Seven reflections on the UK EQ-5D-5L value set – and what comes next

The long-awaited EQ-5D-5L value set for the United Kingdom (UK) has been published in Value in Health. For anyone with an interest in health technology assessment (HTA) in the UK this is a significant moment – one that hopefully draws a line under a rather long-winded saga. 

The publication represents an immense amount of work from the study team and many others, and they deserve real credit for their efforts. Following consultation by the National Institute of Health and Care Excellence (NICE), the future adoption of this new value set will have a meaningful impact on how health-related quality of life (HRQoL) gains are valued in HTA. 

Here are seven early reflections on the value set and its potential adoption, and the questions that it raises.

1. Preferences around different aspects of HRQoL have changed

A value set provides a snapshot of what a population cares about in a specific period. The EQ-5D-3L value set that has existed since before NICE’s inception is based on preferences of the UK general population collected in a study conducted over thirty years ago, in 1995 – back when restaurants had smoking sections, Sunny Delight was considered a “healthy drink”, and Blackburn Rovers won the Premier League. Things change, and clearly that is also true for preferences around different aspects of HRQoL.

So, what exactly has changed? When considering the relative importance of the different dimensions of EQ-5D, there is a clear pattern:

• “Anxiety/depression” and “usual activities” are relatively more important
• “Mobility” and “self-care” are relatively less important
• “Pain/discomfort” remains the biggest concern overall

Treatments that improve mental health, for example, may well benefit from the change in value set. 

However, interpreting these changes is not straightforward – differences in the 3L and 5L descriptive systems and the valuation protocols used may also contribute – and it is not only the relative importance of the dimensions that has changed…

2. Good news for life-extending therapies; less so for those that improve HRQoL

The utility of the worst health states in the two descriptive systems are -0.59 (3L) and -0.57 (5L), so you could be forgiven for assuming that the two value sets might behave quite similarly in practice.

However, an analysis by Biz et al. published in Value in Health alongside the value set has shown that whilst the extreme ends of the scale are similar, the distribution of utilities are substantially different across the two value sets. Based on an analysis mapping existing 3L appraisal data to the new 5L value set, the 5L produces higher utilities on average and compresses them into a narrower range – particularly in the mild-to-moderate part of the scale where most trial data sit. Ultimately, when the 5L value set is adopted, life-extending technologies are likely to fare better on average than those that improve quality of life.

For companies preparing HTA submissions, this change is likely to have a meaningful impact on the cost-effectiveness of their assets. 

3. Preferences under which perspective? More than just an academic question

All else being equal, the analysis by Biz et al. suggests that switching to the new 5L value set from the 3L value set could result in a favourable change in the incremental cost-effectiveness ratio (ICER) for a primarily life-extending end-of-life (EOL) oncology treatment. However, from a priority setting perspective, EOL treatments have been relatively deprioritised by NICE under the severity modifier. The new value set is based on societal preferences, and the shift from the previous EOL criteria to the new severity modifier was also justified on the basis of societal preferences. So why the apparent disconnect?

The answer lies within the type of questions being asked. Value sets are generated from individual-level preference elicitation tasks, in which members of the public are asked to consider their own health from a personal perspective. In contrast, priority setting studies ask people to take a societal view – how should the health system allocate resources? These are genuinely different questions, and it is not surprising that they might come to different answers.

Past research studies – and a recent study of our own – suggest that when people are asked a societal question about EOL care, many express a preference for prioritising HRQoL improvement over life extension. This sits in some tension with a value set change that, as Biz et al. suggest, may advantage survival gains over HRQoL gains. 

These findings are not contradictory in themselves. They do, however, illustrate a broader challenge: ensuring that societal priorities and preferences meaningfully inform HTA frameworks is an important goal, but translating them into consistent policy is not necessarily straightforward.

4. The end of EQ-5D mapping? Not quite…

Analysts will no doubt be pleased that mapping algorithms (or “crosswalks”) from 5L responses to the 3L value set will soon no longer be needed – the new 5L value set can be directly applied to 5L responses without the need for an uncertainty-inducing middleman. But not so fast…

Mapping is likely to remain a feature of many HTA submissions for two reasons. First, EQ-5D-5L is often not collected in key clinical trials. As a result, mapping from condition-specific measures (such as EORTC QLQ-C30 or FACT instruments) to EQ-5D-5L utilities will still be necessary. A further complication here is that direct mapping algorithms from condition-specific measures to the UK 3L value set will no longer point to the right target. Second, even where native EQ-5D-5L data are collected to support a new submission, comparator utilities in cost-effectiveness models are often drawn from historical trials or published literature that have used the 3L. To ensure consistency within a model, NICE committees may wish to see utilities on the same value set, meaning that mapping between 3L and 5L (in either direction) may remain necessary for some time.

