EQ‐5D‐5L health utility scores in Australian adults with inherited retinal diseases: A cross‐sectional survey

Economic evaluations of interventions for ocular disease require utility scores that accurately represent quality of life in the target population. This study aimed to describe the distribution of EQ‐5D‐5L utility values among Australian adults with symptomatic inherited retinal diseases (IRDs) and to assess the relationship between these scores and vision‐related quality of life.


| I N T RODUC T ION
It is vital to consider outcomes that are important to patients when making therapeutic clinical and funding decisions.This is particularly true when therapies are invasive and expensive, as is the case with novel gene-based approaches to treating inherited retinal diseases (IRDs) (Britten-Jones et al., 2022;Britten-Jones, Jin, et al., 2022).However, the heterogeneous nature of IRDs and the low prevalence of individual variants make it challenging to quantify the impact of having an IRD on health-related quality of life (Jayasundera et al., 2022).
Most IRDs are caused by variants in a single gene, resulting in dysfunction in the retinal photoreceptors or retinal pigment epithelium.Almost 300 genes have been recognised within more than 25 phenotypical categories (Daiger et al., 2023).Although each of these individual genetic disease is rare, as a whole IRDs are the most common cause of legal blindness in working-age adults in developed nations (Heath Jeffery et al., 2021).Thus, the impacts of IRD can be devastating for individuals and costly to society (Schofield et al., 2023).In addition, many IRDs are syndromic, with co-existing nervous system, auditory, skeletal, renal, or cardiometabolic conditions (Tatour & Ben-Yosef, 2020).
Ocular gene therapy has recently emerged as a treatment option for IRDs (Britten-Jones et al., 2022;Britten-Jones, Jin, et al., 2022).Given the high cost of administering gene therapy, economic analyses will be vital to inform regulatory and funding decisions.Cost-utility analysis is one type of economic evaluation that involves estimating the impact of an intervention on qualityadjusted life years, a measure that incorporates both quantity and quality of life.To conduct these analyses, validated quality of life instrument scores are converted to utility values via an associated algorithm, where 0 is equivalent to death and 1 perfect health.
Recent economic analyses have relied on assumptions about the relationship between decreased visual function and health utility values when estimating the impact of gene therapy on quality-adjusted life years among people with IRDs (Bhadhuri et al., 2022;Farmer et al., 2020;Johnson et al., 2019;Uhrmann et al., 2020;Viriato et al., 2020).These assumptions have been generated via vignette approaches using clinician elicitation (Lloyd et al., 2019) or extrapolated from people with other ocular conditions (Brown et al., 2001;Zimmermann et al., 2019).Health utilities for people with IRDs have also been reported for a clinically ascertained cohort (Schofield et al., 2022) and following elicitation from the general public (O'Brien et al., 2023).However, this information has not been used to inform published costutility analyses associated with an intervention to improve vision in an IRD cohort to date.
IRD gene therapy trials have been relatively small and there is minimal information on health utility values elicited from people with IRDs.Thus, additional data are required to help inform the models used to assess the potential economic impact of interventions for IRDs (Jayasundera et al., 2022).Therefore, using data from a large cross-sectional survey study, we aimed to describe the distribution of EQ-5D-5L utility values among Australian adults with IRDs.In addition, we explored the association between EQ-5D-5L utilities and visionrelated quality of life as measured via the National Eye Institute Visual Functioning Questionnaire (NEI-VFQ-25).This is an instrument commonly used as a patient-reported outcome measure in vision studies but not designed to derive utilities for the estimation of quality-adjusted life years.

| M AT ER I A L S A N D M ET HOD S
Questionnaire data were collected as part of a crosssectional study that was primarily conducted to investigate knowledge and attitudes towards ocular gene therapy among people with IRDs in Australia (Mack et al., 2021).The instruments were administered in English between January and June 2021, predominantly as an online questionnaire.Paper-based and phone interview alternatives were available on participant request with responses transcribed by study investigators.Deidentified data were securely captured and managed using the REDCap electronic data capture tool, hosted by the Centre for Eye Research Australia.Participants were informed that the study would take approximately 1 h to complete and gave consent to take part in the study after information on the nature of the study was provided, including details about the study purpose, investigators and data retention.This study was approved by the Human Research Ethics Committee of the University of Melbourne ( 2057534) and was conducted in compliance with the Tenets of the Declaration of Helsinki.

