The Significance of Using US-Valued EQ-5D Health States for Comparative Effectiveness and Cost-Effectiveness Studies in US Populations

Abstract

Two articles published in this issue, ‘‘A Comparison of EQ-5D Index Scores Derived from the US and UK Population-Based Scoring Functions’’ by Nan Luo and colleagues and ‘‘The Implications of Using US-Specific EQ-5D Preference Weights for Cost-Effectiveness Evaluation’’ by Katia Noyes and colleagues, substantially advance our understanding of the relative performance of USand UK-based EQ-5D preference scores. The study by Luo and others extends the previous work of its authors, who conducted the 2002 US valuation survey for the EQ-5D, using methods largely consistent with those of the 1993 UK valuation survey. Their new study compares the full complement of 243 EQ-5D health state scores predicted by the statistical models developed from the US and UK valuation surveys (both of which directly valued the same 42 common EQ-5D health states). Not surprisingly, like the authors’ earlier comparison of the directly valued health states, the US statistical model consistently predicted health state scores that were higher than those predicted by the UK model. Over the full set of EQ-5D health states, the difference in mean scores was 0.23, as compared with the 0.11 difference found for the smaller number of directly valued states in the earlier comparison. More notably, the new study found a mean difference in transitions between all pairs of health states of 0.25 for the US valuation model and 0.35 for the UK model: The range of all EQ-5D health state valuations in the US model was more compressed than in the UK model. The authors reasonably surmise that although incremental cost-effectiveness ratios (ICERs) calculated using the US model values would likely be less favorable than those using the UK model values, this result would depend on the nature of the condition under study, the severity of the health states, and the size of the health effect. Noyes and colleagues take another step and examine the differences in results between the USand UK-valued EQ-5D statistical models in a costeffectiveness analysis. Using the US values, they reanalyze data from a clinical trial comparing 2 alternative drug therapies (initial pramipexole v. levodopa) in patients with Parkinson disease, which originally used health state values from the UK EQ5D model (see references 15-17 in the article by Noyes and others). The comparative performance of the USand UK-valued EQ-5D models was consistent with the simulated results that Luo and others obtained: Measured over time, the USand UKspecific EQ-5D scores for patients with Parkinson disease followed similar trajectories, except that the US scores were higher on average. As in the simulation exercise, the range of scores using the US values was more compressed than that with the UK values. In particular, Noyes and others found greater differences between the US and UK estimates among subjects who reported problems with usual activities, pain/discomfort, or anxiety/depression than between the estimates for subjects not reporting these problems. As the authors conclude, the implications of the more compressed range of scores using the US valuation model are that the estimated 4-year gains in quality-adjusted life years (QALYs) of the experimental over the control therapy are less than half as large as those using the UK model, a much less favorable ICER ($108,000/QALY v. $43,000/QALY) and a lower probability that the experimental therapy is cost-effective at any dollar-per-QALY threshold. Taken together, these 2 studies offer insight into how valuations of health states based on the USand UK-valued EQ-5D statistical models compare From the Department of Health Policy, George Washington University School of Public Health and Health Services, Washington, DC.