These issues will need to be carefully considered as NICE plan the implementation process for the new 5L value set.

5. The impact will extend to the severity modifier

The new value set will also impact estimates of absolute and proportional QALY shortfall (AS and PS) that are used to determine severity classifications under NICE’s severity modifier.

AS and PS are determined by the gap between quality-adjusted life expectancy (QALE) – derived from population norm utilities – and QALY expectations under the standard of care. Both inputs will change under the new value set. The direction and magnitude of the net effect on shortfall estimates will depend on which changes more, and more research is needed to understand the likely dynamics. For now, manufacturers whose therapies sit close to a severity modifier threshold, will need to model these scenarios carefully.

At a higher level, there is a broader implication for the “opportunity cost neutrality” of the severity modifier. The modifier was calibrated against 3L-based cost-effectiveness data – recalibrating it under a 5L world is not straightforward, and whether the modifier would remain (more or less) “opportunity cost neutral” is an open question.

Irrespective, ideally the modifier would correspond well with societal preferences, and our research – published in Value in Health in December – has suggested that the current iteration of the severity modifier does not. Adjustments to the modifier seem increasingly justified.

6. What about informal carers?

NICE allow for the HRQoL of (informal) carers to be captured within economic models when the impact is “substantial”, enabling potential “spillover” effects of new treatments (i.e., “spillover” benefits to informal carers of a treatment for patients) to be reflected. In these cases, NICE prefers HRQoL to be captured using EQ-5D to ensure comparability, and therefore the new value set may also have an impact here. 

On the one hand, the two dimensions that have a higher weighting in the new value set (“anxiety/depression” and “usual activities”) may be two of the more likely dimensions to be impacted for carers by new treatments, potentially increasing carer QALY gains. However, our research (which we are currently updating and expanding!) has shown that carers often report being in mild-to-moderate EQ-5D-5L health states, which may offset this for two reasons. First, those health states sit in the upper part of the utility scale, leaving relatively little room for improvement. Second, it is this region of the utility scale that is most compressed under the new value set. Together, improvements in carer HRQoL on EQ-5D-5L with the new value set may translate into smaller QALY gains than under the 3L value set. For most HTA submissions the practical impact may be minor, but for submissions where carer spillover effects are already marginal, it is worth being aware of – and in the most extreme cases it could interact with the carer QALY trap.

Ultimately, until more research is conducted, it will be hard to predict the net impact on carer QALY gains.

7. What comes next?

Soon, NICE will launch a consultation on the recommendation to switch to the UK EQ-5D-5L value set, and adoption is likely to follow thereafter. However, this is not the only UK EQ-5D value set on the horizon.

For a long time, the UK EQ-5D-3L value set has been applied when the EQ-5D-Y-3L – the child/adolescent version of EQ-5D – has been collected. This has been shown to be inappropriate, but it has at least had the benefit of ensuring consistency with other appraisals. That will no longer be the case when the 5L value set is adopted.

The good news is that a value set for EQ-5D-Y-3L is on the way. Over the past several years, our team has been working on the methodological foundations that underpin how child health states should be valued – including whether adult and adolescent preferences differ, how different valuation methods and perspectives affect results, and how to anchor preference data onto the QALY scale. This work has fed into value set development internationally, including the German EQ-5D-Y value set. We are now collaborating on this ongoing UK EQ-5D-Y-3L valuation study, which is being led by Prof Oliver Rivero-Arias (University of Oxford), in collaboration with Prof Donna Rowen (University of Sheffield) and Dr Koonal Shah (NICE). Further updates on that study will be available later this year.

These are busy times for outcomes measurement in UK HTA, and the questions raised by the new value set will keep researchers, HTA teams, and HEOR functions occupied for some time.

OHE is committed to furthering outcomes in UK HTA;  the questions raised by the new value set will keep researchers, HTA teams, and HEOR functions occupied for some time. You can also find out more about our Measuring and Valuing Outcomes (MVO) work on our webpage.

---