| Eligibility and recruitment
Eligibility criteria included people with an IRD (including syndromic forms) living in Australia who were aged 18 years and above.Participants self-selected their ocular diagnosis from a pre-defined list of IRDs, and no independent confirmation of genotype or clinical characteristics was obtained.People who were carriers of IRD variants without an ocular phenotype and those with other polygenic retinal conditions with complex genetic risk factors (such as age-related macular degeneration) were excluded.Parents/guardians and carers of individuals with an IRD were eligible to participate in the original study.As the people responsible for making healthcare decisions, these parents, guardians and carers were asked to give their own views and perspectives (rather than respond as proxies for their dependents) and, therefore, have not been included in the current analysis.
This was an 'open' survey (i.e. could be completed by anyone who met the eligibility criteria) conducted among a convenience sample.Potential participants were invited to participate via email, postal mail, or social media distributed by the Australian Inherited Retinal Disease Registry and DNA Bank (De Roach et al., 2020), patient support groups, specialist ophthalmic hospitals, the ophthalmology and clinical genetics departments of the major metropolitan teaching hospitals and the private practices of the study investigators.

| Demographic and ocular characteristics
For the original study, data were collected on 15 items relating to demographic and ocular characteristics including age, gender and highest level of education.Given that the impact of vision loss on activities of daily living is likely to differ depending on whether that loss is in the peripheral and/or central visual field (Rentz et al., 2014), self-reported IRD diagnosis was classified by an IRD subspecialist ophthalmologist (HGM) as being associated with widespread retinal involvement versus predominantly central, as described previously (McGuinness et al., 2022).No IP (Internet Protocol) address data were collected, allowing participants to remain anonymous.Thus, the number of unique visitors to the REDCap platform could not be counted.

| Quality of life instruments
Permission to use the EQ-5D-5L was granted by the EuroQol Research Foundation.The EQ-5D-5L is a general health utility instrument with five dimensions: mobility (walking around), self-care (washing or dressing), usual activities (work, study, housework, family or leisure activities), pain/discomfort and anxiety/depression (Herdman et al., 2011).Participants were instructed to respond according to their health on the day the instrument was completed.Response options for each dimension were: No problem, Slight problems, Moderate Problems, Severe Problems and Unable/extreme problems.To calculate the utility values, we used an Australian value set which was developed using a EuroQol-endorsed standardised protocol (Norman et al., 2023).Possible utility values range from −0.301 (worse than death) to 1.0 (perfect health) with scores <0 considered to represent a health state 'worse than dead'.Given no Australian population normative values have been published using EQ-5D-5L utility scores derived from this Australian value set, utilities for our participants were also calculated using a UK value set and compared to norms from a South Australian population (Devlin et al., 2018;McCaffrey et al., 2016).
The EQ visual analogue scale (EQ-VAS) was used to assess general health on a scale from 0 ('worst health you can imagine') to 100 ('best health you can imagine').Participants were instructed to move the slider online or mark an X on the paper-based vertical scale to 'indicate how your health is TODAY'.
The NEI-VFQ-25 is a 25-item instrument with 12 domains (Mangione et al., 2001).Each item has five or six response options and was scored from 0 (representing poorer visual function) to 100 (better function).As described in the instrument manual, items with nonmissing values in each subscale were averaged to calculate the subscale score and the composite NEI-VFQ-25 score was derived as the average of the subscale scores, excluding the general health question (Mangione, 2000).
Participants also completed the 22-item Attitudes to Gene Therapy-Eye (AGT-Eye) (Mack et al., 2021) and the 22-item Patient Attitudes to Clinical Trials (Jenkinson et al., 2005) questionnaires.Results from these two questionnaires have been published previously and have not been included in the current analyses (Mack et al., 2023).Each instrument was presented on a separate webpage and valid responses to all items were required for each page to be marked as complete.Participants could return to previous instruments to change responses if required.

