Measuring adaptation in colonic cancer patients' concepts of quality of life (QL) using multilevel models: Should we reconsider use of QL instruments which ignore adaptation?

Abstract

6008 Background: If psychological adaptation to disease brings about major changes in what is important to the patient's quality of life (QL), most instruments in common use are poorly suited to longitudinal QL measurement, especially over periods encompassing substantial life changes such as occur during the course of many cancers. Little is known about this, and there is a lack of suitable methods for assessing its importance. We introduce a new application of multilevel models which directly addresses this question, and present results from colonic cancer patients in a phase III clinical trial. Methods: 126 colonic cancer patients summarised domains of QL and overall QL using linear analogue scales after three months of adjuvant treatment or observation, and again 21 months later. Changes in importance of domains were estimated using multivariate multilevel models. Results: There was strong evidence of a change in the relative importance of the measured domains to patients' QL (p=0.001); functional performance decreased markedly in importance and energy increased in importance, whereas the importance of pain did not change. Conclusions: The factors which these patients considered most relevant to their QL changed profoundly during the course of disease. Conventional QL instruments which apply a fixed set of weights to domain scores assume that no such changes occur. By ignoring adaptation, they might seriously misrepresent changes in subjective QL in these patients. If this effect is also found among other cancer patients, QL instruments which allow for adaptation should be favoured over those which ignore it. Multilevel models offer a simple, transparent way to study this phenomenon, and can readily be applied to existing QL datasets collected for other purposes. No significant financial relationships to disclose.