The EQ-5D Does Not Have the Same Validity Properties as Other Condition Specific Health Related Quality of Life Instruments: Is It Time We Stopped Pretending It Does?

OBJECTIVES: Health Related Quality Life (HRQoL) measures undergo rigorous validity prior to formal use. Indeed, regulatory agencies stress the importance of validation of HRQoL instruments based on properties such as construct and discriminant validity, reliability and repeatability in the target disease area. The EQ-5D should not be exempt from this validation process. We evaluate available HTA Submissions and specifically examine evidence from the EQ-5D along with disease specific measures (DSM) of HRQoL. We conclude that the EQ-5D (in all its versions) is not suitable for determining whether treatment benefit exists in the same way as a DSM does. We argue that the EQ-5D is a second rate instrument, subsidiary to other DSMs of HRQoL and should be replaced by more sensitive (generic) HRQoL measures.

METHODS: HRQoL data from over 400 HTA reports, from several countries (UK: NICE, CADTH: Canada, PBAC: Australia) across diseases were extracted. Patient level clinical outcome data from a randomized controlled trial (DIAMONDS) in ophthalmology was also used to derive disease specific measures of treatment effect (between treatments) as well as over time and contextualize this against those from EQ-5D, using clinical outcomes as anchors.

RESULTS: All available HTA reports were classified according to re-imbursement authorities decisions: ‘not recommended’ /’recommended’. We conclude in over 70% of available reports, DSMs of HRQoL report important HRQoL gains when the EQ-5D did not: that is, the level of concordance between EQ-5D and DSM was low (p<0.05). In addition, data from the DIAMONDS trial also showed poor discriminant properties of the EQ-5D when using patient level data.

CONCLUSIONS: The EQ-5D should not be recommended for measuring HRQoL as an alternative to DSMs. The EQ-5D may grossly under estimate treatment benefits leading to erroneous re-imbursement decisions. Further analyses of patient level data from multiple disease areas is required to investigate this concern.