| Statistical methods
Participants who completed all sections of the survey and had non-missing data on demographics, EQ-5D-5L and composite NEI-VFQ-25 scores were included in the analyses.Given the average time to administer the NEI-VFQ-25, participants who completed the NEI-VFQ-25 online in ≤90 s were deemed unreliable and excluded from analyses (Mangione, 2000).
Participant characteristics and responses to each EQ-5D-5L dimension were compared according to the type of retinal involvement (central vs. widespread) via Fisher's exact test.Given the skewed distribution of utility and EQ-VAS values, distribution was summarised within categories of participant characteristics as median and interquartile range (IRQ); mean and standard deviation (SD) were also given to allow comparisons to previous studies.These values were not given for categories with <4 respondents due to disclosure risk.Instrument scores were compared between categories of participant characteristics via median (quantile) regression, adjusting for age as a continuous variable and gender.These covariates were chosen a priori as they have been identified as potential predictors of health-related quality of life (McCaffrey et al., 2016).
Associations between instruments were visually inspected using locally weighted regression (LOWESS) plots and quantified via Spearman's correlation coefficient.As recommended for examining associations with health-related quality-of-life measures when heteroscedasticity is present (Austin, 2002;Payakachat et al., 2009), the censored least absolute deviations estimator was used to assess the association between NEI-VFQ-25 composite score and median EQ-5D-5L utility value.An upper limit of 1 was specified as the censoring point and 1000 bootstrap replications were performed to produce bias-corrected confidence intervals.
The sample size was not selected to detect a prespecified effect size and analyses were considered exploratory; thus, no adjustment has been made for multiple testing (Rubin, 2017).Statistical analyses were conducted using Stata/BE v18.0 (StataCorp, College Station, TX).

| EQ-5D-5L dimensions
The majority of respondents reported no problems with self-care (n = 520, 80.4%) and no pain or discomfort (n = 397, 61.4% see Table 2).However, almost half reported at least moderate problems doing usual activities (n = 309, 47.8%) and over half the participants reported at least slight anxiety or depression (n = 381, 58.9%).Although the majority of people with IRDs associated with widespread retinal involvement reported at least slight problems walking around (n = 351, 69.9%), less than half of those with central retinal conditions reported this problem (n = 66, 45.5%).
Two hundred and four combinations of dimension responses were observed, with the most common combination of [1 1 1 1 1] (i.e.no problems in any dimension) being observed in only 74 (11.4%) participants (see frequency of health states Table S1).Dimension responses stratified by age group and type of IRD are presented along with previously published population values in Table S2.

| EQ-5D-5L utility values
EQ-5D-5L utility values ranged from −0.18 to 1.00 and were skewed towards higher scores (indicating fewer problems) with a median of 0.88 [IQR 0.76, 0.96] and mean of 0.82 ± 0.21.The lowest EQ-5D-5L values tended to be observed among participants with IRDs associated with widespread retinal involvement (adjusted p = 0.029, see Figure 1 and Table 3).The distribution of utility values within age categories stratified by type of IRD and gender are presented in Tables S3  and S4

| EQ vertical visual analogue scale (EQ-VAS)
The EQ-VAS score for general health ranged from 0 to 100 with a mean of 73.1 ± 18.2 (see comparison to population norms stratified by age and gender in Table S6).
The median [IQR] was slightly higher among those with predominantly central ) compared to those with widespread involvement (75 [61-85]) even after adjusting for age and gender ( p = 0.003, see Table S7).

