A Review of Health-Utility Data for Osteoarthritis

Abstract

The objective of this review was to describe the performance of health-utility measures in valuing the quality-of-life (QOL) impact of changes in osteoarthritis (OA)-related chronic pain when administered within a clinical trial setting. Because the collection of utility data within a clinical trial is not always feasible in the development of health economic models, utility data from prior non-randomised studies conducted among patients with OA were also summarized.We conducted a literature review using the MEDLINE, EMBASE and PsycINFO databases. We selected studies employing validated direct and multi-attribute measures of health utility: the standard gamble, time trade-off, EuroQol index, Health Utilities Index, SF-6D, 15D and the Assessment of Quality of Life measure.We identified four randomized controlled trials and 17 observational studies. The results of prior clinical trials in which these health utility measures were used in evaluating OA are summarized and attributes of the utility measures such as the clinical importance and statistical significance of the results obtained are noted. Furthermore, the sensitivity of the utility measure to changes in co-administered non-utility based measures of health-related quality of life (e.g. visual analogue scale for pain, WOMACtrade mark) are also reported. Five findings emerged.First, the EQ-5D system was the most widely used metric to derive utilities. Second, for whatever utility measure was used, reported mean utilities for patient groups spanned a rather wide range of values across studies, potentially reflecting variation in illness severity, patient co-morbidities and/or patient treatment. Third, when studies reported more than one utility-based statistic, the utility valuations frequently differed by measure, suggesting that the choice of metric can potentially have an effect on QALY calculations. However, there was no consistent pattern as to which measure yielded the highest and lowest utility valuations. Fourth, changes in health-related QOL (HR-QOL) and utility measures displayed the expected relationships. When HR-QOL declined, the utility values also moved in this direction. The reverse was also true. In some instances, statistically significant changes in QOL measures were not mirrored by statistically significant changes in utility measures, suggesting that some studies may have been underpowered for the latter purpose. Finally, the body of clinical trial-based utility literature in OA was found to be relatively modest, with considerably more observational studies collecting utility data.Based on the limited number of trial-based health-utility evaluations in OA to date, there can potentially be divergent findings with respect to clinical and statistical significance of changes in utility measures and corresponding measures of health status. Analysts should carefully evaluate issues of statistical power and clinical sensitivity in utilizing these measures in clinical trials of OA interventions.