Abstract Objective The objective of this study was to evaluate the psychometric properties of the EQ-5D (3L and 5L) and SF-6Dv2 in a group of Chinese patients with late-on Pompe disease (PD), and compare their performance in this patient group. Methods The data used in this study were obtained from a web-based and cross-sectional survey conducted in China. All participants completed the 3L, 5L, and SF-6Dv2. Information about their sociodemographic status and health conditions was also collected. The measurement properties were assessed by examining ceiling and floor effects, evaluating convergent validity, known-group validity, and test–retest reliability (Intraclass correlation coefficient [ICC] and Gwet’s AC). Results A total of 117 PD patients completed the questionnaire. All dimensions of the 3L showed strong ceiling effects, ranging between 17.1 and 42.7%. All three measures showed good test–retest reliability, with ICC values ranging from 0.85 to 0.87. The Gwet’s AC values showed that four out of five dimensions of the 3L showed very good agreement. All hypothesized correlations between the 3L, 5L, SF-6Dv2, and items of WHODAS were supported, indicating satisfactory convergent validity. The 5L showed stronger correlations (|r|= 0.53–0.84) with WHODAS than the other two measures. The outcomes of ANOVA indicated that the 5L had higher F-statistics than the 3L and SF-6Dv2, indicating a stronger discriminant ability to differentiate most condition groups. Conclusion The 5L demonstrates lower ceiling and floor effects, higher discriminant ability, and better convergent validity than the SF-6Dv2 and 3L in patients with PD. In addition, the 5L may generate a larger utility gain compared to the other two instruments when conducting cost-effectiveness analysis.