| DI SC US SION
In this large survey-based study, we found that EQ-5D-5L utility values vary widely among Australian adults with IRDs.This variation was observed even between people with similar levels of self-reported visual function.Thus, the EQ-5D-5L may not be the most suitable instrument for deriving utility values from studies of ocular interventions if it is not sufficiently sensitive to difference in visual outcomes.Lower levels of mobility and higher levels of anxiety/ depression were described among people with IRDs associated with widespread retinal involvement compared to those with IRDs that predominantly affect the central vision.Thus, the lowest health utility values were observed for participants with widespread retinal involvement.
Impaired vision has the potential to impact mobility, usual activities, self-care and anxiety/depression, and this was reflected in the correlation observed between the EQ-5D-5L and NEI-VFQ-25.However, the EQ-5D-5L does not include a vision-specific dimension and a substantial number of participants reported no difficulties in any of the EQ-5D-5L dimensions, even when reporting reduced vision-related quality of life.This suggests that EQ-5D-5L utility values may not be sensitive to the reduced quality of life associated with all types and stages of vision impairment due to IRDs.This lack of discrimination has also been noted when assessing the EQ-5D-3L among people with neovascular age-related macular degeneration (Payakachat et al., 2009), primary open-angle glaucoma (Browne et al., 2012), diabetic macular oedema (Kay & Ferreira, 2014) and other causes of low vision (Lin & Yu, 2012;Macedo et al., 2022), along with a lack of responsiveness to low vision rehabilitation (Malkin et al., 2013).To our knowledge, the EQ-5D-5L and NEI-VFQ have not been compared in people with IRDs to date.

| Comparison to previous research
On average, lower health utility values were observed among our participants than previously reported in the general population (McCaffrey et al., 2016;Rentz et al., 2014).Among people with widespread retinal involvement, EQ-5D-5L utility values were consistently lower than reported in the same age bracket of the general population, with differences ranging from 0.10 to 0.18.Among people with predominantly central IRDs, a greater difference from population norms was observed for those younger than 35 years old (≥0.17 difference) than for older participants (−0.01 to 0.08 difference).This divergence from population norms is more modest than reported in a recent study that measured health utility among Australian adults with clinically diagnosed IRDs and a similar age profile (Schofield et al., 2022).They found mean Assessment of Quality of Life 8-Dimension (AQoL-8D) utility values of 0.58, 0.22 below the population norm.This may in part be due to the inclusion of a vision-specific and psycho-social items in the AQoL-8D, which capture the reduced quality of life due to visual impairment more directly than the items in the EQ-5D-5L (Richardson et al., 2014).
EQ-5D-5L utility values used in several economic evaluations of voretigene neparvovec-rzyl for RPE65mediated IRD originated from a health utility vignette study (Bhadhuri et al., 2022;Farmer et al., 2020;Johnson et al., 2019;Viriato et al., 2020).That study used expert clinician elicitation to derive utility values based on level of visual impairment defined by visual acuity and visual field rather than directly questioning patients (Lloyd et al., 2019).The vignette study assigned profound visual impairment (visual acuity 6/150-6/375 or visual field radius 2°-6°) a utility value of 0.52.This is similar to the mean observed for people in our study who were in the lowest category of NEI-VFQ-25 scores (0 to <25 points).The poorest health state (vision of hand movement only or no perception of light) was allocated a utility of 0.15 in the vignette study; 10 of our participants scored lower than this level.A later study elicited health utility values from members of the general public, using definitions of the same health states which had been expanded to describe the impact of retinitis pigmentosa on activities of daily living and social, family and work life (O'Brien et al., 2023).Mean utility values from that study were higher than those originally elicited from clinicians for four out of the five health states presented.Utility values used in other economic analyses of voretigene neparvovec-rzyl have been informed by the relationship between health utility and visual function observed in other retinal disease populations (Uhrmann et al., 2020;Zimmermann et al., 2019).Utility values ranged from 0.35 to 0.75 for these analyses (Huygens et al., 2021).
Average NEI-VFQ-25 composite scores of 45-47 have previously been reported among people with IRDs, similar but slightly lower than our findings (Schofield et al., 2022;Shojaei et al., 2022).A higher mean composite score (69) was observed following administration of the Japanese version of the instrument among adults with retinitis pigmentosa (also known as rod-cone dystrophy), a progressive IRD with widespread photoreceptor loss (Sugawara et al., 2011).We found the association between the NEI-VFQ-25 and EQ-5D-5L to be weaker than between the NEI-VFQ-25 and AQoL-8D utility values among people with clinically diagnosed IRDs (Pearson's ρ 0.77) (Schofield et al., 2022).The strength of that relationship likely stems from the similarity in wording between the vision-specific AQoL-8D item and the General Vision item of the NEI-VFQ-25, and from the inclusion of psycho-social items in both instruments.Among people with age-related macular degeneration, median EQ-5D-3L utility values have been estimated  to increase by 0.038 for every 10-point increase NEI-VFQ-25 composite score after adjusting for the NEI-VFQ General Health item (Payakachat et al., 2009), which was similar but slightly lower than our estimates for median EQ-5D-5L scores among people with predominantly central IRDs.

