Measuring neuropsychological change following breast cancer treatment: an analysis of statistical models

This article considers the quantitative techniques currently in use in the evaluation of cognitive impairments associated with chemotherapy treatment for breast cancer. To illustrate differences among analytical approaches, all analyses were applied to baseline and posttreatment scores on neuropsychological tests obtained from Stages I and II breast cancer patients receiving either chemotherapy or hormonal therapy; a healthy control group with similar demographics to those of the treatment groups was also included. Conventional group analyses were compared with individual-based analyses (standardized regression-based and reliable change methods). Both univariate and multivariate techniques with and without covariates produced negligible effects. In contrast, results of the individual-based analyses identified a subset of participants in the chemotherapy group who experienced a severe decline in function on two or more tests. Differences between the control and treatment groups were greater than differences between the treatment groups alone. The standardized regression-based approach was more sensitive than the reliable change index in detecting chemotherapy and hormonal therapy subjects whose performance was different from baseline scores on two or more tests (roughly 80% vs. 50% of participants). From a clinical perspective, the degree of impairment determined on the basis of the individual-based methodologies could have a major impact on quality of life for those affected. On the whole, we argue that the standardized regression-based approach, allowing for the assessment of individual practice effects and evaluation of moderator variables, is the method of choice in this context.