Comparing EQ-5D-3L and EQ-5D-5L performance in common cancers: suggestions for instrument choosing

To compare the performance of three-level EuroQol five-dimensions (EQ-5D-3L) and five-level EuroQol five-dimensions (EQ-5D-5L) among common cancer patients in urban China.

A hospital-based cross-sectional survey was conducted in three provinces from 2016 to 2018 in urban China. Patients with breast cancer, colorectal cancer, or lung cancer were recruited to complete the EQ-5D-3L and EQ-5D-5L questionnaires. Response distribution, discriminatory power (indicator: Shannon index [H'] and Shannon evenness index [J']), ceiling effect (the proportion of full health state), convergent validity, and health-related quality of life (HRQoL) were compared between the two instruments.

A total of 1802 cancer patients (breast cancer: 601, colorectal cancer: 601, lung cancer: 600) were included, with the mean age of 55.6 years. The average inconsistency rate was 4.4%. Compared with EQ-5D-3L (average: H' = 1.100, J' = 0.696), an improved discriminatory power was observed in EQ-5D-5L (H' = 1.473, J' = 0.932), especially contributing to anxiety/depression dimensions. The ceiling effect was diminished in EQ-5D-5L (26.5%) in comparison with EQ-5D-3L (34.5%) (p < 0.001), mainly reflected in the pain/discomfort and anxiety/depression dimensions. The overall utility score was 0.790 (95% CI 0.778-0.801) for EQ-5D-3L and 0.803 (0.790-0.816) for EQ-5D-5L (p < 0.001). A similar pattern was also observed in the detailed cancer-specific analysis.

With greater discriminatory power, convergent validity and lower ceiling, EQ-5D-5L may be preferable to EQ-5D-3L for the assessment of HRQoL among cancer patients. However, higher utility scores derived form EQ-5D-5L may also lead to lower QALY gains than those of 3L potentially in cost-utility studies and underestimation in the burden of disease.