| Strengths and limitations
Strengths of this paper include the sample size, which is much larger than other studies that have previously captured health utility data among people with IRDs.
A variety of conditions were represented which allowed comparisons to be made between IRDs.However, IRD diagnosis was self-reported in this predominantly online study and we are unable to confirm the genotype or phenotype of participants.Furthermore, no clinical data on visual function was available.In addition, miss-classification of the location of retinal involvement is likely for some individuals, as certain IRDs manifest with central retinal involvement and/or vision loss for some people and widespread changes in others (Tanna et al., 2017).This study only included adults, 80% of whom were 40 years of age or older.This is in contrast to trial participants for current IRD gene therapies who have tended to be children, teenagers and younger adults (Britten-Jones et al., 2022;Britten-Jones, Jin, et al., 2022).
We assessed health utility using the EQ-5D-5L which has improved sensitivity and decreased ceiling effects over the EQ-5D-3L.We used a newly developed Australian value set to derive utility values from dimension responses and there may be subtle differences in how health states are valued in other countries (Clemens et al., 2014).Sensitivity analyses conducted using utility scores derived from a UK value set showed that the UK-derived scores were lower than the utilities derived from the Australian value set for most participants, although utility scores calculated with the Australian value set tended to be lower for people with extreme impairment in any of the five EQ-5D-5L dimensions.Although we compared average utility values derived from the UK value set between our participants and members of the general public in South Australia, normative values for the wider Australian population calculated using the Australian value set are not yet available (Norman et al., 2023).Given the height of the EQ-VAS was not standardised across different electronic device types, EQ-VAS scores may not be directly comparable to studies which have used paper administration (Jiang et al., 2021).
Although the official EuroQol version of the EQ-5D-5L does not have a vision-specific dimension, a 'bolt-on' vision item has been proposed (Gandhi et al., 2020;Longworth et al., 2014).Inclusion of that item may have increased the correlation with the NEI-VFQ-25 and more accurately captured differences in quality of life associated with impaired vision.Despite the NEI-VFQ-25 being among one of the most widely used patient-reported outcome measures used in ophthalmic research, suboptimal psychometric properties have been reported (Braithwaite et al., 2019).Newer vision-related quality-of-life instruments, such as the Visual Symptom and Impact Outcomes patient-reported outcome (Fischer et al., 2023) and the Michigan Retinal Degeneration Questionnaire (Lacy et al., 2021), and a computerised adaptive testing system (Prem Senthil et al., 2019) have been proposed for use among people with IRDs, and are likely to more accurately capture the disease-specific traits of interest.However, none of these have an associated utility algorithm and thus do not serve the function of the EQ-5D-5L.Most visionrelated quality-of-life instruments that are used as patient-reported outcome measures in vision research were not initially designed to be used in economic evaluations.Conversely, the Vision and Quality of Life Index (VisQoL) (Peacock et al., 2008) and the Visual Function Questionnaire-Utility Index (VFQ-UI) (Rentz et al., 2014) are examples of instruments designed to generate utility values that could be used in the calculation of quality-adjusted life years for cost-utility analyses.However, neither of these instruments directly captures some important impacts of vision loss such as anxiety, a significant factor in IRD (Sainohira et al., 2018).The AQoL8D (Richardson et al., 2014) has been proposed as an instrument to derive utilities that would be more suitable for patients with vision disorders (Schofield et al., 2022).
Another strength is the direct assessment of healthrelated quality of life by people with IRDs rather than by healthcare professionals.However, non-random sampling was used for this survey.Despite online, paper and telephone options available, people with severe visual impairment or blindness due to their IRD may have been less likely to participate, biasing results towards people with better visual function.In addition, we recruited for the survey through patient support groups (Retina Australia, Vision Australia, UsherKids Australia, Cure Blindness Australia), the Australian Inherited Retinal Disease Registry and DNA Bank, and our ophthalmology colleagues.This may mean that people with IRD who are not currently engaged with support or medical services would not have heard about the survey, limiting responses from those more disengaged with the IRD community.

