Symptom Cluster Research in Women with Breast Cancer: A Comparison of Three Subgrouping Techniques

Abstract

AIMS To examine how symptom cluster subgroups defined by extreme discordant composite scores, cut-off scores, or a median split influence statistical associations with peripheral cytokine levels in women with breast cancer. BACKGROUND Systemic cytokine dysregulation has been posited as a potential biological mechanism underlying symptom clusters in women with breast cancer. Symptom characteristics may play an important role in identifying cytokines of significant etiological importance, however, there is no consensus regarding the ideal subgrouping technique to use. DESIGN A secondary analysis of data collected from a cross-sectional descriptive study of women with stage I-II breast cancer was used to examine and compare the relationships between peripheral cytokine levels and symptom subgroups defined by extreme discordant composite scores, cut-off scores, or a median split. METHODS Participant symptom scores were transformed into a composite score to account for variability in symptom intensity, frequency and interference. Cytokine levels in subgroups defined by composite scores within the highest and lowest 20% were contrasted with those composed from cut-off scores and a median split. RESULTS Subgroups defined by the composite score or cut-off scores resulted in similar statistical relationships with cytokine levels in contrast to the median split technique. The use of a median split for evaluating relationships between symptoms clusters and cytokine levels may increase the risk of a type I error. CONCLUSION Composite and cut-off scores represent best techniques for defining symptom cluster subgroups in women with breast cancer. Using a consistent approach to defining symptom clusters across studies may assist in identifying relevant biological mechanisms.