florencia Jon, Kristopher Dennis, and Edward Chow.
• Rapid Response Radiotherapy Program, Odette Cancer Centre, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, ON, Canada.
• Palliat Med. 2012 Sep 1;26(6):826-33.

BackgroundAdvanced cancer patients often experience multiple concurrent symptoms, which can have prognostic effects on patients' quality of life. Including patients who did not experience all of the symptoms measured by an assessment tool may interfere with accurate symptom cluster identification. Varying statistical methods may also contribute to inconsistencies of cluster results.AimsTo compare symptom clusters in a subgroup of patients reporting exclusively non-zero ESAS scores with those in the total patient sample. To examine whether using different statistical methods results in varied symptom clusters.DesignPrincipal Component Analysis (PCA), Hierarchical Cluster Analysis (HCA) and Exploratory Factor Analysis (EFA) were performed on the 'non-zero' subgroup and the total patient sample to identify symptom clusters at baseline and weeks 1, 2, 4, 8 and 12 following palliative radiotherapy.Setting/ParticipantsA previous single-centre study used Principal Component Analysis to explore symptom clusters in 1296 advanced cancer patients. The present study analyzed this previously reported data set.ResultsNotably different symptom clusters were extracted between the two patient groups regardless of the statistical method at baseline, with the exception of a cluster composed of drowsiness, fatigue and dyspnea using Principal Component Analysis and Hierarchical Cluster Analysis. At follow-ups, different statistical methods yielded significantly varied symptom clusters. Only anxiety, depression and well-being consistently occurred in the same cluster across methods and over time.ConclusionsThe composition of symptom clusters varied depending on if patients with non-zero scores were excluded at baseline and on the statistical method employed. Identifying valid clusters may prove useful for bettering symptom diagnosis and management for cancer patients.

9 keywords...

hide…