| Further research
Neither the EQ-5D-5L nor the NEI-VFQ-25 are diseasespecific and neither was developed for paediatric populations.Further investigation is needed to determine the most appropriate health-related and vision-related quality-of-life instruments to inform future economic analyses for IRD interventions (D' Amanda et al., 2020;Jayasundera et al., 2022).
Low vision-related quality of life and utility values were observed for many participants with widespread retinal involvement.This suggests that interventions that can successfully target the causative genes have the potential to dramatically improve outcomes that are important to patients.However, the impact of existing vision impairment and associated therapy on social and psychological well-being persist following treatment, even in the presence of functional improvement.
Longitudinal health utility data from people with IRDs is needed to assess whether assumptions about values used in recent cost-utility analyses are valid.Registries such as the Victorian Evolution of inherited retinal diseases NaTUral History Registry (VENTURE) are likely to be a rich source of information to inform future analyses (Britten-Jones, O'Hare et al., 2022;Britten-Jones, Jin, et al., 2022), but long-term data is also required from people who have undergone therapy for IRDs.

| CONC LUSIONS
Despite measuring different constructs, EQ-5D-5L health utility values were moderately correlated with self-reported visual function as measured by the NEI-VFQ-25 among adults with IRDs.However, health utility scores varied widely among people with IRDs who reported poorer visual function.Careful consideration should be given to the choice of instrument used to collect health utility data from participants with IRDs in future studies to ensure the full impact of vision impairment on quality of life is captured.

AC K NOW L E DGE M E N T S
Open access publishing facilitated by The University of Melbourne, as part of the Wiley -The University of Melbourne agreement via the Council of Australian University Librarians.

T
A B L E 2 EQ-5D-5L dimension responses according to type of retinal involvement.

F
Distribution of EQ-5D-5L utility values according to diagnosis.Utility values derived from an Australian value set.Each marker represents one participant.Boxes indicate median and interquartile range with whiskers extending to 5th and 95th centiles.n it is p ig m e n to s a C o n e − r o d d y s tr o p h y C o n e d y s tr O th e r w id e s p r e a d c o n d it io n S ta r g a r d t d is e a s e M a c u la r d y s tr o p h y O th e r c e n tr a l c o n d it io n n it is p ig m e n to s a C o n e − r o d d y s tr o p h y C o n e d y s tr o p h y O th e r w id e s p r e a d c o n d it io n S ta r g a r d t d is e a s e M a c u la r d y s tr o p h y O th e r c e n tr a l c o n d it io n Widespread retinal involvement Predominantly central Type of inherited retinal disease T A B L E 3 Distribution of EQ-5D-5L utility values derived from Australian value set.

F
Distribution of EQ-5D-5L utility values according to the National Eye Institute Visual Functioning Questionnaire (NVF-VFQ-25) composite score among people with inherited retinal diseases associated with widespread retinal involvement (n = 502) and predominantly macular involvement (n = 145).Utility values derived from an Australian value set.Grey line indicates locally weighted regression line of best fit.ρ, Spearman rank correlation.
T A B L E 1*p-value comparing characteristics between inherited retinal diseases associated with widespread retinal involvement to those with predominantly central involvement calculated via Fisher's exact test.