OBJECTIVES: Currently, 2 EQ-5D-5L value sets are available in the UK: one based on a direct valuation of EQ-5D-5L health states (Devlin et al), and another using a crosswalk approach mapping the 5L descriptive system data onto the 3L valuation set (van Hout et al; NICE recommended). This study evaluated the psychometric properties of these 2 EQ-5D-5L utility scores in metastatic Merkel cell carcinoma (mMCC).

METHODS: Data collected in the phase 2, single-arm avelumab trial JAVELIN Merkel 200 (part A; NCT02155647) were analysed, including 88 patients with previously treated mMCC. Criterion validity was assessed by Pearson correlations between utility and EQ-VAS, FACT-G and FACT-M scores at week 7. Responsiveness was assessed by correlating change from baseline (CFB) scores in these measures. Effect sizes were calculated to assess test-retest reliability and responsiveness in subgroups of patients with improving, stable, or deteriorating health on the basis of clinically relevant changes in the above quality-of-life scales considered external anchors.

RESULTS: For any given time point during the 18-month follow-up, Devlin utilities were ≥0.10 points higher than van Hout crosswalk utilities. Both algorithms demonstrated notable correlations (from 0.45-0.72; p<0.01) between utility and the EQ-VAS and FACT-G total, physical, and functional scores. CFB scores also demonstrated large correlations (from 0.5-0.6) between these measures. In patients with health improvements, effect sizes were larger for the van Hout utilities, whereas the Devlin utilities had larger effect sizes for health deteriorations, probably due to different levels of baseline utility.

CONCLUSIONS: Both Devlin and van Hout utility scores had good and comparable validity, reliability, and responsiveness in patients with mMCC. Differences in mean utility between the scoring algorithms may affect cost-effectiveness estimates for treatments improving overall survival, where higher utility translates to more quality-adjusted life-years.